Cognitive Psychology and Emotional Disorders Second Edition
J. Mark G. Williams Fraser N. Watts Colin MacLeod Andrew Ma...
114 downloads
4726 Views
4MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Cognitive Psychology and Emotional Disorders Second Edition
J. Mark G. Williams Fraser N. Watts Colin MacLeod Andrew Mathews
JOHN WILEY & SONS 1997
Page v
Contents
About the Authors
vii
Series Preface
viii
Preface to the First Edition
ix
Preface to the Second Edition
xi
1. The Cognitive Approach to Emotional Disorders
1
2. The Information-processing Paradigm
16
3. Cognitive Impairments
44
4. Attention to Emotional Stimuli, I: Causes and Correlates
72
5. Attention to Emotional Stimuli, II: Mechanisms Underlying Bias
106
6. Memory
133
7. Thoughts and Images
169
8. Judgement
190
9. Schemata
210
10. Nonconscious Processing
231
11. Theoretical Overview
276
References
317
Author Index
377
Subject Index
393 Page vi Page vii
About the Authors
Mark Williams is Professor of Clinical Psychology and Director of the Centre for Medical and Health Sciences, University of Wales, Bangor. He trained as a clinical psychologist at Oxford, qualifying in 1976, and completed his doctorate there in 1979 on learned helplessness in humans. There followed periods as Lecturer in Applied Psychology at the University of Newcastle upon Tyne and as Scientist (later Senior Scientist) at the Medical Research Council's Applied Psychology Unit in Cambridge, until he took up his present appointment in 1991. Professor Williams' other books include The Psychological Treatment of Depression (1992) and Cry of Pain: Understanding Suicide and Self Harm (1997).
Fraser Watts was for many years a Senior Scientist at the MRC Applied Psychology Unit in Cambridge, and initiated the Unit's programme of research on the cognitive psychology of emotional disorders. He is now Starbridge Lecturer in Theology and Natural Science in the University of Cambridge, and a Fellow of Queens' College.
Colin MacLeod trained initially as a cognitive psychologist at Oxford University, completing his DPhil under the supervision of the late Donald Broadbent within the area of natural language processing. Developing an early interest in the patterns of cognitive distortions associated with emotional abnormality, he then went on to undertake formal training in clinical psychology at the Institute of Psychology, London, in the early 1980s. After spending some years working as a Research Fellow within the University of London, and as a Clinical Psychologist within the UK Health Service, he relocated to Australia in the late 1980s. At present he is Professor of Psychology at The University of Western Australia, where he also has been Head of the Psychology Department since 1994.
Andrew Mathews is a Senior Scientist at the Medical Research Council's Applied Psychology Unit in Cambridge, and visiting Professor of Psychology at the Institute of Psychiatry, University of London. After completing his PhD in London, he held a research post at the Department of Psychiatry in the University of Oxford, moved to become Professor of Psychology at St George's Hospital Medical School in London, and then worked for several years in the United States before returning to Britain to join the Applied Psychology Unit. His research on psychological treatment mechanisms, and cognitive factors involved in anxiety, have been published in over a hundred papers, chapters and books; and he is currently Co-Editor of the journal Cognition & Emotion.
Preface to the First Edition
This book has grown out of regular meetings which have taken place over the last few years between the authors to discuss matters of mutual research interest. The study of emotional disorders from the standpoint of cognitive psychology is very new, and it is difficult to know the appropriate time to put down in writing the current state of the evidence and current speculations about causation and mechanism. The disadvantages are obvious: the field has not yet reached a steady state. The danger is that today's speculation will seem naive tomorrow in the light of further evidence. Yet already the field has some coherence and clarity. First, it is now more obvious what are the tractable questions (for example, to what extent do different types of patient differ in the types of bias in information-processing they show?) Second, it is clearer which experimental paradigms are useful, and which are less so for examining such biases (for example, depth of processing paradigms have yielded a great deal of evidence, but lexical decision paradigms have not). The experiments which allow us to come to these preliminary conclusions are scattered throughout the literature. This book attempts to draw them together in order to be able to observe the regularities that may be emerging. Having drawn them together the book aims to make explicit what the connections may be between them.
We shall concentrate on disorders of anxiety and depression, not only because they are the commonest forms of psychopathology, but also because it is these emotional disorders which have been most studied by experimental cognitive psychologists. Other disorders, such as mania and hypomania, aggression and psychopathy, have not been subject to so detailed and systematic an enquiry.
Following a general introduction to the integration of cognition and emotion (Chapter 1) and to the information-processing paradigm (Chapter 2) the book reviews the core areas where experimental cognitive paradigms have been applied to emotional disorders. Consideration is given in Chapter 3 to research on cognitive impairmentshow emotional disturbance affects performance on neutral tasks. This represented the earliest attempt to apply information-processing concepts, and laid the groundwork for much of the later research.
Page x
The more recent work has been largely concerned with investigating the processing of emotionally
valent information. Chapter 4 examines the research on attentional biases, and Chapter 5, that on memory biases. Such biased processing has often been explained within schema theory, and this contention is critically examined in Chapter 6.
Chapters 7 and 8 move away from considering underlying biases and focus on how the phenomenology of emotionally disordered individuals can best be understood. In Chapter 7 this involves a discussion of current theories of mental imagery, and their implications for emotional pathology. In Chapter 8, the way in which normal heuristics and biases contribute to judgements made by patients with emotional disorders is considered.
The final two chapters give a theoretical overview in order to integrate the research. Many of the findings discussed in the book raise the issue of the relation between conscious and nonconscious processing. Chapter 9 therefore discusses this in detail. Its conclusion is that, though there can be a dissociation between conscious and nonconscious processing, experimental investigation of the latter is not precluded. The final chapter builds upon this conclusion, suggesting the need to distinguish between two different types of processing operating at encoding (an automatic priming process and a strategic elaboration process) with a parallel distinction at retrieval (a 'remembering without awareness' aspect and a 'strategic recollection' aspect). Evidence that different emotional disorders affect different aspects within this processing model is reviewed.
In writing this book, there has been much discussion of all the chapters by all the authors. However, specific members of the group took main responsibility for particular chapters as follows: Chapters 1, 3 and 7, Fraser Watts; Chapters 2, 6 (with AM) and 9, Colin MacLeod*; Chapters 6 (with CM) and 8, Andrew Mathews; Chapters 4, 5 and 10, Mark Williams.
We gratefully acknowledge helpful discussions at various stages in the production of the book with several colleagues: Alan Baddeley, John Duncan, Phil Johnson-Laird, Andrew MacLeod, Anthony Marcel and John Teasdale. Thanks are also due to Tim Dalgleish for help with proof-reading and indexing the manuscript. Much of our own research which is reported in the book has been supported by the Medical Research Council of Great Britain.
MARK WILLIAMS FRASER WATTS COLIN MACLEOD ANDREW MATHEWS August 1987
*Colin MacLeod is now at the Department of Psychology, University of Western Australia, Nedlands, W. Australia.
Page xi
Preface to the Second Edition
At the time of writing the first edition, the field of Cognition and Emotion was relatively new. The book helped mark the territory of this new field, defining the areas where experimental psychology could expect to improve our understanding of emotional disorders, and where cognitive models of normal functioning might be informed by data on how emotional systems break down.
Since its publication, there has been an explosion of interest in using the experimental techniques of which we wrote. In preparing this edition, we found that in the field of mood and memory bias, there had been a further 77 studies to take into account; for the chapters on attentional bias, a further 72 studies. In the field of nonconscious processing there were no less that 337 new references to take into account. Many of these studies were relevant to the hypotheses we advanced in that book. We had suggested that previous studies on cognitive aspects of emotion had been too dependent on questionnaire measures of cognition. Though adequate for some purposes, a questionnaire cannot assess aspects of cognition of which the person is not aware. We reviewed a great deal of evidence suggesting that individuals are often not aware of the heuristics and biases that affect their everyday thinking, feeling and behaviour. This conclusion was grounded in evidence from studies of nonconscious processing, from studies of the way people make judgements, and from studies of the concept of schema (a concept which had itself been widely used but loosely defined in clinical cognitive work). In this edition we revisit each of these domains in turn. The further work that has been done justifies our early optimism that the application of experimental cognitive psychology would yield valuable insights into emotion and its breakdown.
A salient message of the first edition was that the field should move away from generalist theories of cognition and emotion (such as schema or network theories) that failed to distinguish between different aspects or stages of information processing, and failed to distinguish between different types of emotion. Using the example of differences between anxiety and depressive disorders, we showed how biases in information processing in one disorder were difficult to demonstrate reliably in the other. We hypothesised that this might result, in part, from different emotions having their predominant biasing effects on different aspects of processing, anxiety mainly affecting relatively automatic processes of integration or priming, and depression mainly affecting relatively strategic processes of elaboration. Now, eight years on, we are in a better position to see if that early heuristic model has been confirmed or refuted, and the current position is laid out in the final chapter.
As in the first edition, we have collaborated in the writing of this book, but each of us has taken special responsibility for particular chapters: FNW for the Introductory chapter, together with those on Cognitive Impairments and on Thought; CM for the chapters on Information Processing, and Nonconscious Processing; AM for the chapters on Judgement and Schemata; and JMGW for the chapters on Attention, Memory and the Theoretical integration in the final chapter. We thank Gordon Brown, Tim Preece, Mike Oaksford, Tim Dalgleish, Anne Richards, Brendan Bradley, Karin Mogg, Rich McNally, Andy MacLeod, John Teasdale and John Duncan for help in discussing many aspects of this now burgeoning literature on Cognition and Emotion.
Particular thanks go to Alan Baddeley, who recently retired from being Director of the Medical Research Council's Applied Psychology Unit in Cambridge. It was he who in the early 1980s took the decision to gather together a group of scientists that would address the under-researched links between experimental cognitive psychology and emotion. The field of Cognition and Emotion, as a coherent domain of study, was largely Alan Baddeley's vision, and his tireless enthusiasm nourished its growth. In token of our gratitude, we are pleased to dedicate this book to him.
Page 1
Chapter 1 The Cognitive Approach to Emotional Disorders
Recent years have seen the development of a body of psychological work, both empirical and theoretical, on the application of cognitive psychology to the investigation of emotional disorders. Our intention in this book is to review that body of work, and to consider its clinical and scientific implications. We hope that clinical psychologists whose primary concern is with the assessment and treatment of patients with emotional disorders will value the opportunity to appraise a body of fundamental research on these conditions that is now well developed and has attracted considerable interest. However, we believe that the scientific work with which we are concerned is potentially of broader interest. The cognitive psychology of emotional disorders is only one aspect of an increasing general interest in the connection between emotion and cognition. Also, psychologists with a primary interest in the information-processing paradigm have become more interested in applying concepts and paradigms to everyday phenomena, and emotional disorders represent one such application.
The purpose of this introductory chapter is to set the cognitive investigation of emotional disorders in the broader context of cognitive approaches to emotional phenomena. This will proceed through
various stages:
First, we shall place our own interest in the cognitive investigation of emotional disorders in the context of recent general interest in the connection between cognition and emotion. We shall also present a brief conceptualisation of emotion and emotional disorders, raising issues which are pertinent to cognitive investigations.
Next, we shall describe some of the cognitive processing phenomena suggested by clinical observations, and indicate the experimental paradigms available for their investigation. In this context we shall introduce the two cognitive theories of emotional disorders of greatest importance in the development of the field, Beck's schema theory and Bower's network theory. Page 2
We shall emphasise the need for theories to be able to make differential predictions, for example to handle the differences between anxiety and depression, and discuss the value of distinguishing between different levels of the cognitive system in the way that recent hierarchical theories of emotion have done.
Cognition and Emotion
Discussions of the relationship between cognition and emotion tend to focus either on the effects of emotion on cognitive processes or on the role of cognitive processes in the genesis of emotional states, and these will be considered in turn.
Work on the effects of emotion on cognitive functioning can introduce 'emotion' at three different points. One concerns the materials used. Attention and memory for emotional materials diverge at many points from those that apply with neutral materials even in 'normal' subjects who are not emotionally aroused. The effects are often quite subtle, such as the well-known phenomenon reported by Kleinsmith & Kaplan (1963) that emotional materials may be less well remembered at short retention intervals but better remembered at longer intervals.
Emotion can also be introduced through the subject, and there are two possible ways in which this can happen. 'Normal' subjects can experience a temporary state of emotional arousal, or subjects can have chronic emotional arousal and preoccupations. Either 'state' or 'trait' emotion in subjects can affect cognitive processing, the clearest results being seen in subjects who have both, i.e. people with chronic emotional vulnerability experiencing acute emotional arousal. Cognitive processing in subjects with emotional disorders or in states of emotional arousal is generally less efficient, and the evidence relating to this will be reviewed in Chapter 3.
Finally, there can be interaction between the emotional content of materials and the emotional state or preoccupations of subjects. Most of the cognitive phenomena considered in this book fall into this category, and involve the use of experimental materials of relevance to patients with particular emotional disorders. It appears that the effects of emotion on cognition are particularly powerful in such paradigms.
Effects of cognitive processes on emotional states have also been studied. For example, the theories of Lazarus (1991) and Scherer (1984) emphasise that whether events lead to emotional reactions depends on how they are appraised. Scherer's theory postulates a series of five stimulus evaluation checks for (a) novelty, (b) intrinsic pleasantness, (c) goal/need significance, (d) coping potential, and (e) norm/self compatibility. Cognitive Page 3
representations, such as thoughts and images, can also affect emotion (see Chapter 7) and clinical theorists have often emphasised that there is a reciprocal relationship between cognition and mood, in which mood is affected by cognition as well as vice versa.
However, the assumption that cognition influences or determines emotion became controversial with Zajonc's argument that emotion is potentially independent of cognition (Zajonc, 1980, 1984; Zajonc, Pietromonaco & Bargh, 1982). Though the heat has gone out of the debate generated by Zajonc's paper, the issues it raises are still important. The claim that emotion can be independent of cognition did not go unchallenged (e.g. Lazarus, 1982, 1984; Mandler, 1982), and the debate was fuelled by different views about scientific theories and their verification. Zajonc suggested that Lazarus's view that emotional reactions are mediated by cognitive appraisal was effectively unverifiable because his definition of cognitive appraisal was so broad as to include phenomena that are unobservable. Lazarus regarded this as an unnecessarily positivist stance that did not sufficiently allow for scientific constructs which are tied, not to a single measure, but to a network of observations which jointly support the construct concerned.
It is now widely agreed that this debate depended, to a large extent, on how the key terms of 'cognition' and 'emotion' are defined, especially by the meaning of cognition (e.g. Watts, 1983; Leventhal & Scherer, 1987). One empirical fact that Zajonc regarded as a 'crucial' piece of evidence for his position (Zajon, Pietromonaco & Bargh, 1982, p. 216) concerned the phenomenon of previously exposed stimuli being preferred to those that have not been exposed. Zajonc was impressed by the fact that this does not depend on 'awareness of recognition of the object as familiar' and took this as support for his position; but clearly at best it only shows that affective changes are not necessarily based on conscious awareness. There would thus be fairly wide support for a reformulated version of Zajonc's thesis that emotion can be independent of conscious cognitive processes (Mandler, 1982). Some of the evidence to show that emotional reactions can be produced by stimuli of which people are not consciously aware will be reviewed in Chapter 10. However, to conclude from this that 'cognition' is not involved in the production of emotional reactions is to use 'cognition' in a limited and idiosyncratic way.
The general debate about the relationship between emotion and cognition led to a parallel debate in the clinical literature. Rachman (1981) accepted Zajonc's conclusion about the relative independence of affect and cognition and drew out its clinical implications. Though noting that Zajonc had not suggested that affect is wholly independent of cognition, Page 4
the clinical conclusions that Rachman drew seem to go beyond what followed legitimately from the actual evidence Zajonc marshalled. Among these is the proposition that behaviourally based treatments should be preferred to cognitive ones. This might be correct, but it does not follow from Zajonc's thesis. The fact that affect can be independent of conscious cognition justifies caution about whether or not cognitive treatments for emotional disorders will be effective. However, it certainly does not by itself justify the conclusion that they will be ineffective; this is an empirical matter. It should also be noted that many forms of cognitive therapy, such as Beck's treatment of anxiety, use corrective experience as well as verbally mediated work on conscious cognition. Rachman's paper stimulated a subsequent debate in the clinical literature which is beyond the scope of this chapter, but which can be followed in Watts (1983), Greenberg & Safran (1984a), Rachman (1984), Mahoney (1984), and Greenberg & Safran (1984b).
One of the unfortunate legacies of debate about Zajonc is a tendency to regard cognition and emotion as parallel psychological systems that may or may not be relatively independent. This is by no means a necessary way of approaching things. One can regard 'cognitive' theories as a generally applicable way of theorising in psychology. Indeed any theory that includes a consideration of what used to be called mental processes is in this general sense a cognitive theory. If cognitive theories are seen as a general framework for formulating psychological theories of emotion or whatever else, there no longer appears to be anything particularly controversial in formulating a psychological theory of emotion in broadly cognitive terms. Whatever the involvement of different levels of cognition turned out to be, and that is an empirical matter, they could be captured within the general framework of a cognitive theory.
It is also important to emphasise that the cognitive approach does not exclude other approaches to emotional disorders, such as social or biological ones. For example, connections can be made between some of the cognitive dysfunction found in depression and the disruption of circadian rhythms (Healy & Williams, 1988). There are also indications that social context can have important effects on cognitive processing, and one of the encouraging developments of recent years has been the attempt to apply information-processing methodologies to social processing. There have also been significant attempts to integrate the cognitive and social perspectives, such as in Gotlib & Hammen's (1992) approach to depression.
The cognitive approach also promises to provide a broad integrative framework within which the effects of various approaches to the treatment of emotional disorders can be considered. Theories of the cognitive Page 5
features of emotional disorders have an obvious relevance to cognitive therapy, but they are also relevant to other approaches such as behavioural and psychodynamic ones.
For example, it is widely recognised that an important feature of the behavioural treatment of phobic anxiety is systematic exposure to the phobic stimulus. One of the issues that arises is how the patient processes the phobic stimulus: there is a danger of a patient showing 'cognitive avoidance' which is likely to interfere with the therapeutic value of exposure treatment (Foa & Kozak, 1986). The information-processing perspective is relevant to investigating the kind of processing that takes place during stimulus exposure.
Similarly, the information-processing perspective is relevant to psychodynamic approaches to emotional disorders. Because the informational perspective focuses primarily on processes that can proceed without phenomenal experience, it can provide a meeting point with the psychodynamic interest in non-conscious processes. The concepts of the unconscious in information-processing and psychodynamic psychology are, of course, not identical. The important issue is whether people remain unconscious of material, not just because of the informational limits of consciousness, but because they are motivated by anxiety to do so. There are currently signs of a growing integration of information-processing and psychodynamic psychology (Erdelyi, 1985; Wegman, 1985; see Chapter 10). It is beyond our purpose here to consider this issue in detail, but such trends at least suggest that an information-processing approach to emotional disorders should not be regarded as incompatible with a psychodynamic one.
Emotion and Emotional Disorders
A clear understanding of the role of cognition in emotion also depends on a clear conceptualisation of emotion. Important issues include: what are the range of phenomena to be included in 'emotion'?, what are the differences between emotions in the role of cognition?, and what are the differences between short- and long-term aspects of emotion?
Emotion is a multifaceted phenomenon and it has been widely accepted that comprehensive theories of emotion must include its different facets. Lang (1985) specified behavioural, physiological and cognitive components. Scherer (1984) proposed a similar model with five components: cognitive processing of stimuli, motivation, physiological activation, motor behaviour and subjective feeling state. The information-processing approach to emotion does not exclude non-cognitive components of emotion or regard them in any way as secondary.
Page 16
Chapter 2 The Information-Processing Paradigm Consider the following transcript from a hypothetical patient, MF, attending a psychology outpatient clinic as a GP referral: I feel so low and depressed that everything is just too much effort. I can't even manage very simple jobs without getting distracted or confused. I'm just unable to concentrate any more. And I feel afraid all the time too. Society is so dangerous nowadays. Everything I read in the papers is about some terrible accident, or about violence. It's not even safe on the streets everywhere I look there are dangers. But I can get anxious even when I'm just sitting quietly, thinking about nothing. Anxiety just seems to come out of the blue and sweeps over me. I am doing less and less nowadays, and seldom see friends any more. Whenever I'm out socially I get uncomfortable. People find it difficult to accept me now. It's clear that they either find me boring or embarrassing, so I prefer to avoid these situations . . .
What is wrong with MF? A simplistic answer may involve formulating a diagnosis for this patient, such as mixed anxiety and depressive neurosis. However, such a label does nothing to increase our understanding of his problems. Each abnormal symptom indicates that something about MF is functioning maladaptively, and we can only understand the nature of this patient's condition if we can accurately describe those fundamental processes which are giving rise to his specific problems. But what kind of model should we construct? A careful look at the transcript above will reveal a number of important cognitive operations which may potentially contribute to the emotional distress suffered by MF. For example, his description suggests the possibility not only of some general impairment in cognitive functioning, but also of processing biases, such as enhanced perceptual salience for threat-related stimuli, distorted probability judgements concerning the likelihood of dangerous events, or inferential biases in the way ambiguous situations are interpreted. Each of these possible types of processing biases, and more, will be considered in greater detail within the main chapters of this book. In this current chapter we offer a very general introduction to one particular approach to psychological modelling, based Page 17
on the information-processing paradigm, and indicate its potential for generating plausible hypotheses to explain various symptoms associated with the affective disorders. Before beginning, however, we should stress the necessary limitations of this overview. Cognitive psychology is a young and diverse discipline which is continually evolving. With each shift in emphasis new levels of explanation and corresponding new theoretical concepts emerge, sometimes only briefly before their utility is effectively challenged. For the purposes of this general introduction it would be inappropriate to focus upon the developing frontiers of cognitive science. Rather, we shall attempt to introduce some of the central themes which have traditionally characterised the information-processing paradigm and which have survived in some guise or other, throughout its various stages of development. This inevitably will result in a somewhat historic orientation, but should familiarise the reader with the kinds of enduring issues which are most pervasive within applied cognitive research. The Nature of Scientific Paradigms Theorists typically draw on a particular set of conceptual structures and processes, and rely upon a conventional series of assumptions, when formulating hypothetical models. In effect, they utilise the components of a pre-existing 'construction kit', which both facilitates model design and constrains the range of possibilities. The construction kit employed to develop any model
represents the scientific paradigm adopted by the model builder. A scientific paradigm is neither right nor wrong, and so we do not normally attempt to prove or disprove it. However, it can be judged more or less fruitful on the basis of its ability to produce models which can accommodate and predict relevant data. Indeed, Kuhn (1962) argues that major scientific advances are marked not simply by the emergence of innovative models, but by the adoption of new, more fruitful paradigms to generate such models. In its short history, the discipline of psychology has been marked by the development of several alternative paradigms. Psychoanalytical models construe man's activities as attempts to reduce tension between conflicting internal drives, moral values, and the external constraints imposed by reality. Behavioural models consider man primarily as a basic learning device, and try to account for current activity in terms of an individual's classical and instrumental conditioning history. Cognitive models, however, characterise man as a complex information-processing system, and attempt to describe human experience and activity in terms of the integrated operation of fundamental processing mechanisms which act Page 18
upon, and are themselves acted upon, by the flow of information through the system (cf. Massaro & Cowan, 1993). Importing the concept of information from communication science in the 1950s freed theoretical psychology from the constraints of S-R behaviourism. By demonstrating that even simple behaviours, such as profiles of choice reaction-time responses, were dependent upon such variables as the number of stimuli in the set (Hicks, 1952) stimulus frequency (Hyman, 1952), and the provision of context (Crossman, 1953), the behaviourist notion that all responses represent conditioned reactions to a discrete stimuli was discredited. This recognition that the brain deals with information rather than simply eliciting conditioned responses to specific stimuli, has allowed psychologists to discuss internal representations without commitment to any particular neurological or biochemical framework. Regardless of their physical nature, such internal representations can now be considered in terms of the type and amount of information which they contained. Over the past two decades, there has been a growing recognition that cognitive processes may play an important role in the development and maintenance of neurotic disorders. Thus, from the mid1970s clinical theorists often have proposed that idiosyncrasies in the way information is processed may lead to styles of thinking which characterise, and perhaps constitute, the affective neuroses (e.g. Beck, 1976; Beck, Emery & Greenberg, 1985). It is only in the more recent years, however, that such accounts have begun to exploit fully the rich potential offered by the informationprocessing paradigm to model the nature of these hypothetical individual differences (e.g. Eysenck, 1992, 1988; Teasdale & Barnard, 1993). The remainder of the current chapter introduces some of the important concepts which have historically characterised this paradigm. Capacity Limitations One common characteristic of all information-processing devices is that they are to some degree constrained by available capacity. Accordingly, information-processing models of the brain typically characterise it as a limited capacity system, and certain features of human experience and activity are considered to reflect constraints on available processing capacity. While it may be selfevident that we cannot do all things simultaneously with full efficiency, there are a number of ways in which capacity limitations can be conceptualised. One option has been to postulate that cognitive processes require 'resources' to operate (or 'effort' or 'attention': e.g. Kahneman, 1973; Shiffrin, 1976; Roediger, Knight & Kantowitz, 1977; Hasher & Zacks, 1979; Kahneman & Treisman, 1984), and that these Page 19
resources are allocated in continuous modulated quantities from some common undifferentiated reservoir. Subject and task parameters will determine the amount of resources required for a
particular process to operate. Processes which make relatively small demands upon the resource pool may therefore proceed simultaneously, but whenever the overall requirements exceed the capacity of the resource pool, this will not be possible. While this simple formulation of capacity limitation is appealing, and has been incorporated in a number of useful models, it is inconsistent with the observation that interference between tasks often appears to reflect task structure rather than task difficulty (e.g. McLeod, 1977, 1978; Kinsbourne & Hicks, 1978). That is, one task may cause severe interference when performed simultaneously with one other kind of task, yet little interference when performed simultaneously with a different task of equal difficulty. This has led a number of researchers to put forward structural accounts of capacity limitations (e.g. Keele, 1973; Fodor, 1983; Broadbent, 1984; Pashler & Johnston, 1989; McCann & Johnston, 1992; De Jong, 1993; Ruthruff, Miller & Lachmann, 1995). According to such formulations, capacity limitations reflect competition between tasks for specific information-processing mechanisms or structures necessary for performance. Tasks which need to use a common processing structure cannot be performed at the same time, and hence their execution necessitates time-consuming task switching, while tasks which do not require the use of common processing structures can efficiently proceed simultaneously. In recent years, many theorists have argued that one cognitive operation for which tasks commonly compete for access is the response selection mechanism. Consistent with this proposal, there now is considerable experimental evidence to suggest the presence of a response selection bottleneck within the cognitive system (e.g. Pashler, 1993, 1994; Carrier & Pashler, 1995). Certain alternative accounts of capacity limitations have integrated resource-based and structuralbased formulations into a common conceptual framework (e.g. Kantowitz & Knight, 1976; Kinsbourne & Hicks, 1978; Navon & Gopher, 1979a, 1979b; Sanders, 1979; Wickens, 1979, 1984). The suggestion is that tasks draw for their performance upon a number of structure-specific reservoirs of processing resources. That is, the processing system incorporates a number of mechanisms, each having its own capacity, and these structures can share or divide their limited processing resources between concurrent activities. This approach to modelling capacity constraints has become known as 'multiple resource theory', and has proved to be both powerful and popular in explaining a wide variety of dual task interference patterns, including those observed in the Page 20
processing of complex visual displays (e.g. Wickens & Boles, 1987), speech signals (e.g. Payne et al., 1994) and media broadcasts (e.g. Leigh, 1991). Wickens (1984) suggests that discrete reservoirs of cognitive resources can be differentiated along at least the three following dimensions: (a) stimulus characteristics, such as sensory modality; (b) internal codes, which may be structural, phonetic, semantic, etc.; (c) response mode, which may, for instance, be manual or verbal. There is indeed considerable evidence that resources can be differentiated on the basis of such factors as input modality (e.g. Baddeley & Liberman, 1980; Rollins & Hendricks, 1980), requirements for short-term storage (Baddeley & Hitch, 1974; Baddeley, 1986), means of response (e.g. Wickens, 1984) and even the hemisphere of the brain primarily involved in processing an input (e.g. Friedman et al., 1982; Spielman, 1992). A rather different approach has been taken, however, by researchers who attempt to characterise capacity limitations in terms of neither resource limitations nor structural competition, but as a consequence of skill limitations in the parallel co-ordination of multiple cognitive processes (e.g. Hirst, 1986; Hirst & Kalmar, 1987; Kieras & Meyer, 1994; Meyer & Kieras, 1994, 1995). According to such accounts, limited skill in segregating two or more messages may result in inappropriate transformations to either or both inputs. Such 'cross-talk' between tasks leads to the breakdown in processing which characterises overloaded cognitive systems. Thus, these models
place primary emphasis upon the high level executive control procedures that operate to schedule the performance of multiple tasks by flexibly assigning perceptual processors, cognitive processors, and motor processors, and they attribute dual task interference to the inappropriate scheduling of such operations (cf. Meyer et al., 1995). Consistent with this central premise, there is evidence that the provision of explicit instructions to guide cognitive control processes can serve to reduce the degree to which cross-talk is observed between simultaneously performed tasks (e.g. Dutta et al., 1995). How might such concepts be employed to illuminate a possible basis for some of the symptoms reported by our hypothetical patient MF? One feature of his condition was that simple tasks seemed excessively arduous, and concentration difficult to maintain. The information-processing paradigm, by highlighting the limited capacity nature of the cognitive system, suggests the possibility that such symptoms may be a consequence of severe restrictions in available processing capacity. More specific hypotheses can be derived from resource-based, structural or skill-based accounts of such capacity limitations. Indeed, a considerable literature now exists which attempts to characterise the cognitive deficits associated with depression and anxiety in precisely this way (e.g. Ellis & Page 21
Ashbrooke, 1988; Eysenck, 1992; Sarason, Sarason & Pierce, 1994). Such issues will be discussed more fully in Chapter 3. Selective Processing Regardless of their precise nature, capacity limitations within the cognitive system will inevitably lead to competition, and frequently to selectivity in processing. This differential processing of simultaneous sources of information commonly is termed selective attention, and many researchers have devoted much work to the study of this cognitive ability. The reader is directed to reviews of this research provided by Johnston & Dark (1986), Kinchla (1992), and Plude, Enns & Brodeur (1994). While certain aspects of selectivity may reflect simple stimulus characteristics such as saliency or intensity (e.g. Muller & Rabbitt, 1989; Theewues, 1994), it also is clear that selectivity can be governed by internal control. The cognitive system can, through various priming techniques, be set to process information selectively in a particular sensory modality (e.g. Hillyard & Kutas, 1983; Oatman, 1984; Woods, Alho & Algazi, 1993), concerning a particular stimulus dimension (e.g. Walley, McLeod & Weiden, 1994), from a particular semantic area (e.g. Johnston & Dark, 1982, 1985; Teder, Kujala, & Naatanen, 1993; Dark, Vochatzer & Van Voorhis, 1996), or from a particular spatial and/or temporal domain (e.g. Snyder & Uranowitz, 1978; Eriksen, 1990; Yantis & Jones, 1991; Alho, Woods & Algazi, 1994). A number of important controversies have emerged within the field of selective attention. For example, some theorists have emphasised mechanisms of selectivity based on physical differences between relevant and irrelevant inputs, which operate early in the processing continuum (e.g. Broadbent, 1958). There is, however, considerable evidence that an irrelevant unattended message may undergo substantial semantic processing (e.g. Treisman, 1960; Deutsch & Deutsch, 1963; Broadbent, 1971, 1982). On the basis of such observations other theorists have emphasised mechanisms of selectivity operating late in the processing continuum on the basis of semantic information. Such early controversy over the locus of selectivity has, however, largely been replaced by the view that it is pervasive throughout the processing continuum (Erdelyi, 1974). Another underlying controversy, emphasised in reviews by Johnston & Dark (1986) and by Dark, Vochatzer & Van Voorhis (1996), but recognised by William James as early as 1890,
distinguishes those approaches which Page 22
construe selective attention as the cause of differential processing (e.g. Lundh, 1979; Duncan, 1980; Broadbent, 1982; Marcel, 1983b; Kahneman, Treisman & Burkell, 1983; Shiffrin, 1985) and those which construe it as a consequence of differential processing (e.g. Neisser, 1976; Hochberg, 1978; Johnston & Dark, 1986). The former approach requires that a selective attentional system operates upon the records of some separate, large-capacity, nonconscious system to determine which lines of processing are pursued to completion and/or result in a conscious representation. One problem with such an attempt is that it 'explains' mechanisms of selective attention by postulating a selective attentional system which itself possesses the capability to selectinviting the infinite regress characteristic of homunculus-based theories. The alternative view is that differential processing occurs through the mechanistic priming of processing structures. Such priming may be relative chronic, leading to semi-permanent processing advantages for certain domains of information, or may be a transient result of current cognitive activity. According to this account, selectivity in a current input is the passive result, through priming, of prior processing activity, and attention is simply the resultant phenomenal correlate of such priming. In recent years, several hybrid models have been proposed, which represent a compromise between cause and consequence accounts of selective attention. While these hybrid models endorse the view that selective attention results from the internal priming of certain informational domains, they also permit that such internal priming itself can be controlled by intention, meaning that selective attention can be both the cause of differential processing, and the consequence of intentional set (e.g. Duncan & Humphreys, 1989; Bundensen, 1990; Cave & Wolfe, 1990). Whatever account of selective attention is adopted, one major requirement for an informationprocessing model of any particular cognitive activity is to accommodate the observed patterns of selectivity which commonly occur. Why is one thing noticed more easily than another? Why is one particular interpretation more obvious than an alternative? Why is one item of information stored more efficiently than another? Why is one fact recalled more easily than another equally well-learned fact? Acceptable cognitive models must be capable of answering such questions. The emphasis which the information-processing paradigm places on selective factors in cognitive processing once again suggests a possible hypothesis to accommodate some of MF's symptoms. His comments appear to indicate a gross over-awareness of certain aspects of his environment; particularly those aspects which convey threat-related information concerned with real or imagined dangers. One could postulate that fundamental biases in the selective functions that determine which Page 23
inputs are most completely processed may underlie this subjective experience. There is a longstanding, though not unchallenged, view expressed within the cognitive literature that selective functions may operate to inhibit the subsequent processing of emotionally threatening inputs (e.g. Broadbent & Gregory, 1967; Bruner & Postman, 1947a, 1947b; Erdelyi, 1974). If such 'defensive' patterns of selectivity are inoperative for MF, or if his selective functions operate using different criteria which favour the processing of emotionally threatening information, then such information ultimately will become over-represented in his conscious experience. There is now considerable evidence to indicate that people do indeed show basic individual differences in the way their selective functions operate when dealing with emotionally threatening inputs (e.g. Byrne, 1961; Miller & Mangan, 1983), and a number of studies now suggest that particular patterns of selective processing, favouring affectively negative information, may underlie various features of anxiety and depression (cf, MacLeod, 1990; MacLeod & Mathews, 1991; Mathews & MacLeod, 1994).
Activation vs Inhibition For many years, information-processing models of selective attention endeavoured to explain observed effects by postulating the presence of internal mechanisms that can operate to increase the activation of selected mental representations, while leaving activation levels of other representations unchanged. According to such selective activation accounts, the representations of unattended information, including stimuli that individuals may wish to ignore, simply are not activated. Across the past decade, however, there has been growing recognition that the cognitive system also incorporates mechanisms capable of selectively inhibiting the internal representations of mental constructs, and inhibition-based models of selective processing have become increasingly influential in recent years (cf. Fox, 1995; May, Kane & Hasher, 1995). The study of inhibitory processing has its principal origins within the priming literature. It has long been known that the processing of a target item, such as a word or a picture, is facilitated when either this same item, or a related item, has been processed shortly before its presentation. This robust phenomenon has become known as the 'priming effect' (e.g. Meyer, Schvaneveldt & Ruddy, 1975; Warren & Morton, 1982). In a series of experiments, Tipper compared the effects of presenting the initial priming stimuli under a condition that required subjects to attend to this item, or under a condition that required subjects to actively ignore the ge 24
item. When the prime was attended, Tipper found the standard priming effect, characterised by the facilitated processing of the related target item. However, when the prime was ignored, Tipper observed not just the absence of such facilitation, as would be expected if the representation of the ignored prime simply had not been activated, but a substantial impairment in the processing of the related target, suggesting that the representation of the ignored prime must actually have been inhibited (e.g. Tipper, 1985; Tipper, Brehaut & Driver, 1990). Since this early demonstration of the 'negative priming effect', such inhibitory mechanisms have been incorporated not only into models of selective attention (e.g. Neumann & DeShepper, 1992), but also into models of language processing (e.g. Gernsbacher & Faust, 1991; Simpson & Kang, 1994), memory functioning (Anderson, Bjork & Bjork, 1994; Anderson & Spellman, 1995), reasoning (e.g. Spellman & Holyoak, 1993) and executive control procedures (e.g. Houghton & Tipper, 1996). Therefore, it is possible that the unusual patterns of selective processing reported by our patient MF may arise either because of idiosyncratic patterns of representational activation, or because of certain deficiencies in the inhibition of certain mental representations. Early cognitive theories of the selective processing idiosyncrasies associated with anxiety and depression most commonly have implicated activation mechanisms. For example, Bower's influential network model of emotion and cognition attributes such selective processing to the spread of activation within memory, originating from an activated emotion node but serving to prime other nodes within semantic memory that contain mood-congruent representations (Bower, 1981). In contrast, more recent accounts of cognitive abnormalities in clinical populations often have implicated impairments in the inhibitory mechanisms that govern cognition. There now is considerable empirical evidence to suggest that deficiencies in cognitive inhibition may indeed contribute to the reductions in processing efficiency observed in some clinical populations (e.g. Merrill, Cha & Moore, 1994; Sullivan, Faust & Balota, 1995). Furthermore, it has been observed that the patterns of intrusive thinking that characterise certain anxiety conditions, such as obsessivecompulsive disorder, are associated with impairments in the functioning of those cognitive mechanisms underpinning inhibitory processing (e.g. Enright & Beech, 1990, 1993a, 1993b). Stages of Processing
So far we have shown that cognitive models characterise the brain as an overcrowded, limited capacity system designed to process the most Page 25
important aspects of the available information. We have not yet discussed how this processing might take place. Numerous models have been proposed to account for many different kinds of cognitive activity. Let us consider, however, some of the common features which often appear in models developed from the information-processing paradigm. The cognitive approach typically is reductionistic. Attempts to formulate a model for any complex process usually proceed by identifying and organising the simpler component subprocesses which are involved. The simplest models which emerge can be characterised as linear sequences of independent processing stages. Each stage accepts information from the preceding stage, conducts some operation on it, and produces an output to be acted upon by the following stage. One attraction of this stages-ofprocessing approach lies in its potential to reveal basic, invariant, component subprocesses which may constitute the building blocks of cognition. If these can be exhaustively identified, it has been argued, then models of higher order processes may simply involve assembling those invariant subprocesses in an appropriate sequence, much like constructing a complex circuit from simple electrical components (e.g. Sanders, 1980, 1990). Given this incentive it perhaps is not surprising that a great deal of research effort has been directed towards identifying the component stages of processing which underlie complex cognitive operations. Frequently, this has involved the application of additive factors and subtraction methodologies to reaction-time data (e.g. Sternberg, 1969, 1984; Pachella, 1974). The additive factors method involves introducing various experimental manipulations to a particular task, which plausibly will increase the time taken to perform the whole task. If two different manipulations produce additive effects on performance time then, it is argued, they must affect different processing stages, whereas if their effect is interactive they are considered to influence the same stage. By imposing a large number of such manipulations on any task, proponents of this approach endeavour to identify the number of processing stages involved and, by considering the kinds of manipulation which affect each, to delineate the types of processing that these stages most plausibly reflect (e.g. Miller, 1988; Roberts & Sternberg, 1993). The subtraction method can be used to refine further an already formulated stages-of-processing model by employing task variants which eliminate the need for one particular theoretical stage, and calculating the duration of this 'subtracted' stage from the magnitude of the time saving (e.g. Theios & Amrhein, 1989). The precise influence of subsequent novel task manipulations can then be predicted on the basis of the model and these parameters. Both methods are based on important assumptions which are vulnerable to criticism, and therefore the techniques have proved controversial (cf. Page 26
Broadbent, 1984; Smid et al., 1991; Ridderinkof, van der Molen & Bashore, 1995). The most important assumptions are that information should be processed through stages in a strict serial order, and that the functioning of each stage (in particular its temporal duration) must be entirely independent from the processing outcomes of any other stage. As will be argued later in this chapter, there are plausible alternatives to both these possibilities. While models based on strictly linear, and fully independent, processing stages may be vulnerable to criticism, the principle of reducing complex mental operations to discrete component processes has many attractions, and has remained a pervasive characteristic of cognitive research. Contemporary theorists commonly conceptualise the cognitive system as a collection of processing modules, each module concerned with a particular type of computation or transformation (e.g. Fodor, 1983; Johnson-Laird, 1988; Shallice, 1994). Some modules are
thought to receive information directly from the environment, whereas others are assumed to take their inputs from other modules. Some are considered to be relatively self-contained, or 'cognitively impenetrable', while others may be under volitional control. The ways in which such modules are organised will determine the cognitive functioning of the individual, though it seems likely that the range of organisational possibilities may extend beyond simple linear sequences. Identification of fundamental component operations increases the potential specificity of hypotheses concerning the cognitive basis of MF's symptoms. We have suggested that his concentration difficulties may reflect restrictions in available processing resources, and that his preoccupation with threat-related information may arise from biased selection procedures, themselves perhaps reflecting idiosyncratic patterns of activation or inhibition across cognitive representations. Perhaps we can also identify the fundamental processing stages which may be involved in the mediation of these cognitive idiosyncrasies. Indeed, attempts to localise the processing operation underpinning these cognitive characteristics of anxiety and depression have met with some success. Many theorists now acknowledge the strong likelihood that the cognitive impairments experienced by anxious and depressed individuals may reflect a specific reduction of processing capacity within the working memory subsystem, a component of the cognitive system which serves as a temporary store for intermediate processing results (Ellis & Ashbrooke, 1988; Darke, 1988; Eysenck & Calvo, 1992). There also is evidence that, for anxious individuals, the selective biases which favour the processing of negative information may be located specifically within early perceptual stages of processing, while for depressed individuals these biases may be restricted to the later stages of processing involved in storage and Page 27
retrieval operations (cf. Mathews & MacLeod, 1994). Thus, the stages-of-processing framework supports a level of analysis capable of shedding further light on the possible cognitive basis of MF's symptoms. However, one should be aware of the potential problems which accompany too rigid an adherence to the simplistic application of the stages-of-processing approach. For example, there is the problem of determining which processes represent the invariant building blocks of cognition. Just because two separate types of complex process each involve, let us say, a binary decision stage, this does not mean that this component binary decision process is identical in both tasks. The subtraction method sometimes can be applied usefully in such situations, to determine whether the duration of this operation is comparable in both tasksa necessary but insufficient condition for the argument that the component process is identical in each task. A sophistication of this approach has been suggested by Sternberg (1977), which involves using the subtraction method, but also measuring individual differences in the time taken to perform the component stage. If the same pattern of individual differences is replicated when examining the relevant processing operation in both tasks, then this adds considerable strength to the argument that the same component process is implicated in the two cases. More recently, advances in brainimaging technology have provided another avenue for the study of processing stages. Using measures such as event-related brain potentials (ERPs), brain electrical source analysis (BESA), and regional changes in cerebral blood flow (rCBF), researchers have delineated the precise patterns of neural activity associated with hypothetical processing stages such as stimulus preprocessing, feature extraction, identification, response selection, motor programming and response execution (cf. Mulder, Wijers, Lange, Buijink, Mulder, Willemsen, & Paans, 1995). The identification of identical neural correlates, when subjects perform particular components of two different tasks, can provide converging support for the claim that the same fundamental cognitive process has been employed in each case. Discrete vs Continuous Processing
The classical stages of processing approach to psychological modelling assumes that, while the initiation of later stages will be logically contingent upon the success of earlier processing stages, each stage of processing functions discretely. That is, an early stage first acts upon its input and provides its output, and only then does the next processing stage commence, by receiving this output and acting upon it. The assumption Page 28
that processing stages operate discretely, in this manner, is central to both additive factors and subtraction methodologies. However, a number of theorists have questioned the view that one processing stage must be completed before a contingent second processing stage can start. An alternative possibility is that the functioning of processing stages may overlap in time, with each stage constantly processing the continuously available output from the preceding stage (cf. Posner & McLeod, 1982; Miller, 1993; Massaro & Cowan, 1993). For example, Norman & Bobrow (1975) suggest that the continuous output of each processing stage may be a set of quantities, each indicating the current probability that one of a number of possible conclusions about the input is correct. If so, then a subsequent processing stage can begin to make use of this output even before the earlier stage is completed. Of course, the accuracy (and possibly the processing rate) of the output from later stages would remain contingent on the informativeness of the changing output from earlier processing stages, but this kind of model considers the operation of processing stages to be continuous rather than discrete. Models implicating the continuous, rather than the discrete, operation of logically contingent processing stages have been termed parallel contingent accounts (Turvey 1973), cascade accounts (McClelland, 1979), or continuous flow accounts (Eriksen & Schultz, 1979). This hypothetical relationship between processing stages is implicit in various theories of language comprehension (e.g. McClelland & Rumelhart, 1981), pattern identification (e.g. Selfridge, 1959; Turvey, 1973; Massaro & Cowan, 1991), word recognition (e.g. McClelland, 1976; Henderson, 1977), memory retrieval (e.g. Wickelgren, 1976; Ratcliffe, 1988) and motor control (e.g. Miller & Hackley, 1992; Osman et al., 1992). McClelland (1979) has reported an analysis of cascade models, which suggests that additive factors and subtraction methodologies may not always be capable of delineating stages within these types of theoretical account. For example, he has demonstrated that experimental manipulations which increase processing difficulty by influencing the asymptotic output of particular processing stages all may produce interactive effects, even when they affect different stages. More recent simulation-based research has shown that the degree to which the additive factors methodology is capable of identifying processing components within overlapping stage models actually depends critically upon the precise parameters that are assumed to govern the transmission of information from one stage to the next (e.g. Miller, van der Ham, & Sanders, 1995). Therefore, this recognition that information processing within logically sequential stages may be performed continuously rather than discretely has served to enrich the potential complexity of stages-of-processing models, to extend their potential diversity, and to increase the need for precision in their specification. Page 29
Serial vs Continuous Processing Traditional stages of processing accounts clearly assume that the cognitive system processes information sequentially, or serially. Overlapping stage models depart from the strong version of this assumption, but still emphasise the sequential transmission of information from one stage to the next in a serial manner. However, while certain experimental findings suggest that serial processing may indeed take place within the cognitive system, there also is now very substantial
evidence to indicate that much information processing proceeds in a parallel manner, with the encoding of multiple perceptual inputs and the retrieval of multiple memory representations taking place simultaneously rather than sequentially. The most common experimental approach taken to determine whether information processing is serial or parallel in nature has involved manipulating the number of items to be processed, and observing the effect of this manipulation on processing speed (e.g. Treisman, 1986; Cheal & Lyon, 1992). If processing speed does not vary as a function of set size, then one can assume that parallel processing of set items has occurred. In fact, under many conditions, processing speed is reduced when set size is increased, suggesting that serial processing is taking place. For example, Sternberg (1966) found that the time taken for subjects to decide whether a target item was a member of a memory set increased linearly as the number of items in this memory set was increased from 1 to 6, leading him to conclude that the items in the memory set were accessed serially. Similarly, using a visual search task, Treisman has found that subjects take longer to detect, within a field of distracters, the presence of target defined by the conjunction of two critical features, as the number of distracters is increased (e.g. Treisman, 1988; Treisman & Sato, 1990). This led her to conclude that subjects serially check each item in the visual display for target status. However, in other circumstances, no such increase in reaction time is observed as set size increases, indicating that parallel processing of all set elements must be taking place. For example, extensive practice often has the effect of eliminating the set size slope, both in memory search tasks (e.g. Logan, 1988), and in visual search tasks (e.g. Steinman, 1987), suggesting that serial processing can become parallel with such practice. Also, certain types of stimulus array appear to permit parallel processing even without such practice. In visual search tasks, when a target differs from distracters on only one dimension then often it can be detected with equivalent speed as the number of distracters is varied, without any need for practice (e.g. Steinman, 1987; Treisman, 1988). This suggests that all elements in such arrays are processed for this criterial feature in parallel. The Page 30
same is true when targets differ from distracters along certain fairly complex dimensions; for example in terms of their three-dimensional orientation (e.g. Enns, 1990), or their apparent directional lighting (e.g. Enns & Resnick, 1990). Also, when targets and distracters differ in their categorical status, as is the case when letter targets appear among digit distracters or vice-versa, then target detection speed is unaffected by an increase in the number of distracters, suggesting that all display items are processing for categorical status in a parallel manner (e.g. Jonides & Gleitman, 1972). Such observations indicate that the cognitive system may be capable of both serial and parallel processing. Jonides & Gleitman (1972) provide striking evidence to suggest that the mode of processing adopted may be determined not simply by stimulus characteristics, or by level of practice, but also by the processing goals of the subject. These researchers had subjects search for the target symbol O, among a set of visual distracters consisting of the digits 1 to 9. Half of the subjects were instructed to search for the number '0', and half were told to search for the letter 'O' (both possible labels equally well describing the O symbol). When the target was labelled as zero, longer detection latencies were observed as the number of digit distracters increased, suggesting that serial search was taking place. In contrast, when identical displays were presented, but the target symbol was simply labelled as the letter 'O', then detection latencies were independent of the number of digit distracters, indicating that parallel search was occurring. Thus it seems that both serial and parallel processing may both be viable, and that the chosen mode may depend, at least in part, upon volitional factors.
The set size methodology, for dissociating parallel and serial processing, has attracted occasional methodological criticism. Almost without exception, such critics have questioned the need to conclude that serial process ever takes place. Theorists such as Townsend (1974, 1990) note that, if parallel processing suffers from capacity constraints, then the time taken to execute the simultaneous processing of many items may be greater than the time taken to simultaneously process a few items, within this reaction-time function implying sequential processing. While the need to postulate the existence of serial processing has been questioned, there has been a general consensus that the absence of set size effects must indicate the occurrence of parallel processing. Indeed, across the past decade or so, some of the most compelling and powerful computational models of human cognition have conceptualised the cognitive system as massively parallel in nature (e.g. McClelland & Rumelhart, 1985, 1986; Hinton, McClelland & Rumelhart, 1986; Smolensky, 1989; McClelland, 1995). Page 31
Connectionist vs Symbolic Architectures The most radically parallel accounts of human cognition collectively have been termed parallel distributed processing (PDP) models, and reflect a set of assumptions about the architecture of the cognitive system that has become known as connectionism. Relatively new on the scene, PDP models stand in striking contrast to the types of classical symbol-processing accounts that have been popular since the advent of cognitivism in the 1950s. Recent reviews of this distinction between symbolic and connectionist approaches to cognitive modelling have been provided by Estes (1991) and MacDonald (1995). The classical tradition has been to construe mental representations as symbols within the cognitive system, with each such symbol corresponding directly to a semantically interpretable entity. Information processing, according to this tradition, involves the application of rules to manipulate these symbols into new configurations. Therefore, although the types of rules that are assumed to control the cognitive system differ greatly according to the specific model under consideration, this classical approach conceptualises cognition essentially as a symbol manipulation process, and places principal emphasis upon the study of symbolic representations and the rules which operate upon them (e.g. Newell & Simon, 1972; Broadbent, 1984; Anderson, 1992). Within the connectionist cognitive architecture, however, neither symbolic representations nor rules are assumed to exist. Rather, the system is construed as an enormous collection of homogeneous nodes, linked through a rich network of connections. These nodes do not correspond to external referents. Instead, external states of affairs are internally characterised by profiles of activation across extensive areas of the network. The only property of any individual node is a level of activation, and this level of activation itself is determined dynamically by the inputs received through its connections with other nodes. The most critical factors determining the profiles of activation that occur across nodes are the weights attached to the connections within this network. These weights not only can vary along a positive continuum, with high and low positive weights causing a given level of activation in one node to result in the transmission of high or low activation across the connection, but also can vary along a negative continuum, with high or low negative weights causing a given level of activation in one node to result in the transmission of high or low inhibition across the connection. The structure of proposed connectionist networks varies across different models, and become extremely complex in many formulations, often with connections from weighted nodes not only directly affecting the activation of other nodes, but also serving to modify the Page 32
strength of the connections between other pairs of nodes. Usually, however, such networks are
considered to be organised as several layers of nodes, with the profile of activation within lower layers being determined principally by inputs from perceptual channels, and the upward spread of such profiles of activation through the weighted connections ultimately resulting in profiles of activation within higher layers that correspond to complex mental states, and to response configurations. Common across all connectionist models is the assumption that all learning is represented by changes to the weights associated with the connections within the network. Therefore, although learning within such models does result in patterns of information processing that appear consistent with the application of rules to mental symbols, no such rules or symbols actually are represented within the connectionist architecture. Rather, all knowledge is distributed throughout the system, represented wholly within the same matrix of weights that governs the transmission of activation across the massively parallel connections comprising the network (cf. Rumelhart & McClelland, 1986; Clark, 1989; McClelland, 1995). In spite of the elegant simplicity of their architecture, parallel-distributed processing models have provided powerful explanatory models for many highly complex cognitive phenomena, including language comprehension (e.g. Elman, 1995), reasoning (e.g. Levine, 1994; Sun, 1994) and cognitive development (e.g. McClelland & Jenkins, 1991; Plunkett & Sinha, 1992). Nevertheless, their introduction has been marked by a considerable degree of controversy. Some theorists have argued that connectionist models should not be considered as alternatives to symbolic models but, rather, only as neurally plausible implementations of such models (e.g. Broadbent, 1985; Courrieu, 1994). According to this position, connectionism represents a different level of explanation to that offered by symbolic approaches, but does not contradict the validity of such models (just as neurochemical models of information processing do not invalidate cognitive models in general). Others dispute this view, pointing out that the predictions generated by connectionist models often are inconsistent with those that result from symbolic models, and arguing that connectionism properly represents an alternative conceptual hypothesis to that offered by symbolism (e.g. Smolensky, 1995a). Not all theorist who take this latter position speak in favour of connectionism. For example, Fodor & Pylyshyn (1988) have argued vigorously against the plausibility of an exclusive connectionist cognitive architecture, pointing out that some of the properties such as a system would have are inconsistent with human experience. Recognising the advantages and limitations of both connectionist and symbolic architectures, many theorist now have formulated hybrid conceptions of the cognitive system within which rule-governed Page 33
symbolic systems and connectionist networks both make complementary contributions to information processing (e.g. Cooper & Franks, 1993; Hadley, 1993; Eskridge, 1994; Smolensky, 1995b). As will be seen through the main chapters of this book, most of the information-processing models that have been proposed to account for the types of cognitive symptoms reported by our hypothetical patient MF have been formulated within the classical symbolic framework. However, within the past few years, some clinical theorists have begun to draw upon the connectionism framework, to formulate distinctively different models not only of the cognitive distortions associated with emotional pathology (e.g., Williams, Mathews & MacLeod, in press), but also of the cognitive change processes associated with successful therapy (e.g. Caspar, Rothenfluh & Segal, 1992). Bottom-Up vs Top-Down Processing Many cognitive capabilities are substantially more complex than can be accounted for by models that implicate simple input-output chains, performed either serially or in parallel. When one considers many complex cognitive abilities, it becomes clear that low-level operations not only
influence higher level operations, but also are likely to be influenced by the results of these higher level processes. Most cognitive activity is not simply a passive response to an input but, rather, is goal directed. We try to understand, to reach decisions, to solve problems, to remember particular details, and so on. Thus, typically, the completion of each processing operation should take the system closer to some desired goal. However, it may only be when certain final operations are performed that it becomes apparent whether or not this goal has been achieved. If not, then earlier processing operations must be modified adaptively. Thus, feedback from the certain processing operations modify, or may even cause the reorganisation of, those processing stages which provide their own input. Most aspects or cognitive functioning therefore, are less accurately characterised as simple linear chains of processing stages, performed passively in either a serial or parallel manner to transform an input into a corresponding output, than as non-linear processing loops in which information is transferred in both directions. Those models which concentrate upon the way in which low-level basic processes lead to higher order representations are often termed 'bottom-up' models. Those which emphasise how higher order representations influence basic low-level operations are termed 'top-down' models (cf. Sanford, 1985, 1990; McCauley, 1988). Page 34
The distinction between top-down and bottom-up approaches can be illustrated with reference to many kinds of cognitive operation, including lexical processing (cf. McClelland & Rumelhart, 1981; Chase & Tallal, 1990); understanding text (Adamowicz, 1989; Kelly, 1995); solving mathematical or other problems (e.g. Carroll, Thomas & Malhotra, 1980; Silver, 1981; Sanford, 1983; Ranney, 1987); comprehending complex situations (e.g. Clark & Marshall, 1981; Sanford & Garrod, 1981; Shadbolt, 1983); and performing social judgements (e.g. Brewer, 1988; Fielder, Semin & Bolten, 1989). As an illustrative example let us consider perceptual processing. The question most basic to research on visual perception is this: how do we process information to yield an accurate internal representation of the external world impinging on the retina? Bottom-up models are typified by workers in scene analysis, a branch of artificial intelligence concerned with establishing a visual grammar to explain how basic elements are combined to yield complex percepts (cf. Frisby, 1980; Marr, 1980, 1982; Stent, 1981). Such accounts attempt to explain how the perceptual system ultimately identified objects, simply through the computational analysis and integration of the light input (cf. Kolers, 1983). A completely bottom-up account of perception requires a massive amount of processing for every perceptual input and, furthermore, is not entirely consistent with certain available data. For example, it is possible to construct quite different objects which, when viewed monocularly from one particular angle, will project an identical retinal image. If one of these objects is a familiar structure, such as a chair, whereas the other is a meaningless structure, then both objects will be perceived as this familiar structure when viewed in this way (cf. Haber & Hershenson, 1973). Thus the pre-existing higher order representation, in this case the representation of the object chair, controls the manner in which the current input is structured by the cognitive system. There are a great many similar examples where familiar or expected higher order preexisting concepts will determine the way in which lower level aspects of a current input, such as size (e.g. Bolles & Bailey, 1956; Gogel & Newton, 1969), colour (e.g. Duncker, 1939; Bruner, Postman & Rodrigues, 1951; Harper, 1953) or form (e.g. Leeper, 1935) are perceived. Those models of perception which consider low-level, more basic processes to be influenced by higher order representations typify the top-down approach. It is more realistic to accept the existence of both bottom-up and top-down operations, and to produce models which account for the relationship between these two different aspects of processing. Neisser (1976), for example, characterises the interaction between bottom-up and top-down processes in perception as involving a cycle. An initial, brief, bottom-up Page 35
phase will generate a hypothesis about what is being perceived. Expectations derived from this hypothesis will guide the next intake cycle, in a top-down manner, towards particular aspects of the stimulus. The selected aspects of the stimulus then will be briefly processed in a bottom-up manner, with the result possibly modifying the hypothesis, and leading to altered expectations which guide the next intake cycle in a top-down manner. Through this cycling of bottom-up and top-down operations, each may influence the other in a way which maximises efficiency. As has already been indicated, top-down processing is certainly not limited to perception, and models implicating a similar cyclical relationship between top-down and bottom-up processing have been proposed to account for a diverse range of cognitive abilities, such as occulomotor control (e.g. Colby, 1991), linguistic analysis (e.g. DeMarco & Harrell, 1995; Faust, Kravetz & Babkoff, 1993), and the processing of non-verbal communicative signals (e.g. Smith, 1995). Indeed, even the comprehension of very simple event descriptions would appear to rely heavily upon the top-down influence of pre-existing knowledge. Consider, for example, the processes involved in understanding this short text: Jane was invited to Jack's birthday party. She wondered if he would like a kite. She went to her room and shook her piggy bank. It made no sound. (Charniak, 1972)
While the passage is easy to follow, it cannot reasonably be argued that our comprehension of the overall situation simply reflects our cumulative understanding of its elements in a bottom-up manner. Nowhere is it stated that Jane wants to buy Jack a kite for his birthday present, or that she has no money to do soyet that is most people's understanding of the story. Indeed, our complete understanding of each of the explicitly presented elements is only possible in the light of our overall understanding of the situation. We can only interpret the reasons for shaking the piggy bank, and the consequences of its silence, if we already appreciate the point of the story. Thus pre-existing information about birthday parties, present purchasing and piggy banks must be brought to bear in a top-down manner in order to extract the true meaning of this input. The common understanding that Jane is a child, for instance, is not supplied by the text, but reflects the top-down influence of previously stored sets of assumptions. Clearly, the kinds of higher order representations capable of influencing comprehension processes in a top-down manner must be quite complex. One common suggestion is that much knowledge of the world is organised into convenient 'packets', 'modules' or 'schemata', which represent Page 36
conventional stereotypical situations and activities (e.g. Bartlett, 1932; Minsky, 1975; Schank, 1982; Sanford & Garrod, 1982; McKoon, Ratcliffe & Seifert, 1989). These prototypical representations are associated with a considerable amount of default information; that is, information which is generally true in a particular situation, but which may be modified by the input. They also contain empty 'slots' which require a certain kind of information to be supplied. The comprehension process initially involves identifying the most appropriate module to accommodate the available information, perhaps on the basis of an initial bottom-up analysis. Thereafter this module will exert a top-down influence, determining the way in which the information must be organised and interpreted in order to be most efficiently incorporated into this schematic structure. Therefore, the way in which the input is ultimately understood will be largely determined by pre-existing knowledge structures, and these structures will supply the additional information which allows inferencing and general elaborative processing. Similar schematic-based accounts have been suggested for the comprehension of stories (Rumelhart, 1975), the appraisal of social situations (e.g. Srull & Wyer, 1988) and the retrieval of information from long-term memory (Bara, 1983).
A fuller discussion of schemata can be found in Chapter 9. For the present, however, the most important point to note is that the way a complex situation is interpreted or recalled may depend upon the range of prototypical schemata stored in long-term memory which are capable of accommodating the available details. Such representations presumably are acquired through direct or vicarious learning experiences, and thus it is highly probable that there will be substantial individual differences in the nature of the schemata which are stored, and in their relative accessibility. If so, then the same event may be processed in highly dissimilar ways by two individuals who differ in the particular schemata available to organise the input. Disproportionately numerous or accessible schemata which function to organise information in particular ways may distort processing, resulting in the construction of biased appraisals or idiosyncratic mental models of ongoing events (Johnson-Laird, 1983; Morrow, Bower & Greenspoon, 1989). Of particular relevance to our current considerations, some clinical theorists have proposed that individuals may differ in the degrees to which they can access 'danger' or 'loss' schemata which operate in a top-down manner to impose threatening or depressing interpretations upon incoming data (e.g. Power & Champion, 1986; Beck & Clark, 1988). Perhaps, therefore, MF's observation that in social situations people tend to reject him and find him boring or embarrassing may reflect the schemata which he has stored to accommodate prototypical social situations. Those schemata may contain default information that is consistent with such a view, and which Page 37
become incorporated into his subjective appraisal of any particular social interaction. Additionally, his prototypical schemata may contain slots which operate to collect evidence that he is being rejected and judged negatively. Innocuous or ambiguous aspects of social interactions, such as pauses in conversation or half-smiles, may therefore be actively misinterpreted due to the top-down influence which the schemata exert on the organisation of the input. Control Hierarchies We have moved from a consideration of simple models involving linear sequences of processing stages to more complex accounts involving the transfer of information in both directions, and more closely resembling processing loops. According to such theories, complex representations both result from and also affect more basic processes. We have discussed this type of approach with reference to both perception and comprehension, and analogous models exist for a wide range of processes, such as motor skills learning (e.g. Poulton, 1957) and problem solving (e.g. Wason & Shapiro, 1971; Wason, 1971). However, such accounts typically deal with the relationship between the component cognitive operations involved in one particular kind of processing activity. It is unrealistic to conceptualise the brain as an assortment of wholly unrelated processing systems. Rather, it seems evident that quite different kinds of cognitive activity often interact closely with one another. For example, motor skills learning is closely associated with perceptual processing, problem solving often relies upon the retrieval of appropriate information from memory, certain kinds of memory retrieval sometimes appear to require a degree of problem solving, and so on. How should we conceptualise the relationship between all the different kinds of cognitive activity of which the brain is capable? The most compelling view, initiated by Bartlett (1941) and Craik (1966), restated by Broadbent (1977b), and richly instantiated within many contemporary models of the cognitive architecture (e.g. Bainbridge, 1993; Shallice, 1994; Lord & Levy, 1994; Houghton & Tipper, 1995; Chmiel, Totterdell & Folkard, 1995), is that multiple cognitive activities may occur in parallel, organised through a hierarchical structure of control. The processes at higher levels in this hierarchy are concerned with plans and behaviour which extend over wider ranges of space an time than is the case for those operating at lower levels. Those processes operating at higher levels also act to control those processes operating at lower levels, so that the pursuit of short-term goals
contribute effectively towards the execution of these long-term plans. That is, the lower levels of the cognitive system are organised and regulated by the higher levels. Page 38
It is important not to confuse this notion of levels with the concept of stages discussed earlier in this chapter. Within any cognitive level, self-contained processing may proceed quite independently, with their component operations perhaps interacting in a bottom-up and topdown manner within this level as required. However, this cognitive activity usually will be initiated by processes operating on a higher level, and governed by more strategic considerations. These higher level processes may not only initiate the lower level processes, but also may utilise the results of the lower level operations to refine the strategic pursuit of longterm goals. The idea that complex human behaviour may involve a hierarchy of processing subsystems, with each level in this hierarchy controlling the operation of lower levels, can be illustrated by considering a specific complex human activity such as, let us say, driving a car (cf. Ranney, 1994). Higher level processes will determine the long-term goals, such as the destination, thus setting sub-goals for subordinate processes involved in maintaining the appropriate route. At this next level of control, even shorter term subsidiary goals will be set for still lower-level processes, which execute upon the appropriate manoeuvres in particular locations, and these processes in turn set the subordinate goals for the motor programmes, operating at yet lowerlevels in the hierarchy, which control pedal pushing, gear changing and steering. The goals of these lowest level motor processes may change several times a second, whereas the goals represented at the highest level of control will be maintained throughout the full duration of the journey. This complex skill will be poorly performed if any of the fundamental low-level processes are not efficiently conductedfor example, if the steering is loose, gear changing stiff or the brakes spongy. Alternatively, difficulties may be encountered even when all those components' processes function properly, if they are not efficiently co-ordinated by higher levels within the control structure. If superordinate level processes set inappropriate goals for lower level processes, then the car driver may set out for the wrong destination, fail to negotiate junctions correctly, or mix up elementary manoeuvres by changing gear before depressing the clutch, and so on. Therefore, either faulty low-level component processes or a poorly organised control structure would each produce a deficit in travelling ability. Exactly the same reasoning can be applied when we consider cognitive deficits, such as the confusion, distractibility and poor concentration reported by our hypothetical patient, MF. Such impairments possibly may reflect dysfunctional abnormalities in basic cognitive processes such as memory access, or perceptual analysis. Alternatively, it may be the case Page 39
that all the basic processing mechanisms are quite intact, but the higher levels of cognitive control, which should operate to formulate long-term goals and delegate appropriate short-term sub-goals to the lower-levels, are not being implemented efficiently. Some accounts of depression-linked cognitive deficits emphasise the former possibility, while others place greater weight on the latter type of explanation. Thus, certain researchers have argued that these depression-linked deficits reflect a basic impairment in long-term memory retrieval (e.g. Sternberg & Jarvik, 1976; Henry, Weingartner & Murphy, 1973), while other researchers claim that such deficits are caused by depression-linked idiosyncrasies in cognitive control procedures (cf. Weingartner & Silberman, 1982; van den Elzen, 1996). In reviewing the literature, Weingartner and Silberman (1982), for example, note that cognitive deficits tend only to occur when depressed patients must process information in a manner that requires a high degree of
strategic control over low-level cognitive operations. Van den Elzen (1996) provides evidence to suggest that depressives may fail to implement cognitive control effectively because of their increased tendency to disengage from the long-term plans represented within the upper levels of the cognitive control hierarchy. Automatic vs Strategic Processes Early researchers often assumed that they ultimately would be able to explain cognitive functioning in terms of the unchanging stereotyped sequences of operations underpinning mental activity. Recognition that the cognitive system may more accurately be characterised as a complex hierarchical organisation of control procedures, however, gives rise to the possibility that the same cognitive tasks may be performed in qualitatively different ways by different individuals, or even by the same individuals on different occasions. Indeed, the evidence to support this possibility is now overwhelming. For example, individuals appear to differ in the strategies they employ to perform spatial tasks (e.g. Cooper & Marshall, 1985; Fischer, et al., 1994), to process language (e.g. Carpenter, Miyake & Just, 1994), to store and retrieve information from memory (e.g. Miller, 1994; Conway & Engle, 1994) and to learn complex skills such as mathematics (e.g. Geary, Frensch & Wiley, 1993; Lemaire & Siegler, 1995). It is clear, then, that certain aspects of information processing can be performed in a strategically flexible manner, that is capable of adapting to changing conditions in order to more effectively pursue the same processing goals (cf. Pressley, 1994). Page 40
A useful distinction is often drawn between those processes which are 'automatic', or incapable of such flexibility, and those which are 'controlled', or strategically modifiable (Hasher & Zacks, 1979; Schneider and Fisk, 1984; Nelson, Schreiber & Holley, 1992; Yonelinas & Jacoby, 1995). An automatic cognitive process involves a sequence of operations which are executed in an invariant manner, without the need for attention or conscious effort, in response to particular internal or external input configurations. It usually is assumed that such processes operate through relatively permanent sets of associative connections, which link their component operations, and these connections may either be innately determined or result from extensive and consistent training. Because automatic processes do not require attentional resources in order to function, their execution is unconstrained by capacity limitations, and numerous automatic processes may operate in parallel. In contrast, a controlled process is a temporary sequence of cognitive operations activated under the volitional control of, and maintained through the maintenance of attentional effort by, the subject. Controlled processes are tightly constrained, therefore, by capacity limitations, and only one such sequence at a time may occur without interference, except when two or more controlled processes are so slow that they can be serially interlaced. Despite this apparent disadvantage, however, the inherent flexibility of controlled processes allows them to be set up, altered and adaptively refined to effectively handle novel situations for which automatic cognitive processes have never been developed. Many investigators have employed cognitive load manipulations to dissociate the properties of automatic and controlled information processing (e.g. Shiffrin & Schneider, 1977; Strayer & Kramer, 1990; Granholm, Asarnow & Marder, 1991; Banquet et al., 1994). For example, Shiffrin & Schneider (1977) required subjects to determine whether an item from a particular memory set was present in a visually presented display. Initially, performance was relatively slow and enormously affected by memory load, indicating a limited capacity serial-comparison process, which these researchers construed as a controlled process. Provided that the same set of items was maintained as the search targets, however, this process appeared to change qualitatively over several thousands of practice trials. Detection speed and accuracy not only improved greatly, but also became independent of memory load, indicating the development of capacity independent parallel-
comparison processes, which was classified as an automatic process. When a new set of target items was then introduced, and the old set of items employed as distracters in the visual display, performance not only deteriorated sharply but actually dropped below the level obtained at the start of training when controlled search Page 41
was first utilised. Eventually, these inappropriate automatic-attention responses could be unlearned, and new sets of appropriate automatic responses learned, but only after a very considerable amount of retraining. On the basis of these and similar observations, researchers often argue that automatic and controlled operations may serve quite different informationprocessing functions. Schneider, Dumais & Shiffren (1984), for instance, propose that the function of automatic processing is to perform consistent and invariant component operations, including the interruption of ongoing controlled processing in order to reallocate attention, and to bias and prime cognitive operations. In contrast, Schneider and colleagues construe controlled processing as instrumental in the development of new automatic processing, and as essential to deal with novel tasks. Therefore, a common view is: 1. Automatic processes, once established, proceed independently from controlled processes. Thus, after reversal, although subjects are actively trying to respond to the new target stimulus set, they cannot easily prevent the occurrence of responses to inappropriate set items which have become automatised. 2. Related to this, while controlled processes cannot easily modify automatic processes, the reverse is not also true. Once an automatic-attention response has been established this can direct attention (i.e. can direct controlled processing) automatically towards the targets, regardless of concurrent input or memory load, to enable conscious detection of these stimuli to occur. This dichotomy between automatic cognitive processes which proceed without effort or conscious attention, and controlled information processing which reflects the application of effortful cognitive strategies, seems to reappear within the recent distinction often drawn in the learning and memory literature between 'implicit' and 'explicit' information processing. While explicit learning reflects the extraction of rule knowledge through the application of effortful strategies, implicit learning is said to have occurred when such rule knowledge is acquired without the subject's effort or awareness (cf. Reber, 1992; Seger, 1994). Similarly, explicit memory refers to the experience of retrieving stored information as a result of a deliberate and effortful memory search, while implicit memory refers to the process whereby previously encoded information exerts an influence on current performance without the subject either attempting to retrieve this information, or knowing that such retrieval has taken place (cf. Schacter, 1992; Graf & Masson, 1993). It seems highly plausible that this differentiation of implicit and explicit information processing may map rather closely onto the distinction between automatic and controlled processing. Page 42
It has been argued that evaluation of the emotional quality of a stimulus input is an automatic process which can proceed entirely in the absence of awareness, and may be little affected by controlled operations (Zajonc, 1980). Furthermore, there is evidence (reviewed in Chapter 4) which indicates that patients suffering from anxiety disorders appear to show an attentional response to threat-related stimulus items, which is entirely automatic in nature (e.g. Mathews and MacLeod, 1986; MacLeod & Rutherford, 1992; van den Hout et al., 1995). Consider the implications of this kind of information-processing for our patient MF. He reports experiencing episodes of anxiety 'out of the blue' even when he is thinking about nothing at all. However, he can only comment on those thought processes which are attentionally mediated or controlled. According to the information processing paradigm, he will be processing much information
automatically, and this appears to include emotional appraisal processes. The emotions he experiences may not, therefore, be independent of his environment, but rather may represent an appropriate response to evaluations which have been made quite automatically. Clearly, this argument depends upon the limitations of conscious awareness, and Chapter 10 will consider this topic in more detail. Another comment, made by MF, may also be illuminated by the above considerations. He notes that 'everything I read in the papers is about some terrible accident, or about violence', and that 'everywhere I look there are dangers'. Shiffrin & Schneider's hypothetical automatic-attentional responses may be highly relevant here. If, at some time in the past, MF has actively attended to certain cues which indicate potential dangers, then such attentional biases may have become automatised. Perhaps a more accurate description of MF's condition would involve rephrasing his comments as follows: 'Every newspaper report of terrible accidents and violence, I will read,' or 'Everywhere there are dangers, I will look.' The unlearning of such automatic responses may only be possible after a very extensive period of practice. Concluding Remarks This chapter has avoided examining detailed models for particular cognitive activities, and instead has considered some of the more general conceptual issues central to the informationprocessing paradigm. We have seen how the brain can be characterised as a system dealing with internal representations that convey information. Capacity limitations, however they are conceptualised, often cause bottlenecks which necessitate selective processing of certain representations in preference to others. Such Page 43
selectivity can be accomplished either by the preferential activation of these preferred representations, or by the inhibition of competing representations. Complex cognitive operations can more easily be modelled by reducing them to their simpler components or stages although certain methods for achieving this reduction are dependent on controversial assumptions. Certainly, it is unlikely that all cognitive operations involve linear sequences of stages which transmit information in a single direction. For example, some processes will involve the simultaneous activity of several operations in parallel rather than in series. Also, the transfer of information from later stages to earlier stages may be a characteristic of much cognitive activity, which thus will involve the cyclic interaction of bottom-up operations driven by the stimulus and top-down operations driven by higher order, long-term representations in memory. Just as any particular kind of cognitive process will involve component operations, much complex human activity must involve the co-ordination of many different cognitive processes. We have seen how the brain may be conceptually organised as a hierarchical system of processing levels, with the higher levels setting superordinate goals that hold over extended time periods, and delegating appropriate shorter term sub-goals to those processes operating at lower-levels. Certain aspects of this system therefore may be strategically and adaptively modifiable, though various low-level operations and certain well-learned processing sequences may be performed in an invariant and automatic manner, being triggered in response to particular external or internal input configurations. Each possibility brings its own advantages and limitations. The information-processing paradigm offers rich potential for modelling individual differences. According to this explanatory framework, human experience and activity will be mediated by available processing capacity, and by the priorities of selective functions which allocate this capacity. It will be affected by the nature and efficiency of the processing stages which analyse information, and constrained by the range of available long-term representations or schemata which exert top-down control over perception, comprehension, memory retrieval and other
cognitive processes. The organisation of the control structure, and the goals set at each level, almost certainly will be idiosyncratic, as will the degree to which certain processing sequences may have become automatised. In the main chapters of this book we shall discuss how those particular individual differences associated with the affective disorders can be illuminated by the adoption of the information-processing paradigm.
Page 44
Chapter 3 Cognitive Impairments Having set out the general approach to cognitive functioning that has been taken in current experimental psychology, we shall now consider how these functions are affected in anxiety and depression. In this chapter, we shall consider how performance is affected with 'neutral' materials (i.e. those that are of no obvious relevance to the patient's condition). Subsequent chapters will consider the processing of materials relevant to the patient's particular emotional preoccupations. There are two main reasons for considering the performance on 'neutral' tasks as a preliminary to examining tasks with emotional materials. One is theoreticalthe interpretation of effects found on tasks involving emotional materials is assisted by knowing what effects obtain with neutral materials. For example, it can be argued that interpretation of the memory performance of depressed patients with positive and negative materials has been impeded by the frequent omission of neutral materialsa point that we return to in Chapter 6. There has also often been a failure to ground theories of hedonic memory bias in depression on an understanding of how general memory functions are affected by depression. The other reason is practical. Most patients are concerned about their performance on 'neutral' tasks. It is therefore of clinical relevance to understand how and why they have problems with them. In turn, this may have implications for remedial strategies that they could adopt. The potential scope of a chapter on the cognitive performance of patients with emotional disorders is wide, and some limits will need to be set. Only performance on tasks that lend themselves to analysis in information-processing terms will be considered, though this boundary is not a clear one. For example, studies of the speed of motor performance in emotional disorders will be considered only in as far as they are relevant to information-processing theories. Little or no attention will be given to linguistic or social performance. It is chiefly the effects of clinical depression and clinical anxiety that will be considered. The literature on the effects of anxiety is extensive, but Page 45
most of this is not based on research with patients. It will be considered briefly for the implications it may have for the performance of patients with anxiety disorders, but not reviewed in the detail that would otherwise have been appropriate. In any case, good reviews of this non-clinical literature on the effects of anxiety on cognitive performance are available elsewhere (e.g. Eysenck, 1992). Performance deficits in depression have been studied quite extensively and will be a central focus of the chapter. Again, less attention will be given to studies based on mood-induction procedures (which are well reviewed by Ellis & Ashbrook, 1989) or on variations in mood within the normal population. A useful review of the early literature on the effects of depression on
performance was published by Miller (1975), and a more recent one by Cassens et al. (1990). Also, reviews which concentrate chiefly on memory deficits have been published by Johnson & Magaro (1987) and Ellis & Ashbrook (1988). Reference will also be made to the growing literature on cognitive performance in obsessional compulsive neurosis (e.g. Reed, 1985; Watts, 1995a). There have also been a few studies of cognitive performance in mania which will not be considered here; they have been described by Johnson & Magaro (1987). Particular interest in experimental abnormal psychology attaches to studies which test for interaction effects (i.e. that emotional disorders affect performance on one task more than another). Such studies contribute a great deal more to the exact specification of the performance deficit in clinical disorders than studies which use a single performance measure. However, there is a methodological problem that besets most studies testing such 'differential deficit hypotheses'. The two tasks may differ in psychometric features, such as discriminating power and reliability, and it may be these differences, rather than substantive differences in the functional demands made by the tasks, that are responsible for groups differing on one task more than another (e.g. Chapman & Chapman, 1973). For example, depressed patients show relatively greater impairment on tests of free recall than of recognition. However, this might not be the result of the additional psychological functions required in free recall, such as the generation of retrieval cues, but simply because free recall is psychometrically better able to differentiate between subjects. In fact, as will be argued in the course of the chapter, the latter explanation is probably not correct in this particular case. Nevertheless, it is a general methodological problem that is serious and pervasive, and raises questions of interpretation that often remain unresolved. It would be tedious to refer to it in connection with every study to which it applies, so it should be taken as a general caveat that applies to many of the studies referred to in this chapter. Unfortunately, there is not yet a consensus Page 46
about how this problem should be dealt with. The technique of 'matching' tasks advocated by the Chapmans has not been universally endorsed (e.g. Baron & Treiman, 1980; Watts, 1989), partly because matching tasks on difficulty may make them different in other ways, producing fresh problems of interpretation. For example, Calev's (1984) matched tests of recall and recognition depend on the recall test being shorter. Before reviewing empirical work in detail, it may be helpful to indicate briefly some of the broad explanatory frameworks currently available for explaining performance deficits in emotional disorders, though a more detailed discussion of these frameworks will be postponed until later. There are at least three general explanatory frameworks that need to be considered. The first is couched in terms of processing resources. It is an assumption of many theories of cognitive performance that capacity is limited, especially in tasks requiring conscious awareness. The evidence for capacity limitations comes primarily from studies in which subjects are required to perform two tasks, though there is no consensus about whether the interference effects of one task on another are due to their both requiring general cognitive resources, or to their overlapping in specific perceptual-motor requirements (see 'Capacity limitations' on p. 18). A resource allocation theory has been advanced for anxiety by Eysenck (1992), and for depression by Ellis & Ashbrook (1989). They will be reviewed in detail later in the chapter. The next theoretical framework can be seen as a more specific version of the first, in that it involves a specific proposal about why resources are depleted. People with emotional disorders are often preoccupied with negative or anxious thoughts, and it can be proposed that these take up cognitive resources that are then not available for other tasks. This is the prevailing view about how anxiety can result in an impairment of test performance (Wine, 1980; Eysenck, 1992) and a
parallel theory can be advanced for other emotional disorders. A third perspective that needs to be considered is that at least some of the effects of emotional disorders on cognitive performance are mediated by high levels of physiological arousal. The effects of arousal on performance are complex and depend on the level of arousal and the nature of the task (e.g. Eysenck, 1982); though high arousal is often associated with impaired performance. Many emotional disorders are characterised by high arousal, but there is an important exception. There is good evidence on a range of measures of arousal that 'psychotic' (or 'endogenomorphic') depressives are under-aroused (e.g. Byrne, 1975a). If such patients show deficits similar to those found in other emotional disorders characterized Page 47
by hyper-arousal, the possibility that those deficits are mediated by hyper-arousal can be refuted. Unfortunately, arousal levels are seldom reported, and have to be inferred from the type of depression. First, however, the relatively sketchy research on deficits in cognitive speed, attention, thinking and problem solving will be considered. This will be followed by a review of the fuller literature on memory deficits. Speed, Attention and Thinking Cognitive and Motor Retardation Speed of performance is such a basic aspect of performance that it is a good place to begin. There is no doubt that depression results in retarded performance on a wide range of tasks (Payne & Hewlett, 1960; Miller, 1975); the interest comes in the details of the effect. Retardation has been reported more consistently for endogenous-type than for non-endogenous type depression, but this depends on the task used. A distinction can be made between cognitive and psychomotor facets of retardation, and there is some evidence that whereas psychomotor retardation can be found in all depressed patients, cognitive retardation is found only in endogenous-type depression. One operational approach to this distinction is to divide performance times on a task into (a) a preparatory or decision phase and (b) an execution phase. Byrne (1976a) did this for choice reaction-time performance and found that 'psychotic' depressives were retarded in both phases, whereas 'neurotic' depressives were retarded only on movement time. The relationship between the two phases was also disrupted in depression. In normals, if decision time is artificially extended, there is a compensatory reduction in movement time, but depressives do not show this (Byrne, 1975b). Cornell, Suarez & Berent (1984) have confirmed, using a different methodology, that only psychotic depressives are retarded in the cognitive component of a motor task. Three tasks were constructed: (1) a simple reaction-time task, (2) a version involving additional motor components and (3) a version involving additional cognitive components. An additional motor component slowed down both groups of depressed patients significantly more than normals. In contrast, an additional cognitive component slowed down patients with 'melancholic' depression more than normals, though it affected non-melancholic patients only slightly Page 48
It is not clear what is the critical feature of endogenous-type depression that produces cognitive retardation. Miller (1975) suggested that cognitive retardation in depression was due to a motivational deficit based on learned helplessness. Even if the motivational deficit is more severe in endogenous depression, it is not clear why this should affect the two types of retardation differentially. Payne & Hewlett's (1960) hypothesis that cognitive retardation is caused by depressive thoughts and worries can be adapted more readily to explain why it occurs chiefly in
endogenous depression. We know that ruminative thinking (i.e. the tendency to dwell on the same thought or theme repetitively) is more than four times as common in 'melancholic' than in 'nonmelancholic' depressives (Nelson & Mazure, 1985), and that the negative thinking of endogenous depressives is less concerned with recent events and their implications (Matussek & Luks, 1981). It is possible that cognitive speed is retarded by processing resources being taken up by a type of ruminative thinking specific to endogenous depression. Another possibility is that cognitive retardation is the result of lowered arousal which, as we have seen, is specific to endogenous depression. Clearly, both these accounts would need to attribute motor retardation to some dysfunction found in all depressives, possibly the motivational deficit. Secondary Tasks and Depressive Retardation There have been several reports that depressive retardation can be reduced by adding a secondary task, which is intriguing because it is rather rare for a secondary task to result in improved performance on a primary one. The effect has usually been referred to as the 'distraction' effect, but this terminology will be avoided here because it is not helpful descriptively and it embodies doubtful assumptions about the mechanisms involved. Foulds (1952) showed that a simple secondary task (repeating digits after the experimenter at approximately every 2 seconds) increased the speed of maze performance in depressives, anxiety states and obsessionals, but not in hysterics or psychopaths. Similarly, Campbell (1957) found that the same secondary task (repeating digits), and also letter cancellation, speeded up the maze performance of both depressives and schizophrenics but not normals. Blackburn's (1975) results were more complicated. She found that an 'internal' distraction task (counting upwards at a rate of one digit per 2 seconds) produced a non-significant trend towards faster maze performance in all groups, whereas external distraction (listening to a prerecorded news item) tended to speed up manics and bipolar depressives, but to slow down unipolar depressives. There was no normal control group. Page 49
Foulds's explanation of the facilitating effect of a secondary task was that it disrupted depressive preoccupations and thus left processing capacity available for the primary task. We know from other work (e.g. Fennell et al., 1987) that requiring subjects to do a task that takes up information processing capacity reduces the frequency of distracting thoughts, and this may be the basis of the effect. However, it is not clear that this can explain why the secondary task does not impair performance as much as the depressive preoccupations which it replaces. Another problem is that it may not be able to explain the specific form of the effect, as this theory appears to predict a more general improvement in performance as a result of a secondary task than actually occurs. Beneficial effects have been shown only for the speed of maze performance, and there is a case for giving this greater prominence in an explanation of the phenomenon. Errors, in contrast, tend to be increased by distraction. Foulds (1952), for example, found that the clinical groups whose speed improved most also showed the most substantial increase in 'crossed lines' in the distraction condition. It is plausible that a secondary task produces changes in performance strategies. There may be a tendency for secondary tasks to speed up performance at the expense of level or accuracy of performance, an effect which also occurs when subjects are given incentives (Eysenck, 1982, ch. 5). It then remains to be explained why a secondary task should speed up depressives, but not normal controls. This could be the result of depressives and normals adopting different strategies before the addition of a secondary task. Depressives may normally adopt a conservative strategy, maintaining accuracy at the expense of speed, as Glass et al. (1981) have suggested. This would leave them more scope to increase speed without conspicuous loss of level of performance. Normals, in contrast, would be more likely to adopt a speed already close to their maximum
before the addition of a secondary task. Attention Investigations of general attentional dysfunction in depression have been limited, but there are a few experiments that have opened up interesting lines of enquiry. The first (Byrne, 1976b) distinguished between the dysfunctions in vigilance performance of psychotic and neurotic depressives. As the former are considered to be under-aroused, and the latter over-aroused, it is not surprising that they perform rather differently at a vigilance task. Psychotic depressives Page 50
showed the poorest overall level of correct detections, and the steepest deterioration over time. Neurotic depressives showed a different pattern. Their overall detection rate was affected less (though still significantly) and their errors mostly took the form of false positives. Byrne argues that these differences are consistent with what might be expected on the basis of differences in arousal, though they have not been replicated (Byrne, 1977). The second experiment bears on cognitive retardation, but locates it more specifically. Depressives and normals do not differ in the length of exposure necessary for a stimulus to be correctly identified. However, when backward masking is used, depressives show disruption of recognition at longer interstimulus intervals (ISI) than normals (Sprock et al., 1983). Extension of the ISI from 60 to 120 ms improved the performance of both depressives and normals, but further extension from 120 to 300 ms helped only depressives. It seems that sensory registration is not retarded in depression, rather it is the subsequent encoding that follows registration that is retarded. It is the latter that backward masking disrupts. Thus Sprock and co-workers suggest that depressives take longer to form a representation of the stimulus, but not to register it. The third attentional experiment is based on a paradigm developed by Broadbent to test his filtering hypothesis (Broadbent, 1971). He argued that where filtering is possible (i.e. where relevant stimuli can be distinguished from irrelevant ones on the basis of a clear physical cue), pre-stimulus instructions about which stimuli to select should be advantageous. This obtains, for example, if subjects are presented with a list of digits, some of which are spoken in a male and some in a female voice. The performance of normals is improved by telling subjects before rather than after the series which set of digits they will be required to recall. However, depressives do not show this effect (Hemsley & Zawada, 1976). This has been interpreted as demonstrating a filtering deficit in depression. Griffin, Dember & Warn (1986) noted the extent to which vigilance performance is normally affected by prior training with a particular signal frequency. Normal subjects who are trained at one signal frequency tend to go on detecting signals at that frequency, even when the frequency is changed. However, Griffin and colleagues found that this was not the case with depressed subjects, and they interpreted this in terms of a lack of adequate motivational involvement in the task. Most recent studies of attention in anxiety focus on bias in attention rather than general deficits in attention performance. The possibility (discussed in connection with attentional bias results in Chapter 5) has been raised that anxious subjects have a general difficulty maintaining attentional focus (see Eysenck, 1992, for a further discussion). For example, Broadbent, Broadbent Page 51
& Jones (1986) found that highly anxious subjects experienced more difficulty in focusing
attention, even when neutral materials were used. The fact that this disruption occurred only when the target location was unknown suggested that, given warning, anxious people can use top-down strategies to focus attention, but have difficulty controlling attention within the sort of data-driven task that assesses speeded reaction to unexpected stimuli. In a similar study that was mainly concerned with biases in attention, Mathews et al. (1990) found that patients suffering from General Anxiety Disorder (GAD) had more difficulty detecting a letter at an unexpected location when words were appearing elsewhere on the screen (either threatening or neutral) suggesting a general difficulty with attentional control. Fox (1994) investigated the possibility that highly anxious people have a general inability to maintain attentional focus, using both a Stroop paradigm, and a related paradigm in which colour patches and semantic stimuli were spatially separated. She also used Tipper's negative priming procedures (Tipper, Brehaut & Driver, 1990) to demonstrate that high anxiety subjects differ from low anxiety in showing reduced negative priming. Negative priming, in this context, refers to the increase in latency for a subject to respond to a target location, if, on the previous trial, that location was one which had to be ignored. In Fox's experiment, however, following presentation of a to-be-ignored distracter at one location, highly anxious subjects did not show the expected retardation on the following trial. That is, highly anxious subjects faded to show negative priming. She concluded that high anxiety is associated with a general deficit in inhibiting distracting information. Chapter 5 considers these data to see how far they can explain attentional bias, i.e. explain why highly anxious people are relatively more prone to interference on some stimuli and not others. The studies reviewed in Chapters 4 and 5 will show that emotional information distracts such people more than non-emotional information, especially where it is related to a current concern, or primed by such a self-related concern. While each of these effects may be made more salient in anxious people by a general deficit in their ability to inhibit to-be-ignored material, such a general deficit in attentional control is not yet able to explain the precise pattern of observed results. Nevertheless, these failures to control attentional focus may well explain more general deficits in concentration and memory complained of by anxious subjects. Abstracting Ability The most investigated question regarding general thinking function in depression is whether or not there is a failure of 'abstracting' ability. Most Page 52
of the available studies indicate that there is. The most commonly used test has required the interpretation of proverbs (Gorham, 1956). Several studies have shown that depressives obtain lower abstraction scores than controls matched for IQ, though scores for concrete interpretations tend not to be affected (Braff & Beck, 1974; Sprock et al., 1983;). Similar results have also been reported by Caine, Yerevanian & Bamford (1984), Clark et al. (1985) and Raskin, Friedman & DeMascio (1982). Donnelly et al. (1980) used the Category Test from the Halstead-Reitman battery requiring subjects to abstract features such as orientation from geometric figures, and obtained significantly lower scores in depressives. Similar results were reported by Savard, Rey & Post (1980). The deficit in abstracting ability in depression is a specific one that exceeds that found in general verbal tests such as vocabulary (Braff & Beck, 1974). An unresolved issue is whether abstracting and conceptual functioning are impaired only in depressive episodes or whether they represent a stable trait of people who are vulnerable to depression. Andreasen (1976) failed to find differences between depressives at admission and at discharge in conceptual functioning, but Donnelly et al. (1980) found that scores on the Category Test improved with recovery from
depression. Cassens, Wolfe & Zola (1990) have cautioned that deficits in abstracting ability seem to be found more commonly in older depressed subjects. If there is an age-related trend of this kind, it emphasises the difficulty of drawing secure conclusions from studies that do not match groups for age. Problem Solving Studies of problem solving in depressed students have given inconsistent results (e.g. Gotlib & Asarnow, 1979) but a study of depressed patients (Silberman, Weingartner & Post, 1983) has provided a rich analysis of the nature of the problem-solving deficit in depression. The task, based on a paradigm developed by Levene (1966), involved hypothesis-testing behaviour in a 16-trial two-choice visual discrimination problem. Feedback was provided on only three trials, which would be sufficient, if problem solving was perfectly efficient, to narrow down to a single correct hypothesis. A measure of 'focusing' was employed which reflects the extent to which subjects narrowed their hypotheses down, and it was found that depressives showed significantly poorer focusing. All subjects performed the task a second time in which they were required on each trial to list all the hypotheses that they considered might be correct. This improved the performance of all subjects, but helped depressives more Page 53
than controls. Depressives had no difficulty in listing the four possible hypotheses at the start of the series. Their deficiency seemed entirely attributable to failing to discard invalid hypotheses and thus retaining an over-large set of possible ones. The helpful effect of the requirement to report hypotheses is important from a clinical point of view as it suggests a remedial strategy that could be used clinically. Recent studies have focused on interpersonal problem solving, and have confirmed a deficit in depressed patients. Marx & Schulze (1991) located a particular deficit in generating actionoriented strategies, even though they showed an adequate definition of both the problem and the target. Though there is thus reasonable confirmation of an objective deficit in problem solving in depressed patients, one recent study (Blankstein, Flett & Johnston, 1992) did not find this. However, it did obtain evidence that depressed patients have a lower self-appraisal of problemsolving ability. Another issue that has recently received attention is the possible role of impaired problem-solving ability in the genesis of depression. Here again the distinction between objective and selfappraised problem-solving ability may be relevant. Schotte, McNally & Turner (1990), on the basis of a longitudinal study, argued that the objective in interpersonal problem solving is a concomitant of depression rather than a cause. (However, Dixon et al. (1993) found evidence that self-appraised ineffectiveness in problem solving is an antecedent of depression that may play causal role. The literature on problem-solving deficits in depression has been well reviewed by Nezu, Nezu & Perri (1989), who are particularly concerned with the clinical implications of problem-solving deficits. Attention and Thinking in Obsessional Patients There have been a number of investigations of attentional dysfunction in obsessionals from which a coherent story is beginning to emerge, see Watts (1995a). Though negative results may be obtained if obsessionals and controls are compared on a single attentional task (Gordon, 1985; Sher, Mann & Frost, 1984), a more interesting picture emerges when the analysis focuses on
differences in functioning between two paradigms making rather different attentional demands (Broadbent, Broadbent & Jones, 1986). However, this work has so far been based on 'normal' subjects with high scores on a questionnaire measure of obsessionality, and has not yet been extended to patients. Broadbent and colleagues had two similar tasks, one intended as a measure of 'filtering', the other as a measure of 'pigeon-holing. In filtering, Page 54
stimuli are selected on the basis of some basic sensory feature, whereas in pigeon-holing they are classified on the basis of meaning or experience. In both cases, choice reaction time to stimuli (A or B) was measured, but only in the filtering task did they know where on the screen the stimulus would appear. Obsessional subjects performed less well on the pigeon-holing task than would be expected on the basis of their performance in a filtering task. If this is replicable with patients, it would provide an account of some aspects of the kind of attentional functioning that obsessional patients display when checking. It is noticeable that they often do not rely on 'scanning' when checking. Rather than assuming that their attention will be captured by something amiss, they try consciously to identify and inspect each object as part of the checking procedure. The difficulties that obsessionals have with the more complex demands of 'pigeon-holing' is paralleled by their approach to more intellectual tasks. Obsessionals have been shown to operate with narrow conceptual categories. Reed (1969a) found that patients with obsessional personality were underinclusive in a test of features essential to particular categories; and Reed (1969b) similarly found that they sorted blocks into more circumscribed categories. Persons & Foa (1984) also obtained evidence for underinclusiveness in obsessionals from performance in a card-sorting paradigm. This finding of narrow conceptual categories in obsessionals, which has emerged from studies using different tasks, thus seems reasonably secure. Regarding problem solving, Reed (1977) has suggested that obsessionals are impaired in inductive but not in deductive reasoning, and supported this with evidence that patients with obsessional personality perform worse than controls on number-series tasks, but are better at arithmetic. This is an interesting dissociation, but clearly the tasks differ in many ways, and so the results lend themselves to other interpretations. Decision-making is also retarded in obsessionals, primarily due to procrastination and a desire for additional information (Volans, 1976). Sher has approached the study of cognitive function in checking by studying normal subjects with high scores on a checking questionnaire (Sher, Frost & Otto, 1983; Sher, Mann & Frost, 1984; Sher et al., 1989). Negative results were obtained on several measures, though checkers were shown to underestimate their performance on a 'reality monitoring task in which subjects were presented with a series of word pairs (e.g. hot: cold). In some cases the second word was printed in full; in others only the first letter was supplied and subjects generated it themselves. Subjects were subsequently required to distinguish which words were actually presented, and which were self-generated. The underestimation effect appeared to be specific to Page 55
obsessional 'checkers' and was not shown by obsessional 'cleaners'; it was also specific to the reality monitoring task and no comparable underestimation was shown on a recognition task. Perhaps checking is in part a consequence of lack of confidence in the capacity to distinguish whether events actually occurred or were merely thought to occur. Another deficit that appears to be associated with subjects obtaining high checking scores is a poor memory for actions. This was reported first for non-patients with high levels of checking (Sher, Frost & Otto, 1983) but has been confirmed in patients with compulsive checking (Sher et
al., 1989). At the end of a testing session, subjects were asked to recall the tasks that they had performed, and checkers listed significantly fewer than controls. The effect was specific to free recall and did not appear with recognition memory for actions. Further review of cognitive deficits in obsessional-compulsive disorder, and their implications for treatment can be found in Tallis (1993). Memory Memory functions in anxious and depressed people have been investigated quite extensively, though the literature on anxiety is mostly non-clinical. Investigations of clinically depressed subjects have left no doubt that their performance on free recall tasks is poor. The evidence for this, which has been reviewed by McAllister (1981), Johnson & Magaro (1987), Ellis & Ashbrook (1988) and Watts (1993, 1995b), is based on (a) comparisons of depressed patients with controls matched on intelligence or educational level, (b) comparisons of the performance of depressed patients when currently depressed and when recovered, and (c) correlational studies of the association between memory performance and the severity of depression. The work of Cronholm & Ottosson (1961), which was one of the first clear demonstrations of a memory deficit in depression, remains one of the best studies in the field, and has been substantially replicated by Sternberg & Jarvik (1976). Cronholm & Ottosson (1961) compared a sample of patients with endogenous depression with surgical patients. The groups were matched on educational level (as well as age, sex and area of residence) and were also shown to have comparable vocabulary levels. Three specially devised tests were used, examining memory for word pairs, for simple figures, and for personal data about fictitious people. Page 56
Subjects were tested immediately, and again after three hours. Significant differences were found on all three tests at both occasions of testing, though there was no evidence of more rapid forgetting in depressives. Despite the fact that there is clear evidence that depressed patients are impaired in performance on memory tasks, the objective impairment is not always quite as great as memory complaints would suggest. Not all the studies that have reported memory complaints in depression have been able to demonstrate corresponding objective impairments. Watts (1993) has suggested that it may be particularly in people who are both depressed and elderly that memory complaints outstrip objective memory impairment. This chapter will not review evidence for the existence of a depressive memory deficit comprehensively, but will focus on qualitative issues about the nature of the deficit. Responsiveness The first possibility that needs to be considered is that depressives can access material from memory as well as normals, but perform less well on memory tasks for other reasons. Johnson and Magaro (1987) have argued that the depressive memory deficit is partly due to response bias. There are two related hypotheses that need to be considered: (a) that depressives lack confidence in their memories, and (b) they cannot make the effort to report all the memories accessible to
them. Investigation of the first possibility depends on studies of recognition memory and focuses critically on the level of false alarms. If depressives have low levels of both hits and false alarms, it suggests that the low level of hits is due to a cautious response criterion. On a signal detection analysis, this would result in them differing from controls on rather than d'. One impediment to investigating this is that tests of recognition memory are often less sensitive to memory problems than tests of free recall, with the result that free recall usually differentiates groups significantly, whereas recognition memory tests may show only a non-significant trend to differentiate groups. This is probably partly just a psychometric artefact due to the different discriminatory power of the two types of rest, and applies just as much to other groups, such as the elderly, as to depressives. However, it is not to be explained wholly in these terms. Even when free recall and recognition tests are matched for difficulty in normals, depressives do less well on free recall than on recognition (Calev & Erwin, 1985). Despite these general problems in demonstrating group Page 57
differences on tests of recognition, there are several studies that show significant effects of depression on 'hit' rates in recognition memory (e.g. Miller & Lewis, 1977; Silberman et al. 1983; Watts, Morris & Macleod, 1987). Another problem that has beset studies of recognition memory in depression is that the level of false alarms obtained in depression seems to depend on procedural variables. The clearest demonstration of this is in the study of Watts and colleagues (1987) in which some groups vocalised words on presentation while others did not. Depressed patients showed more false alarms than controls with vocalisation, but fewer without vocalisation. Consistent with the view that processing conditions are crucial, Zuroff, Colussy & Wieglas (1983), whose subjects performed a self description task when words were presented, found more false alarms in depression; whereas Miller & Lewis (1977) and Dunbar & Lishman (1984), who used no encoding task, found fewer false alarms. The implication is that, whether or not the lower level of hits characteristic of depression appears to be due to cautious response criteria, depends on procedural variables. However, it is clear from the study of Watts and colleagues that depression reduces hits even in conditions in which false alarms tend to be increased. The effect on hits cannot therefore be explained entirely in terms of response bias. Overall, Watts and co-workers found that depression affected d' rather than . There has been some indirect evidence, from analyses of patterns of errors, to support the hypothesis that depressives' memory performance appears to be deficient only because they do not make the effort to produce all the responses they are capable of. Henry, Weingartner & Murphy (1973) noted that the errors of depressives tend to be of omission rather than 'commission'. Similarly, Whitehead (1973, 1974) found that depressives were characterised by high rates of omission errors but low rates of transposition errors in a serial learning task. Dannenbaum, Parkinson & Inman (1988) also found that omission errors were particularly associated with depression. However, studies which have attempted to control amount of output experimentally suggest that the depressive memory deficit cannot be explained wholly in terms of poverty of output. Watts & Sharrock (1987) tested memory for prose, first by free recall and then by a form of cued recall that required much less output (i.e. answers of only one word or a short phrase to questions about the prose passage were sufficient). If the poor performance of depressives on memory tests is due to an output problem, they should have been less impaired on the 'low-output' test, Page 58
whereas in fact the low-output test differentiated groups slightly better than unaided free recall. Convergent evidence comes from a study of student volunteers in whom a depressed mood had been induced (Leight & Ellis, 1981). A forced recall paradigm was used in which subjects were required to guess if they did not know the answer, but depressed students nevertheless obtained lower recall scores. This also suggests that the effects of depression on memory are not wholly explicable in terms of poverty of output. Short-term Memory One of the few studies of sensory memory in depression is that of Colby & Gotlib (1988). They used a modified version of the Sperling task in which subjects are presented with a 3 X 4 array of characters, and partial recall is presumed to reflect sensory store processes and complete recall to depend on short-term store. Depressed subjects showed normal partial recall, but impaired total recall. Also relevant to sensory memory are studies of the memory comparison stage (i.e. scanning and retrieval) in Sternberg's 'additive factors' approach. Some studies have found that depressed subjects show normal scanning speed, but impaired performance on non-scanning components such as encoding and response output (Glass et al., 1981). However, Brand & Jolles (1987), using a redesigned version of the memory comparison task, have been able to demonstrate slower scanning in depression. It remains to see whether this can be replicated. A very well established short-term memory task is, of course, digit span. This is generally not affected by depression, even when used in the 'supra-span' form, i.e. recall of a series of digits two longer than span (Kopelman, 1986). However, the introduction of delays of 20 or 30 seconds into the digit span task does seem to produce an impairment in depression. Another study that supports the significance of introducing short delays is that of Cohen et al. (1982) who found that depressed subjects showed relatively less impairment at zero delays in the recall of trigrams. It is interesting to note that the critical effect seems to come from introducing even a small delay such as 20 seconds. Studies with a variety of materials have found that it makes relatively little difference whether the delay is further extended by a period such as 30 minutes (e.g. Sternberg & Jarvik, 1976; Stromgren, 1977; Kopelman, 1986). Page 59
Though digit span is one of the few memory tasks that is not usually affected by depression, it can be affected by anxiety. The literature here is somewhat conflicting, and it is important in making sense of it to distinguish between trait, which generally does not affect digit span, and situationally induced (state) anxiety which does affect it (see Eysenck, 1979a). A good example of the impact of situationally induced anxiety on digit span is that of Idzikowski & Baddeley (1987) on parachuting anxiety. Level of Encoding Turning to hypotheses about the effects of anxiety and depression on encoding one specific hypothesis has been that anxiety biases people away from encoding in terms of semantic features and towards encoding in terms of more superficial (acoustic or visual) features. This represents a direct application of the 'levels-of-processing' approach to memory (Craik & Lockhart, 1972). However, there have been a number of relevant experiments on the effects of anxiety in nonclinical samples which have given only weak support to the hypothesis. It is particularly doubtful whether anxiety enhances phonemic encoding, though it may well impair semantic encoding. An initial finding by Schwartz (1975) that neurotic introverts were adversely affected by
phonemically similar response words in paired associate learning, whereas stable extroverts were adversely affected by semantically similar response words, has not been replicated (Craig et al., 1979). Mueller (e.g. Mueller, 1976, 1977, 1978) has published several tests of the hypothesis that anxious subjects show less semantic clustering but more acoustic clustering than normals, but has found that they show less clustering of both kinds. Another relevant approach is the falserecognition paradigm. When subjects incorrectly say that a word has appeared before, it is sometimes because it is acoustically confusible with a word that has actually appeared previously and sometimes because it is associated with a previous word. Depressed students show significantly fewer semantic associative errors, but do not show significantly more acoustic errors as might be predicted from the encoding-bias hypothesis (Hasher & Zacks, 1979). There is other evidence that anxiety and depression affect people's ability to structure material semantically at encoding. When a list is constructed of words that are capable of being grouped into several semantic categories, but are presented to the subject in random order, normal subjects tend to recall them in semantic clusters. It seems that this may be disrupted in Page 60
emotional disorders. This has been reported in non-clinical studies of neurotic introverts (Schwartz, 1975), and high-anxiety subjects (Mueller, 1976). The same is true of depressed patients (Koh, Kayton & Berry, 1973; Russell & Beekhuis, 1976; Weingartner et al., 1981, Exp 3; Calev & Erwin, 1985). However, caution must be exercised over assuming that this necessarily represents an explanation of the lower recall scores; as in the studies of Schwartz (1975) and Mueller (1976), lack of clustering was not associated with impairment of recall. It may be that emotional states have separate and independent effects on clustering and level of recall. Research with laboratory tasks has been supplemented by analyses of the approaches that students take to studying. A pioneering study was reported by Fransson (1977). In-depth interviews were used to explore the methods students employed in studying textbook passages, on the basis of which they were classified as having used 'deep' or 'surface' methods (see Marton, Hounsell & Entwistle, 1984). A strong association was found between high levels of anxiety and use of a surface approach. There was also a negative correlation between state anxiety and performance on a test of factual knowledge for the passage. However, the effects of anxiety only emerged with materials related to the subject the student was studying, which Fransson (1977) interprets as having induced stronger 'intrinsic' motivation. Watts et al. (1986a) carried out a related study but, instead of an interview, used Entwistle's questionnaire measures of 'meaning' and 'reproducing' approaches to studying. As would be expected, there were significant positive correlations between the reproducing orientation and several measures of anxiety tension and neuroticism. In contrast, hostility and depression were the emotional states that interfered with (i.e. showed significant negative correlations with) 'meaning' orientation. It thus seemed that different emotional states may bias subjects towards deep processing and bias against surface processing. Cognitive Effort Evidence has also come from a variety of paradigms suggesting that depression and anxiety reduce the level of effort expended at encoding, though again, most of the evidence has come from studies of student volunteers. Ellis, Thomas & Rodrigez (1984) reported two relevant experiments on normal subjects who had undergone a depressive mood-induction procedure. Both used paradigms designed to manipulate the level of effort involved in encoding. In the first, sentences were presented with a missing target word, together with two possible words Page 61
to fill the gap. In 'low effort' sentences it was obvious which was the correct one. In the 'higheffort' sentences it was less obvious. Examples for which the missing word is 'dream' are 'The girl was awakened by the frightening' (low effort) and 'The man was alarmed by the frightening' (high effort). At recall, subjects were required to write down the target words. Subjects in neutral mood replicated previous work, showing that 'high-effort' words were better recalled, but depressed subjects did not show this. Though there was a main effect of mood, paired comparisons showed that this was almost wholly attributable to poorer performance of depressives on the 'high-effort' sentences. Another experiment examined the recall of target words that were embedded in sentences at presentation. At test, sentences were presented with the target word omitted. In one condition the original 'base' sentences were used; in the other condition 'elaborated' sentences were used which included an extra phrase that was semantically related to the target word (e.g. 'The hungry child opened the door' and 'The hungry child opened the door of the refrigerator'). Elaborated sentences were better recalled in normals, but the advantage of elaborated sentences was not significant in depressed patients, again suggesting that they are at a relatively severe disadvantage in situations requiring (or at least presenting the opportunity for) elaborated encoding. Though these experiments opened up a promising line of enquiry, there is now doubt about the replicability of the findings, with forms of depression other than depressed mood. Potts, Camp & Coyne (1989) were not able to replicate Ellis and colleagues' effects for subjects scoring highly on the Beck Depression Inventory (BDI). Hertel & Rude (1991a) replicated the effects for mood induction subjects comparable to those used by Ellis and colleagues but, like Potts and co-workers, were not able to demonstrate them for high BDI subjects. So far, this is not too great a cause for concern as experiments on subjects varying in BDI scores within the normal population have often proved a relatively weak way of manipulating depression. Much more serious is the failure of Hertel and Rude (1991b) to replicate the effects of Ellis and colleagues on depressed patients. At the present time the possibility remains that the kinds of interaction Ellis and colleagues found between depressed mood and sentence type are for some reason specific to experimentally induced mood. Nevertheless, there are experiments using other paradigms which have provided support for the general idea that depressed patients are impaired in effortful processing tasks. Roy-Byrne et al. (1986) reported two experiments on depressed patients designed to test the hypothesis that 'effortful' tasks are more impaired than 'automatic' ones. This Page 62
important distinction was discussed in the previous chapter. In one experiment, subjects heard a word list in which some words were repeated. The effortful task was again to recall the words; the automatic task was to say whether particular words had been heard once or twice. Depressives were impaired on the effortful task, but not impaired at all on the automatic task. In similar vein, Golinkoff & Sweeney (1989) found that depressed subjects were impaired in the high-effort task of paired associate learning but not in the low-effort one of making frequency judgements. Finally, Calev, Nigal & Chazran (1989) found that depressed subjects were more impaired in the high-effort task of generating words belonging to a semantic category than the low-effort one of producing words beginning with a given letter. Yet another related line of research has compared the effects of depression on implicit and explicit indices of memory. With depressed patients, as with a number of other clinical groups who show memory impairment, deficits are more likely to be found in explicit than implicit memory tasks (Hertel & Hardin, 1990; Danion et al., 1991; Peters, 1992). The only experiment
so far to have found an effect of depression on implicit as well as explicit memory is that of Elliott & Greene (1992). It is a puzzling discrepancy and we shall return to it in the final chapter. Structuring of Materials A related line of research has looked at whether depressed patients are impaired in their capacity to structure materials in memory tasks. This is the kind of high-effort task where a particular deficit might be expected. One approach has required subjects to sort words into categories at presentation, though results have been disappointing. Russell & Beekhuis (1976) found that depressed subjects showed no differences from controls in sorting, though they showed lower levels of recall and less clustering at recall. Weingartner et al. (1981) found differences in sorting for lists of random words, with depressives using fewer categories. This might be thought to be contrary to the hypothesis of a structuring deficit in depression, though Weingartner et al. interpret it, without much discussion, as showing that depressives are impaired in the 'kind of organisation' they impose on material. In any case, level of recall and the number of sorting categories used were uncorrelated in depressives. Page 63
A more promising approach to the hypothesised structuring deficit in emotional disorders is to vary the amount of structure in the material, and to see how this interacts with anxiety or depression. Ellis et al. (1985) and Ellis & Ashbrook (1988) have argued that highly structured materials make fewer encoding demands and are therefore less vulnerable to the disruptive effects of emotional disorders. Ellis & Ashbrook (1988) cite a number of non-clinical studies, especially those based on mood induction, which have failed to show effects of depressed mood on prose and other highly organised materials. However, caution needs to be exercised over assuming that this will also be true of clinical depression. Among clinical studies, Weingartner et al. (1981, Exp 3) used lists composed of words from various different semantic categories and found the effect of depression on performance was relatively slight when words from the same category were grouped together at presentation. Presumably, less processing effort was required when words were presented in semantic clusters. However, opposite results have now been obtained with a variety of different forms of structuring of materials, i.e. depressed subjects have shown greater deficits with more structured materials. Watts et al. (1990) found this when comparing clustered with unclustered word lists. Peters (1992) found it when comparing semantically related with unrelated words. Levy & Maxwell (1968) also found it with word lists showing varying degrees of approximation to text. Yet other studies have found that depressed patients show their greatest deficits with intermediate levels of structure. Watts and colleagues found this for varying approximation to text and Channon, Barker & Robertson (1993) found a greater deficit for words which were semantically related but presented in random order than for words which were either presented in semantic clusters or which were not semantically related at all. Clearly not much can be concluded from such a complex set of results. One problem of interpretation is that different kinds of structure may interact with depression in different ways. Another methodological difficulty is that levels of structure are difficult to calibrate, and several studies have used only two or three levels. However, if it turns out that intermediate levels of structure do produce the greatest memory deficits, it can be interpreted along the following lines. Lists with no structure at all do not repay subjects' efforts to structure them at encoding, and so it makes little difference whether or not they have the resources to do this. Equally, lists that have a structure that is very easily grasped can be processed effectively, even by depressed subjects with limited processing resources. However, it may be with materials that have an intermediate level of structure, i.e. where there is a latent Page 64
structure but one that takes effort to grasp, that depressed subjects are most impaired. Memory for prose in clinical depression is of considerable interest, not only for its relevance to the hypothesis of a structuring deficit at encoding, but also because of its greater relevance to everyday life. Watts & Cooper (1989) studied memory for a passage of prose in depressed patients, though the passage chosen, the 'Circle Island' story (Dawes, 1964), can itself be criticised for its artificiality. Subsequent analyses have focused on whether a structuring deficit could be demonstrated for prose that paralleled that already described for word lists. With prose passages, selectivity of recall provides a useful index of the degree of organisation imposed on the passage. The units of a prose passage vary in how central they are to its structure, and it can be inferred that subjects who show a strong bias towards recalling central units have performed more structuring than those who recall a more random selection of units. Depressed patients showed significantly less bias towards recall of central units than did controls, whether centrality was identified on the basis of a story grammar analysis or on ratings of centrality to the gist of the passage made by a separate group of subjects. However, depressives showed as strong a bias as controls towards selective memory for highly imageable units, a non-structural variable used as a control. At present, there is thus considerable support for the view that structuring of materials is deficient in the memory processes of depressives. This may be a specific example of a general deficiency in 'effortful' cognitive processes (Johnson & Magaro, 1987). However, it is not clear for clinical studies whether depression operates most powerfully on encoding or retrieval processes. Evidence from mood induction studies has a unique value in disentangling effects of mood at encoding and retrieval, and indicates that both encoding (Ellis, Thomas & Rodrigez, 1984) and retrieval (Ellis et al., 1985) are affected, though the encoding affect is probably more powerful (Leight & Ellis, 1981). One recent mood induction study, that opens up a promising line of enquiry on the effects of structure, manipulated mood at encoding and also varied whether or not stories were presented with titles. Depressed subjects were at a particular disadvantage when stories were presented without titles (Ellis et al., in press). General Issues The final section of this chapter will be concerned with three general issues: formulations of cognitive impairments in emotional disorders in Page 65
terms of processing resources; the role of worry and ruminations in impairing performance; and the possibility of using remedial strategies to improve performance. Processing Resources As suggested at the outset, an account of the performance deficits found in depressed patients can be offered in terms of depleted cognitive resources. Ellis & Ashbrook (1988) set out the assumptions of such an account. 1. Depression can affect the amount of capacity that can be allocated to a given task, through some capacity being tied up in thinking about one's sad state (extra-task processing), and perhaps also because processing capacity is directed towards task-irrelevant features (taskirrelevant processing). 2. The encoding of information, both in everyday and laboratory tasks, normally requires allocation of capacity (i.e. 'cognitive effort') and performance will be positively correlated with the effort allocated.
3. Where depression is mild, or where the capacity required by a task is only moderate, the task performance may be unimpaired; where depression is severe and the task is demanding, performance will be affected by depression. Krames & McDonald (1985) have also advanced the hypothesis that the performance deficits found in depression are comparable to those produced in normal subjects by adding a secondary task. Eysenck (1979a) and Eysenck & Calvo (1992) have advanced a capacity theory of the effects of anxiety but, following Kahneman (1973), they use the concept of effort somewhat differently. They start with the parallel assumption that capacity is taken up with extra-task processing (e.g. worry), an assumption shared by other theorists (Wine, 1980; Sarason, 1988). It is assumed that this reduces the capacity available for task performance. However, it is also assumed that potential adverse effects on performance may be compensated for by increased effort. This contrasts with the implicit assumption of Ellis & Ashbrook (1988) that in depression the reduction in the capacity available is paralleled by a reduction in the capacity allocated to the criterion task. Eysenck & Calvo's position leads them to locate the primary effect of anxiety, not on the quality of performance, but on 'effectiveness', which is the ratio of performance to effort. This distinction between performance and effectiveness raises the parallel question of whether the more appropriate capacity theory explanation of Page 66
the effects of depression on performance should be that fewer resources are available (perhaps because of depressive ruminations) or that fewer resources are deployed (perhaps because of a motivational deficit). Capacity theory has difficulty in satisfactorily measuring or operationalising resource allocation with sufficient precision to resolve this point. As Navon (1984) has pointed out, most of the current paradigms from which inferences about processing resources are made are capable of alternative explanations. It is significant that the different formulations of Eysenck & Calvo (1992) and Ellis & Ashbrook (1988) were concerned respectively with anxiety and depression, and it seems plausible that cognitive effort is affected differently in these two states. Eysenck & Calvo (1992) have assembled a variety of evidence that is consistent with the hypothesis that anxiety is associated with increased effort, and that anxious people have fewer undeployed resources in reserve. First, and perhaps most compelling, are dual-task studies. A valuable lead was given by Hamilton's finding that secondary tasks are adversely affected by anxiety more than primary ones (e.g. Hamilton, 1983). The design used by Calvo and his colleagues starts by finding a task that is performed comparably by anxious and non-anxious subjects when it is the only task. A further task is then added that has to be performed to a certain standard. If anxious subjects are maintaining performance on the first task at the cost of expending greater cognitive effort, their performance should be more adversely affected by the addition of an extra task. Eysenck & Calvo (1992) assemble evidence from several different studies to support this prediction. Also relevant to Eysenck & Calvo's theory about cognitive effort in anxiety are studies using self-report indices of subjective effort (e.g. Dornic, 1977, cited by Eysenck, 1992). Yet another supporting line of evidence is the fact that, though ego-involving instructions tend to improve the performance of less anxious people, they have little effect on the more anxious subjects (e.g. Nicholson, 1958). This also suggests that anxious subjects are already making cognitive effort close to their maximum. Finally, the suggestion that anxiety is associated with greater
effort can also form part of an explanation of the intriguing finding that anxiety is often associated with poor performance on hard tasks, but with better performance on easy tasks (see Eysenck, 1982). Though the evidence suggesting that anxiety leads to enhanced effort is not conclusive, the case is a plausible one and the conclusion probably correct. Because of increased effort, anxiety may often not bring about an immediate reduction in cognitive performance. However, there may be Page 67
longer costs in terms of the strain of having to maintain performance on the basis of high effort. Hockey (1986) has presented a similar theory of the effects of stress on performance in terms of increased effort to preserve performance levels in the short term, but with longer-term physiological costs. Increased effort seems much less likely to apply in depression. It is an interesting question how depressed subjects would perform in the kind of dual-task studies Calvo has used with anxious subjects. Unfortunately, there seem to be no studies in which depressed and anxious subjects have been compared directly. However, Krames & McDonald's dual-task study of depression produced quite different results from what would be expected with anxiety. In normal subjects, the addition of digit load to a word-learning task produced the expected deterioration in performance. In contrast, depressed subjects showed improved word learning performance as a result of adding a digit load. This is reminiscent of the dual-task studies on retardation reviewed in 'Secondary Tasks and Depressive Retardation' on page 48. The motivational deficit often assumed to be associated with depression would lead to the view that reductions in available capacity are exacerbated by poor allocation of even such resources as are available. Though direct assessments of this are difficult, the prediction would be that the kind of evidence Eysenck cites for increased effort in anxiety could not be paralleled in depression. Nevertheless, a commonality between anxiety and depression may be that both chronically overestimate the difficulty levels of tasks (see Kukla, 1972). The likely effect of this is that they overestimate the amount of effort that will be required to achieve any given level of success. Such an Expectancy x Value account has been successfully used by Williams & Teasdale (1982) to account for some laboratory helplessness deficits, and this is a matter that may repay further investigation. Experiential Correlates of Performance Deficits It has been noted that current theories of the effects of anxiety and depression on performance assume that these states lead to ruminations and worries that occupy cognitive resources, leaving less for task performance. There is general agreement that anxiety and depression are associated with an increased incidence of task-irrelevant thoughts. Some of the studies claiming to support this have used rather indirect measures of task-irrelevant thoughts, but it has also been supported by studies using Page 68
more direct measures (e.g. Ganzer, 1968). The evidence is stronger for anxiety than for depression, and for non-clinical than clinical studies, though Watts & Sharrock (1985a) reported that depressed patients showed considerably more mind wandering than normal controls. What is less clear is whether these task-irrelevant thoughts mediate the deleterious effects of anxiety and depression on performance. Support for this idea came initially from evidence that scores on a questionnaire measure of
worry correlated negatively with task performance, whereas other aspects of emotionality did not so correlate (Doctor & Altman, 1969). However, there are now sufficient studies finding a lack of correlation between task-irrelevant thoughts and task performance to justify a reconsideration (e.g. Galassi, Frierson & Sharver, 1981a, 1981b; Galassi, Frierson & Segal, 1984; Bruch, Jaster & Kaflowitz, 1983). Bruch, Kaflowitz & Keuther (1986) found that what correlated with test performance was not the frequency with which task-irrelevant thoughts occurred, but the subjective meaning attached to them. The 'worry' scale used in most of the early studies did not make this distinction. This suggests a reformulation of the hypothesis in terms of the meaning rather than the frequency of emotional thoughts. Some researchers have worked on the parallel assumption that the poor performance of depressed patients is due to their negative thoughts. In a rare study, Seibert & Ellis (1991) showed that unhappiness was associated with a high level of task-irrelevant thoughts, which were in turn correlated with poor performance on a memory task. However, somewhat puzzlingly, happiness was equally associated with a high level of task-irrelevant thoughts. Watts, Macleod & Morris (1988a) examined the correlation between task-irrelevant thought and objective performance in clinically depressed patients. The results indicated that it may be important to distinguish between tasks. Poor performance on some tasks was correlated with task-irrelevant thoughts, but other tasks showed different phenomenal correlates. Watts and colleagues distinguished between two phenomenally distinct forms of lapse of concentration which were uncorrelated with each other, (a) mind wandering and (b) going 'blank'. Subjects who reported high levels of mind wandering had poor performance on memory for prose, as might be expected if Wine's (1980) theory of test anxiety were extended to depression. However, performance on a planning task, the 'Tower of London' (Shallice, 1982), showed no correlation with mind wandering but was instead related to patients' reports of going 'blank'. The different phenomenal correlates of the two tasks could have been partly due to their having different structural characteristics (such as more 'subroutines' in the planning task). However, they could also have been due to the differences in the amount of effort required. When memory for prose was made more effortful by adding a requirement to form imagery, it showed a tendency to correlate with 'blanking' rather than 'mind-wandering'. Possibly, people who are vulnerable to 'blanking' have particular difficulty with effortful cognitive tasks; they may try to avoid them because attempting to apply effort tends to result in their going 'blank'. Whatever the truth in such speculations, it is clear that, though the assumption that performance deficits are associated with the presence of task-irrelevant thoughts may be correct for some tasks, it is not correct for all. This is a conclusion that is readily consistent with recent trends in evidence regarding dual-task performance. It has become clear that the extent to which one task interferes with another depends not just on the amount of general resources they both require, but in large measure on the extent to which they overlap in the specific resources required (Allport, 1980). It is therefore not at all surprising that mind wandering should be more incompatible with some tasks than others, depending on their specific properties. Remedial Effects of Processing Strategies Finally, we shall consider the use of strategies to remedy the cognitive impairments found in emotional disorders. Even where other treatments for general mood state are being used, there may be a case for using cognitive remedial strategies where problems of cognitive performance are a particular concern. For example, treating examination anxiety in students generally only lowers subjective anxiety but does not also combat the poor academic
performance associated with anxiety (Finger & Galassi, 1977). Separate cognitive interventions are needed for that. Similarly, there may be depressed patients in whom memory problems are unresponsive to more general treatments, so that specific remedial strategies will be of practical importance for the patient. Reducing cognitive deficits may thus contribute to a patient's general clinical improvement. Concentration problems are pervasive in depression and cause considerable frustration (Watts & Sharrock, 1985a). Reducing them may thus cut out one of the factors that maintain negative mood state, just as treating negative thoughts cuts out another such factor. There have been numerous demonstrations (e.g. Cermak & Craig, 1979) that strategies designed to promote 'deeper' processing improve memory in normals, and it seems that they may also do so in subjects who are Page 70
anxious or depressed. The characteristic description that depressed patients give of their difficulties in reading is that they cannot 'take in' what they are reading, with the result that when they get to the bottom of a page they have no idea what they have just read. This seems to be a phenomenological account of poor semantic processing. However, the application of processing strategies to help depressed patients concentrate differs from their standard laboratory application. For example, most laboratory research has examined the effects of processing strategies on incidental rather than intentional learning. In normal subjects incidental learning is probably more sensitive to the effects of processing, though the incremental benefit of processing strategies in intentional learning may be more marked for depressed than non-depressed subjects. When required to learn, most non-depressed subjects may spontaneously adopt good processing strategies while depressed patients fail to do so. Though there is a good deal of clinical interest in such studies, it seems unlikely that comparisons of the effectiveness of such strategies in normals and emotional groups will contribute much to a theoretical understanding of the nature of the processing deficit associated with emotional states. The reason for this is that, as with structure in materials, it is unclear what prediction ought to be made from the hypothesis of a processing deficit in emotional groups. On the one hand, depressives might benefit more from processing instructions because such instructions would remedy a natural processing deficiency. On the other, it might be predicted that depressives would benefit less because of their inability to do deep processing. In general, the evidence confirms that processing instructions improve the memory performance of anxious and depressed subjects. Two experiments have investigated the effects of imagery instructions in high-anxiety students. Edmunson & Nelson (1976) found that all subjects performed better at paired associate learning with instructions to form images than with instructions to repeat the word pairs. However, puzzlingly, Scott & Nelson (1979) failed to replicate this. Conflicting results have also been found in experiments on depressed subjects. Ellis, Thomas & Rodrigez (1984) carried out a similar study using a group of normal subjects who had been given a depressive mood induction and found that semantic processing instructions helped subjects in sad mood as much as subjects in normal mood. However, the results of Weingartner et al. (1981, Exp 1), in a study on depressed patients, cast doubt on this conclusion. Their results are incompletely reported, but it appears that, with cued recall, semantic processing instructions helped both depressives and controls, whereas, with free recall, they helped normals but not depressives. However, the latter seems Page 71
likely to have been due to a floor effect. Watts, Macleod & Morris (1988b) examined the effects of imagery instructions on memory for prose (descriptions of a department store) in
relatively severely depressed patients. Imagery instructions substantially improved memory performance, especially in non-endogenous depressives, though it did not affect the number of times patients actually reported losing concentration while listening to the passage, or their later ratings of the concentration problems they had experienced while listening to it. Recent work of Hertel and her colleagues has focused on the value of seeking to improve the performance of depressed patients by providing them with effective processing strategies. The assumption is that, whereas normal subjects are more likely to use these spontaneously, depressed patients need to be guided to use them. Hertel & Hardin (1990) found that the poor performance of depressed patients disappeared when they were supplied with cognitive strategies. This finding obtained both for a spelling task and for recognition memory. Hertel & Rude (1991b) also found it for a sentence completion task. Where subjects were required to repeat the word at the end of the sentence as a way of holding their attention to the task, the performance deficit of depressed subjects disappeared. None of these studies has looked at anything more than immediate effects of processing instructions, and much more work needs to be done to investigate whether they have any clinically useful long-term effects. One key issue here will be the potential applicability of the processing strategies to everyday situations. Clearly, many commonly used 'semantic' processing instructions, such as rating the meaningfulness of words, have very limited application. However, imagery is more flexible. For example, a patient who was finding it difficult to concentrate on job instructions at work might find it helpful to imagine the operations while they were being described to him. The evidence that depressed patients are deficient in structuring material at encoding, which was reviewed earlier, suggests that they might benefit from strategies designed to ensure adequate structuring. Structuring strategies can be applied to a variety of intellectual tasks, and Watts (1985) has discussed their application to study counselling. The main practical issue is the extent to which people who are anxious or depressed will actually apply remedial strategies in everyday life, a problem that is analogous to extending the effects of social skills training outside the training situation. It can hardly be expected that depressed patients will make consistent use of relatively effortful strategies such as imagery formation. Perhaps the best that can be hoped for is that they will find it useful to have strategies available to lift their performance in situations in which poor performance would be particularly frustrating.
Page 72
Chapter 4 Attention to Emotional Stimuli, I: Causes and Correlates One of the themes emerging from both Chapters 1 and 2 was the extent to which the cognitive system demands selectivity in processing. In Chapter 3 we reviewed evidence that one effect of emotional disturbance was to bring about a general impairment in cognitive performance, for example, concentration and memory deficits, even on neutral materials. Part of that effect could be explained by the task-irrelevant processing being done by the patient concurrently with the 'primary' task. It was not simply the amount of such secondary processing but its affective meaning and valence which explained impaired performance on neutral tasks. In this chapter, we concentrate on this tendency of emotionally disturbed people to find their attention drawn towards emotional stimuli. Introduction Consider the cases of two patients referred to a clinical psychology department for treatment of their anxieties. MM had a phobia of birds. She had lived with this fear for some years, but lately it was inhibiting her work. She was employed by a government department in a converted small aircraft hangar, and birds occasionally flew in and became trapped inside the building. She could no longer disguise her fear, and became extremely upset and tearful. On assessment it was found that her fear extended to both live and dead birds, her fear of dead birds inhibiting her from walking through markets or past butchers' shops where there might be poultry hanging for sale. The more that butchers' poultry looked like a dead bird (e.g. neck and head with feathers remaining) the more fearful she was. For our present purposes the interesting aspect of her fear was her extreme sensitivity to any stimulus in her environment which looked like a bird. When walking down the street, she would notice a dark, flapping shape Page 73
on the road many metres away and avoid it lest it should turn out to be a dead bird. She would notice live and dead birds more frequently than did any of her family and friends, though sometimes she would be wrong, and a shape would turn out to be pieces of black plastic flapping in the wind. TS was a welder in a shipyard and had been off work for a considerable time owing to a chronic anxiety state. This would manifest itself in tremors and sweating when in the company of coworkers. Coffee-breaks and lunch time in the canteen were a special problem. The anxiety symptoms also affected his work as a skilled welder. As this began to deteriorate, it had an inevitable vicious circle effect on his mood and then again on his work performance. Of all his symptoms, what was particularly distressing for himself and his wife was the way in which he noticed bad news in the newspaper. Only a brief scan of a page would make his eye rest on the stories of muggings, assault and suicide. These were occasionally small articles at the bottom of the page, but his eyes seemed to find them first, setting up a long ruminative anxiety about harm coming to himself or his wife. He would discuss these issues with his wife who, though patient at first, began to tire of his constant worrying. This only served to make him more anxious. A feature shared by these two patients is their sensitivity to stimuli in their environment which represent their fear. We shall examine this phenomenon of 'attentional bias' in this chapter, first outlining some of our basic assumptions, then by examining how investigators have sought to bring the phenomenon under experimental control in clinical and subclinical groups. In the case
of each experimental investigation of the clinical phenomenon, we shall see if there exist parallel experiments on non-clinical groups in the cognitive psychological literature which might help to explain what the underlying processes are. The Nature of Attentional Bias First let us outline our assumptions. We assume that attentional bias can be said to have occurred when there is a discrete change in the direction in which a person's attention is focused so that he/she becomes aware of a particular part or aspect of his/her stimulus environment. We also assume that such a change (a) may take place in any sense modality; (b) is perceived as being passive or involuntary but can operate voluntarily; and (c) is normally perceived to be contingent upon a discrete change (onset or offset) in the 'internal' or 'external' environment of the person. Page 74
These assumptions serve only to begin the process of studying the phenomenon in greater detail. They beg many questions. For example, what is 'direction of attention'? Is it always 'passive and involuntary'? Does there always have to be a discrete change in the perceptual environment, or can an individual suddenly become aware of something that was there all along? Nevertheless, they do highlight three aspects that we assume (at least at the outset of this chapter) are present. First, that the 'attentional bias' with which we shall be concerned is something of which the person is aware. *There may be other sources of disruption to the person which are related to such attentional bias and which indeed may share common mechanisms. For example, one's mind may go blank while trying to read, a phenomenon which might be explicable in terms of one's attention being caught by something of which one is not aware. However we reserve 'attentional bias' to refer to an aspect of phenomenal awareness. The individual notices something, and he or she knows that he or she has noticed it. Second, the 'environment' from which the stimuli are 'picked up' may be internal or external. This simply allows that one may be particularly sensitive to pain in the body, as well as shapes in the external environment. Third, although vision and audition have been the most studied, we should allow that similar phenomena may occur in any sense modality. So we may be sensitive to the smell of gas, the taste of cooked tomatoes, the sight of a dead bird, the sound of a spouse's car, or the bodily sensation of chest pain. Notice that examples of 'attention to salient material' may be drawn from everyday life and not just psychopathology. If one is buying a house, one seems to notice 'For Sale' signs. If one buys a new car, the number of people driving just that type of car suddenly seems to increase. The most natural analogy for such shifts in attention has been to liken attention to a beam of light, narrowly focused on one aspect of the world. The narrower the attention, the more concentrated or 'brighter' the light; the broader the attention, the less concentrated or 'dimmer' the light. Some monitoring also takes place outside the bright 'focus' (a dim illuminationa sort of 'night light'!). If any stimulus related to one's current concern occurs in this dim light, this peripheral light is brightened slightly at the expense of the 'focused' beam, or the entire focus may *Although we are concerned to explain a phenomenon of which individuals are aware, some of the psychological research which may elucidate the processes underlying the phenomenon do not themselves involve subjects being aware of what is distracting their attention. Where this research is
relevant, it is reviewed despite its apparent breach of our assumptions. Page 75
shift, so that what was once peripheral is now central. This latter shift is the attentional switch towards salience with which we are concerned. Such a spatial analogy has many ambiguities and there may be other better analogies (discussed by Broadbent, 1982), but it does suggest at least two ways in which individuals may differ from each other (or in which a single individual may vary over time). First, the degree of threat of the current concern represented in the periphery may vary. Second, the strength of resistance to switching may varyhow flexible or rigid, how easily distracted the system is. These two characteristics may not be mutually exclusive, but each needs to be borne in mind. For example, overtiredness may make concerns seem more distracting because they become more salient, and/or because switching becomes more probable even in the absence of concern-related stimuli. There is one more point that needs to be made before considering the experimental work. We have defined attentional bias partly in terms of the phenomenal awareness of the individual. But phenomenal awareness may involve both perceptual and response aspects of processing. Consider the stimulus array in Figure 1. In each of the three arrays, the word 'dead' appears in the same location. One can vary the discriminability of the word by varying the number and type of different irrelevant letters surrounding the target stimulus. One might expect a patient, such as TS mentioned above, to be quicker than average at spotting the word 'dead' even when others find it relatively difficult (such as in array (a)). Phenomenally, it appears that the 'perceptual' side of the cognitive system is 'tuning in' more quickly. However, the response demanded in such an experiment (saying the embedded word) allows an alternative explanation. That is, TS may have a 'response set' or 'response bias' towards particular words such as 'dead', 'suicide', 'accident'. His perceptual system may extract no more or less information from the stimulus array, but the output mechanisms may be biased to respond to a wide variety of stimuli with the concern-related word. We shall see later how some experiments have been unable to disambiguate the perceptual-bias from response-bias explanations
Figure 1 Stimulus arrays varying in discriminability of the negative word 'dead' (see text) Page 76
Two broad strategies for investigating attention bias have been use experimental investigations. The first is to show how the tendency to attend to salient stimuli may facilitate performance. The second has been to show how the same tendency may debilitate performance. So, for example, TS might pick out salient stimuli more
quickly if offered stimulus arrays such as those in Figure 1. But an alternative would be to ask him to pick out an alternative word (for example, a word containing a letter 't''fact' in array (b)). In this case we might expect TS to show debilitated performance because, the assumption will be, his attention has been drawn towards the concern-related stimulus 'dead'. Attentional Bias and Facilitated Performance Attentional Bias and Lowered Auditory Thresholds Parkinson & Rachman (1981) used a task in which subjects listened to taped music within which were embedded words related to a major current concern. Two groups of mothers were used in the experiment, matched for age. Mothers in the experimental group each had a child who had been admitted to hospital for tonsillectomy on the day of testing. Control mothers had a child the same age but their child was not being admitted. The two groups each listened to music in which three types of word (10 of each) were embedded at randomconcern-related (e.g. bleeding, injection, operation, pain); auditorily confusable neutral words (e.g. breeding, inflection, operatic, pine); and dissimilar words (e.g. newspapers, bird, pass, uniform). Subjects were asked to repeat each word as they heard it, the entire list of 30 words being played five times, starting at a low volume, and gradually increasing in volume until all words were clearly audible. Results showed that the experimental mothers, whose children were having operations, reported more concern-related words than controls at times 1, 2, 3 and 4 but in the fifth (loudest) presentation of the words this difference disappeared. However, the two groups did not differ on how many auditorily dissimilar neutral words they were able to report. An interesting aspect of the data is whether auditorily confusable neutral words (e.g. breeding) would behave like the concernrelated words or like the neutral words. Results showed that these words fell between the two. At the lowest volumes (times 1 and 2) the experimental group reported more of these words, but as the words were played more loudlytimes 3, 4, 5this difference disappeared. Page 77
This result is reminiscent of the patient who feared birds and would confuse any black, flapping material on the road with a dead bird when at a distance. Unfortunately, information about the experimental procedure given by Parkinson & Rachman is not very detailed. They do not, for example, report whether a response to an auditorily confusable item was counted only if the subject was accurate or if the subject merely heard anything. There are other unsatisfactory aspects of this experiment, such as the apparent non-matching of frequency of stimulus words. A further problem is that subjects were allowed to set the volume of music to their own comfortable level. It is possible that the more anxious mothers set themselves a lower volume, which would, of course, allow them to hear more words. Fortunately for the investigators, the fact that the experimental and control mothers did not differ in perception of neutral words allows them to escape some of these criticisms. There is one further problem with this paradigm, however, which has been alluded to in relation to Figure 1, and to which we shall returnthe task demanded of the
subjects that they respond by giving the word they had heard, so the data are consistent with a response bias interpretation. Klinger, Barter & Maxeiner (1981) report a number of experiments examining subjects' sensitivity to material related to current concerns. In one experiment, college students completed a Thought Sampling Questionnaire which had been found to reflect an individual's concerns. Some days later the subjects were given a 15-minute dichotic listening task. During the 15 minutes, twelve 25-second sections were inserted such that the material within the section would refer to the current concern of the individual. Subjects did not have to shadow either message, but were asked to move a toggle switch to indicate which message they were listening to at any moment. At intervals during the tape, a tone sounded and the subjects reported on their thoughts of the moment. Results showed that subjects spent more time listening to concern-related material than would be predicted by chance, and were more likely to report that they were thinking about their current concern when the tone had called for a thought report after the concern had been alluded to on the tape. The implication of this experiment was that something in a channel not currently being attended to was being 'noticed' and causing a switching of attention to that channel. The perception of concern-relevant target words embedded in either message in a dichotic listening task has been studied by Burgess et al. (1981). Six socially anxious and agoraphobic patients were compared with six control subjects and 12 highly fearful non-patients (i.e. people who identified having significant fears on the Fear Survey Schedule, six of whom were extroverts and six introverts). Both shadowed and non-shadowed Page 78
messages contained 10 occurrences of the neutral word pick. On another tape, shadowed and non-shadowed messages contained 10 occurrences of a word or a phrase chosen to be personally relevant for each subject (e.g. seminar, failure, shopping alone) on each channel. All target stimuli were embedded out of context. After some initial practice at shadowing, subjects were instructed to tap whenever they heard the specified target on either channel. They were told not to switch attention between ears but rather to tap 'if (s)he should happen to hear the target word'. The results showed that fearful subjects and patients differed from controls in their response to the targets in the shadowed message. The main hypothesis of the authors was confirmed; individuals were able to perceive and respond to concernrelated stimuli in the non-shadowed message without any loss of performance on responding to targets in the shadowed message. A problem with interpreting this result is that the stimuli that were chosen were likely to be words that are frequently in use by the subjects and patients. Thus frequency (which can itself be responsible for distracting attention) and emotional salience are confounded in this experiment. The issue of fearfulness vs familiarity of stimuli was examined by Foa & McNally (1986) who repeated Burgess and colleagues' design with 11 obsessionalcompulsive patients undergoing a three-week exposure and response-prevention treatment. Patients showed greater sensitivity on a behavioural (button-pressing) and skin conductance response to salient stimuli (e.g. urine, cancer, rabies) embedded in the unattended channel, than to neutral stimuli in that channel. The authors reasoned that the treatment would make these stimuli more familiar but less
feared, so that the fear hypothesis would predict less and the familiarity hypothesis greater difference in response to salient and neutral stimuli, respectively. Following treatment the difference between response to salient and neutral stimuli did not reach significance, which the authors took to be confirmation of the fear hypothesis. In fact, the proportionate differences between responses to salient and neutral stimuli were almost identical before and after treatment, and the appropriate statistics to test directly pre- to post-treatment changes were not performed. Their results are somewhat ambiguous, therefore. However, since a familiarity hypothesis may have predicted greater sensitivity to the salient stimuli, and this was not evident, the conclusion that it is the fear of the stimuli which renders it salient seems relatively safe. A similar issue regarding fear vs. familiarity will arise in relation to Stroop effects, to be discussed later. In fact, it will be seen later that it is not only 'fear' but relevance to current concern which is responsible for attentional bias. One indication of this was shown by Schotte, McNally & Turner (1990) in a dichotic listening task. They found that bulimic subjects were more able to detect the target fat
Page 79
than the alternative item pick in the unattended channel compared to control subjects. Despite these results, the dichotic listening paradigm had been criticised by Trandel & McNally (1987) following their difficulty in finding evidence of attentional bias for Vietnam-related words in combat veterans suffering from PTSD. All subjects in their experiments were equally disrupted by the threatening material. If the paradigm was to be useful, it needed modifying to make it more reliably sensitive to the specific fears or concerns of the clinical population being studied. One possibility is to use a shadowing paradigm in which there are no targets to be responded to in the non-shadowed message and in which there is an independent way of assessing attentional focus. If an individual is constantly monitoring stimuli in the non-shadowed message, and if when this monitor picks up threatening or concern-related material that attention tends to shift, then there may exist an attentional 'draining' phenomenon in which attention to focal stimuli is impoverished, though not yet shifted away completely. How should such subtle attention shifts be measured? A paradigm incorporating a secondary 'probe' stimulus (Posner & Boies, 1971; Bargh & Pietromonaco, 1982) was used by Mathews & MacLeod (1986) to examine attentional 'draining'. They compared the performance of anxious patients and matched normal controls on a dichotic listening task where one message was shadowed. The other, unattended, was made up of lists of words which were either neutral or threatening. In addition to assessing the effects of type of unattended material on shadowing performance, subjects' RT to a visual probe on a VDU screen in front of them was measured. Results showed that subjects were unaware of the identity of the words in the unattended channel as assessed both by random interruptions of the tape with instructions to recall the non-shadowed words, and by a surprise recognition test following the tape. Despite this, anxious patients were slower to respond to the tone when the words on the unattended channel were threatening, though not when they were neutral. What is the relevance of such phenomena for the sort of attention to salient stimuli we have been considering? It has been known for some time that unattended material is processed semantically. Lewis (1970) showed that when an unattended word is a synonym of a simultaneously presented word in the shadowed ear, shadowing slows down. Further work by Underwood (1977) showed that a context
presented in the unattended ear facilitates shadowing of target words which mesh with the context. There have been many interpretations of these sorts of result (see LaBerge, 1981, for a review). One possibility is that Page 80
the meaning of stimuli are processed without drawing on attentional capacity, but that perceptual identification of the source or location of that information demands follow-up by attentional processing. This suggests that the means of distributing attentional resources within the cognitive system, though normally under voluntary control, can be overridden by the occurrence of salient stimuli. If the salient stimuli are under current attentional focus, voluntary attempts to switch away will tend to fail. By contrast, if the salient stimuli are outside current attentional focus, voluntary attempts to remain focused will tend to fail. This analysis shows up an ambiguity in paradigms which use dichotic listening tasks. For maintaining attention in the light of distraction may require greater effort, which in turn affects shadowing and RT performance, rather than these being affected by depleted attentional resources themselves. It may, of course, amount to the same thing in the case of 'attentional draining', but we should be aware of the ambiguity lest it has implications for other experiments or for possible remedial strategies. Attentional Bias and Lowered Visual Thresholds In reviewing the studies of perceptual pick-up of concern-related stimuli in the auditory system, we saw how a threshold paradigm (Parkinson and Rachman's (1981) experiment on mothers whose children were having tonsillectomies) suffered from not being able to distinguish between perceptual and response-bias explanations. In an analogous 'visual' paradigm, Small & Robins (1988) studied thresholds for identifying emotional words following a depressing Mood Induction Procedure (the Velten). They found lower thresholds for depression-related words (e.g. gloomy, sad) and elation-related words (e.g. adore, lively) compared to neutral words. The fact that the effect was shown on elation words strongly suggests conceptual level priming of the antonyms of the depression words. The authors also point out that, whichever interpretation is adopted, the result might be due to response-bias effects, rather than reflecting a genuinely perceptual effect. Similar experimental paradigms on visual modality have been used in experiments on 'perceptual defence', though most tend to share the same interpretive problems. In the next section we shall review some attempts to use experiments in the visual system which overcome these ambiguities, but first we shall consider the question to what extent the visual perceptual defence literature is relevant. In the typical visual perceptual defence experiment, threatening or non-threatening words are presented for successively longer and longer Page 81
intervals in a tachistoscope until they are successfully identified. Typically, taboo words or threatening words have been used, and 'defence' is said to have occurred when a threat word has a higher recognition threshold. (See extensive reviews by Erdelyi (1974) and Dixon (1981).) A variety of arguments has been proposed which attempt to attribute this perceptual defence phenomena to factors other than the emotionality of the stimuli, but most such criticisms have proved difficult to sustain. One early suggestion was that frequency differences between taboo and non-taboo material may account for the findings (e.g. Howie, 1952; Solomon & Postman, 1952; Postman, 1953; Howes, 1954). However, this argument was largely based on comparisons of frequency counts for such items in standard word-frequency texts, such as the ThorndikeLorge (1944), which has subsequently been shown grossly to underestimate the frequency of taboo items (e.g. Eriksen, 1963). Furthermore, many studies which specifically controlled for frequency or eliminated potential frequency confounds by conditioning affect to novel
meaningless stimuli, have reported a significant influence of negative affective tone on perceptual threshold (e.g. Dulany, 1957; Levy, 1958; Sales & Haber, 1968). An alternative explanation for the perceptual defence phenomenon was that it simply reflects a subject's relative expectancies for encountering taboo and neutral stimuli, rather than representing a specific perceptual reaction to affective tone (e.g. Postman, Bronson & Cropper, 1953; Freeman, 1954). Again, however, many studies find a relationship between affective tone and perceptual threshold when differential expectations are directly controlled, or made unlikely (e.g. Bootzin & Natsoulas, 1965; Dorfman, Grosberg & Kroecker, 1965). The most influential, and initially compelling, criticism of the perceptual defence hypothesis has been based on the argument that the disproportionately low correct report rates found for emotionally threatening stimuli arise because subjects are less willing to report perceiving such stimuli until they are very sure they actually occurred. Thus the anxiety-provoking nature of the stimuli may produce a response bias rather than a perceptual bias (e.g. Goldiamond, 1958, 1962; Eriksen, 1963; Minard, 1965). While such a response bias may indeed exist, appropriate experimental and statistical techniques have confirmed that perceptual sensitivity appears to be genuinely influenced by the affective tone of the stimulus (e.g. Broadbent & Gregory, 1967; Dorfman, 1967; Bootzin & Stephens, 1967). It has been suggested that the degree to which such defence actually occurs may be subject to individual differences, with various writers proposing specific personality dimensions which are supposed to underlie such variationse.g. repression-sensitisation (Byrne, 1961, 1964); Page 82
deniers-non-deniers (Silverman & Silverman, 1964); blunters-monitors (Miller, 1987a, 1987b). Of particular clinical interest is the potential relationship between this personality dimension and manifestations of anxiety. Miller (1987a), for example, specifically argues that blunters, who engage in cognitive avoidance of threat-related cues, will show less stress and arousal than monitors. Indeed the correlation between Byrne's repression-sensitisation scale, and the Spielberger Trait Anxiety Inventory has been found to be 0.80 (Wason & Clark, 1984), which has led to the suggestion that this scale is probably simply a measure of anxiety (Cromwell & Levenkron, 1984). This would imply that anxiety may be associated with a weakening or reversal of perceptual defence, and there is indeed recent evidence that this is so in anxious patients (e.g. Mathews & MacLeod, 1985) though the data for depressed patients remain ambiguous (Powell & Hemsley, 1984). More recent work in this field has defined repression using a combination of low scores on a trait anxiety scale, together with a high score on a scale assessing social desirability motives (such as the Marlowe-Crowne). Most of this work has used normal college students, so it is not clear which results will apply to emotionally disturbed groups, but there is clearly a great deal of interesting research to be done using this new operational definition of repression. The research is reviewed in Singer (1990). Disruption and Facilitation: The Visual Dot-Probe Paradigm In the research on attentional bias reviewed above, the effect of concern related or threat stimuli is assessed by their potential to facilitate another task. In this section we examine a paradigm in which both facilitation and disruption due to threat material can be demonstrated. MacLeod, Mathews & Tata (1986) presented 16 'generally anxious' patients and 16 matched controls with pairs of words appearing simultaneously towards the top or bottom of a VDU screen (3 cm apart). The words appeared for 0.5 second and each subject's task was to name the top word each time out loud. On some randomly specified trials, a small dot appeared in the place where the word had been, and when this occurred, subjects were to press a button as quickly as possible. On half the trials on which the dot probe appeared, it replaced the top word, and on half it replaced the bottom word. (Navon & Margalit (1983) have confirmed that detection latency for such a probe
is a sensitive measure of visual attention. Subjects take longer to respond to a probe if their attention has been drawn elsewhere.) The interesting question that this paradigm can Page 83
answer is whether anxious patients differ from controls in this measure of visual attention when threat words occur at the top or bottom location. The results showed that anxious patients and control subjects showed a very different pattern of response, depending on whether the threat word was at the top or bottom, and whether the probe replaced the word at the top or bottom. If the probe replaced a threat word at the top, anxious subjects responded quickly relative to their response if the probe replaced a neutral word at the top and a threat word had occurred at the bottom. If, however, the probe occurred at the bottom after a threat word had occurred at the top, anxious patients were relatively slower to detect and/or respond to it. This pattern of results for the anxious patients suggests that they orient towards the location at which threat has occurred. Control subjects tend to show the opposite pattern, suggesting that they orient away from the location at which threat has occurred. Thus this experiment was able to demonstrate individual differences in response to threat stimuli, using a neutral response to a neutral probe stimulus. In order to investigate the specificity of these effects, the patients were divided into two groups on the basis of whether their concerns were predominantly physical or social in nature. The responses of these patients to threat words showed no tendency for different degrees of facilitation or disruption by stimulus components with different predominant concerns. What could be concluded from this first experiment using this paradigm? It is important to realise what it did not show. It did not show that anxious people were any more 'sensitive' to threatening material at the most peripheral level. Controls were 'picking up' the threat stimuli as early as the anxious patients. What differed was the attentional allocation pattern in response to these inputs. This experiment was one of the first to demonstrate that allocation of attention towards threatening stimuli found in attentional bias may be explicable without involving response bias. It would have been unparsimonious to suggest that the anxious subjects in this experiment had 'output logogens' for threat words which were nearer threshold, the 'firing' of which then activated an attentional response which could account for the differences in response latency to a neutral dot-probe. This experiment was influential in deriving the model we proposed in Williams et al. (1988). It showed that we needed to assume the existence of a decision mechanism which (a) was at a preattentive level, (b) was sensitive to general differences in threat, (c) allocated attention to different parts or aspects of the environment, and (d) was independent of response bias. In one of the first published replications of this effect, MacLeod & Mathews (1988) used the dotprobe task with medical students approaching an exam Page 84
ination. They were interested in the way in which trait anxiety might act as a vulnerability factor, showing up individual differences in attentional bias only when state anxiety was high, when the exam was near. The results were as predicted. A long time before the exam, there were no differences between subjects high or low in trait anxiety. A week before the exam, however, high trait anxious subjects were biased towards words to do with exams (speeded reaction to dot-probe following exam word in that location), whereas low trait anxious subjects showed the opposite tendency, being slower to react to the dot-probe when it replaced an exam-related word. Note that this interaction with level of trait anxiety occurred in the context of long-term stressa longawaited exam. The question remained whether equivalent effects would be obtained for shortterm, acute stress. This question was examined by Mogg et al. (1990). Mogg and colleagues (1990) used difficult
and insoluble anagrams to induce stress in medical students, testing them first on the Stroop task (see later) and on the visual dot-probe using general threat, specific achievement-related and neutral words. The stress succeeded in inducing higher levels of state anxiety, but the degree to which state anxiety had been raised did not correlate with attentional bias. However, the highstress group did show biased allocation of attention, retarded detection of the upper probe after a threat word had appeared in the lower position, and retarded detection of the lower probe after an upper threat word. That is, the acute stress had brought about an attentional bias towards threat. Although this result analogised the performance of anxious patients, there was no main effect nor interaction involving trait anxiety, suggesting that such short-term stress has a direct effect which does not require a predisposition to worry to show its effect. Trait anxiety effects are more likely to be present, on this interpretation, when there is sufficient time for worry to incubate, allowing this characteristic of high trait anxious people to show itself. Once the tendency for anxious people to worry has been successfully treated, however, will the attentional bias still be present? According to a study by Mogg, Mathews & Eysenck (1992) the answer appears to be that the bias is no longer present after recovery. Eighteen GAD patients were compared with the same number of recovered patients and age and IQ-matched controls. The results replicated the original MacLeod, Mathews & Tata, (1986) findings for the difference between patients and controls. Anxious patients were faster to detect probes in the location of threat. Furthermore, the extent to which this bias occurred correlated with levels of trait anxiety, confirming the chronic nature of their psychopathology. That is, when currently suffering clinical levels of anxiety, it is those most prone to worry who show the largest bias. This bias was most apparent for those who had social anxiety on words representing social Page 85
threats. The extent of self-rated social concerns correlated (r = 0.50) with bias on social threat words but not with physical threat words (r = -0.27). However, there was no significant difference between the performance of recovered patients and controls. Once current anxiety has dissipated, the bias is no longer apparent. This, of course, leaves open the possibility that further stress and worry will bring about a recurrence of clinical anxiety, and that the high trait anxious individuals will be more likely to respond to stress with attentional changes around the theme of their predominant worry. Attentional Bias and Debilitated Performance in the Stroop Test So far we have considered experiments in which the dependent variable was chosen so that attentional bias would facilitate performance, at least on some aspect of the task. In the remainder of the chapter we consider the strategy which has been most frequently used to study attentional bias: a task where performance on the dependent variable suffers as a result of the allocation of attention to alternative, more salient aspects of the stimulus array - the emotional Stroop test. We focus on the Stroop not only because it has been the most widely used task in the cognition and emotion literature (thereby yielding more data on its causes and correlates) but also because there is a larger body of research on the non-emotional paradigm, which allows more extensive discussion of possible mechanisms. In the original version of the task (Stroop, 1935), a subject is required to name the colour of ink in which an item is printed while attempting to ignore the item itself. The items in question are meaningless stimuli such as rows of 'X's, or actual names of colours. In the latter case, a word such as 'red' would appear in green ink, the word 'brown' in red ink, and so on. The most common finding has been that it takes subjects longer to name the colours when the base items are antagonistic colour names than when they are rows of meaningless stimuli (see Jensen & Rohwer, 1966; MacLeod, 1991).
Subsequent research has found that most words produce some interference (Klein, 1964), especially if the word is associated with a colour (e.g. sky, grass; Scheibe, Shaver & Carrier, 1967). Warren (1972, 1974) found that more interference was produced in naming the colour of a word if it or its associate had recently been presented auditorily to the subject. Geller & Shaver (1976) found that more interference was produced in the colour naming of self-referent words if a subject had to perform the task in front of a camera and mirror which acted to increase self-awareness. Page 86
During the 1980s, several investigators started to use Stroop-like tasks to examine cognitive processing associated with emotional disturbance. The Stroop was used to measure 'construct accessibility' (Gotlib & McCann, 1984; Williams & Nulty, 1986), 'activation of danger schemata' (Mathews & MacLeod, 1985), 'emotional salience of words' (Watts et al., 1986b), or simply 'distraction by emotional stimuli' (Ray, 1979; Williams & Broadbent, 1986b). Despite these differing labels, each of these studies had in common the measurement of latency to name colours of negative affect words (either series of individual words presented in a tachistoscope or columns of words presented on a single card). Studies used both neutral and positive words as control stimuli, compared performance in 'disturbed' groups with non-disturbed controls, and selected emotional words which are specific to the psychopathology under study. Using this paradigm, Gotlib & McCann (1984) found that mildly depressed subjects were significantly slower to name the colours of negative words than positive or neutral words. Mathews & MacLeod (1985) found that anxious patients were slower in colour naming threat words, with those with physical worries being particularly disrupted on the physical threat words. Watts et al. (1986b) found that spider-avoidant subjects, although showing little disruption on general emotional words compared to control subjects, nevertheless showed large disruption in colour naming spider words. Similarly, Williams & Broadbent (1986b), studying overdose patients, found greater disruption in naming words which were more specifically related to psychopathology (e.g. overdose, drug) than for more general emotional words (e.g. immature, helpless). Despite the finding that disruption was greater for words referring to specific rather than generally negative stimuli, it remained unclear to what extent emotional Stroop disruption reflected state or trait emotion. Williams & Nulty (1986) had found some evidence for a trait effect in mildly depressed subjects, suggesting that in some circumstances the emotional Stroop may be assessing residual effects of previous emotional disturbance. These results contrasted with Mathews & MacLeod's (1985) finding that colour-naming disruption on emotional words was predicted best by current state emotion levels. However, since the Williams & Nulty study was concerned with depression, and the Mathews & MacLeod study with anxiety, it was possible that both the degree of colour-naming disruption and the stability of the disturbance may have depended on an interaction between type of emotional disturbance (e.g. anxiety vs depression) and types of word used in the Stroop task. Table 1 summarises the studies using the emotional Stroop. Further details of the effect sizes in these studies, together with an extended discussion of Page 87 Table 1 Studies on anxiety and depression which have found colour-naming to be disrupted by words concerned with theme of the psychopathology Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
Anxious (12)
Anxious
High trait anxiety Dawkins &
Anxious Furnham (1989)
'Repressor'(12) Low anxious(12) (no detail on level of anxiety)
Repressor
Richards & Millwood (1989)
Anxious (16) Low anxious (16)
Anxious (threat words)
Mogg et al. (1990)
Anxious (18) Low anxious (19) Ss either received stress vs no stress
Anxious High stress Ss on achievement words
Richards & French Anxious (13) Low anxious (14) (1990)
Anxious (threat words)
Mogg & Marden (1990)
Anxious (12) Low anxious (12) Rowing club members (12) Non-rowers (12) (matched for anxiety)
Anxious (on all emotional words, including positive)
Martin, Williams & Clarke (1991) Experiment 1
High anxious (12) Medium anxious (12) Low anxious (12)
No significant effects involving group
Richards et al. (1992) Study 1
Anxious (but on threat words on Anxious (20) Low anxious (20) (Ss allocated to blocked presentation only) threat or positive mood induction groups)
Richards et al. (1992) Study 2
Anxious (a) on +ve words Anxious (20) Low anxious (20)(Ss allocated to following +ve MIP; (b) on -ve threat or positive mood induction words following -ve MIP) groups)
MacLeod & Rutherford (1992)
Anxious (23) Low anxious (24) (before and 6 weeks after exam)
Anxious Ss (prior to exam on masked trials only) Page 88
(table continued from previous page) Table 1 Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
High trait anxiety (continued) Mogg, Kentish & Bradley (1993)
Anxious (20) Low anxious (20) + Stress vs relax mood induction
Mood-induced 'relaxed' Ss (a) more interference on subliminal positive words; (b) more interference on supraliminal negative words
Fox (1993)
Anxious (18) Low anxious (18)
Anxious Ss on (a) traditional stim and (b) separated stim
Fox (1994)
Anxious (18) Low anxious (18) 'Repressor' (18)
No significant effects involving group
Anxious patients Mathews & Anxious out-patients (24) MacLeod (1985) Controls (age, sex matched)
(Anxious (threat words)
Mogg, Mathews
Anxious (threat words)
Anxiety state patients
& Weinman (1989)
(ICD-9) (18) Controls (18)
Martin, Williams GAD (DSM-III) patients & Clark (1991) (12) Experiment 2 Matched non-patients (12)
Signif. interaction: anxious patients more interference on threat words (vs controls showed facil, 27 ms)
Golombok et al. (1991)
GAD patients (DSM-III-R) (24) Controls (24) (age, sex, IQ) tested before and after either placebo or 10 mg diazepam
Marginally significant interaction: anxious patients more interference overall (NB: problem with analysis: diazepam placebo treated as between S)
Mogg et al. (1993)
GAD (DSM-III-R) patients Anxious patients (negative words, (a) supra and (b) subliminal) (19) Depressed (DSM-III-R) patients (18) Controls (18)
Mathews & Klug Anxiety neuroses (20) (1993) Controls (20) (age, sex, education)
Anxious patients (on (a) negative and (b) positive words related to problems: not on (c) positive and (d) negative words unrelated to problems) Page 89
(table continued from previous page) Table 1 Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
Anxious patients (continued) Mathews et al.(1995)
GAD (DSM-III-R) patients (24) tested before and after psychological treatment (mean 17 weeks' interval) Controls (23) (matched test-retest interval) (age, sex, IQ)
Martin, Williams Anxious patients (12) & Clark (1991) Controls (12) Experiment 4
(a)Anxious patients pretreatment (threat vs non-threat (No difference between interference for positive and neutral words) (b) Post-treatment; anxious patients no interference Anxious patients (on all emotion words, including positive)
Post-traumatic stress disorder McNally et al. (1990)
Vietnam veterans with PTSD (DSM- PTSD group (PTSD words) (no III-R) (15) difference in neutral, positive, Vietnam veterans without PTSD (15) obsessional words)
Foa et al. (1991)
Rape victims with PTSD (DSM-IIIR)(15) Rape victims without PTSD (13) Controls (16) (age/IQ)
PTSD patients (rape words)
Cassiday, McNally & Zeitlen (1992)
Rape victims with PTSD (DSM-IIIR)(12) Rape victims without PTSD (12) Controls (12) (age/IQ)
PTSD patients (rape words vs all other word types) (negative and positive words vs neutral)
McNally, English Vietnam Veterans with PTSD(DSM- Interference on PTSD words & Lipke (1993) III-R) (24) only) (no difference in neutral, Re-tested one week later positive, obsessional words)
Kaspi, McNally & Amir (1995)
Vietnam veterans with PTSD(DSM- Interference on PTSD words III-R) (30) only) (vs negative, neutral, Vietnam veterans without PTSD (30) positive words)
Panic disorder Ehlers et al. (1988)
Panic disorder (DSM-III) patients (24) Controls (24) (age)
Panic patients (physical threat words)
table continued on next previous page) Page 90
(table continued from previous page) Table 1 (continued) Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
Panic disorder (continued) McNally, Reiman & Kim (1990)
Panic disorder (DSM-III) patients (14) Controls; therapist' experts' in panic (14) (age)
Panic patients showed larger interference on all negative words than controls. Both groups showed more interference on catastrophe words
McNally et al. Panic disorder (DSM-III-R) (1992) (24) Obsessional Compulsive Disorder (DSM-III-R)(24) Controls (24) high vs low arousal manipulation (exercise)
Panic patients showed more interference for catastrope words than for other words (Controls, no difference between words)
McNally et al. Panic disorder (DSM-III-R) (1994) (16) Obsessional-compulsive Disorder (DSM-III-R)(16) Controls (16)
Panic patients (threat words)
Obsessive-compulsive disorder Foa et al. (1993)
OCD patients (DSM-II-R) (33; 23 with washing rituals; 10 with checking rituals) Controls (14) (age, sex, IQ)
IQ)
(a) OCD washers (on contamination words) (b) OCD checkers (on general threat words)
(a) OCD washers (on contamination words) (b) OCD checkers (on general threat words)
Lavy, van Oppen & van den OCD patients (DSM-II-R) (33) Controls (29) (age, sex, Hout (1994) education)
OCD patients (OCD negative words)
Social phobia Hope et al. (1990)
Social phobics (DSM-II-R) (16) (a) Social phobics (social Panic Disorder (15) threat words) (b) Panic patients (physical threat words)
Mattia, Heimberg & Hope (1993) Study 1
Social phobics (DSM-II-R) 29 Controls (50) (age, education)
(a) Social phobics (social threat words)
(table continued on next page) Page 91
(table continued from previous page) Table 1 Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
Social Phobia (continued) Social phobics (DSM-II-R) (29) treatment study 12 weeks phenelzine, placebo, or CBT)(17 responders, 12 nonreponders across three groups
(a) Treatment responders showed reduced interference on social threat words; (b) non-responders did not
Watts et al. (1986) Study 1
Spider-avaidant Ss (35) Control participants (18) (age, sex, education)
Spider phobics (spider words)
Watts et al. (1986) Study 2
Spider-avoidant Ss (28 of Ss from Study 1). N=14 in treatment group; N=14 in no treatment control
(a) Both treatment and (b) no treatment groups showed reduction in spider Stroop interference. Groups significantly different at second testing only
Martin, Horder & Jones (1992)
Spider-avoidant children (24) Spider-avoidant children (spider Control children (24): 3 age bands: 6/7 words) years; 9/10 years; 12/13 years
Mattia, Heimberg & Hope (1993) Study 2 Specific phobia
Mathews & Snake avoidant (18) Sebastian (1993) Controls (18) Experiment 1 All Ss exposed to 18 inch boa constrictor
No significant effects
Mathews & Snake avoidant (18) Sebastian (1993) Controls (18) different from Experiment 2 Experiment 1). No snake
Snake avoidant (snake words)
Mathews & Snake avoidant (40) (different from Sebastian (1993) Experiments 1 and 2) allocated to be Experiment 3 exposed to tarantula spider (20) vs no (20)
(a) Spider absent (equivalent to Experiment 2): significant Stroop interference for snake words. (b) Spider present: no effect
(table continued on next page) Page 92
(table continued from previous page) Table 1 Study
Subjects (N)
Group(s) showing interference
(emotional vs control words) Specific phobia (continued) Lavy, van den Spider phobics (DSM-II-R) (36) Pre-treatment interference (spider Hout & Arntz Controls(30) (age/education), both words) reduced at post-treatment (1993) groups tested twice (2.5 hours apart, during which phobics given exposure treatment). Depression Gotlib & McCann (1984)
Mildly depressed (Beck scores > Depressed Ss (depressive words) 9)(15) Non-depressed (Beck scores < 4)(15)
Williams & Nulty (1986)
Moderately depressed (19) Non-depressed (19) tested one year before Stroop and concurrently
Depressed (past) (negative words)
Gotlib & Cane (1987)
Depressed in-patients DSM-III-R) (34; major depression, N = 27; dysthymic disorder, N = 7) tested at admission and after discharge Non-depressed controls (14) (age) retested after matched interval.
Depressed at pretreatment (negative words) (No main effects or interactions for second testing session)
Segal & Vella Depressed patients (RDC) (18) (1990) Controls (14) Controls + self focus nouns manipulation (14)
Depressed group: self descriptive words more interference (a) vs neutral nouns (b) vs non-self descriptive adjectives (unrelated prime condition)
Klieger & Cordner (1990)
Mild-moderately depressed (21) Non-depressed (14)
Depressed group (negative words)
Hill & Knowles (1991)
Mildly depressed (>11 on inventory to diagnose depression)(12) Non-depressed (< 5 on IDD) (12)
No interaction of group and word valence
(table continued on next page) Page 93
(table continued from previous page) Table 1 Study
Subjects (N)
Group(s) showing interference (emotional vs control words)
Depression (continued) Segal et al. (1995)
Depressed (58)(RDC) Controls(44) (education,IQ) Idiographic selection of prime and target to be self-relevant vs non selfrelevant
Depressed group; interference for self-descriptive adjective primed with self-descriptive phrase (vs other adjective and prime conditions)
Self-relevant threat Ray (1979)
Final year students, 4 weeks before examinations (38) Low state anxiety (12) Mid anxiety(14) High anxiety(12)
MacLeod & Gynaecological out-patients Hagan (1992) (31)(awaiting appointment for
High anxiety (exam words)
Concurrent correlation between trait and state anxiety and
colposcopy to diagnose cervical pathology) (N=15 later received diagnosis of pathology)
interference on subliminal condition only. Subliminal (but not supraliminal) threat interference predicted dysphoric reaction to diagnosis (0.54, p < 0.05)
Riemann & McNally (1995)
45 students allocated to anxiety, elation or neutral film mood induction
High self-relevance (positive and negative) more interference than low self-relevance and neutral
Dagleish (1995) Experiment 1
Anxious(12) Low anxious(12) Ornithologists
(a) Anxious (threat words) (b) Ornithologists (bird words) (No difference between semantically related and unrelated neutral words)
Dagleish (1995) Experiment 2
Anxious (14) Low anxious (14)
Anxious (emotional words, both (a) threat and (b) positive)
Page 94
the processes underlying them, may be found in Williams, Mathews & MacLeod (1996). Together the studies provide impressive confirmation, across a wide variety of psychopathologies, that latency to name the colour in which an emotional word is printed is increased in people who are emotionally disturbed. Yet although the basic emotional Stroop phenomenon is now beyond doubt, precisely why it occurs is still far from clear. Understanding the causes of attentional bias in this task may give important clues to why such biases occur on other tasks. We shall therefore review the question 'why' in some detail. There are two sorts of 'why' question that can be asked. First, what are the cognitive mechanisms underlying the emotional Stroop phenomenon; what components or stages within the information processing system are giving rise to the interference? Second, what causes these mechanisms to be disrupted (e.g. characteristics of the person, such as level of trait or state emotion, or some interaction between person and situation). Most research has been concerned with demonstrating Stroop interference as a marker of attentional bias. Some studies have also tried to identify the emotional causes and correlates of interference. Very little research has addressed the question of mechanism. We shall therefore focus on the larger sample of studies which have examined the causes and correlates of the emotional Stroop in this chapter, and leave the question of mechanism until the next chapter. Causes and Correlates of Emotional Stroop Interference Three sources of variance contributing to degree of emotional Stroop interference have been studied: variance due to trait or state emotion; variance due to the specific nature of the negatively valenced word stimuli used; and variance due to the particular situation in which the task is performed. In these studies, both clinical and analog groups have been examined. For example, some have focused on contrasting patients with diagnosed General Anxiety Disorder (GAD) against non-disordered control subjects. Others have contrasted high trait anxious against low trait anxious non-clinical subjects. The issue of clinical pathology specifically, and trait vulnerability in general, are considered together in this chapter, since often both sources of individual difference have been examined in the same
studies (e.g. Martin, Williams & Clark, 1991). Trait versus State Emotion Mathews & MacLeod (1985) had found that Stroop interference was correlated directly with state anxiety rather than with trait anxiety, but Page 95
conclusions from other studies were not so clear cut. Subsequent studies have continued to investigate the relative involvement of state and trait variables by (a) the further study of trait and state inventories, (b) the study of recovered patients (assumed still to have the elevated trait emotion but to have reduced state emotion) and (c) the study of subjects tested prior to examinations or following experimentally induced failure experience. Studies using anxiety inventories. Richards & Millwood (1989) found that high trait anxious normal subjects were disrupted in colour naming negative words (compared to colour naming neutral or positive words) but these researchers did not report analysis of state anxiety effects (the high and low trait groups differed on state anxiety also). As part of a study of repression, Dawkins & Furnham (1989) compared high and low trait anxious groups. The stimuli used in their version of the emotional Stroop task were words representing strong emotional expression chosen to be particularly salient to repressors (cry, sob, scream, kick, punch, tense, plead, blush, beg, etc.). Colour-naming latencies shown on these words were compared with those shown on neutral words (no positive words were used). The high trait anxious subjects showed interference on these negative words while the low trait anxious showed none. However, once again, no analysis was conducted to determine whether state anxiety may have mediated this group difference, In a clinical study which took account of both trait and state anxiety, Mogg, Mathews & Weinman (1989) contrasted emotional Stroop performance of patients with a diagnosis of GAD against that of matched normal controls. The stimuli were either non-threatening (holiday, contented), physically threatening (disease, mutilated) or socially threatening (failure, inadequate). Results showed the GAD patients to display slower colour-naming responses on the emotionally negative words, with those patients having greater social concerns showing greater disruption on socially threatening words, and those patients having greater physical concerns showing greater disruption on physically threatening words. In relation to the distinction between state and trait, Mogg and co-workers found that Stroop interference correlated with trait anxiety, but not with state anxiety, This appears consistent with the conclusion that the trait variable, at least for anxiety, most reliably correlates with the degree of colour-naming interference shown on negative words in the emotional Stroop task. However, Martin, Williams & Clark (1991) compared high vs low trait anxious normals and found no difference in emotional Stroop interference between these groups. In the same study, they examined emotional Page 96
Stroop performance in clinically anxious patients who had equivalent levels of trait anxiety to the non-clinical 'high trait' controls. The patients showed significant colour-naming interference on both negative and positive words relative to neutral words. This study therefore called into question whether it is high trait anxiety rather than clinical status which is associated with emotional Stroop interference.
In fact, there is now additional evidence indicating that Martin and colleagues' failure to find emotional Stroop performance differences between high and low trait subjects is unrepresentative. In addition to the studies reviewed above, Mogg & Marden (1990) and Mogg et al. (1990) replicated Richards & Millwood's (1989) finding that high trait anxious normals show greater colour-naming interference for threatening than for neutral words. Mogg, Kentish & Bradley (1993), in an experiment which examined both subliminal and supraliminal presentation of emotional stimuli, found that degree of interference in colour naming correlated with trait anxiety for subliminally presented, but not supraliminally presented stimuli. This raises the possibility that non-clinical, high trait anxious subjects may be able to override the tendency to be distracted by emotional material, accounting for Martin and colleagues pattern of differences between clinical and non-clinical subjects. Studies of recovered patients. In a study of recovery from emotional disturbance, Gotlib & Cane (1987) tested 34 depressed in-patients (with diagnoses of major depression or dysthymia), comparing them with 14 non-depressed controls. They found that the depressed patients showed greater colour naming interference on depressive words (compared to neutral and positive words) than did controls. However, when the depressed patients were tested again following clinical improvement that had resulted in their discharge, there was no difference between the time taken to colour name the negative words and the neutral or positive words. Therefore, for depression at least, emotional Stroop interference appeared to be due to current depressed mood state, rather than to a chronic vulnerability to depression. Mathews, et al. (1995) studied Stroop performance before and after psychological treatment for GAD. A previous study from the same research group (Mogg, Mathews & Eysenck, 1992) had found no difference in emotional Stroop interference between nonanxious controls and recovered anxious patients, but that earlier study was cross-sectional, so it remained uncertain whether the recovered patients had shown emotional Stroop interference when they had been anxious. The later study was longitudinal, thus it was possible to follow the same patients from pre- to post-treatment. Mathews and colleagues found (as Gotlib & Cane, Page 97
1987, had found for depressed patients) that pre-treatment differences between patient and controls in the degree of emotional Stroop interference disappeared by the end of treatment. However, it remains possible that elevated trait anxiety may predispose people to react to elevated state anxiety with disproportionately greater emotional Stroop interference effects. No study has yet given a stress test to recovered anxious patients to examine this hypothesis. The nearest experimental investigations of this issue have made use of trait and state measures alongside naturally occurring or experimentally induced stress manipulations. It is to these type of studies that we now turn. Manipulation of stress levels. Mogg et al. (1990) manipulated current state anxiety by the use of experimental stress. Subjects were given difficult or insoluble tasks, and the emotional Stroop task administered after this failure experience. No effects of state anxiety on pattern of colour-naming interference were found, and even the effects of the short-term stress were not mediated by levels of state anxiety. That is, colour-naming interference was greater for words representing achievement concerns following the failure stress, but this
effect was not correlated with the increases in state anxiety which followed the stress. By contrast, the level of trait anxiety was associated with the presence of emotional Stroop interference. Irrespective of whether people had undergone failure or not, high trait anxious individuals showed more interference for all threatening words (both general threat and specific achievement words). Therefore, in normals, there appears to be more evidence of trait involvement in emotional Stroop interference. The effect of the current context (failure experience) was not mediated by state emotion, but rather by the situational stress itself. MacLeod & Rutherford (1992) also examined the patterns of colour-naming observed as a function of trait and state anxiety, in a slightly different version of the emotional Stroop task. Their task employed very rapid word presentations followed in each case with a coloured pattern mask that subjects were required to colour name. Subjects were high and low trait anxious students who were tested on two occasions, once at a point in a semester when their state anxiety was low, and once in the week prior to their end of semester examinations when their state anxiety was high. These researchers found that the tendency to display colour-naming interference on negative words related to examinations, relative to positive words related to examinations, on this masked version of the emotional Stroop, was an interactive function of trait anxiety and stress. When state anxiety was low, the low and high trait groups showed equivalent patterns of colour-naming. When state anxiety was elevated Page 98
by exam proximity, however, the high trait anxious subjects displayed increasing colournaming interference on the negatively toned examination words. In contrast, the low trait subjects showed no such effect. It was not clear from this experiment whether the stressful event of exam proximity, or the emotional consequences of that event, was responsible for eliciting this differential pattern of colour-naming interference across these two groups. In summary, the balance of evidence suggests that individual differences in trait emotion (as assessed by questionnaire) is associated with individual differences in Stroop interference. However, trait differences appear to require some activation by current emotion or circumstances to show the disruption. This interaction is more likely to occur if the current circumstances have had time to incubate (for example, anticipation of having to take an examination) rather than if the current emotion disturbance is short term (for example, failure on an experimental task). The treatment/recovery studies with clinical groups are consistent with this conclusion, however, the lack of test-retest untreated clinical control group limits the weight that can be put on these data. Finally, we have seen that a complete model of the emotional Stroop may need to explain why high trait subjects sometimes do not show interference, whereas no study has failed to find interference in clinical patients. Emotional Stroop Specificity and Current Concern Martin, Williams & Clark (1991) found that anxious patients show interference in colournaming on both positive and negative words. Similarly, Mogg & Marden (1990) (in anxious subjects) and Cassiday, McNally & Zeitlen (1992) (in rape victims who had PTSD) used matched positive words as well as a range of threatening words. Both studies found increased colour-naming latencies on the positive words, compared to the neutral words in their emotionally disturbed groups, though Cassiday and co-workers also found that, for these subjects, threat words were even more interfering than were the positive
words. Can any strong emotional stimuli interfere with colour-naming in emotionally disturbed populations? If so, it may call into question those explanations of the emotional Stroop which rely on notions such as 'danger schemata'. One possibility is that it is not the emotionality of the words per se which is causing the effect, but rather the degree to which the words are semantically related to the 'schemata' of the individuals being tested. This would be consistent with the common finding of greater degree of interference on words which refer specifically to the theme of the emotional Page 99
disturbance, whether the disturbance relates to suicide (Williams & Broadbent, 1986b) spiders (Watts et al., 1985); rape (Foa et al., 1991; Cassiday, McNally & Zeitlin, 1992); social or physical threat (Mogg, Mathews & Weinman, 1989) panic (McNally et al., 1992); 'Vietnam trauma', (McNally, English & Lipke, 1993) or food (Lavy, van den Hout & Arntz, 1993). In each case, words of equivalent emotional valence, unrelated to the theme, were not so likely to produce interference as were specific theme-related words. Furthermore, in McNally et al.'s (1992) study, panic patients showed more interference on catastrophe words despite rating the positive words as more emotional. Clearly, level of emotionality of the stimuli will not by itself explain interference. Two separate issues arise from these studies. The first is whether relatedness of words to a person's current concern explains the colour-naming interference on emotional stimuli. To examine this we need a study which independently varies the degree of specific relevance to personal concern and the emotional valence of the word stimuli. The second issue is whether, if relatedness to current concern is found to be the more important variable, this effect merely reflects people's 'expertise' in processing information within that domain. Such individuals may use these Stroop words more frequently, or they may experience more intercategory priming within lists of such words due to their common membership of an idiosyncratic personal category. To examine this, we need to consider studies which provide therapy to emotionally disturbed individuals, thus reducing the degree to which the emotionally negative words will relate to currently active personal concerns while increasing a subject's exposure to, hence expertise in, processing information relating to the object of their fear. We examine each of these issues in turn. Relatedness versus Emotionality. Mathews & Klug (1993) tested anxious patients (a mixed diagnostic group of general anxiety disorder, panic, and social phobic patients) and normal controls using words which varied independently in their hedonic tone (positive/negative) and degree of relatedness to anxiety (related/unrelated). Thus some words were emotionally negative and related specifically to anxiety (shaking, crazy, nervous, panic); others negative but unrelated to anxiety symptoms (negative, sin, quarrel, destructive). Similarly, positive words were either related to the concerns of anxious people, by referring to traits that anxious individuals strive for (fearless, competent, relaxed, confident), or were unrelated to such concerns (beauty, brilliant, pride, mercy). There were also a set of matched neutral words, giving five sets in all. Each set of eight words were printed on cards, repeated 12 times on each card (96 words per card). Page 100
Results showed that it was the relatedness of the words to anxiety (either positive or negative) and not the emotional valence of the words which accounted for the patterns of
colour-naming interference. Anxious patients showed significantly more interference on anxiety-related words than anxiety-unrelated words, irrespective of hedonic tone, whereas normal controls showed no difference in colour-naming latencies across the different word lists. The authors suggest that differences in whether or not subjects in previous experiments showed interference, and whether such disruption was shown for positive words as well as for negative words, depends on whether the words happened to relate to the concerns of the subjects (either by referring to their feared situations, their anxiety symptoms, or to traits they actively desired due to their anxiety symptoms). Further evidence to support this conclusion has been found by Riemann & McNally (1995). They produced idiographic lists for their subjects, choosing words for a computer version of the emotional Stroop for each subject based on the results of an earlier administration of Cox & Klinger's (1988) Motivational Structure Questionnaire (MSQ). The MSQ lists 15 content areas (family and home; friends; marriage, relations, love and sex; physical health; mental and emotional health; employment, job and money, education; organisations; religion; government, politics and labour union activities; hobbies and pastimes; sports and recreation; entertainment; travel; and crime). Subjects chose the two most positive and the two most negative content areas for them at that time, and were then asked to write a brief explanation of the particular concerns they had within each domain. In this way, the researchers were able to choose words which varied in degree of relatedness to current concern (low and high) and in valence (positive and negative). Results showed that the words related to high negative current concern and high positive current concern produced more colour-naming interference than did low negative and low positive current concern and neutral words. Relatedness not a complete explanation. However, there is some evidence that different types of patient may show differences in the extent to which relatedness of material to personal concerns, rather than emotional valence of these materials serves as the source of colour-naming interference. McNally et al. (1994) found that emotional Stroop interference in panic patients was greater for specific panic words (fear, dizzy, anxious) than their near antonyms (safe, steady, carefree). These positive words were not associated with any greater degree of colour-naming interference than were neutral words. It could be argued that these words had not been chosen to represent the traits of particular relevance to individual patients. Page 101
However, other studies have chosen the words idiographically, yet found negative personally relevant stimuli to be more disruptive than positive personally relevant material. Lavy, van Oppen & van den Hout (1994) attempted to replicate Mathews & Klug's (1993) experiment with obsessive-compulsive patients. They used a 2 x 2 matrix of words; related/unrelated to obsessivecompulsive disorder and positively/negatively valenced. They also found colour-naming latencies to be increased on negatively valenced words for the anxiety patients, but specific relatedness of words to their disorder was not associated with any further elevation of colour-naming latencies in these patients. Similarly, Segal et al. (1995), following up a previous study (Segal & Vella, 1990), studied Stroop performance of depressed patients using positive and negative trait adjectives (such as 'trustworthy' or 'quarrelsome') that had been selected by the patients as self-descriptive. These words were presented to patients or matched non-depressed controls following primes denoting interpersonal situations such as 'able to feel close' or 'I often feel judged', a manipulation intended to activate the individual's self-schema. The results replicated Pratto & John's (1991) finding that negative trait
adjectives were in general more disruptive to colour-naming than were positive adjectives. Further, trait adjectives that had been rated as self-related took longer to colour name than trait adjectives that were not self-related, consistent with Mathews & Klug (1993) and Riemann & McNally's (1995) finding. But Segal and colleagues also found that, for depressed patients, being exposed to a prime which was self-relevant disproportionately delayed colour-naming of a self-relevant trait word if it was negative. They suggested that, although relatedness to personally relevant concerns is an important variable, the greater interconnectedness of negative current concerns for emotionally disturbed people may make words related to these concerns particularly liable to interfere with colour-naming. This study raises the second issue mentioned earlier: the possibility that this effect reflects expertise in processing such information. Is interference due to 'expertise'? In the standard emotional Stroop experiment, emotionally disturbed patients, who are known to ruminate on certain themes, are presented with words related to these themes which are likely to be frequently thought about, and which may come to form a highly interconnected category for these individuals (Segal et al., 1995). In short, they may have become experts in processing information related to their problem. Many of the above research findings might reflect simply this expertise: the increased frequency with which certain concepts are used, or the increased intercategory associations for the particular domain of interest. Page 102
Some studies which have examined expertise have not found that it contributes to colournaming latencies. Mogg & Marden (1990) found no more colour-naming interference on words related to rowing for members of a Boat Club (e.g. sculling) than unrelated words (e.g. teacup). However, this may have been because the rowing category was not very salient for these individuals; the rowing words may not have been used very frequently by these subjects, nor might the words have formed a very tight category for them. Consistent with the suggestion that the group to be examined must be genuine experts, Dalgleish (1995) has found that keen ornithologists show colour-naming interference for the names of rare birds. However, the problem with studying experts who are very committed to their field, and show high frequency of expert word usage and high intercategory associations among such materials, is that these stimuli are likely to be emotionally significant for such individuals. However, it is possible to separate emotional salience, or relevance to current concern, from expertise effects by studying the effect of therapy. In Watts et al.'s (1986) study involving therapy for spider phobia, for example, individuals were exposed to many more spider stimuli than they would have naturally exposed themselves to. Encouraged to observe and attend to the size, shape and other characteristic of spiders, such subjects were likely to have been much more expert on spiders at the end of treatment than at the outset. Yet colour-naming interference for spider stimuli within this phobic group was significantly reduced as the fear of spiders reduced across the course of treatment. A recent study by Mathews et al. (1995) found the same pattern of results in GAD patients undergoing group Anxiety Management Training. Colour-naming interference for threat words, which had been present in these patients before treatment, had normalised at posttreatment. Both these studies are consistent with findings from a treatment study of obsessive-compulsive patients which used a dichotic listening task to assess attentional bias towards threatening stimuli (Foa & McNally, 1986). These authors found that subjects were more likely to detect threatening words than control words before, but not after
treatment, despite the fact that treatment most likely made them more expert in processing information related to anxiety disorders. A problem with these experimental designs is that we do not know the extent to which practise effects on the emotional Stroop task may reduce colour-naming interference on the second occasion of testing. There is little doubt that considerable test-retest practise effects occurred in Watts et al.'s (1986) study and the lack of a no treatment control group in Mathews et al.'s (1995) study means that this issue cannot be addressed in their experiment. In the absence of such controls, contrasting emotional Stroop Page 103
performance in those individuals who respond to treatment against that of individuals who do not respond to treatment is a useful substitute. Such a comparison was carried out by Mattia, Heimberg & Hope (1993). Mattia and colleagues examined emotional Stroop interference on words representing social threat, in social phobics who were undergoing a 12 week group cognitive behavioural programme or phenelzine treatment. They found that treatment responders (as rated blind on a Clinicians Severity Rating) displayed a reduction in the amount of colour-naming interference shown on social threat words, whereas the non-responders displayed slightly increased colour-naming interference on such stimuli. Whether the effects of treatment on emotional Stroop performance are due to the effects of treatment on mood state, or to practise effects on the task, therapy studies remain a helpful way of addressing the 'exposure' vs 'expertise' account of interference. In each of the treatment studies, patients were exposed to stimuli related to the theme of their emotional problem. The fact that colour-naming interference on these stimuli decreased, along with the decrease in the emotional salience (or relevance to current concern) of the previously feared stimuli strongly suggests that frequency of usage or intercategory association due to expertise cannot provide a complete explanation for Stroop interference in emotional disorders. Situational Specificity of Effects The earlier discussion of the trait-state issue suggested that situational context should make a difference to degree of colour-naming interference shown on the emotional Stroop. We have already referred to the study by Mogg et al. (1990) in which students performed the task following a failure experience. This increased stress resulted in greater colour-naming interference on words related to an achievement theme. Such findings suggested that, consistent with current concern theory, any environmental manipulation which activates the concern should result in a greater degree of disruption of colour-naming concernrelated words. Similarly, manipulation of the 'internal' context of subjects by having them fast for 24 hours increases the amount of disruption shown in naming colours of food words (Lavy, van den Hout & Arntz, 1993). However, an important exception to these findings has been reported by Mathews & Sebastian (1993) who varied the context by having the object of subjects' fear either present or absent. They tested snake-avoidant subjects and controls on words related to snakes, or words related to a general threat or a neutral household theme. All subjects were tested in the presence of a boa constrictor (and were told there would be a behavioural test after the emotional Stroop task to see how close they could get to the snake). Under these conditions, they did not find, within the colour-naming latency data, the predicted interaction
between word type (snake-related, general threat and control words) and group (snake avoidant vs control). In a further experiment without the snake present, the predicted effect was obtained, snake-avoidant subjects showing greatest colour-naming interference on words related to snakes. In a third experiment focusing only on snake-avoidant subjects, the situational stress was re-imposed, but a large spider was substituted for the boa constrictor for half the subjects. When the spider was absent, subjects showed the predicted Stroop interference on snake words. Subjects tested with a spider present showed no such interference. The different pattern of results between this study and previous studies that have manipulated the situational context, may have arisen from a number of sources. For example, Mathews & Sebastian's study involved anxiety for specific threatening stimuli (snake and spider) associated with specific avoidance responses, which is not true of experimental or exam failure, nor of food deprivation. Ultimately, we need a study which uses more than one type of situational manipulation in the same experiment to examine such variables. Even so, the disappearance of an otherwise robust effect in Mathews & Sebastian's study using such a situational manipulation is a potentially significant finding to which we shall return. It provides further evidence that, in certain circumstances, colour-naming interference can be overridden by subjects. A model of emotional Stroop needs to be able to explain how this occurs. Causes and Correlates of Emotional Stroop Effects: Conclusions In examining the causes and correlates of emotional Stroop interference, we have seen that, whatever the mechanisms underlying the phenomenon, progress has been made in determining the conditions which activate those mechanisms. First, some explanations can be ruled out on the basis of the recent research. Before this work, it might have been argued that the emotional Stroop effects were due to momentary disruption in processing produced by momentary elevation of anxiety caused by seeing a threatening word. The fact that Stroop interference can be shown with positive words shows that this is unlikely to be the case. Second, the state vs trait issue has moved beyond simply asking whether it is one or the other. Current level of state anxiety appears not to be consistently linked with level of colour-naming interference, perhaps because state anxiety is a fairly crude measure of the state of a person's current concern. Elevated trait anxiety has more reliably been associated with colour-naming interference for threatening Page 105
material, and has been found to be associated with the magnitude of such interference both in clinically anxious patients and in subjects varying along normal dimensions of anxiety. Level of trait anxiety did not interact with either state anxiety or with the state induced by experimental failure to predict degree of interference in Mogg et al.'s (1990) study, though it does make subjects more likely to show interference with the longer-term stressor of anticipated examinations (MacLeod & Mathews, 1988; MacLeod & Rutherford, 1992). Recovery from clinical status is associated with decreased colour-naming interference, though no stress test has been done with such recovered patients to see if emotional Stroop performance might be a marker of latent vulnerability. However, colour-naming disruption is not the universal finding for non-clinical individuals differing in trait anxiety (Martin, Williams & Clark, 1991; Mogg, Kentish & Bradley, 1993). These discrepancies, together with the variation in situational determinants of Stroop disruption discussed in the previous section, yield data which a model of emotional Stroop needs to accommodate, and which will be discussed in the next chapter.
Page 106
Chapter 5 Attention to Emotional Stimuli, II: Mechanisms Underlying Bias Why do some emotionally disturbed people selectively attend to emotional stimuli? As we have seen in the previous chapter, this is an important question. We saw how a patient with a fear of birds found herself noticing any stimulus that looked like a bird, and then panicking, avoiding certain places and situations. We saw how another patient would notice suicide stories in the newspaper, which would start a prolonged rumination about death. Introduction We have distinguished between two related questions that can be asked about such effects: the question of cause, and the question of mechanism. The 'cause' question asks: in what type of person, in what mood, in what situation, or following what event is attentional bias shown? This question was the subject of the previous chapter. The 'mechanism' question asks: on what part of the information-processing system do these causes act to bring about attentional bias? This is the subject of the present chapter. We again use the Stroop as a prototypical example, but consider other paradigms where they contribute to our aim of narrowing down the possible mechanisms underlying attentional bias. Most investigators assume that emotional words in tasks such as the Stroop have their effects because they 'capture attentional resources'. The most common explanatory frameworks have been Beck's schema theory (Beck, Emery & Greenberg, 1985) or Bower's network theory (Bower, 1981). Emotional material is said to attract disproportionately more processing resources due to the activation of specific knowledge structures ('danger schemata') representing personal threats (Mogg, Mathews & Weinman, 1989). The fact that trait anxiety is more reliably Page 107
associated with interference than state anxiety is taken to indicate that a long-term schema for threat information is involved in producing interference rather than short-term, current arousal. However, we have seen that short-term threat caused by failure on a task can produce greater interference on words related to achievement, though the extent of this interference does not interact with level of trait anxiety (Mogg et al. 1990). The fact that this was found on both the Stroop task and (for general threat words) on the dot-probe task is reassuring about the generalisability of the phenomenon of stress-induced attentional bias across different paradigms. We therefore need a model which takes account of both long-term and short-term activation of emotional material which may sometimes interact, but which may sometimes have a simple additive effect on interference. Some Possible Explanations In the following section, several possible mechanisms will be discussed. Three of these, explaining interference as due to (a) distraction by task-irrelevant processes, (b) the cognitive effort expended in preventing emotional material from becoming conscious and (c) a general deficit some people may have in maintaining attentional focus, will be rejected as insufficient. We shall consider data showing that attentional biases do not necessarily lead to memory biases, and discuss the implication that interference arises from mechanisms earlier than, or independent of those involved in mnemonic encoding. Finally, we shall discuss whether the recent PDP model of the colour conflict Stroop (Cohen, Dunbar & McClelland, 1990) may provide a heuristic model within which
emotional Stroop phenomena may be explained, and suggest that opponent processes within such a model may account for dissociation between attention and memory biases when these occur. Task Irrelevant Processing Dawkins & Furnham (1989) suggest that emotional stimuli slow colour-naming by activating taskirrelevant self-preoccupying processes that consume attentional capacity. Although this explanation may be thought to contradict the finding that positive stimuli can sometimes cause as much interference as negative stimuli, the fact that any emotional stimulus can interfere if it is relevant to a person's current concerns (Mathews & Klug, 1993; Riemann & McNally, 1995) is compatible with a 'task-irrelevant' processing explanation. Further, it is consistent with the Page 108
finding that the amount of interference correlates with the intrusion subscale of the Impact of Event Scale in people who have post-traumatic stress disorder following rape (Cassiday, McNally & Zeitlen, 1992). However, two findings weaken this explanation. First, if emotional Stroop interference were due to the negative words triggering such task-irrelevant processes, one would predict that blocked presentation (where all the negative words are presented one after the other, followed by a block of words of another type) would lead to a greater degree of interference than when presentation involves single words randomly varying in hedonic tone. Cassiday, McNally & Zeitlen, (1992) included a comparison of blocked versus random presentation and found no difference in the degree of interference on the threat words. Second, if colour-naming interference were due to such task-irrelevant processes consuming cognitive capacity, one would expect interference on supraliminal versions of the task to be greater or more persistent than interference on subliminal versions of the task. This does not occur, and the evidence we have reviewed suggests that in some circumstances subjects may override the tendency for supraliminal (but not subliminal) stimuli to interfere with colour naming (Mogg, Kentish & Bradley, 1993). The Cognitive Effort of Repression/Avoidance Ruiter & Brosschot (1994) suggest that emotional Stroop interference is due to the greater cognitive effort which is required to shut out the perception of negative stimuli or to render such stimuli unconscious (cf. Holmes, 1974). There are two factors which make this sort of theory unlikely to account for the data. First, it is difficult to see how it could account for interference on all stimuli (including positive stimuli) related to current concerns. 'Repression' cannot be defined so widely that individuals are motivated to avoid all such material. Second, such an account is incompatible with the patterns of memory performance observed following emotional Stroop experiments. Pratto & John (1991), for example, studied an unselected group of student subjects' performance on the emotional Stroop task using desirable and undesirable trait adjectives. They found more colour-naming interference for negative trait words than for positive trait words. Subjects were then given a free recall task for the trait adjectives used in the Stroop task. Performance was generally very poor, but significantly more negative than positive traits were recalled. Pratto & John concluded that, since the emotional material had not been relegated to the unconscious, emotional Stroop interference could not have been caused by the effort involved in such repression. Page 109
The Difficulty in Maintaining Attentional Focus Many investigators have found that more emotionally disturbed individuals (high trait anxiety or clinically diagnosed anxiety disorder) have generally been slower to colour name all stimuli, even
neutral words. This raises the possibility that such individuals have a general difficulty maintaining attentional focus (see Eysenck, 1992, for discussion). The research on this topic has used tasks other than the Stroop, but any conclusions based on such research would be relevant to discussion of mechanisms underlying attentional bias in general, so it will be reviewed here. Mathews et al. (1990) found that GAD patients had more difficulty detecting a word at an unexpected location when there were words appearing elsewhere on the screen (either threatening or neutral), suggesting a general difficulty with attentional control. This replicated an earlier study by Broadbent, Broadbent & Jones (1986) who found that highly anxious subjects experienced more difficulty in focusing attention, even when neutral materials were used. The fact that this disruption occurs only when the target location is unknown suggests that, given warning, anxious people can use top-down strategies to focus attention, but have difficulty controlling attention within the sort of data-driven task that assesses speeded reaction to unexpected stimuli. Fox (1994) investigated the possibility that highly anxious people have a general inability to maintain attentional focus, using both a Stroop paradigm, and a related paradigm in which colour patches and semantic stimuli were spatially separated. She also used Tipper's negative priming procedures (Tipper, Brehaut & Driver, 1990) to demonstrate that high anxious subjects differ from low anxious in showing reduced negative priming. Negative priming, in this context, refers to the increase in latency for a subject to respond to a target location, if, on the previous trial, that location was one which had to be ignored. In Fox's experiment, however, following presentation of a to-be-ignored distracter at one location, highly anxious subjects did not show the expected retardation on the following trial. A further experiment using threat stimuli as prime and either threat or neutral stimuli as probe suggested that highly anxious subjects failed to show negative priming following threat. She concluded that high anxiety is associated with a general deficit in inhibiting distracting information. While these observations of general distractibility are important, they do not explain why highly anxious people are relatively more prone to interference on some stimuli and not others. The studies reviewed earlier show that emotional information will distract such people more than nonemotional Page 110
information, especially where it is related to a current concern, or primed by such a self-related concern. While each of these effects may be made more salient in anxious people by a general deficit in their ability to inhibit to-be-ignored material, such a general deficit in attentional control is not yet able to explain the precise pattern of observed results. For example, consider the pattern of data from another type of experiment that has found the location of the threat stimuli to be important; that of Mogg et al. (1991b) using a modification of the lexical decision task. Although Bower's theories would have expected biased lexical decision to occur for anxious people on threat words because such concepts ought to be in a more highly activated state, such effects had not been shown. It has also been difficult to show priming effects on lexical decision tasks for depressive words in depressed patients (Matthews & Southall, 1991). Mogg & al. (1991b) hypothesised that difficulty in showing effects of emotional disturbance (in this case, anxiety) on lexical decision tasks was due to the fact that such tasks do not involve any competition between attentional resources. They used a paradigm in which two letter strings occurred simultaneously, one above and one below the fixation point. Some were non-words, others were words (some of which were threatening). The subjects' task was to press the 'yes' button if either letter string on the screen was a word. Twelve GAD patients were matched on age and vocabulary scores with 12 control subjects.
Results showed that there was a bias, but it was only found when the word occurred in the lower position. Although they had not been expressly told to do so, subjects appeared to be reading the top word preferentially: all subjects were generally faster if the top letter string was a word. It was only if attentional search was involved that the bias was found such that anxious patients were faster to respond to threat than neutral words, and controls were faster to respond to neutral than threat words. The degree of bias towards threat correlated with trait anxiety. So although spatial location may contribute to attentional bias, showing up biases which might not be apparent on single location tasks, the pattern of results cannot be explained simply by the assumption that highly anxious subjects have a general deficit in controlling their attention. The Limits of Attentional Bias Interference and the Lack of Memory Bias In attempting to identify the processes involved in Stroop interference, it is helpful to consider the limits of the cognitive bias. Several studies have Page 111
shown that subjects can display emotional Stroop interference without showing later bias in a memory task (e.g. high trait anxious subjects (Richards & Millwood, 1989); GAD patients (Mathews & MacLeod, 1985; Mogg, Mathews & Weinman, 1989); panic patients (Ehlers et al., 1988); snake-avoidant subjects (Mathews & Sebastian, 1993); and subjects who have undergone a 24 hour fast (Lavy, van den Hout & Arntz, 1993). Such a dissociation between attention and memory has been found on other tasks too. Watts, Trezise & Sharrock (1986) tested spider phobics, examining the prediction that the emotional properties of spiders might result in poorer recognition memory for them. Dead spiders mounted on cards were used. It transpired that the effect depended on the size of the spiders, being found only with the larger onespresumably because they aroused more anxiety. Because of this unpredicted moderator variable, a second study was undertaken which again found that phobics' recognition of spiders tended to be poorer only for the larger ones. In this study, a processing manipulation was introduced in which subjects were required to describe the spiders while they were looking at them. The effect reached significance with this manipulation, but fell short of significance without it. This is relevant to the interpretation of the effect; if it can occur when subjects have to describe the spiders, the nature of the cognitive avoidance involved is clearly more subtle than simply not looking at the spiders. At first sight, poor memory following attentional bias towards a threat stimulus seems inconsistent with the emotional Stroop results of Pratto & John (1991), discussed earlier. They found better memory for negative words (which had produced greater colour-naming interference). However, Pratto & John had not divided their subjects into high and low anxious, so were unable to check whether subjects showing most interference also showed greatest recall of negative words. The independence of emotional Stroop interference and memory for the Stroop words in the studies mentioned earlier was demonstrated by the fact that emotionally disturbed individuals show large degrees of emotional Stroop interference on some words but do not recognise or recall those words any better than other subjects who have not shown such a pattern of colour-naming latencies. In several of these experiments, however, there is a general tendency for all subjects to recall or recognise the negative material disproportionately well. For example, all subjects in Richards & Millwood's (1989) experiment (both high and low trait anxiety), recognised more negative than neutral words, and more neutral than positive words (d' of 0.82, 0.59 and 0.45 for negative, neutral and positive respectively). Similarly, both snake-avoidant and control subjects in Mathews & Sebastian's (1993) snake Page 112
Stroop study recalled more snake words. Without these findings, it might have been possible to argue that differences between individuals on the Stroop, which are not then reflected in subsequent memory tasks, are simply due to the differential sensitivity of the tasks. Instead, the results show that the memory tests were sensitive enough to have been able to pick up differences in performance had they been present, and that the dissociation is not artefactual. This pattern of dissociation guides us in indicating the type of mechanism underlying attentional bias. Either the activation giving rise to such bias must be such that it dies away rapidly, before the mnemonic processes which would help encoding into memory begins, or else the type of activation involved in mediating a bias such as colour-naming interference must be different in kind from that involved in explicit mnemonic encoding. Mathews & MacLeod (1994) conclude that the data are best explained by a prioritisation model. Following Oatley & Johnson-Laird (1987), they suggest that the role of emotion is to signal a juncture in the pursuit of desired outcomes or avoidance of undesired outcomes. Emotions elicit changes in information-processing modes to deal with a potentially new situation which may require action to be taken. Anxiety represents a shift into a mode of hypervigilance, such that the person scans the environment for any threatening stimulus, particularly those which have been associated with threat before. In this mode, the cognitive system prioritises the initial automatic encoding of threat stimuli, but not the strategic rehearsal of such material for explicit encoding into memory. This is analogous to the physiological response to threat, which involves blood being pumped to the muscles to prepare them for action, but away from the digestive system. If true, the question arises how early in the processing system these scanning biases can be detected. Evidence that emotional Stroop interference occurs at the very earliest stage of information processing comes from studies requiring colour-naming responses to displays within which semantic stimuli are presented subliminally, and we review these in the next section. Interference and the Subliminal Presentation of Stimuli MacLeod & Rutherford (1992) found that high anxiety students showed disproportionate colournaming interference to threat words on a 20 ms subliminal Stroop task. Using the same task in a prospective study, MacLeod & Hagan (1992) observed that level of distress in women after Page 113
they had received the results of a cervical examination was significantly predicted by the degree to which their colour-naming latencies had been longer to threat words presented under subliminal exposure conditions at the time of initial gynaecological screening. Mogg, Kentish & Bradley (1993) found that the degree of interference in colour-naming stimuli, where the negative information was only presented for 14 ms, correlated significantly with trait anxiety. No such correlation emerged from the interference produced by the same type of material presented supraliminally. Finally, Mogg et al. (1993) found pre-attentive bias towards negative information presented subliminally (14 ms, pattern masked) in anxious patients, but not in depressed patients (as predicted by Williams et al., 1988). Further, they found no evidence of such pre-attentive bias towards positive material, either in the supraliminal or subliminal condition. They concluded that preattentive bias in anxiety operates selectively to favour negative information specifically rather than for emotional information in general. This leaves open, however, the question of how colournaming interference may arise for specific theme relevant stimuli, and Mogg and colleagues speculate that the use of computer presentation of individual words (rather than presentation of all words concerning a single theme on a single card) prevents elaboration of specific theme-relevant
information. Consistent with this suggestion, McNally et al.'s (1994) study of the specificity of emotional Stroop interference in panic patients used a computer presentation of individual words, and found that panic patients did not display differential colour-naming interference to specific panic words versus general threatening words. Further, in Mattia, Heimberg & Hope's (1993) study of social phobics, which found that such patients did show colour-naming interference specifically for social threat words but not for physically threatening words, all the words relating to a single theme appeared on a computer screen together, and subjects read the material as if it were a card presentation. In summary, we have seen that the data allow us now to go beyond general explanations of the emotional Stroop as due to emotional stimuli 'capturing attentional resources' or attracting more processing resources due to the activation of threat related knowledge structures. Several models which might account for both long-term and short-term activation of emotional material have been considered, including two distraction explanations (due to task-irrelevant processes and due to cognitive effort used to prevent emotional material from becoming conscious), and a third, general deficit explanation. These were rejected as insufficient to account for the recent data. The data showing that attentional biases in Stroop do not lead to memory biases pointed to the likelihood that interference arises from mechanisms earlier than, or independent of, those Page 114
involved in mnemonic encoding, and studies using subliminal Stroop confirm that such early biases can be found. At what point in the cognitive system do such biases arise? In order to help answer this question, in the next section we shall examine how a recent connectionist account of the original colour-conflict Stroop may accommodate the emotional bias results. Modelling Attentional Bias Within a PDP Framework The PDP Model of Cohen et al (1990) Investigators who have used the emotional Stroop to examine the mechanisms that underlie attentional bias have in the past been unable to make use of explanations of the original Stroop effect since these naturally emphasise competition between incompatible colour stimuli, meanings or responses. Since there is no such direct competition between the word's meaning and its colour in the emotional Stroop, such explanations are likely to be severely limited in accounting for the data we are concerned with. However, MacLeod (1991), after reviewing 50 years of colourconflict Stroop research concludes that the recently developed PDP model by Cohen, Dunbar & McClelland (1990) provides the best account of the data, and Williams, Mathews & MacLeod (1996) have recently suggested how it might explain emotional Stroop results. There are three variables within such a connectionist framework that are relevant to explaining emotional effects: the strength of processing of a pathway, the resting level of activation of individual input units, and the neuromodulatory control of certain classes of input units. Cohen, Dunbar & McClelland (1990) model conflict in the classic Stroop task in terms of two pathways, one for colour-naming, the other for word reading (see Figure 2). Each has input units (representing colours or words respectively), intermediate units, and output units (for the responses to name the colour or read the word). These pathways interact, interfering and facilitating the action of each other. Interference takes place when dissimilar patterns of activation converge on a single point of intersection, at any point in processing after sensory registration has occurred. Cohen et al. suggest that attention is a modulator (cf. Treisman, 1960). They model its effects by having attention alter the responsiveness of the processing units in a pathway, with 'task demand units' (one representing the demand to name the colour, the other representing the demand to read the word) able to modify the behaviour of the intermediate units
Page 115
(see Figure 2). Attention is thus seen, not as a qualitatively different entity within the informationprocessing system, but as an additional source of input within a processing pathway. Performance of any task that requires a particular processing pathway involves establishing a pattern of activation in the relevant sensory modules to generate the appropriate pattern of activation in the relevant output
Figure 2 Diagram of Cohen, Dunbar & McClelland's network showing connections between input units, intermediate units and response units. Task demand units connect to intermediate units, modulating processing by adjusting the resting levels of activation (and thus the responsiveness) of intermediate units (from Cohen, Dunbar & McClelland, 1990, p. 336) (Copyright, American Psychological Association) Page 116
modules. A pathway is defined in terms of the connections between the units that make up the modules used in a task. The strength of these connections determine the speed and accuracy of activation flow along the pathway which, in turn, determine the speed and accuracy of a response. Since individual modules send and receive information to several other modules, each can participate in several different pathways. According to this model, interference on the colourconflict Stroop task arises from the action of a common, underlying variable which they term strength of processing. Word reading interferes with colour-naming because it is more highly practised than colour-naming. Accounting for Emotional Stroop
This suggests the first way in which this model might account for some effects in the emotional Stroop. In some contexts, attention is drawn towards stimuli which are of long-term concern to individuals simply because they are more highly practised. This might be the basis of the 'expertise' effect noted in the previous chapter in which ornithologists showed colour-naming interference for words denoting rare birds (Dalgleish, 1995). However, we saw in the previous chapter that expertise effects were insufficient to account for all the data. Following therapy, in which patients become more expert at processing information related to their psychopathology, Stroop effects diminish rather than increase. However, there are two other ways of accounting for colour-naming interference by emotional words within Cohen, Dunbar & MacClelland's model*. First, input units for such words may have a higher resting level of activation. Second, any such input unit, if associated with threat, might be subject to neuromodulatory control affecting the responsivity of those units. To understand how this occurs, consider an input unit for the word 'spider'. Whether the unit is active depends on its net input (a weighted sum of the inputs) and the unit's activation function. Different models employ different activation functions (see Williams & Oaksford, 1992). Following Cohen and co-workers, we assume a sigmoid activation function (shown in Figure 3), which produces a continuous output between 0 and 1 for various levels of activation input. At very low or very high input levels, small changes in levels of input activation make little difference to output levels. However, in the middle range of activation, relatively small changes have a larger effect on output, the unit is at its most sensitive. *We are grateful to Gordon Brown for first making us aware of these possibilities. Page 117
Figure 3 Activation function under conditions of differing neuromodulatory control
Returning to our example of the input unit for the word 'spider', let us assume that its resting level of input activation is higher for spider-avoidant people (say 0.6, compared to non-phobic's 0.4).
Let us also assume that the resting level of input activation for a neutral word (e.g. 'fireplace') is 0.3 for both phobics and non-phobics. The result is chronically higher activation output when a phobic sees the word 'spider' increasing the activation level throughout the word-reading pathway, and resulting in greater interference with colour-naming at the response module. The phobic subject will normally show little interference for the word 'fireplace' and the non-phobic will show little interference for either word. Higher trait anxiety involves frequent rumination on negative themes and worries, which increases the resting activation level for input units representing these worries. In a similar way, relatively short-term goals Page 118
increase resting activation levels of the input units representing them. So if a person is currently searching for a fireplace for his or her new house, the resting activation level for 'fireplace' will increase. Any stimulus 'input' in the environment to do with fireplaces (magazine advertisements, etc.) will have a greater relative output, and interfere more with other ongoing task demands. The activation function shown in Figure 3 also makes clear the third way in which emotion may affect colour-naming interference. For the gain parameter itself can be changed, steepening or flattening the activation function. A similar methodology was adopted in Cohen & ServanSchreiber's (1989) model of cognitive deficits in schizophrenia where the gain parameter was turned down in a subset of units in the model. This intervention was treated as analogous to the biological effects of dopamine deficiency. We assume that there is similar neuromodulatory control of those input units that have been associated with threat for an individual. (The neurotransmitter norepinephrine may act in just such a way in anxiety. It is distributed throughout the brain with a major site of origin being the locus coeruleus. Lesions at this site lead to decreased levels of norepinephrine and impair a number of tasks which depend on the organism's sensitivity to punishment or non-reward (Gray, 1990).) Figure 3 shows how changing the gain parameter alters the output of all units for any given input activation level. We suggest that any input unit which has been associated with punishment or fear in the past acquires a threat 'tag' becoming thereby susceptible to such neuromodulatory influence. The result of such neuromodulatory increase in the gain parameter will be that all such tagged units will have increased output for any given level of input. Thus, the effect of increased neurotransmitter for anxiety will be a general increase in activity in pathways representing words related to threat for that individual. Note that this does not require the assumption that these words be associated with each other, avoiding the need to assume increased inter-item association in anxious people. So if a person has a history which has given rise to a fear of spiders, dark closets, and German Shepherd dogs, each of these will be separately tagged for threat. Increased neurotransmitter activity for anxiety will increase the activation function for each concept without the need to assume they are part of the same 'danger schema'. The result for all such input units is an increase in responsiveness such that, for any given input, the output of activation from the unit is enhanced, increasing the level of activation of units in the rest of the Page 119
pathway. The result is that colour-naming interference occurs despite the fact that the task demand unit'name colour'is operating. The task demand unit places the intermediate units for naming the ink colour in the most responsive part of their dynamic range. But the greater activation level from units for emotion words mean that more activation accumulates at the intermediate units despite the task demand units for 'name colour'. The result is that although the subject is
attempting not to attend to the word, information flows along this pathway anyway. This information flow is not enough to make the subject read the word by mistake (though this sometimes happens in emotional Stroop experiments), but enough to affect the speed of the colour-naming response. In telling subjects to name the colour (attend to colour-naming, or 'activating the task demand unit for colour') information in the irrelevant (word reading) channel is not allocated strategic attention by the subject. The fact that it can produce interference, without relying on such attentional allocation, is consistent with the notion of the bias being 'pre-attentive' or 'automatic' (i.e. one that does not require a strategic switch of attention). However, as Cohen and colleagues indicate, automatic processes are rarely completely independent of attentional control. They simply vary in how susceptible to control they are. In the emotional Stroop, like the colour-conflict Stroop, word reading interferes with colour-naming, but subjects still name the colour and not the word. Explanatory Power of the Model Can the model explain the results that have been reviewed earlier? The alternative models each failed because they could not account for the pattern of data across a number of experiments. Unlike the 'task-irrelevant processing' hypothesis, the proposed framework can account for interference in colour-naming of subliminally presented stimuli since it views automatic and strategic processing as a continuum. Unlike the cognitive effort (repression) hypothesis, it can account for the interference shown when colour-naming positive words, so long as they are salient to a current concern of the individual (and are thereby either highly practised, or the input units representing the item has a higher resting activation level). Unlike the general distractibility hypothesis, it can model the complexity of the data whereby individuals are prone to interference on specific stimuli and relatively immune to interference on others. We have seen that although high trait emotion is usually associated with colour-naming interference for emotional stimuli, this is sometimes not observed (Martin, Williams & Clark, 1991; Mathews & Sebastian, 1993; Page 120
Mogg, Kentish & Bradley, 1993). The lack of effect was unlikely to be due to the insensitivity of the measures used in these experiments, since identical stimuli produced the predicted interference in other experimental conditions or with other groups in each study. Mathews & MacLeod (1994) have suggested that this pattern of results is due to the ability of non-clinical subjects to override Stroop interference on some salient stimuli. (Note that such override has never been observed in clinical patients, a finding whose clinical implication we explore later.) This ability of non-clinical subjects to override the tendency for emotion words to interfere is further evidence against the task-irrelevant and cognitive effort hypotheses, since the work required to override the effect of the salient stimuli would be predicted to produce greater, not less, interference. However, it is explained by the Cohen and co-workers' model by increased activation of the task demand unit. Of all the 'override' phenomena, that observed in snake-avoidant subjects in the presence of a snake or large spider (Mathews & Sebastian, 1993) is most counterintuitive, on the basis of previous models. As Mathews & Sebastian indicate, any explanation which assumes that attention was captured by the actual snake still leaves unanswered why colour-naming latencies on snake words was reduced in interference when the snake was present, since reduced capacity for the Stroop task should have reduced capacity for both colour-naming and word reading, thus preserving interference (though possibly slowing the whole task down). If we assume that (a) the input unit for 'snake' and related concepts becomes most active with the snake present; and (b) the word 'snake' will tend to be output unless the colour-naming task demand unit is strengthened,
then the reduced interference is explained by the increase in effort that subjects had to expend in following the task instructions 'name the colour'. Cohen, Dunbar & McClelland (1990, p. 351) show that when the task demand unit associated with colour-naming is increased, interference is reduced. However, the model also predicts that overall latency to respond will reduce under these conditions. We therefore returned to the data of Mathews & Sebastian (1993) to see if such a reduced latency to respond occurred. The data, shown in Figure 4, show that this is precisely what happened in their experiment. Further, a PDP model predicts that there will be reduced interference for non-emotional words as well, with the feared object present. Figure 4 shows that this, too, occurred. Williams et at. (1996) examine other such 'override' phenomena, and suggest that such speeded responses may be the signature that such strategic override is occurring. Page 121
Figure 4 Mean latency to name snake-related or matched neutral words by snake-avoidant subjects with or without snake (Expt 1) or spider (Expt 3) present (data from Mathews & Sebastian, 1993).
Further experiments will be needed to determine the precise conditions under which exposure to such manipulations increase or decrease interference, and to answer the question why increased anxiety in other contexts does not also increase the task demand unit activation. One possibility is that reduced interference occurs when (as in the case of direct exposure to a feared object) completion of the task at all demands large increases in task demand unit activation. Although this remains speculative at this time, the PDP framework offers a useful heuristic within which experimental predictions may be made. PDP and Attentional Bias on Other Tasks Can the PDP framework account for biases in attention associated with emotional disturbance in the visual dot-probe task (MacLeod, Mathews & Tata, 1986) and dichotic listening tasks (Mathews & MacLeod, 1986)? The PDP model sees attention as the modulation of processing, causing a shift Page 122
in the responsiveness of units in competing processing pathways. It views attention as simply an
additional source of information providing a sustained context for the processing of information in a particular pathway. Few changes are therefore necessary in order that the same model of pathway modulation which explains task selection (colour-naming or word reading) in Stroop can be used to account for spatial allocation of attention in the dot-probe tasks or channel selection in dichotic listening tasks. For these paradigms, the feature dimensions of the Stroop need only be replaced by networks of units corresponding to, for example, different spatial locations. In such a system, attention bias units are assumed to represent different locations, with attention able to be 'allocated' to different locations such that information in those locations is processed to a greater degree. The ability of the PDP model to generalise to these other selective attention tasks is explicitly seen as one of its major advantages (Cohen, Dunbar & McClelland, 1990, p. 355). Modelling Dissociations Priming and Elaboration One of the most stable aspects of the data reviewed earlier was the finding that attentional bias towards threat is not necessarily followed by increased ability to recognise or recall the material which has received such preferential attention. A complete account of the mechanism underlying Stroop must explain why this occurs. In our 1988 book, we linked this dissociation with related findings in the implicit/explicit memory literature. In this book, we shall go into further detail about these distinctionsespecially in Chapter 6 (on memory) and in the final chapter. For now, we give a preliminary introduction to this related work, and begin to explore some ways in which attention bias might be independent from memory bias. The body of work that is most relevant is that on perceptual memory reviewed by Jacoby & Witherspoon (1982) and Graf & Mandler (1984). In this research, memory which is based on awareness is contrasted with memory which appears not to depend on awareness. The need to make such distinctions is often traced back to the nineteenth-century neurologist Claparede. As part of an experiment to discover what an amnesic patient could remember, he shook hands with such a patient while holding a pin in the palm of his hand. The next day, the patient was, as usual, unable to recall ever having met Dr Claparede. However, he showed an 'inexplicable' reluctance to shake hands. The patient clearly 'remembered' Page 123
something, but was not aware of any memory. Similarly, research by Warrington & Weiskrantz (1974) shows how the advantage of previous cueing on a figure identification task or facilitated performance due to previous practice on jigsaw puzzles is retained in amnesics despite their having forgotten ever doing the tasks. Procedural memory appears relatively intact; declarative memory is by contrast very impaired. Discussion of the procedural-declarative distinction is beyond the scope of this chapter. It is how such distinctions are elucidated by perceptual memory research that is relevant for our present purposes. In perceptual memory experiments the presentation of words as part of the first phase of an experiment makes them more likely to be identified when presented for very brief (e.g. 35 ms) intervals in phase two of the experiment. The important aspect of this work, however, is that this perceptual facilitation can be shown to be independent of more awareness-based recollection procedures. That is, after learning a list of words, some will be recognised later in a memory task; some will be identified more easily in a tachistoscope; but these will not necessarily be the same words. Graf & Mandler (1984) used a depth-of-processing paradigm in which different questions were asked of the to-be-encoded words, emphasising either structural features (how many 'T' junctions in this word?) or semantic features (e.g. is this a pleasant word?). In this paradigm one would expect better memory for the words encoded under semantic conditions because the encoding and retrieval procedures involve more awareness-based strategies of elaboration. (See
Klein & Kihlstrom (1986) for an alternative explanation of such depth of processing effects. However, the explanation for these phenomena does not concern us here.) Graf & Mandler's results confirmed the superior recall and recognition of semantically encoded items, but also showed that depth of processing made no difference to a task chosen to assess perceptual memory (word-stem completion in which subjects are given three letters of a six-letter word and asked to complete the word; e.g. FOR---: forbid, forget, forest, etc.), The numbers of stem completions of previously presented words in the semantic condition was only 1.2 times the number of such stem completions following structural encoding. This figure can be compared to a semantic : structural ratio of 2.7: 1 for recognition performance and 4.0 : 1 for recall performance. Graf & Mandler's account of their results may have significant bearings on the interpretation of attentional bias phenomena in clinical conditions. They distinguish two processes which operate on mental representations: integration and elaboration. Integration is the mutual activation among different components of a single schema (defined for their purposes as a cluster of perceptual and semantic variables which represents a word or Page 124
concept in the cognitive system). Integration makes a word more accessiblethat is, the word will be more likely to come to mind, be noticed or identified when only some of its features or components are presented to the subject. This clearly closely resembles attentional bias as we have defined it. Elaboration occurs when schemata are activated in the presence of other mental events, making further new relationships with those events or reactivating previously established relationships. Such elaboration is required for a subject to become aware of relations among sets of previously unrelated words (e.g. clustering in free recall), and to relate a word to its context (e.g. encoding specificity phenomena, and the depth of processing effects on free recall and recognition noted by many investigators). Graf & Mandler (1984) use the term 'retrievability' to describe the better recall of material mediated by the activation of relationships between schemata, that is, by the generation of new, and reinstatement of old, paths for retrieving the words. Williams et al. (1988) suggested that, if there were two distinct processes underlying recall performance and perceptual memory performance, then different types of emotional disturbance may be associated with disruption of only one or other aspect of processing. Some patients may show facilitated performance on a perceptual memory task because the underlying mental representations have been active recently and are therefore in a more integrated state (the internal components are mutually activating each other). Others may show facilitated recall of negative words or personal memories because their mood has caused the underlying mental representation to form new, or reactivate old, links. Let us consider how these formulations may help in elucidating the examples of attentional bias considered in this chapter. Our 1988 review suggested that, assuming that an individual's cognitive system allocates attention initially to any stimuli that are better 'integrated' in Graf & Mandler's sense (but not necessarily better elaborated), then the greater perceptibility of threat words is understandable. A threat stimulus has similar effects as a recently presented stimulus, and differences in elaboration between stimuli make little difference to the integration process. We suggested that the visual dot-probe task has more in common with the perceptual memory tasks of Jacoby & Witherspoon (1982) and Graf & Mandler (1984)that is, it is concerned with integration of subcomponents of individual schemata. We are now able to see that the 'integration' of which we wrote then may be modelled within a PDP framework by the three processes we have described: increased strength of processing in the word-reading pathway (which captures expertise effects), increased resting level of activation of
Page 125
input units for concepts of current concern (which captures specificity effects), and increases in input activation due to neuromodulatory influence on the activation function of those input units associated with threat (which captures general mood effects). We suggested in 1988 that attention biases do not lead to memory bias because integration and elaboration were distinct processes. Can we be more specific about why this is? Why is it that, when elaborative processes start to operate, they do not find the products of the integration process in a readily accessible state, ready to be incorporated into elaborative processes? Learning Rates and Opponent Processes There are two possible solutions to this problem. Both are preliminary and require more research, but we mention them here to give an idea about the possible direction of future work. The first solution was suggested by Williams & Oaksford (1992). They suggested that the neuromodulator which affected units associated with threat also affects learning. Within PDP models, all learning rules have the basic form:
where y is the learning rule and h is the learning rate parameter. They suggested that a side effect of high concentrations of the neurotransmitter involved in anxiety is to turn down h, thereby suppressing the learning of new associations. Since only the units in the representation of threat stimuli are potentiated, these within-representation connections will be enhanced more than elaborative connections to other active representations. Thus, a stimulus representation may be primed, without such priming necessarily leading to enhanced elaboration. Although it was not the purpose of Williams & Oaksford to detail the pharmacological evidence for these speculations, they did point out that such a theory was biologically plausible. Neurophysiological evidence had shown that norepinephrine can modulate neuronal plasticity in the rat hippocampus, a phenomenon which has been simulated by varying the same learning rate parameter h (Munro, 1986). A second possible solution to the frequent lack of effect of attention bias on later recollection is to suggest that the units representing the threat words are no longer in a sufficiently activated state at the time when elaborative processes would normally take place. A model which would predict activation followed by rapid inhibition of the same unit is opponent process theory (Houghton, 1994). We consider it here in some detail, for it promises not only to explain dissociations between attentional and Page 126
memory tasks, but also the finding of differential allocation of attention within tasks which demand competition for attention. Houghton points out how neural systems are potentially unstable. They must have control mechanisms built into them to maintain both low level stability (at the level of responsiveness of neural systems) and high level stability (at the level of control of voluntary selective attention), In short, such a system must be able to inhibit its activity. One such form of control, lateral inhibition, is well described, providing for competition for activation. But it is too non-specific to provide the sort of stability required. For example, it does not inhibit the unit which is itself the source of activation, leading to a situation of instability or locking the system into selfreinforcing perseverative interlock. Houghton (1994, p. 109) describes a neural network basis of another source of inhibition in the systemopponent processeswhich allow the system to rapidly return to baseline.
For a specified excitable process (call it the a-process or primary process) there exists an opposing process (call it the b-process, or opponent process) which is in opposition to the a-process, such that the activation of the b-process can attenuate or quench the a-process.
The activation of the b-process can be driven by the a-process. In common with a previous account of opponent processes (e.g. Solomon & Corbit, 1974), it is assumed that (a) there is a lag between the onset of a and b processes; and (b) the b-process is slower to decay (see Figure 5). Solomon & Corbit's (1974) model explicitly linked the occurrence of such excitation inhibition mechanisms to motivation and emotion. They argued that any affective state automatically generated, within the CNS, an opposing state which acted to return the system to equilibrium. Because of the time course of excitatory and inhibitory mechanisms, specifically the fact that the b-process takes longer to decay, when the input causing the hedonic stimulation stops, the opponent process is shown as an opposite state. In Houghton's model, the performance of anxious subjects is explained if it is assumed that the time-course of the inhibitory (b-process) has a longer latency to onset and/or a slower 'rise time' for threat stimuli. (Note that it makes little difference whether we see neuromodulatory control of attention as facilitating the response of excitatory units, or delaying the response of inhibitory units. The attraction of the latter is that it also explains the delay in dissipation of inhibition.) In the case that the b-process is delayed, the a-process will not be inhibited or adapt as quickly. Page 127
Page 127
Figure 5 Diagram showing opponent processes of excitatory and inhibitory units, and resultant activation level Page 128
The result of this is that the peak of the primary response will be higher, so that at stimulus offset, the peak of the rebound will also be greater (i.e. more opposite to the primary process). So initial delay in the onset of the opponent process has resulted in a larger rebound with the inhibitory effect of the opponent process taking longer to dissipate. After these automatic processes have operated, the subject may, of course, switch in strategic search processes to inspect that aspect of the environment which is of current concern*. But at the point at which a stimulus or concept (in the experiments reported here involving threat) are likely to be elaborated to make mnemonic links to related concepts, the opponent switch-off is still operating with the result that reduced mnemonic elaboration will take, place. The subject will have to put greater effort into strategic elaboration and inspection of those stimuli which have been subject to more prolonged inhibition. The variable memory performance of some emotionally disturbed patients on stimuli which relate to their current concern may reflect this balance between the
inhibition of material resulting from the operation of the opponent process, and the subject's higherorder goal to approach or avoid that material (see Klinger, Barter & Maxeiner, 1981). Attentional Bias and Psychopathology How far have these cognitive paradigms such as dichotic listening, visual dot-probe and emotional Stroop been able to meet their goal of explaining attentional bias in emotional disorders? The most common use of the tasks has been to demonstrate simply that attentional bias exists in a disorder. We have seen, especially in relation to the Stroop, that attentional bias is to be found across a wide range of disorders; it is clearly not specific to anxiety or depression. However it is in the domain of anxiety that most research has been carried out on the questions of the causes and mechanisms mediating the bias. Two issues have dominated the research. First, whether any bias is a trait or state feature, or an interaction of the two. Second, to what extent bias is shown for specific concernrelated material, or more generally for any threat- or loss-related theme. In relation to the first question, it has been found that trait mood is most likely to interact with current mood or circumstances to determine attentional bias when there has been time for worry to have incubated on a *Evidence from a study by Klinger, Barter & Maxeiner, (1981), detailed at the end of this chapter, suggests that whether this strategic inspection takes place depends on different factors than those involved in basic integration and elaboration processes considered here. Page 129
theme. Hence the interaction between trait mood and examination stress in producing biases in the dot-probe paradigm when an exam is imminent (MacLeod & Mathews, 1988). Subjects have had time to ruminate, and high trait subjects will have ruminated more about the negative consequences of the exam. By contrast, experimental failure, while successfully biasing subjects towards specific or general threat themes (on Stroop and dot-probe respectively), does not interact with trait anxiety in predicting the extent of attentional bias. Trait anxiety being the tendency to worry about things over the long term, it appears not have a role to play in the sudden exacerbation of threat-related themes. Neither need the acute stress of experimentally induced failure rely on state anxiety for its effects. Although experimental stress did increase state anxiety, the extent to which it did so did not predict degree of attentional bias on either Stroop or dot-probe tasks. Achievement-related concerns are activated directly by such a failure experience, without the need for mediation by raised state anxiety. This raises the question of the specificity of the bias for current concerns. It is clear that attentional bias can be shown for any concern-related theme. Experimental failure activates achievement themes just like buying or selling a house makes 'For Sale' signs salient, or buying a new Renault car makes all other Renaults in the district salient. As Mogg, Kentish & Bradley (1993) observe for the emotional Stroop, such specificity is most likely to be observed when the paradigm blocks words of the same theme together, allowing some 'training' of the system within the task. However, specificity has also been found with the dot-probe task with GAD patients (Mogg, Mathews & Eysenck, 1992) with extent of self-rated social concern correlating with bias on social threat words but not on physical threat words. Indeed, there is no reason in principle why the presentation of single words representing very salient idiographic current concerns should not disrupt performance. After all, much discussion of selective attention starts from the observation that the most idiographic of all stimulione's own namehas the power to capture attention from most alternative ongoing tasks. We have suggested that Cohen, Dunbar & McClelland's (1990) model of the original Stroop can explain the attentional bias associated with emotion. Attentional bias for emotion themes is
explained within a PDP model in terms of (a) increased practice affecting the strength of processing in a pathway underlying performance of a task, (b) differences in resting activation levels of various input units representing current concerns and (c) neuromodulatory control of input units associated with past threat. We suggest that it is these neuromodulatory influences which come into play when stimuli in the environment signal a juncture in pursuit or avoidance Page 130
of a desired or undesired outcome. This produces emotion, signalling a change in processing priorities. Input units associated with the new (threat) processing mode become activated. Depending on the frequency and variety of contexts in which such units have been activated in the past (including the frequent rumination of the high trait anxious person), this neuromodulation may have relatively general or specific effects. Differences in input activation levels (differences between people, between units, and over time) produce a state (albeit sometimes temporary) within the processing system which is functionally equivalent to increased strength of processing within the pathways responsible for word reading (in the Stroop), location (in the dot-probe) or for unattended channel (in dichotic listening). The model explains the ability of subjects to override attentional bias in some circumstances by increasing effort expended in performing the primary task (increasing the task demand units for naming the colour, shadowing, etc.). Whereas such an override has been found in non-clinical subjects, such as high trait anxious normals and snake-avoidant students (Martin, Williams & Clark, 1991; Mathews & Sebastian 1993), it has not been observed in clinical groups. This is consistent with the suggestion by Mathews & MacLeod (1994) that breakdown occurs when an individual can no longer expend the extra effort required to override the tendency for concernrelated stimuli to capture attention. Such breakdown is characterised within the PDP model as the task demand units' greater activation eventually gives way to the increased strength in the competing processing pathway. Since this strength from input units denoting (especially negative) current concerns will have been increasing for some time (though compensated for by increases in effort), removal of the influence of the task demand units will result in an abrupt and catastrophic increase in input from the processing pathways associated with the concern. The misinterpretation of such stimuli (for example, bodily sensation in panickers) may be due to the abruptness and size of the switch of processing pathways when the effort to override input from concern-related units reaches its limit. As far as the Stroop is concerned, the bias in colour-naming found in subliminal versions of the task is not amenable to such override strategies, and may therefore be a more sensitive measure of current concern. Consistent with this, MacLeod & Hagan (1992) found that degree of bias on a subliminal Stroop (representing physical health concerns) performed before receiving results of a physical examination predicted the degree of a patient's emotional disturbance when they received the results. Degree of disturbance on the supraliminal version of the same task did not predict emotional disturbance. Page 131
Concluding Remarks At the outset of Chapter 4, we defined attentional bias as an aspect of phenomenal experienceas a process in which people are aware of something novel, threatening or related to their current concern. What are the clinical implications of the experiments we have examined? Let us look again at the bird phobic MM, and the generally anxious TS. Both found their attention being caught by concern-related, threatening material: dead birds or similar shapes on the one hand; suicide stories on the other. First, we now have much more evidence that this tendency occurs at an early stage in cognitive processing, at a point where the meaning of the stimulus has been accessed but its elaborative links with other associated stimuli have not been made. Whereas, at this point, a non-anxious person's attention might
orient away from the suicide story in the newspaper, the anxious person's attention tends to switch towards the unpleasant story. The two patients referred to above were similar in their 'noticing', but very different in how much they subsequently ruminated about the stimuli. This difference may have been due to the difference in how circumscribed the stimuli werea bird is a fairly concrete, well-circumscribed stimulus, whereas an overdose story is less so. Or the difference could be due to the comparative self-relevance of the stimuli. A bird stimulus has arguably less personal implications than do stories about suicide, muggings and murderthe former less likely to be encodable along many self-schema dimensions, the latter is more likely to be so encodable. Nevertheless, it is important that the same amount of attention to emotional stimuli should have different consequences. Further corroboration for the distinction between the 'noticing' and 'ruminative' aspects of attentional bias comes from some research reported by Klinger, Barter & Maxeiner (1981). They used a Concern Dimension Questionnaire in which subjects listed a number of 'things they had thought about most today and yesterday', followed by listing a number of things which were significant in their lives which they had not thought about (or very little) during the same period. Subjects then rank-ordered all items according to how much they had been thought about, and characterised them on a number of scales assessing value, expectancy and time-course of the concern. Analysis of the questionnaire revealed four dimensions: the valence of the goal (positive or negative), the time available before something has to be done, the intensity of valence (how positive or negative), and the probability of success in the goal being achieved. The valence of the goal Page 132
was in general a weak predictor of the amount of time people reported thinking about things (confirmed by time-sampling methods in subjects' daily lives). The time available for action, intensity and expectancy of success were better predictors of time spent thinking about the goal, but were predictive only in natural settings. They did not predict 'switching' in the dichotic listening task reported earlier in this chapter. This is consistent with the possibility that subjects' tendency to notice concern-related stimuli were not mediated by the same processes that determine whether they spend further time on thinking about the concern. Although these may be independent, the evidence reviewed in Chapters 4 and 5 has shown how orientation towards sources of threat may disrupt performance whether or not subjects are consciously aware of the threat stimuli. Some subjects, even if high on trait anxiety, are able to override the tendency for their attention to be drawn towards threat. Subliminal tasks show that such subjects are still orienting to threat at the earliest stage of processing, but are thereafter able to direct resources away from threat on later processing cycles. Clinical conditions represent a breakdown in this override, resulting in further rumination or worry. As Klinger, Barter & Maxeiner's (1981) experiment showed, variables that determine such rumination may differ from those mediating pre-attentive 'noticing'. The various ways in which 'noticing' and 'rumination' are connected remains for further research to determine. The final chapter will go into further detail on this question, distinguishing between memorial and non-memorial elaboration. For the moment, we move away from discussing attention. As we have seen, memory can also be independent from 'noticing' and it is to biases in remembering that we now turn.
Page 133
Chapter 6 Memory In September 1980, Gordon Bower began an address to the American Psychological Association on 'Mood and Memory' with a stark example of the way memory may be affected by mood. He referred to the findings of Diamond, the forensic psychiatrist who had examined Sirhan Sirhan, the assassin of Senator Robert Kennedy in 1968. Apparently Sirhan had carried out the deed in a greatly agitated state and afterwards remembered nothing of the event. Diamond hypnotised Sirhan and helped him to reconstruct from memory the events that had occurred in the kitchen of the Ambassador Hotel in Los Angeles. Under hypnosis, as Sirhan became more worked up and excited, he recalled progressively more, the memories tumbling out while his excitement built to crescendo leading to the shooting. At that point Sirhan would scream out the death curses, 'fire' the shots, and then choke as he reexperienced the secret service body guard nearly throttling him after he was caught. Despite the fact that Sirhan would have liked to have felt that he did the deed (in the cause of Arab nationalism) he was never able, through conscious effort in a non-hypnotised state, to remember doing so. Introduction As seen in the above example, mood can act as a distinctive 'state' or 'context' which can have powerful effects on memory. Once one has changed the state or context it will be harder to remember things that were learned in the original state. However, if the context is reinstated, the material again becomes available to normal memory. There are two issues raised by this example. The first is whether context-dependent memory is a reliable phenomenon; the second is whether, even if it is, mood states can act as such contexts. If we find that they can, we need to ask some further questions about when they are most likely to do so, and what clinical implications, if any, result. Page 134
The first question is more easily answered. Ability to successfully retrieve from memory appears to be facilitated if one can reinstate the conditions which prevailed when the material was encoded. One demonstration of this comes from the experiments on diving and memory by Alan Baddeley and colleagues. For example, Godden & Baddeley (1975) had divers learn lists of words either on the beach or under approximately 4.5 metres of water. They were asked to try to recall the list learned in one context while in the same context on some occasions and while in the changed context on other occasions. Godden & Baddeley found a decrement in recall of words of over 30% if the context was changed in this way between learning and recall. Other examples of contextdependent learning are reviewed by Bower (1981). The second questioncan mood states act as such a context?will be the main subject of this chapter. On the face of it there appears to be enough prima facie evidence to suppose that it can. An interesting example is given by Williams & Marker (1992) who report a manic patient who hid a considerable amount of money during a manic phase, then could not recall where he had hidden it when his mood returned to normal. A few months later he had another manic phase, and found himself trying to hide something else. Without explicit recall of the previous incident, he chose the
same unlikely hiding place for the second time, and found his money. Williams & Marker suggest this may be a special case of state-dependent memory, in which the reinstatement of a context brings about a re-enactment of previous behaviours and associations. Another example is given by Williams (1992, p. 38). He quotes the case of a patient whose experience of going swimming was reinterpreted later in the light of subsequent bad events. At one therapy session when she was in a good mood she recalled the swimming event as having been a good experience. On a subsequent session when she was in a more depressed mood she remembered different elements of the same swimming event which made it seem much less good, for example that it had been a struggle to get to the pool for fear of making herself look ridiculous, and that it had been embarrassing to see herself in a swimsuit because of her weight problem. If we assume in this example that the original mood was neutral, note that a strict state-dependent explanation (in which positive, negative or neutral material is better recalled if the original context is reinstated) is inappropriate. Rather, this patient would have been showing mood-congruent recallremembering better the unpleasant aspects now she was in a low mood, having remembered the pleasant aspects when in a good mood. She would have been showing a different phenomenon if Page 135
the original mood had been distinctively bad rather than neutral. In that case all the bad aspects would have been encoded and even if later retrieval mood was neutral it would have been the bad aspects which would have been recalled. In this case she would have been showing evidence of mood congruent encoding. The distinction between strict context/state-dependent learning on the one hand, and mood-congruent encoding and recall on the other, should be kept in mind in the discussion which follows. Mood-congruent encoding and retrieval has been much easier to demonstrate experimentally than has mood state-dependent learning. Experiments which have manipulated mood at encoding and retrieval to examine state-dependent retrieval have produced equivocal results. A much cited study by Bower, Monteiro & Gilligan (1978) failed to find that making mood the same as encoding and retrieval facilitated recall of a single list (16 positive or negative abstract nouns). The mood manipulation was hypnosis, which has been used in the same laboratory many times to produce many more robust mood-biasing effects. Only when the authors, in a third experiment, used an interference paradigm where two lists were learned (one in happy mood, the other in sad mood) did they find that reinstatement of encoding mood at the retrieval stage facilitated the recall of the list learned in that mood. The fact that such large effects were found with only four subjects in each of the six groups was surprising, given the failure of the earlier experiments in the series. Indeed, Bower & Mayer (1985) have now attempted to replicate the experiment and failed. Schare, Lisman & Spear (1984) did find a state-dependent learning effect but, like Bower, Montiero & Gilligan (1978), only when two lists were learned in an interference paradigm. The lists consisted of neutral words, one of which was learned when happy and the other while sad (mood having been induced using the Velten (1968) procedure of having subjects read sad or happy self-statements such as 'I am less successful than other people' or 'I can feel a smile on my face'). One procedural feature which distinguishes it from the Bower procedure is that, during the recall phase, words from either list were recalled. (In the Bower paradigm, subjects attempt first to recall List 1, then List 2 in counterbalanced order.) This seemingly minor procedural point may be significant since it allows interference between competing items during retrieval. Early retrieval of items from the list learned in the same mood will tend to cue other words from the same list. This process alone will tend to inhibit the later recall of other actually stored items. Consistent with this, McDowell (1984) found biased recall in depressives only when a mixed list was used.
Page 136
It can thus be seen that mood state-dependent learning is far from easy to demonstrate experimentally. By contrast, mood congruent memory bias has been easier to demonstrate, at least for depression. In what follows we shall review this research, showing how differences between types of patient (e.g. anxiety versus depression) and differences between types of memory assessment (direct versus indirect) have given rise to much discussion about the way specific moods may affect specific aspects of information processing. Depression and Autobiographical Memory Memory Bias and Mood Since a characteristic of depression is the predominance of recall of negative aspects of the past, it is appropriate to examine first whether this phenomenon can actually be observed under controlled conditions. Lloyd & Lishman (1975) used a list of neutral words as stimuli to cue memories in clinically depressed patients. Patients were instructed to think of either a pleasant or unpleasant memory, and to signal to the experimenter when a suitable memory came to mind. The time taken to retrieve memories to each cue word was recorded by stopwatch. They found that the more severe the depression, measured by the Beck Depression Inventory, the quicker the patient retrieved an unpleasant memory. Two major problems in interpreting these results occur. First, the more severely depressed patients may have had more genuinely depressing experiences, so may find it easier to retrieve any one of them. Second, the more severely depressed patients may simply be evaluating more of their neutral or ambiguous experiences as more depressive, thus spuriously inflating the number of memories from which to choose. Both of these were subsequently taken into account in two separate research strategies employed by John Teasdale and co-workers at Oxford. The first strategy used non-depressed volunteers whose mood had been experimentally manipulated using Velten's (1968) self-statements (Teasdale & Fogarty, 1979). Subjects were randomly allocated to 'elation' or 'depression' conditions so that the frequency and intensity of actual depressive life experiences could be assumed to be equal prior to mood induction. They found that latencies to remember positive or negative personal events were biased by the mood induction procedure (MIP), though most of this was due to slowed recall of positive material in depressed mood rather than speeded recall of negative material. Both Page 137
this and later experiments showed that the memories recalled are actually pleasant or unpleasant (when rated in neutral mood), and so the bias could not simply be explained in terms of the overinclusive effects of depressed mood (Teasdale, Taylor & Fogarty, 1980). The second strategy to overcome the possibility that depressed people remember more depressing events because they have experienced more, was to examine clinically depressed individuals selected for the presence of diurnal variation of mood (Clark & Teasdale, 1982). Patients were given words as cues and asked to respond with the first personal memory which came to mind. These were later rated for pleasantness. Two independent though complementary effects could be observed. First, happy memories were less probable (and depressing memories more probable) when patients were more depressed. When the same patients were at the less depressed point in their cycle, this picture was reversed. These within-subject results clearly cannot be explained with reference to different frequencies of actual depressive experiences. Second, negativity ratings of the experiences were also mood-dependent, the more depressed ratings being given, the more depressive the current mood. This effect was not by itself sufficient to explain the memory biasing,
however. Results consistent with the above have been found using mood induction procedures other than the Velten technique (e.g. hypnosis, Gilligan & Bower, 1984, Autobiographical Recollections Method (Salavoy & Singer, 1985)). Gilligan & Bower (1984) reported an experiment in which subjects kept diaries for one week, noting down any positive or negative events (time, place, gist), and giving an intensity rating. One week after handing in the diary, subjects were randomly allocated to two groups. Subjects were hypnotised and either a happy or depressed mood suggested. Percentage recall of happy incidents averaged 32% for both types of recall mood. However, this figure fell to 23% for those subjects recalling negative events in happy mood, and rose to 38% for those subjects recalling negative events in sad mood. In another study, subjects spent 10 minutes generating memories from childhood, giving details of the gist, time and place while either in hypnotically induced happy or sad mood. Next day they returned and rated the memories (in neutral mood) for how positive or negative they were. Ninety-two per cent of memories recalled by 'happy' subjects were positive, compared with 45% for the 'sad' subjects. Reliability of Mood Effects on Autobiographical Memory How reliable are these findings of mood biasing in personal memory? Williams et al. (1988) concluded that the effect was robust. Bias appeared Page 138
to occur whether naturally occurring depression or mood induction was used. Methods of cueing personal memory had also been varied without affecting the nature of the results: Lloyd & Lishman (1975) and Teasdale & Fogarty (1979) used neutral words and told subjects whether to retrieve a pleasant or unpleasant memory to each cue. Teasdale, Taylor & Fogarty (1980) and Clark & Teasdale (1982) used neutral words but allowed subjects to retrieve any personal memory which was later rated for its hedonic valence. Gilligan & Bower (1984) did not use any specific cues, allowing subjects to recall anything that came to mind from their diary or their childhood. However, two aspects of this research have cast doubt on the universality of the mood-congruence effects for autobiographical memory. Some research has raised doubts about the reliability of Gilligan & Bower's conclusion that the bias extends to childhood memory (or other very longterm memories). Second, research has found that, on some occasions, mood incongruence may be found. A third development within mood and memory research, emphasising the specificity or generality of memory represents a more fundamental shift within the field, and will be reviewed in detail later. Mood and Childhood Memories Doubts on the ability of mood to affect childhood memory were originally raised by Salavoy & Singer (1985). Using a guided imagery (autobiographical recollections) method as well as a taped depressing story (see Goodwin & Williams, 1982; Williams, 1992, p. 213 for details of this method) they found no effect of elated or depressed mood on recall of early childhood memories, though they did find the usual biasing effect when memories from the previous week were asked for. The non-replication of the childhood memory bias could have been due to a less effective mood induction being used than that of Gilligan & Bower (1984). However, the lack of mood effects on memory for childhood memory has been replicated by Gerlsma et al. (1994). They assessed childhood memory using the Parental Bonding Instrument (Parker, Tupling & Brown, 1979; in which subjects recall parental rearing practices), in both a community sample and social phobics undergoing treatment. Subjects were tested at two points in time, and the degree of mood change between Time One and Time Two measured. There was little evidence of contribution by mood. For example, although social phobics showed considerable improvement in anxiety and depression during treatment, no mood effect on Parental Bonding Instrument scores were found.
Page 139
Another study to find similar results was that of Fromholt, Larsen & Larsen (1995). They were interested in memory bias in depressed elderly patients. Thirty elderly patients depressed for the first time were contrasted with non-depressed aged-matched controls in the proportion of negative memories in free narrative (responses to the instruction: 'Tell me about the events that have been important in your life'). The narratives were afterwards rated by the subject for the time when the various events had occurred and independently rated for emotional valence. Depressed patients recalled an average of 12 distinct memories (compared to 18 in the control group), and the depressed had more negative memories (though no fewer positive or neutral memories). However, this bias was confined to events which had taken place in the recent past (last five years), a bias that was further exaggerated because of a reduced memory for positive (though not neutral) events for this time period only. Consistent with the studies reviewed earlier, there was no bias for events prior to this recency 'window'. The authors reassessed the depressed patients six months later, by which time their mood was considerably improved. Events recalled from the last five years no longer showed a bias towards recall of the negative, the emotional quality of the recalled events now being indistinguishable from that of the control group. The lack of mood effects on very longterm memory have important implications for the interpretation of such memories recalled during psychotherapy, implications which are explored by Brewin, Andrews & Gotlib (1993). Network theory, which prompted much of the early research on mood and memory, would not have predicted that the age of the memory would affect the ease with which mood might bias it. The most plausible explanation for the effect is that older memories are more likely to be overrehearsed. For example, Salavoy & Singer (1985) required that only five childhood memories were retrieved by their subjects, making it highly likely that these few were the most highly rehearsed. Mood may only be an effective extra cue in the search for memories that do not so readily come to mind (cf. Teasdale & Dritschel's experiment, discussed later). Mood-Incongruent Memory The second limitation on mood congruency in depression concerns its generalisability across situations, particularly when the mood occurs naturally rather than being caused by the subjects themselves attempting to get into a certain mood. On the basis of the study by Clark & Teasdale Page 140
(1982), who had found that naturally occurring diurnal variation in depressed patients was associated with bias in recall congruent with the depth of mood, it has been assumed that natural variation and mood induction procedures were affecting the same processes. While mood induction may indeed analogise 'real' depression, it appears that, for non-depressed subjects, naturally occurring mood changes may sometimes be associated with a different pattern of results. Such 'natural' mood variation has been studied by Parrott & Sabini (1990) who examined the effects of a sunny vs rainy day; happy or sad background music; and mood as the result of exam disappointment or success. Their paradigm involved asking subjects in these situations to recall three memories (e.g. three events from high school). They found mood to vary, as expected, with the weather, the music, or the exam success or failure, but the first memories produced in these contexts showed a pattern opposite to that predicted by mood congruence. The memory response was more likely to be happy when the prevailing mood was unhappy, and less likely to be happy when the prevailing mood was more happy. The production of happy memories when feeling 'down' because of disappointment with exams or with the weather might be explained as 'mood repair'. But why should subjects recall an unpleasant memory when happy? Teasdale & Barnard (1993) suggest that, in conditions of naturally occurring happy mood, the cognitive system seeks out and resolves past discrepancies, unfulfilled goals, etc. This may be more common than we have
thought hitherto. For example, a recent review of Teasdale & Barnard's book (Williams, 1994) included the following extract from a newspaper interview with a British television broadcaster, Mariella Frostrup, who describes how she can listen to certain sorts of music that bring back sad memories when she is happy: I like to listen to . . . really old-fashioned records. A big favourite is Leonard Cohenbut only when I'm very happy. My dadwho died when I was 16used to like Leonard Cohen and when I'm happy I'm able to sit and think about him a little bit. (Independent, 9 March 1994)
We might refer to this as the 'repair of past mood', in contrast to the more common repair of current mood. Such repair of past mood may be critical to the emotional processing which naturally occurs following a trauma (such as bereavement), and seems an important component of much psychotherapy. Note that this is a different explanation from that given for other evidence of mood incongruence found by Rinck and colleagues, discussed later. Page 141
Are these biases due to long-term emotional factors, or due to short-term variation in mood? Few studies have directly addressed the question of whether the effect of depressed mood on autobiographical memory bias is more powerful as a 'state' or 'trait' effect. The fact that Teasdale & Fogarty (1979) were able to demonstrate mood-congruent recall on normals using mood induction, and that extent of memory bias can alter with diurnal variation in mood, shows that it can be a state effect. Both Fogarty & Hemsley (1983) and Fromholt, Larsen & Larsen (1995) found that bias in autobiographical memory disappeared with recovery in mood, again consistent with a state effect. However, these studies do not preclude the possibility that in clinical groups a long-term vulnerability to react to small changes in mood with large degree of memory bias could also be a relatively enduring characteristic. Martin's (1985) demonstration of such biases in subjects scoring highly in neuroticism supports this possibility, and we shall return to this question later when we examine studies on biased recall of material learned within an experimental session. Depression and Memory Bias for Experimental Materials Word Lists and Stories Although the experiments reviewed above examined phenomena which bear a close relationship to clinical aspects of mood, they do not allow us to specify very exactly which factors underlie the bias. In autobiographical recall one can never be sure what the original situation wasneither what the original mood was nor what the original material was that is being recollected. A number of experiments have therefore used lists containing some positive and negative words to examine memory biases in more detail. The use of word lists allows one to control the affective tone of the encoding material as well as the type of mood prevailing during encoding and retrieval. Do depressed patients selectively learn negative words? One way to summarise these results is to examine them by meta-analysis. Matt, Vacquez & Campbell (1992) conducted a meta-analysis of experiments which have examined mood congruence for positive or negative material (adjectives, sentences, whole texts, etc.). They found that normal, non-depressed subjects are biased towards the recall of positive information (recalling on average 8% more positive than negative material). This bias is eliminated in 'analogue' depressed subjects (e.g. students scoring high on the Beck) Page 142
who recall positive and negative material about equally. Depressed patients recall, on average, 10% more negative than positive material. Since they have been reviewed extensively elsewhere (Blaney, 1986), this early research will not
be detailed here. (See also Clark & Martin's (1986) critique of these investigations.) We turn instead to those studies which have addressed the issues of what aspects of information processing contribute to such biases. Encoding, Retrieval and Self-reference Bower, Gilligan & Monteiro (1981) examined the effect of mood-congruent encoding by having subjects read a story about two characters playing tennis (one sad, the other happy). The story contained an equal number of sad and happy statements. Subjects read the stories after they had been hypnotised and a happy or sad mood suggested. In neutral mood on the following day, subjects attempted to recall the events narrated by the story. Eighty per cent of the facts recalled by subjects who had been sad while encoding were sad facts from the story. For subjects whose encoding mood had been happy, only 45% of their recall was of sad facts. A similar result was reported by Bower (1981) in which subjects in happy or sad (hypnotically induced) moods read events from a 'psychiatric interview' with an individual in which both happy and sad events were mentioned. In this experiment retrieval mood was also manipulated for the recall phase which took place 20 minutes later. Results showed a clear interaction between mood at encoding and type of fact retrieved (more sad facts if learned in sad mood; more happy facts if learned in happy mood). However, mood at recall did not affect retrieval. As a further check on this failure to find an effect of retrieval mood, another experiment was performed using the story about sad and happy tennis players. This time the stories were encoded in neutral mood and, six hours later, recalled both in happy and then in sad mood (counterbalanced within subject design). No effect of recall mood could be demonstrated. Isen et al. (1978) used success or failure at a computer game to study mood-congruent retrieval. Subjects memorised positive and negative words in neutral mood prior to playing the game. Success experience caused more positive adjectives to he recalled, but failure did not facilitate recall of negative adjectives. Teasdale & Russell (1983) used a similar design, but manipulated subjects' mood with Velten-type negative or positive self-statements (Velten, 1968). They had previously been presented (in neutral mood) with a list of 36 adjectives (12 positive, 12 negative, 12 neutral). Mood at recall was found to affect significantly which words from the list were recalled. In a later experiment, Clark & Teasdale (1985) replicated this recall moodPage 143
congruency effect, but showed that the effect was more characteristic of females than males, possibly due to the fact that the female subjects were found (even in neutral mood) to rate the stimulus words as more personally useful in their daily lives. Thus they would have been more likely to have used these words when depressed in the past. What could explain this inconsistency in mood-congruent retrieval effects? Isen et al. (1978) and Teasdale & Russell (1983) have found such effects, but Bower (1981) did not. Two possible explanations suggest themselves. The first is that Bower's use of story material means that the individual items are associated with each other (e.g. one tennis player enjoys the sunshine, the other is scorched by the sun). If one is recalled it is likely to cue another item even if opposite in hedonic tone. Mood at retrieval may not be sufficient to override these inter-item associations which are not as strong if a list of words is presented. There is a second reason why retrieval mood may have operated more on the words than the story. Both Isen et al. (1978) and Teasdale & Russell (1983) used personality trait adjectives. It is possible that these items tend to be encoded with reference to the self, unlike the Bower stories which are explicitly about other people. Depth of processing studies of the self schema (see Chapter 9) find that material encoded with reference to the self often shows mood- or personality-
congruent bias in recall, whereas the same item encoded with reference to another person does not do so. It is possible that self-reference critically moderates the strength of mood-congruence effects, though Blaney has argued (1986) that self-reference is only significant to the extent that it prompts evaluative processing. Consistent with these suggestions, Hasher et al. (1985), who used integrated story material which was not self-referenced, failed to find a mood-congruent retrieval effect. Similarly, Nasby (1994) has found evidence for an asymmetry, such that self-reference is important for negative mood, but more general evaluative processing is important for positive mood. Female undergraduates were given elation, neutral or depression MIP, and asked to rate 40 trait adjectives (varying in likeability) on whether each described themselves or their mother. Depressed mood enhanced 'yes' rating (and subsequent retrieval) of negative self rated adjectives, whereas positive mood enhanced 'yes' ratings and recall of both self and other rated positive adjectives. Retrieval bias remained even when recall scores were adjusted to take account of the effect of 'Yes/No' judgements (an adjustment which needs to be made to make the bias results interpretable as due to the mood, rather than as due to the fact that people tend to recall any item they have said 'yes' to). Studies which do not make this adjustment, such as that by Greenberg & Beck Page 144
(1989) on content specificity in anxious and depressed patients, are difficult to interpret for this reason. However it may not be self-reference as such which is the important factor mediating enhanced memory, but the amount of elaborative encoding which self-reference inevitably involves. Teasdale & Dritschel (cited in Teasdale & Barnard, 1993) required subjects to form images of a series of vignettes. They were then given either a depressing or elating mood induction procedure, followed by incidental recall of the image descriptions. They found that descriptions which had elicited images based closely on autobiographical memories were more likely to be generated quickly and to be vivid, yet did not show mood-congruent retrieval. However, images which were less personal did so. Teasdale & Dritschel suggest that the less personal images required more elaborative processing to create the image, thus creating more extensive elaborative links which the later mood induction could influence. The possibility that retrieval (even autobiographical retrieval) may vary in the level of processing required (particularly perceptual versus conceptual processes) is one to which we shall return. Teasdale & Barnard's theory that it is the amount of elaborative processing which is important, rather than simply the amount of self-reference, is also consistent with findings of moodincongruent memory by Rinck, Glowalla & Schneider (1992). They gave subjects a depressing or elating MIP and required them to rate the valence of pleasant or unpleasant words. Half the words were strong in affective tone, and mood congruent encoding for these words was found. However, for the other half of the words (weaker in affective tone), mood-incongruent encoding was found. Subjects who had rated mildly pleasant words while happy had poorer recall of these words later. It appeared that subjects took longer to make their decision on the valence of these mildly affective words, so that, when the words were slightly pleasant and mood was happy, the decision could be made faster, with less need to elaborate the word's meaning. The result would be a counterintuitive poorer memory for the mildly positive words encoded in positive mood. Before considering the relevance of various theories to explain mood-memory bias, however, we briefly review other research on memory and emotion: memory for emotional events in normal subjects, and research on whether emotion itself can be remembered. Memory for Emotional Events in Normal Subjects in Neutral Mood Christianson & Safer (1996) review the research on the effect on normal subjects' memory for
emotional events. Some such events appear to be Page 145
able to be recalled extremely vividly (flashbulb memory phenomena) though Christianson points out some reasons for caution. First we should not believe that the memory is accurate just because it is imageable or reliable over time. Second, there are many clinical examples in which traumatic events are followed by amnesia for the event, and so generalisation from analogue groups may not be appropriate. Nevertheless, when laboratory studies have studied milder forms of stress situations manipulating the emotional content of TBR material (usually varying the unpleasantness of the material through using scenarios such as accidents), they find a tendency for people to be able to recall more central details (or the gist) of the episode, the more emotional the material. By contrast, unemotional material often allows recall of more peripheral detail (Christianson & Loftus, 1987). This pattern of results might be taken to imply that the search strategy used by subjects viewing emotional or unemotional material differs. An explanation in terms of such peripheral processes will not account for all the data, however. It is more the centrality of the content than the centrality of the physical location which is important. Burke, Heuer & Reisberg (1992) found that memory for an emotional story (compared with memory for a neutral version) contained more information about the gist of the event, and more information about the central characters, an advantage for emotional material which lessens over time. Further, when pictorial stimuli are used so that researchers can equate the amount of critical TBR material and control for the number of eye fixations on this information, there is still an advantage for subjects in the emotional condition (versus the neutral condition) in remembering detail information concerning the central part of the picture (Christianson & Loftus, 1991 and Christianson et al., 1991). These data seem to require explanation in terms of qualitatively different types of processing that emotional and neutral information elicits, rather than simply assuming that the same type of processing is varying in the 'bandwidth' of information encoded. Emotional stories would be more likely to demand action if they occurred in real life, and appear to be encoded in terms of their higher level meanings and implications with perceptual details being encoded if they, in their turn, would be relevant to the action that would be demanded (for example, the presence of a gun). Centrality is thus not defined in terms of location, but in terms of centrality to the action necessary. We shall return in Chapter 11 to consider how the concept of 'Implication' has become central to a theory of emotion proposed by Teasdale & Barnard (1993). Page 146
If the encoding of emotional information elicits such facilitated processing for central details ('tunnel memory'), why have the data on anxious people's memory for threatening themes been so inconsistent? Note that the focus of the studies reviewed by Christianson & Safer (1996) was normal subjects. Non-clinical and clinical subjects may differ in a number of ways, for example in personality or as the result of different levels of exposure to extreme traumatic events. The result is that, in clinical groups, emotional material is not simply more aversive, but has different implications. Even in the apparently straightforward case of spider phobia, it appears that phobics have beliefs about the catastrophic effects of being with a spider and being unable to escape (e.g. I might have a heart attack; Thorpe, 1994) which affects their processing of spider materials. Thus the focus of attention for the phobic in a spider memory task may not be on the spider but on themselves, the potential embarrassment of screaming, collapsing and generally failing to cope. Memory for Emotions Themselves Do people remember emotions, or just the events that elicited them? Strongman & Kemp (1991) found that, when asked to recall experiences of 12 different emotions, it was the objective
circumstances surrounding the event which were recalled, rather than their own feelings. Perhaps, then, emotion is like pain: it can be reinstated, but not remembered (Morley, 1993). If so, what are the processes by which emotion is reinstated. Several investigators have suggested that such processes depend on implicit or unconscious processes perhaps reflecting evolutionary early, perceptual processes (Tobias, Kihlstrom & Schacter, 1992; Johnson & Multhaup, 1992). If so, then it should be possible to remember affective aspects of events without recalling the events themselves. The Claparede phenomenon might be classified as a demonstration of this, and Johnson, Kim & Risse (1985) were able to show that 80% of amnesic patients who were unable to remember information about two people (depicted in photographs) over a 20 day interval, nevertheless preferred that picture which had been presented to them with 'biographical' information indicating that the person was a 'good guy' versus the picture of the 'bad guy'. One corollary of the dependence for recalling emotion on implicit processes is that subjects are very bad at accurately estimating their own past mood. This inaccuracy emerges both from diary studies with non-clinical subjects (Fitzgerald, Slade & Lawrence, 1988; Thomas & Diener, 1990) and from depressed patients (Zimmerman & Coryell, 1986; Page 147
Scharder et al., 1990). Christianson & Safer (1996) conclude that no study has shown subjects to be accurate in recalling the intensity or frequency of previously recorded emotions. Using such conscious, mnemonic strategies to make estimates of past mood seems inherently unreliable. The important aspect of mood appears to be how strongly it re-presents itself, and this is not dependent on explicit recall. Psychotherapists are trained to be aware of this re-presentation phenomena, and to observe whether memories elicit emotion in patients despite them saying that an event had not disturbed them. However, psychotherapists may need to be aware of their own capacity for forgetfulness. In one study of psychoanalysts' memory, it was found that the therapists had very poor recall of material they had learned from the patients at initial assessment, and poor recall even for their own initial treatment plan (von Benedek, 1992). The Relevance of Network Theory for Mood-Congruency Effects The evidence reviewed earlier seems to indicate that mood-congruent encoding and retrieval are reliable phenomena: that they can be demonstrated for autobiographical memory and for laboratory word lists or stories with mixed affective content. Difficulties in demonstrating some mood-congruent effects seem plausibly explained by the interconnectedness of the to-be-recalled material, or by the extent to which a subject may encode the material elaboratively. Over the last few years, researchers have questioned whether associative network theory is the most appropriate model to explain the data. It is a major premise of network theory that different emotions are represented as units or nodes in the same semantic network with which memories of events are recorded. But the same semantic network underlies our ability to use emotional words appropriately in any context, not just when the emotion denoted by the word is being experienced. Why is it that any use of an emotion word does not activate the whole emotional apparatus? In Abelson's terms, when is a cognition 'hot' and when 'cold'? In reading this chapter, for example, depression nodes will have been activated repeatedly in the readers, but not (we hope) in a way that has brought to mind all the readers' worst experiences of failure and rejection. What is it that determines when 'depression' will or will not make a person feel sad? Bower's attempt to account for this was to assert that there are different types of node for each emotion word: 'We should distinguish the node corresponding to the concept of fear vs. the node for experiencing fear itself' (Bower & Cohen, 1982, p. 309). Page 148
There are many objections to this development of the theory. First, it is ad hoc, Bower could go on
proliferating nodes indefinitely to get himself out of tight corners. Second, there are doubts whether it is plausible to build hot and cold nodes into the same network as though they are activated in the same way and had the same kind of spreading activation effects. Much of these problems are explored in detail by Teasdale & Barnard (1993) and will not be discussed further here. However, most difficulties for network theory have arisen not from these theoretical problems, but with empirical evidence which has refuted its predictions. In particular, according to network theory, bias should occur on both attentional tasks and memory tasks, on both direct and indirect tests of memory, and for all material congruent with a particular emotion, irrespective of what that emotion is. We have reviewed the evidence on attentional bias in the previous two chapters, showing that material that elicits such a bias is not necessarily recalled preferentially later. In the next sections, we shall review the recent evidence for memory bias in depression, focusing on those studies that examine bias on indirect tests of memory, and then review evidence for bias on both direct and indirect forms of memory test in other moods and clinical conditions, most notably anxiety. Direct versus Indirect Tests of Memory in Depression Williams et al. (1988) suggested that whereas anxiety might be associated with automatic priming of negative stimuli, bias in depression was more likely to derive from bias in resources allocated to elaboration of material (that is, bias in the encoding and retrieval of the relations between items, rather than the integration or priming of item-specific information). Consistent with this, studies of depressed patients that have used direct tests of memory, examining free recall or cued recall of negative material, have found little difficulty in demonstrating a mood-congruent memory bias (e.g. Denny & Hunt, 1992; Watkins et al., 1992; Fromholt, Larsen & Larsen, 1995). Denny & Hunt (1992) examined the performance of 16 depressed patients (DSM-III-R diagnosed major depressives) and 16 matched control subjects. Negative and positive words were presented in a self-reference condition (Describes you? rating 16 on a Likert scale). Depressed patients recalled significantly less positive adjectives (M = 1.3) than negative adjectives (M = 3.1), whereas controls showed the opposite pattern (M = 4.9 and 3.4 for positive and negative respectively). Differences in endorsement on the self-descriptiveness of the words did not affect the retrieval Page 149
bias. It appears that the bias arose mainly from the depressed patients' poor recall of positive words, though the lack of neutral words means that it cannot be judged to what extent the depressed patients' overall level of recall was lower. Watkins et al.'s (1992) study of depressed patients included neutral words, and thus was able to judge the origin of the bias. Seventeen depressed or dysthymic patients (DSM-III-R) were compared with 17 matched controls. Depression-related words (e.g hopeless, guilty, worthless) physical threat words (e.g. assault, collapse) neutral and positive words were presented to subjects, and their task was to imagine themselves in a scene involving themselves and the word. Word stems were presented as cues for recall, and the, proportion of correct retrieval showed depressed patients to be more likely to retrieve negative items (M = 0.49) than control subjects (M = 0.41). The pattern of recall of neutral items showed a reversal of such mood congruence, but no overall memory deficit for depressed patients (M = 0.40 for patients, M = 0.49 for controls). Each of the above studies aimed also to examine whether depressed patients would show a memory bias if an indirect (implicit) test of memory was used. Denny & Hunt (1992) required subjects to complete words from fragments (e.g. the word 'peaceful' from the fragment '-ea-ef--'). Forty-eight such fragments were presented, 24 of which had been seen by the subjects in the study phase. The extent to which exposure to the word in the study phase primed the word was evaluated by subtracting the fragment completion rates of new words from that of the old words.
This showed no significant bias. The mean difference for positive words was 2.69 for depressed patients and 2.92 for controls. The mean difference for negative words was 2.75 for depressed patients and 2.25 for controls. A similar pattern of results was obtained by Watkins et al. (1992). They used word completion both as cues in a cued recall test task, and as cues for an indirect test of memory (complete the first word that comes to mind). As we have seen, the direct test of memory produced a clear mood-congruent bias in the depressed patients. By contrast, the indirect test showed no significant effects. The primed/unprimed difference for positive, neutral and negative words for depressed patients was 0.09, 0.12 and 0.13 respectively. The difference for control subjects was 0.13, 0.13 and 0.07 respectively. The only exception to this pattern of results was reported by Elliott & Greene (1992) who found that depressed patients showed less priming than controls on two indirect tests (first their spelling of homophones such as 'fair, fare' following encoding with a cue to bias towards the nonPage 150
dominant spelling, such as, 'taxi -fare', and, second their word stem completion). However, this study differed from those of Watkins et al. (1992) and Denny & Hunt (1992) in that it used neutral materials that were unlikely to engage the interest of patients. Williams et al.'s (1988) hypotheses are most relevant to memory bias, not general memory deficits. The balance of evidence suggests that, while direct tests of memory show consistent bias in depression, there is no such consistency of such bias when indirect tests are used. Anxiety and Other Clinical Conditions Network theory does not make distinctions between different moods in the effects they are predicted to have on memory. All moods should enhance encoding and retrieval of material congruent with them. In our 1988 book, we indicated that such effects were more difficult to find in anxious groups than they were in depressed groups. Although mood-congruent recall had been found in agoraphobic patients (Nunn, Stevenson & Whalan, 1984) this has not been replicated (Pickles & van den Broeck, 1988) and in a series of experiments on patients with a diagnosis of general anxiety state, Karin Mogg had repeatedly failed to demonstrate mood-congruent recall of threat-related words. In some of her experiments, there appeared to be a reversal of moodcongruent enhancement, with retrieval of threat words gradually diminishing over three learning and recall trials (Mogg, Mathews & Weinman, 1987). How far is this pattern true in other conditions? Spider Phobia Similar results had been found by Watts (1986, p. 229) using words relating to spiders (e.g. web, hairy) in a group of spider phobic subjects. In one experiment, despite enhanced recognition for spider words relative to control words, there was diminished free recall for the spider words. Watts had suggested that the materials for this group may involve 'distaste' or 'disgust' which inhibits rather than facilitates search through memory. A parallel explanation could be advanced for Mogg's findings with threat-related words. Specifying exactly what aspects of encoding or retrieval are involved in this 'inhibition' has become an important issue for the further research which has been carried out since Williams et al. (1988). Further research with spider-avoidant subjects has replicated Watts's earlier results, at least when the material to be recalled directly refers to the Page 151
feared item. Thus Watts & Dalgleish (1991) performed two experiments comparing spider-
avoidant subjects' recall of spider words with recall of matched words to do with babies. They also manipulated whether a spider was present or absent at encoding or retrieval. In both experiments they found less recall of spider words in spider-avoidant subjects. A delay of 24 hours prior to retrieval made no difference to this pattern of results. It is interesting that the one experiment which has found a memory bias in spider-avoidant subjects did not involve the actual word 'spider', nor required subjects to respond with such words. Rusted & Dighton (1991) adapted Anderson & Pichert's (1978) paradigm, a description of a visit to an old house, which might be encoded in different ways (for example, as if one were a burglar). Seven items were inserted which might suggest the presence of spiders (for example, reference to a 'garage, which was empty, and judging by the amount of cobwebs, hadn't been used for a long time'). The results showed that spider avoidant subjects recalled more spider-related detail than controls. One problem with the paradigm was that no record was taken of the subject's individual reading times, so if a spider-avoidant subject took longer an certain passages, it was not monitored. Nevertheless, if the results prove reliable, they will add important information to what we know about retrieval of threat-related material in specific phobia. In the light of Watts's repeated failure to find such bias it will be important for further research to specify what the critical difference between the paradigms might be. Three obvious candidates suggest themselves as the basis for the difference. First, Rusted & Dighton's paradigm does not involve exposure to the actual feared stimuli at encoding or retrieval phases. We shall see that a major problem in examining memory for threat material in anxiety is the requirement of such tasks that individuals respond with the item which they fear. Paradigms that avoid such a direct conflict between memory and response may be more likely to find memory bias. Second, the fact that the items to be recalled were all locations (laundry room, garage, basement, bathroom, his room, Granny's bedroom, under the bed) may be significant. Encoding the possible location of a feared object may be adaptive, even in the absence of encoding details of the object itself. Third, Watts's experiments all used word lists, where intra-list association was strong, and there may have been cue overload effects. Cue overload refers to the decrement in performance found when too many items in a list share the same category. For example, people will recall more items from a list that includes two categories: 'apple, lion, giraffe, pear, grape, tiger' than a list of the same number of items from only one category: 'lion, tiger, giraffe, monkey, elephant, zebra'. The first list allows two category Page 152
cues (fruit and animal) to aid retrieval. The second allows only one cue, which is said to become 'overloaded'. If Watts's subjects encoded all spider-related material in relation only to the cue 'spider, poorer recall would result. In fact, Watts & Dalgleish (1991) did find more spider-related intrusions in the spider-avoidant subjects, consistent with a category-based retrieval strategy which might have impaired performance. In Rusted & Dighton's experiment, however, the TBR material represented items which, although all locations, differed in many respects, one from the other. Laundry rooms have many associations which are distinct from the associations with garages which, in turn, are distinct from the associations most people have to grandmothers' bedrooms! Cue overload is therefore most unlikely to occur in this context. We turn now to the studies which examine memory in conditions or moods other than depression. Following the suggestion that anxious subjects may be more prone to show mood-congruent bias on indirect tests of memory (Williams et al., 1988), several studies have examined both direct and indirect tests in subjects varying on trait anxiety or in anxious patients. (For the moment, we shall use the terms indirect and direct tests of memory interchangeably with implicit and explicit memory, though later we shall discuss the dissociations within each of these categories.) Each of
these will be examined in turn, with special reference to those which have used both direct and indirect tests of memory. Anxiety and Neuroticism in Normal Subjects In an early study of content specificity in normal mood, Ingram et al. (1987) had given depressed, anxious or neutral words to students who were high on a test anxiety scale, comparing their later retrieval of these words with mildly depressed and control subjects. Words were encoded selfdescriptively (Describes you?) or semantically (Does this mean the same as . . . ?). Results showed that the test anxious students were more likely to recall the anxious words (though not the depressive words). Similarly, Claeys (1989) found socially anxious students to be more likely to recall unlikeable trait adjectives than controls. However, Foa et al. (1989) could not replicate these effects with speech anxious students, given threat or neutral words in a self-referent encoding study phase. They manipulated mood (anxiety vs relaxation) at both study and test phases with a view to separating encoding versus retrieval effects. Not only did they find no significant bias on the main analysis of the study, but when they looked only at those subjects who showed increased anxiety from study to test phases (as indexed by increased heart rate), these subjects showed reduced recall of threat words. Page 153
Bradley et al. (1993) examined the effects on memory of levels of neuroticism, dividing their 39 subjects (62% male) into high and low N and giving them a depression or neutral MIP following which subjects rated positive or negative words for self-descriptiveness. Martin (1985) had found that high N subjects showed a recall bias that was independent of depression level. However, Martin and colleagues had used only female subjects, and there are reasons to believe that females are different from males in the frequency of usage of some self-descriptors, which may make them more vulnerable to mood-memory effects (Clark & Teasdale, 1985). Bradley et al. (1993) found no main effect for level of N or MIP on incidental recall. However, there was a significant interaction between them. For those subjects who had a neutral MIP, the higher the N score, the less recall bias was evident. By contrast, for those subjects who had the depression MIP, the higher the N score, the more mood-congruent recall bias was evident. The association of induced depression with memory bias was not unexpected, but the interaction with N suggests that N may act as a vulnerability factor, as Martin (1985) had suggested. However, the fact that in the absence of induced mood, N appears to suppress recall of negative material should remind us of the possibility that mood or personality factors other than depression may give rise to complex results, sometimes showing mood congruence, sometimes the opposite and presumably sometimes cancelling each other out. A similar conclusion arises from a study by Bradley & Mogg (1994). Sixty-two students were required to encode self-referentially positive and negative words, and given an incidental free recall test. In those subjects with low depression (BDI) scores, there was no association between neuroticism scores and recall bias. However, for high BDI scorers (11 and above; N = 20), increased N was significantly associated with greater recall bias. Each of the above experiments focused on explicit recall. Four studies have examined whether high trait anxiety is associated with bias on indirect tests of memory (Richards & French, 1991; Bradley, Mogg & Williams, 1994; Nugent & Mineka, 1994; MacLeod, 1990). Richards & French (1991) gave subjects words to encode either by simply reading them or by self-referenced imagery. In one experiment using an explicit memory test task, no memory bias was apparent. However, using a word fragment completion task, they found that high trait anxious subjects showed a negative bias.
Bradley, Mogg & Williams (1994) used a repetition priming lexical decision paradigm as an indirect test of memory, with primes being either supraliminal (and presented in a list before the LDT started), or subliminal (and presented prior to each target word). They also used an incidental free recall Page 154
task to examine explicit memory. Fifty-three student subjects were presented with words which were either depression relevant (misery, discouraged), anxiety relevant (embarrassed, emergency), positive (adorable, bliss), categorised neutral (carpet, domestic) or uncategorised neutral (geometry, enlarge). Subjects completed trait and state anxiety inventories and the BDI. Trait anxiety was found to correlate with depression (r = 0.70) so subjects were divided (on the basis of trait anxiety scores) into high or low negative affectivity groups. Results for the lexical decision task showed no differential priming for the supraliminal stimuli (all subjects on all types of word showed repetition priming effects). On the subliminally primed words, however, the high negative affectivity group showed priming for depression-related words (F = 4.49, p < 0.05), though there was also a strong suggestion of a priming effect of similar scale for positive words (F = 3.68, p = 0.06). The low negative affectivity group showed subliminal priming for anxious and neutral words. Correlational analysis found that the degree of priming for depression-related words was associated with level of depression on the BDI (even when trait and state anxiety were partialled out). Since this pattern of results was not predicted, it needs replicating before we can assess how much weight to place upon it. This is especially true given previous failures to find mood biases using lexical decision tasks (Martin & Clark, 1985; MacLeod, Tata & Mathews, 1987; Matthews & Southall, 1991; though see Hermans, De Houwer & Eelen, 1994, for use of 'affective decision' paradigm that is sensitive to affective priming). Bradley, Mogg & Williams' (1994) results for the free recall task found that level of depression predicted degree of biased recall (r = 0.26, N = 53), but only after trait and state anxiety were partialled out. This was more consistent with prediction, and the finding that the association emerged only when anxiety was removed statistically is consistent with the studies reviewed earlier on the way neuroticism and depression interact. In discussing the studies of Bradley et al. (1993) and Bradley & Mogg (1994) we suggested that neuroticism acted at low levels of depression to impede mood-congruent encoding and/or recall. When mood becomes acutely more depressed, however, neuroticism acts to amplify the effects of the mood disturbance on memory. The fact that we must remain cautious about these results is confirmed by the study of Nugent & Mineka (1994). In two experiments on undergraduates (N = 69 and 111), they divided subjects into high and low trait anxiety, and exposed them to threat or control (positive or neutral) words. Subjects were required to rate the likeability of the word, and were later given both direct (recognition and recall) and indirect (word completion) Page 155
tasks. Neither direct nor indirect tests of memory showed a mood-congruent bias effect. They point out that their subjects had lower levels of state anxiety than the subjects in Richards & French (1991), so it is possible that a certain degree of acute mood disruption is required to activate subjects' vigilance. Consistent with this suggestion, MacLeod (1990) found that high trait anxious subjects showed biased spelling of homophones only under conditions of high arousal (following exercise). General Anxiety Disorder Following failure to find explicit memory bias (Mogg, Mathews & Weinman, 1987), Mogg (1988) reported a further five experiments examining memory performance in GAD patients, with no evidence of memory bias in four out of the five. Although Mogg & Mathews (1990)
found some evidence of increased free recall of threat words (encoded by self vs other reference) in GAD patients compared to controls, they also found that these patients had more threat word intrusions in the memory task. Further, the bias towards threat was found for both self and other rated words. The results suggest that this was wholly due to response bias. MacLeod (1990) assessed anxious patients and controls for their recognition and recall for items which had been used on an emotional Stroop test. No evidence for explicit memory bias was found. However, the anxious patients did show a lower perceptual threshold for the threat words that had been used in the Stroop task. Mathews et al. (1989) examined both a direct and an indirect test of memory in GAD, recovered anxious patients (recovered for at least six months), and matched controls (N = 18 in each group). Ninety-six words were used, either physically or socially threatening (N = 48; e.g. ambulance, immature) or non-threatening (N = 48; positive or neutral, e.g. applause, carpet). Subjects were required to study a sample of these words in the study phase and imagine themselves in a situation involving the word. Word stems were used in the test phase, either with instructions that the stems should be used as cues to try to recall words from the study phase (direct memory test), or with instructions to complete the stem with the first word that came to mind (indirect test). Each word had been chosen so that the stem had an alternative completion which would make a more frequent word. That is, any possible priming by the study phase would bias the word completion away from the more dominant completion. Results showed that, consistent with previous studies by Mogg, there was no mood-congruent memory bias for the cued recall (the direct test). Page 156
However, on the indirect test, the priming by the study phase resulted in the anxious patients completing more words with threat completions (primed minus unprimed difference score, M = 2.5) than either the recovered patients (M = 1.5) or the controls (M = 1.4). By contrast, for the non-threat items, anxious patients showed less evidence of priming by exposure during the study phase (primed-unprimed difference, M = 1.8) than either the recovered patients (M = 2.7) or the controls (M = 2.6). In interpreting this pattern of results, it should be noted that the larger priming effect for GAD patients for threat words was due to the low rate of completion of the stems by a threat word (M = 0.8) compared to stem completion for a non-threat word (M = 1.5) in the unprimed condition; that is, if the words had not appeared in the study phase. Priming of anxious patients by exposure to threat words equalised the stem completion rate for threat and non-threat words (M = 3.3 for both), which constituted a larger increase (priming effect) for the threat words. The reluctance of anxious subjects to respond to word stems with a threatening word when there is another available (in the absence of priming) is something to be borne in mind in interpreting other data on anxious patients. Even if the difference between threat and non-threat completions for anxious patients were difficult to interpret, the study also found a statistically reliable difference for control subjects in the opposite direction to the anxious patients. That is, controls were significantly more likely to show greater priming for the non-threat words (prime - unprimed difference, M = 2.6) than threat words (M = 1.4). These difference scores result from an increase, for threat words, from 1.2 completions when not primed to 2.6 completions after priming; compared to an increase, for nonthreat words, from 1.5 completions when not primed, to 4.1 after priming. Thus the significant interaction between Group, Word valence and Priming was not only due to bias towards threat in anxious patients, but also due to a bias away from threat in normal controls.
This pattern of results for a word completion test task was not replicated in a subsequent study by Mathews et al. (1995). Although the lowest completion rates were again found in anxious patients on threat words that had not been exposed in the study phase (M = 1.75 versus 2.3 for non-threat words), the increase in completion due to exposure was not different for the different types of word (2.1 for threat words and 2.2 for non-threat words). Neither was the finding for normal controls replicated. Mathews et al. (1989) had found that normals had shown a bias against threat completions, a pattern which had contributed to the overall difference between the clinical and control group. Normal controls in Page 157
the later study showed no such bias favouring non-threat completions after priming. However, the studies differed in the important respect of their encoding procedures. Mathews et al. (1989) had required subjects, in the study phase, to imagine themselves in a situation suggested by each word (a conceptually driven task). By contrast, Mathews et al. (1995) had required subjects to perform only a surface (data driven) encoding of each word by counting the number of letter 'e's in each word. This encoding condition was sufficient to produce a large priming effect on the word completion task for all subjects for all types of word, but did not produce any extra advantage for the anxious subjects on the threat words. One possible explanation is that biased completions require some (at least minimal) conceptual encoding. The experiment of Eysenck & Byrne (1994) which found a word completion bias towards threat in high trait anxious normal subjects had employed an encoding task (subjects being instructed to read the word) which would have permitted more conceptual encoding than the orthographic task of Mathews et al. (1995). Mogg et al. (1992) attempted to examine explicit versus implicit memory processes in GAD patients and matched controls by using the distinction between 'remembering' and 'knowing' that an item has been seen in the study phase. The remember-know distinction arises from the attempt to capture the difference between elaborative/explicit memory processes (i.e. recognition based on explicit recollection of an item) and non-elaborative/implicit memory processes (i.e. recognition based on the perceptual fluency of the item). Subjects were given 32 words (16 threat and 16 non-threat) during the study phase, and were told there would be a memory test. Ten minutes later they were given recognition test, with the 32 old words being embedded in a list of 64. Results showed no recognition bias in the anxious patients. Subjects reports about whether, if they recognised a word, they remembered seeing it in the study phase, or whether they simply 'knew' it was there also failed to show any group differences or interactions. The absence of any overall bias in recognition, however, renders the paradigm weak in looking for further distinctions within the results. With no overall recognition bias, and no bias in 'remembered' items predicted or found for anxious patients, there are no degrees of freedom left for any bias on implicit 'know' items to be detected. Recognition memory appears to be a particularly insensitive method for exploring emotional bias. It has not been successful in showing memory bias even in depression, where other tasks have been successful (Williams et al., 1988), so it would have been surprising if a recognition paradigm Page 158
was successful in showing memory bias in anxious patients where the results have been so much more inconsistent. Autobiographical Memory Bias in Anxiety It is now clear that anxious mood can be associated with biased autobiographical recall. Mayo (1989) gave a cued autobiographical memory test to subjects varying in level of neuroticism and
trait anxiety. They found that increased N was associated with decreased recall of positive events. In addition increased trait anxiety facilitated recall of negative events and inhibited recall of positive events. Burke & Mathews (1992) used the Lishman cue words to prompt retrieval of personal events in 12 anxious patients (GAD) and 12 controls matched for age and IQ. Anxious patients recalled more memories which were later judged as 'nervous' by the patients themselves and by independent judges. Higher scores on trait anxiety were associated with a greater number of 'nervous' memories. Interestingly, there was also a significant association between higher depression levels (assessed using the Zung) and the retrieval of fewer pleasant memories. Even within GAD patients, therefore, depression seems to play a part in memory bias. However, there was little evidence that anxious patients recalled more overgeneral memories (see later). Whether this represents a difference between anxiety and depression, or between how threatening or depressing events are encoded or retrieved, has not yet been determined. These studies suffer the same limitation as the original study by Lloyd & Lishman (1975)that the more anxious subjects may have suffered more real-life threat events, and that it is this actual frequency which contributes to the memory bias, not the anxious mood. However, just as Teasdale & Fogarty's (1979) study was able to use mood induction to demonstrate that depressed mood did indeed affect retrieval, so Richards & Whittaker (1990) have used experimental induction of anxiety to demonstrate the potency of this mood in affecting retrieval from autobiographical memory. Richards & Whittaker (1990) exposed 40 subjects (varying in levels of trait anxiety) to three pleasant pictures (a baby and two scenes) or three violent pictures (severed head, scarred face and football hooligans) from newspapers. Following this induction, they gave six pleasant (e.g. happy, peaceful) and six anxiety-related (e.g. nervous, danger) cue words, allowing subject 60 seconds per cue to respond with an event that the word reminded them of. The negative MIP had a marginally slowing effect on Page 159
latency to recall pleasant events, but a large effect on reducing the latency to recall anxietyrelated personal events. Two further effects should be noted. First, they found no tendency for the memories of the anxious subjects or for the anxious memories to be more overgeneral (see discussion of generality of autobiographical memory later in this chapter). Second, they found a result which paralleled that of Burke & Mathews for the involvement of depression. A multiple regression analysis found that although the latency to retrieve anxious memories was predicted best by level of state anxiety, the latency to retrieve positive memories was predicted best by depression (as assessed by the BDI). Other Clinical Groups McNally, Foa & Donnell (1989) examined memory bias in 22 patients with a DSM-III-R diagnosis of panic disorder (compared to 19 matched controls). Subjects were given 15 anxious words (e.g. uneasy, afraid, fearful) and 15 control words (clever, sociable, cheerful) and asked how much the words described themselves. Subjects were given the task either after an arousal condition (step-ups on a cinder block) or after relaxation. A significant Group X Word interaction confirmed that panic patients were showing biased free recall of threat words, an effect which was almost wholly due to the high arousal condition. In this condition panic patients recalled 4.8 anxiety-related words and 3.9 control words; controls recalled 3.8 anxiety-related words and 4.8 control words. More recent studies have extended this work to look at direct versus indirect tests of memory.
Like the experiments on GAD patients, results have been inconsistent for implicit memory, though for explicit memory there is a greater consensus. Cloitre et al. (1994) compared panic patients and matched controls, and found bias on both explicit and implicit tests of memory, but Becker, Rinck & Margraf (1994) found an explicit memory bias but no implicit bias. These experiments have been consistent in finding an explicit memory bias towards threat in panic patients (though in the experiment referred to earlier by Otto et al. (1994), the bias only emerged in patients with greater relative left-hemisphere advantage. Although the results for implicit memory, like those for GAD patients, are inconsistent, Amir et al. (in press) offer a possible new way of looking for such indirect effects of memory. They used Jacoby's white noise paradigm, in which subjects are required to judge the loudness of noise in which words or phrases are embedded. In the original paradigm, phrases that have been heard before in a study phase are, by virtue of being more Page 160
fluently processed, more easily heard against the noise, and thereby give rise to the judgement that the noise is lower. Amir and colleagues presented threat sentences (e.g. 'the tingling sensations worried the anxious woman') or neutral sentences (e.g.'brass buttons adorned the wool coat') to 14 panic patients against three levels of white noise. In the two loudest noise conditions, no effects were observed, but in the quietest condition they found the predicted effect, with panic patients rating the noise as significantly quieter for the threat sentences. Nugent & Mineka (1994) express several cautions about the interpretation of previous results on implicit and explicit memory. First, emotion words are not always equated with other word categories on word frequency. Second, on implicit memory tasks, subjects are sometimes told that some stems could be completed with words they had seen before, but were instructed to complete the items with the first word that came to mind. This might activate explicit retrieval strategies. In fact, however, these criticisms might not be as damaging as they look at first. The nonmatching for frequency for stress words should act against most investigators' hypotheses, on the grounds that more frequent words produce less priming effect on word completion than less frequent words (MacLeod, 1989). Second, the instructions to subjects might lead to some subjects using explicit retrieval strategies, but the purpose of such instructions is to get around the problem of the subject who notices that a word was in the previous list, and then wonders whether this is allowed or not. This instruction has been used in other studies (e.g. Bowers & Schacter, 1990) where it is considered better practice to anticipate the questions that might arise in such a paradigm, but to do so in a way that is the same for all subjects, rather than allowing variation in subjects' spontaneous strategies to add noise to the data. Having said this, Nugent & Mineka's complaint raises an issue that is common to virtually all the implicit memory tasks used to date: that it is unknown to what extent subjects use explicit strategies to do the task even though not requested to do so. This makes it extremely difficult to interpret those studies that have found memory bias on both implicit and explicit tasks. However, any study (e.g. Mathews et al. 1989) that finds a bias on indirect tasks, but not on direct tasks, is less prone to this criticism. In fact, Mathews et al. (1989) examined the correlation between memory performance on the cued recall and stem completion task and found none. Page 161
Nugent & Mineka (1994) also raise the possibility that the inconsistency in demonstrating direct and indirect memory bias in anxiety may be due to mismatches between encoding and retrieval operations. It is to this issue that we now turn.
Data-Driven versus Concept-Driven Encoding and Retrieval So far, we have examined different patterns of results on explicit (direct) and implicit (indirect) tasks. However, over the past few years it has become increasingly recognised that experiments on the distinction between implicit and explicit memory have confounded the type of test used to assess memory with the type of processing performed at study or test. The implicit/explicit distinction is most clearly demonstrable when a subject who cannot recall having been exposed to a word in the study phase shows evidence of memory in biased spelling of the item, or reduced threshold for identifying that particular word, or increased probability of completing a word stem or word fragment with the word that has been studied earlier. Many of these demonstrations use encoding instructions that require subjects to focus on the perceptual aspects of the to-be-studied item (e.g. how many 'i's in it?). If perceptual tasks are used in the test phase (e.g. perceptual threshold tasks which assess the ease with which subjects are able to identify a word exposed for 30 ms), it is possible that it is the match between encoding and retrieval operations (they are both data-driven tasks) that is responsible for the effects, not differences between different memory systems. The reason why performance on these tasks are unaffected by 'explicit' operations may be because most explicit tasks demand the processing of the meaning of the itemthey are conceptually driven tasks. Performance on typical explicit tests are sensitive to differences in the amount of processing of the meaning of the item, but not the perceptual qualities of the iteme.g. what modality it has been presented in, or whether a visually presented word has been studied in upper or lower case (Roediger & Blaxton, 1987). Such an analysis raises the possibility that there might be implicit conceptual tasks, priming on which was dissociable from priming on implicit perceptual tasks. No longer can we be sure what an experiment showing a dissociation between implicit and explicit memory bias has really demonstrated: biased implicit memory (that might occur for implicit conceptual material also), or biased perceptual processing (that might occur for explicit perceptual tasks also). Page 162
Eysenck & Byrne (1994) attempted to examine the type of processing required by the study and test tasks as well as the explicit/implicit distinction in anxious subjects. They wished to examine the proposition of Roediger & Blaxton (1987) that 'the more similar the processing activities required by the test to the encoding activities, the better the performance on the test' (p. 371). They used two encoding conditions, one mainly data driven, the other mainly conceptually driven. In the first subjects were required simply to read the word (e.g. stockbroker, stupid), and in the second subjects were required to generate the word from its opposite, or from a brief definition (e.g. the opposite of clever: s . . . . . .; a month in winter: D . . . . . .). In both conditions, the word was on the screen for 10 seconds. There were three test conditions, also varying in the processing requirements. A word completion task was used as an implicit task (because subjects are not asked to try to recall a studied item). It is 'data driven' because the task requires matching of a pattern of letters on the page with an item of the same pattern from a subject's vocabulary. The second test task was cued recall, using word fragments (e.g. DSPR for despair). It was an explicit task, since subjects were instructed to use the fragments as cues to recall a word they had seen in the study phase, but it involves datadriven processing because of the requirement to match on the basis of physical characteristics of the word. Eysenck & Byrne also suggest that it involves some conceptually driven processing in that subjects are instructed to use the fragment to help search long-term memory for an appropriate word. Finally, there was a free recall task, an explicit and conceptually driven test of memory.
High, middle and low trait anxiety subjects were tested (total N = 40), the subjects also differing on state anxiety and depression (BDI) levels. A pool of 160 threat and categorised neutral words were used. Each of the words had a unique three-letter stem (within the list), and words were matched for the number of words not in the pool that shared that stem. Results showed that after the 'Generate' encoding condition, high trait anxious subjects showed a bias on the free recall (recalling many fewer neutral words) and cued recall (recalling more threat items), but such recall biases were not found after the 'Read' encoding condition. Partialling out depression from these results showed that the free recall bias disappeared, but the cued recall bias did not. Interest then focused on the word completion task. First, completions of words that subjects had not seen in the study phase were analysed to see if there were naturally occurring biases in word completion of threat Page 163
and neutral words. There were none (a result which is common, but interesting in the light of the ease with which interference paradigms such as the emotional Stroop are able to demonstrate naturally occurring attentional priming of threat material). Following the 'Read' condition, there was a significant interaction between anxiety levels and word valence, but no such interaction in the 'Generate' condition. Once again, both anxiety and depression contributed equally to this bias, with significant effects disappearing when either mood was partialled out of the association with stem completion bias. The authors conclude that negative memory bias on word completion depended on the encoding operation being the less conceptually driven task simply to read the word. The finding of a bias on such an indirect test of memory is consistent with the prediction of Williams et al. (1988), and the result adds the important constraint that such a bias may require data-driven encoding operations. However, the finding of a bias on a direct test of memory (free and cued recall) following elaborative, conceptual encoding appeared inconsistent with the Williams et al. (1988) prediction, though the authors point out that it is possible that depressed mood was contributing substantially to this effect. In particular, the finding that the free recall bias (which disappeared when depression was partialled out) had been due to the reduced recall of non-threat material is consistent with Richards & Whittaker (1990) and Burke & Mathews (1992) who also found that ease of retrieval of neutral or positive material in anxious subjects was associated with level of depression in those subjects. We might make the tentative generalisation that where a bias in anxious subjects is due to increased retrieval of threat material, it will not be due to depressed mood, but when anxious subjects show reduced recall of positive or neutral material, it is most likely due to the co-occurrence of depression. Eysenck & Byrne's study stresses the importance of varying the type of processes required of subjects at study and test. The most commonly used tasks used in study phases in the literature involve self-referent (Describes you?) judgements, clearly requiring conceptually driven processing. Failure to find subsequent bias using a data-driven task (such as word completion) may be due to mismatch in processing requirements between study and test, rather than a failure of implicit memory. Similarly, a failure to find explicit memory biases in anxious patients may reflect their failure to perform excessive conceptually driven processing for threat material when given such words in the study phase of an experiment. We shall return to this issue in the final chapter, and consider whether such distinctions offer a way in which current inconsistencies in the literature may be accounted for. Page 164
General and Specific Memories: The Need to Examine Retrieval Processes Williams & Broadbent (1986a) studied mood and memory in parasuicide patients and found that what appeared to be mood-congruent retrieval bias (delayed recall of positive memories) was in a large part due to the clinical group responding at first with inappropriately general memories. Whereas control subjects would respond to a cue word with a specific memory (some event that had occurred at a specific place and time), parasuicide patients tended to respond with a general category (e.g. to the cue 'happy' the response 'playing squash'). A similar finding emerged from a study of autobiographical memories in depressed subjects by Moore, Watts & Williams (1988), who presented patients and matched controls with 16 situations involving social support or lack of social support (e.g. 'a neighbour helped me with some practical problem'; 'my partner criticised me'). The percentage of first responses to these cues that were inappropriately general was 40% for depressives and 19% for controls (F (1, 32) = 11.07; p < 0.01). Further work has found the memory deficit in clinically diagnosed depressed patients (Williams & Scott, 1988; Puffet et al., 1991; Brittlebank et al., 1993; Kuyken & Dalgleish, 1995) In addition to research on depression and parasuicide, work by Kuyken & Brewin at the Institute of Psychiatry in London has shown that over-general memory is particularly severe in those depressed women who have suffered sexual and physical abuse (Kuyken & Brewin, 1995). Work by McNally at Harvard has found similar deficits in Vietnam veterans who have PostTraumatic Stress Disorder (McNally, Litz & Prassas, 1994, McNally et al., 1995). The pattern of data emerging from these studies suggests that overgeneral encoding and retrieval of (even neutral) events represents a long-term cognitive style which may arise early in development in response to traumatic events. Such a style can also be produced by reduced working memory capacity at the time events are being retrieved. The combination of a long-term overgeneral memory style and reduced capacity at retrieval is particularly disabling. These results make us focus on the processes of retrieval, something largely ignored by theories of memory that focus on relatively passive activation processes. Several investigators (e.g. Williams & Hollan, 1981; Norman & Bobrow, 1979, Reiser, Black & Abelson, 1985; Kolodner, 1985) have suggested that retrieval from autobiographical memory involves an intermediate stage in which a general context or description is framed to aid the search for a specific exemplar. The cue 'happy' will first pose the Page 165
question: 'What sort of activities, people, places, objects make me happy?' These general descriptions are recursively refined until a specific example is retrieved, its appropriateness checked and a suitable response made. It appeared that these patients were accessing an 'intermediate description' in memory retrieval, but were stopping short of a specific example. Williams & Dritschel (1992) further investigated the mechanisms underlying overgeneral retrieval. While the general notion of a hierarchical search strategy remained useful, the theory made no distinction between different types of hierarchy a person might use. Williams & Dritschel (1992) distinguished two forms of superordinate memory. The first, categoric memory, refers to descriptions containing generic summaries of events (such as 'drinking in pubs'). The second, extended memory refers to descriptions containing extended event time lines (such as 'my first term in Oxford'). Williams & Dritschel found these two forms of intermediate description to be independent from each other, and that only categoric memories correlated with neuropsychological tests of Supervisory Attentional Control. Further, they found that overgeneral memory in emotional disturbance was associated only with an excess of categoric memories but no increase in extended memories.
An unmodified descriptions theory explanation of the phenomenon was therefore too general in assuming that retrieval was aborted too early in these patients. If overgeneral memory were simply 'stopping short', depressed and PTSD patients' errors would be equally distributed between categoric and extended memories. Yet it is only when recollection involves a categoric intermediate description that it is likely to become blocked. Successful retrieval of specific episodes requires initial access to categoric descriptions but then rapid inhibition of these descriptions so that contextual (time and place) information can be introduced into the mnemonic search. Williams (1996) examines developmental data to suggest that, prior to the third and fourth years of life, normal children tend to retrieve events in a generic way. The ability to inhibit these relatively automatic categoric description processes develops at around three and a half years, allowing the child greater strategic control over the recollection process. It is this ability to inhibit categoric descriptions which is affected by reduced working memory capacity in ageing (Winthorpe & Rabbitt, 1988) and by brain damage (Baddeley & Wilson, 1986; Williams, Williams & Ghadiali, submitted). However, chronic stress or abuse in childhood may affect the ability to learn fully to control these processes. First, the person growing up in such an environment may learn that specific episodic information is too negative, so Page 166
passively avoids this punishing consequence of recollection. Whenever a mnemonic cue activates categoric intermediate descriptions which begin to retrieve fragments of an emotional episode, the search is aborted. Second, failure to access specific memories has further consequences for encoding and retrieving other (even positive) events. One failure to access specific information results in the retrieval process attempting to make further iterations with alternative intermediate descriptions. After a number of such iterations, a more highly elaborated network of categoric intermediate descriptions will exist. For example, on the first iteration a cue such as 'unhappy' may have elicited the description 'I've always failed', which might normally have helped retrieve the event 'received the letter telling me I had failed the exams'. However, if the search is aborted at this point, an alternative intermediate description will be derived, for example 'I was never good at school'. Several such iterations may result in a host of negative categoric descriptions (e.g. 'I never had many friends', 'I always let my parents down'). The result is an over-elaboration of categories, encouraged by and itself encouraging ruminative self-focus. These categoric descriptions are more likely to be encoded along with future event occurrences. In future attempts at retrieval, an initial cue is likely to activate an intermediate description which simply activates other self-descriptions. Williams (1994, 1996) coins the term 'mnemonic interlock' for this phenomenon, analogous to the more general 'depressive interlock' described by Teasdale & Barnard (1993) to explain the maintenance of entire ruminative cycles in depression. The result of the tendency for negative early experience to affect retrieval processes in this way is that the child will come to have a range of self-descriptions in a chronically activated state, so that new emotionally valent events will be encoded along with many general trait selfdescriptors. These overgeneric encoding-retrieval cycles constitute the cognitive style which we later find in depressed, PTSD and suicidal patients. The exact extent of later difficulties will depend on the existence of facilitatory or inhibitory factors in the retrieval context. For example, if another negative life event has recently occurred, there may be very few positive mnemonic cues in the environment to help the retrieval process. The result will be that the patients will have greater difficulty in retrieving specific positive events, the more they find themselves
currently ruminating about past traumas (Kuyken & Brewin, 1995; McNally, Litz & Prassas, 1994; McNally et al., 1995). This new emphasis on retrieval styles has implications both for theory of mood and memory in general, and for practical work with patients. Page 167
Taking theory first: in the past, network models have explained the prepotency of memories with a certain affective content in terms of the activation spreading from emotion nodes and other associated nodes. It focused on the single event memory as the unit of memory to be investigated. This concern with quantitative aspects of retrievalhow fast an individual could retrieve a memory, or how probable a negative or positive memory washas ignored important aspects of both emotional disorders and memory theory. This has occurred more generally in cognitive psychology. So, for example, the availability heuristic (Tversky & Kahneman, 1973) is premissed on the accessibility of representations of specific events in memory. Yet if specific memories are not always the preferred output of the memory system, it becomes important to re-examine the effects these alternative outputs have on cognition. For example, Williams et al. (1996) have shown that experimental manipulation of memory specificity affects the specificity in which subjects imagine the future. Further research will need to see to what extent other aspects of cognition (e.g. speed of making judgements) is or is not correlated with the probability of retrieval of specific events. Increased emphasis on a broader view of what the memory system normally outputs promises to link cognitive psychology more closely to the reality of recollection in everyday life. This represents an important way in which research on psychopathology has been able to feed back to the theory of normal psychological processing. Turning now to the practical aspects of the findings on general memory, we have seen that mnemonic interlock is particularly strong in people who have suffered past trauma. Although the trauma itself may not predict chronicity, we have data to suggest that generality of memory plays a critical mediating role in assisting or inhibiting recovery. Brittlebank et al. (1993) assessed depressed patients' autobiographical memory on their admission to a psychiatric unit, and found that the extent of overgenerality of their memory predicted whether they remained depressed seven months later. We can contrast this emphasis on retrieval mechanisms with the original network framework. Network theories, and their use of the concept of activation spreading through a network, led to the conclusion that effective therapies must either change the patterns of association in the network, or manipulate mood state to diminish the impact of spreading activation. But if emotional disturbance is associated with qualitative changes in the way events are encoded and retrieved (e.g. in level of specificity), we need treatment strategies that directly address the deficit being shown, and this requires the better description of underlying encoding and retrieval processes that have begun to emerge. Page 168
Many psychotherapeutic strategies depend upon the ability of patients to retrieve specific events from the remote or recent past, or to keep diaries of day-to-day events. Current research is examining predictions from this memory research: that effectiveness of many different types of psychotherapy will in large part be determined by how well they are able to bring about the specific recollection of current and past events (see Williams, 1992, ch. 11). Relevant to this issue are the findings that parasuicide patients who have greatest difficulty in retrieving specific memories show the greatest difficulty in producing effective solutions on the MeansEnds Problem Solving Task (Evans et al., 1992) and the greatest difficulty in bringing to mind specific images of future events (Williams et al., 1996). Specific retrieval predicts good outcome because it allows the recoding of depressive interpretations of the past and the construction of more specific images of the future, and because it enhances problem-solving skills.
Concluding Remarks Frameworks in psychology are often useful in suggesting research which then reveals the framework to be inadequate in interesting ways. Semantic network theory has been such a framework. Whereas models lead to 'predictions', frameworks lead to 'expectations'and it is when these are violated that a phenomenon comes to our attention. Several findings have called into question the sufficiency of network theory to account for mood and memory results: (1) the content of the to-be-recalled material (e.g. threat- or loss-related) makes a difference to memory bias, (2) the mood of the subject (e.g. anxiety or depression) produces a different pattern of results, and (3) the retrieval processes operating upon the material also makes a difference (blocking progress from general to specific memories in some clinical conditions but not others). In our 1988 review of early work, we suggested that these emerging patterns represented the start of new lines of research, each of which was likely to yield much useful data relevant both to theory and clinical practice. The current review has confirmed these conclusions. It has provided more evidence that the dissociation to which Williams and co-workers drew attention between different moods and different processes needs to be made. Whether the distinctions which have now emerged are best captured by distinguishing between automatic and strategic aspects of processing will be considered in detail in the final chapter. In that chapter we shall return to these issues, focusing particularly on the extensions to Williams et al.'s (1988) model that are needed in the light of the data we have discussed.
Page 169
Chapter 7 Thoughts and Images The biases in attention and memory found in anxiety and depression have often been explained in terms of the distinctive cognitive structures presumed to be associated with emotional disorders. These can be formulated in various ways, such as associative networks (Chapter 6) or schemata (Chapter 9). Mental models (Johnson-Laird, 1983) are another possibility, and one that Power & Champion (1986) have argued is well suited to cognitive theories of emotional disorders. However, it is moment-to-moment thoughts and images which are the main focus of therapist-patient interaction and the route by which all other cognitive phenomena are brought onto the therapeutic agenda. It is therefore important to consider their role in emotional disorders. Thinking and imaging probably do not exhaust the range of possible conscious cognitive phenomena, but they are the most ubiquitous. One of the central theoretical issues that has been discussed in connection with imagery is how it relates to underlying cognitive structures. Though this debate has focused on visual imagery, there is scope for a closely parallel debate about the relationship between thoughts and knowledge structures. This theoretical question of the relationship of thoughts and images to knowledge structures has a close bearing on several of the important clinical questions concerned with thoughts and images. The first two are concerned with the properties of emotional imagery. Thoughts and images are used in therapy to explore (or 'access') emotional reactions and preoccupations, and there is a need to give a theoretical account of how such accessing occurs. One of the questions that has been posed is whether images are better able than thoughts to access emotion. Second, there may be special characteristics of emotional thoughts and images. One possibility, for example, is that though they have considerable phenomenal impact, they tend to be lacking in detail. The next two issues are concerned with the role of phenomenal experience in bringing about underlying cognitive change. Processes of natural Page 170
recovery from traumatic events such as bereavement involve a process of 'working through' which appears to be carried on at the phenomenal level of thoughts and images. This raises the issue of how phenomenal aspects of 'working through' relate to more basic changes in knowledge structures. Finally, many therapies use the medium of images and thoughts to bring about fundamental and long-term changes in people's beliefs and reactions. Such changes also appear to depend on changing knowledge representations through the medium of work on images and thoughts. We shall consider each of these issues in turn. However, as a preliminary to this, it will be helpful to examine the general debate in cognitive psychology about the nature of imagery, and in particular to review Peter Lang's extensive body of theoretical and empirical work on imagery which has taken this debate as its starting point. Theories of Imagery The General Debate
There are essentially two major schools of thought about the nature of imagery: the 'pictorialists', such as Kosslyn (1980), Shepard & Cooper (1982) and Paivio (1986); and the 'descriptionalists', such as Pylyshyn (1983, 1984) & Kieras (1978). Block (1981) collected together a convenient collection of position statements in this controversy. The pictorialists emphasise the similarity between images and the objects they represent. Thus, they claim that the relations between the components of an image are analogous to the components of the corresponding external stimulus, and that the experience of an image is substantially similar to the perception of the external stimulus. The pictorialists have demonstrated a variety of phenomena relating to the manipulation and use of imagesfor example, that people can rotate them (Shepard) and use them to facilitate memory performance (Paivio). Studies of mental image acuity and scanning (e.g. Kosslyn, Ball & Raiser, 1978) are perhaps the most compelling evidence for the pictorialist position. The pictorialists assert that the functional properties of images depend on their pictorial qualities. Are these claims denied by the 'descriptionalists'? The problem in answering this is that the focus of the argument has shifted subtly over time, with each side finding a way of incorporating all the available facts into its own position. It may therefore be useful to go back to Pylyshyn's Page 170
recovery from traumatic events such as bereavement involve a process of 'working through' which appears to be carried on at the phenomenal level of thoughts and images. This raises the issue of how phenomenal aspects of 'working through' relate to more basic changes in knowledge structures. Finally, many therapies use the medium of images and thoughts to bring about fundamental and long-term changes in people's beliefs and reactions. Such changes also appear to depend on changing knowledge representations through the medium of work on images and thoughts. We shall consider each of these issues in turn. However, as a preliminary to this, it will be helpful to examine the general debate in cognitive psychology about the nature of imagery, and in particular to review Peter Lang's extensive body of theoretical and empirical work on imagery which has taken this debate as its starting point. Theories of Imagery The General Debate There are essentially two major schools of thought about the nature of imagery: the 'pictorialists', such as Kosslyn (1980), Shepard & Cooper (1982) and Paivio (1986); and the 'descriptionalists', such as Pylyshyn (1983, 1984) & Kieras (1978). Block (1981) collected together a convenient collection of position statements in this controversy. The pictorialists emphasise the similarity between images and the objects they represent. Thus, they claim that the relations between the components of an image are analogous to the components of the corresponding external stimulus, and that the experience of an image is substantially similar to the perception of the external stimulus. The pictorialists have demonstrated a variety of phenomena relating to the manipulation and use of imagesfor example, that people can rotate them (Shepard) and use them to facilitate memory performance (Paivio). Studies of mental image acuity and scanning (e.g. Kosslyn, Ball & Raiser, 1978) are perhaps the most compelling evidence for the pictorialist position. The pictorialists assert that the functional properties of images depend on their pictorial qualities. Are these claims denied by the 'descriptionalists'? The problem in answering this is that the focus
of the argument has shifted subtly over time, with each side finding a way of incorporating all the available facts into its own position. It may therefore be useful to go back to Pylyshyn's Page 171
(1983) early trenchant statement of the descriptionalist position. His view was that 'the representation corresponding to an image is more like a description than a picture'. In his theory, 'seeing the image has been replaced by a set of common and completely mechanical processes'. It 'eliminates all reference to perceptual process'. The impetus to develop such a theory of imagery stemmed from a general problem in cognitive science about how to handle the relationship between process and representation (e.g. Dreyfus, 1979; Searle, 1980; Pylyshyn, 1984). Computational models of cognitive processes that are developed within the functionalist tradition of artificial intelligence have no need for representations of the analogue kind that correspond to the thoughts and images of introspection. Thoughts and images have no place in the symbolic operations of computers, so it is ideologically attractive for those working in the 'strong AI' tradition to argue that thoughts and images have no useful place in human thinking either. This is probably the main motivation behind the attempt to develop a theory of human cognitive processes which does not give any unique functional properties to thoughts and images. One of the descriptionalists' points that would probably be widely accepted relates to storage and construction. How are images stored? There is probably no one in the current debate who would claim that images are stored in pictorial form, or that when people 'form' an image they simply pluck a fully formed visual image from their library of mental pictures. Presumably, previous images are 'retrieved' using the same kind of constructive processes as are involved when new hypothetical images are generated or existing images are manipulated or revised. Currently, most of the pictorialists (e.g. Kosslyn, 1980) would accept this. If this is conceded, then the classification of encodings has to include, in addition to imaginal and linguistic representations, a third form of encoding in which knowledge is stored when it is not in consciousness. JohnsonLaird (1983) among others has adopted such a three-fold classification of encodings into images, propositions and 'mental models', a particular form of non-conscious encoding. For him, images are 'the perceptual correlates of models from a particular point of view' (p. 165). In contrast, the pictorialists have stressed the perception-like character of imagery, and have had some success in establishing this. In his introduction to the controversy, Block (1981) suggests that it is now generally accepted that images and perceptions share many properties, though, of course there are differences too. For example, Chambers & Reisberg (1985) have shown that the properties of images can remain unspecified in a way that is not possible in percepts, and that questions of interpretation do not Page 172
arise for images in the same way as in the perception of external stimuli. Images represent what they are intended to in a way that is not the case for percepts. The key question is probably whether images have functional properties (e.g. play a causal role in information processing) as the pictorialists would claim to have demonstrated, or whether they are merely epiphenomenal (i.e. do no useful work). This is what the descriptionalists appeared to claim at the outset of the controversy, and both Shepard & Kosslyn have taken this to be a key point of disagreement. However, as Block (1981) has pointed out, the focus of disagreement has been refined. Some descriptionalists are now willing to admit that there is something associated with imagery (perhaps even uniquely associated with imagery) that has functional properties. However, they question whether it is the experience of the image that has these properties, as opposed, for example, to the neural substrate associated with imagery. What is uncomfortable about this latter
suggestion is that, unless ways are found of controlling the neural and experiential aspects of imagery independently in experimental work, it will be impossible to decide whether it is the phenomenal image or the neural substrate that has the causal properties. An alternative proposal that the descriptionalists can make is that images contain no information that is not contained in non-imaginal encodings of the same knowledge, but that they encode it in a form that facilitates processes of access and manipulation (Clark & Chase, 1972). This may concede more than some descriptionalists would wish, but it has a good deal of plausibility. It at least allows scope for the assumption of clinical researches that it can make a good deal of difference whether or not people access emotionally related thoughts and images. One striking feature of the controversy about imagery in the general cognitive science literature is that it has taken no account at all of clinical data. The prevailing assumption among those who study emotional disorders is that thoughts and images do have significant effects on the emotional states of those concerned. If this is correct, it poses a challenge to a hard-line descriptionalist position. Lang's Approach to Emotional Imagery Next, we shall consider the extensive programme of theory and research on emotional imagery carried out by Peter Lang, which takes descriptionalist theories of emotional imagery as its starting point. It is unique in Page 173
combining at least some descriptionalist assumptions with a clinical focus. Lang's theoretical views have been set out in a series of papers and chapters (Lang, 1977, 1979, 1984; Lang et al., 1980, 1983). It has also been reviewed in more detail than is possible here by Watts & Blackstock (1987), with a response by Lang (1987). Like Pylyshyn and others, Lang proposes that 'affective images are conceptualised as propositional structures, rather than as re-perceived, raw, sensory representations' (Lang, 1977, p. 863). Following Kieras (1978), Lang proposed that the associative network (see Chapter 5) is the most satisfactory way of formulating the propositional structure of representations, though this is not a necessary aspect of the descriptionalist position. The other key feature of Lang's position is his emphasis on response aspects of imagery, and this is the one that has largely guided his experimental work on emotional imagery. Originally, Lang (1977) claimed that imagery contained two elements (stimulus and response aspects), though subsequently (e.g. Lang, 1979) he proposed that imagery also contained semantic propositions (e.g. the proposition in a snake phobia that snakes are dangerous). The response aspects include perceptual, visceral, motor and verbal response systems. Central to his research has been a procedure for training subjects to include response aspects in their imagery, which he has contrasted with training for stimulus detail in imagery. Response training (but not stimulus training) results in the amplification of physiological responses when phobics imagine their particular phobic stimulus (e.g. when snake phobics imagine a snake, but not when speech phobics do so). Another important consequence of response training is said to be that it promotes greater concordance between verbal and physiological response systems, though it appears that this can be achieved by using response-laden imagery scripts without prior training in response imagery (e.g. Robinson & Reading, 1985). The clinical importance that Lang attaches to concordance between response systems follows from his early finding (Lang, Melamed & Hart, 1970) that it is correlated with a good response to desensitisation. However, there is as yet no experimental demonstration that it improves the effectiveness of desensitisation, and in fact response training seems to impede short-term emotional habituation (see Watts & Blackstock, 1987). Lang also proposes that response training makes imagery more vivid by making it more complete, vividness
being defined as the 'completeness of the evoked propositional structure' (Lang, 1977, p. 872). Lang's view is that response training facilitates accessing and retrieval of the 'deep structure emotion prototype' (Lang, 1984, p. 208), especially the Page 174
efferent aspects. In fact, a variety of different accounts can be offered to explain why response training should increase the level of emotion associated with imagery. First, response training may increase the 'completeness' of imagery more effectively than stimulus training, perhaps due to a kind of ceiling effect for stimulus aspects. Second, if subjects imagine responses, such as heart rate, being activated, this might lead directly to the activation of that response system. Third, response aspects of imagery could be more closely linked to affect in the cognitive structure than the stimulus aspects. Lang's research programme has not yet enabled us to rule out any of these alternative explanations of the effects of response training. There is also the possibility that many of the supposed facilitatory effects of response training on fear imagery should really be seen as inhibitory effects of stimulus training (Zander & McNally, 1988, Watts & Blackstock, 1987). There are also interesting questions arising within Lang's research programme on the different properties of emotional imagery in various clinical groups. For example, it seems that agoraphobics may differ from the animal phobics and speech anxious subjects who have been used in most of Lang's experiments. There seems to be a tendency to respond verbally but not physiologically to personally relevant fear scripts. There is also no relationship in agoraphobics between a general questionnaire measure of imagery ability and physiological responsiveness to scripts (Cuthbert et al., 1987; Zander & McNally, 1988). Lang sees a connection between his general descriptionalist position about the nature of imagery and his emphasis on its non-stimulus aspects. However, it is not clear that there is any necessary link between the two. There would be no dispute between descriptionalists and pictorialists about the fact that imagery can have response aspects. This fact does nothing to settle the theoretical debate about the nature and function of imagery. For example, motor responses to a phobic stimulus can be 'pictured' as readily as the phobic stimulus itself. Among the issues that any descriptionalist theory of imagery has to face is whether the activation of visual imagery has any functional value not shared by, say, verbal representations of what is presumed to be the same basic propositional structure. Though it is not entirely clear where Lang stands on this, his implicit assumption seems to be that the evocation of emotional imagery is functional. He is descriptionalist in his assumption that the value of imagery in desensitisation depends on how well the corresponding cognitive structure (the 'emotion prototype' as he calls it) is accessed, but he seems to be unorthodox among descriptionalists in at least tacitly assuming that generating response-laden experiential imagery is a valuable way of accessing the prototype. Page 175
Properties of Emotional Thoughts and Images The Relative Effectiveness of Images and Thoughts in Accessing Emotion One of the most interesting issues that has been raised about the emotion-inducing properties of images and thoughts is their relative effectiveness in accessing emotion. Sheikh & Panagiotoa (1975) among others have claimed that images have a greater capacity than language for attracting and focusing emotionally loaded associations. Klinger (1980), Horowitz (1983) and Sheikh & Jordan (1983) have made similar claims. It has also been suggested that material that is associated with painful or complex associations can be retrieved more readily in pictorial than in linguistic form. However, these claims have been based largely on anecdotal reports of clinical observations. The methodological problem is that this clinical literature may be based on an unrepresentative
minority of people for whom imagery is unusually powerful. There are probably marked individual differences in the functions served by different kinds of representations. Further, the results of the few available experimental comparisons of the emotive properties of imaginal and linguistic representations are conflicting. Reyher & Smeltzer (1968) presented subjects with words and instructed them to generate in response either a word-associate or an image. The imagery condition produced significantly greater skin conductance reactions, and also higher scores on a measure of 'primary process' responses. One of the limitations of the study is a lack of information about the volunteer sample used; there is no assurance of its representativeness. Baker & Jessup (1980) report results which suggest that emotions differ in how they are best accessed. Subjects were required either to verbalise or visualise a series of scenes. Physiological responses were higher for verbalisations. Rating scales tended to show interactions with the hedonic properties of the scenes: depressive scenes were rated as stronger and more vivid when verbalised, whereas neutral and pleasant scenes got higher ratings when visualised. Neither of these studies used self-report ratings of emotional reactions to the linguistic and pictorial representations, which makes them somewhat tangential to the hypothesis that pictorial representations are more emotive. However, it is clear that this hypothesis is one that could readily be investigated experimentally. Despite this conflicting evidence, let us assume that it is correct that pictorial representations are more emotive than linguistic ones. How could this be explained? There are two main possibilities: either we must postulate Page 176
that pictorial representations can incorporate information about emotional reactions more readily than linguistic ones; or that pictorial representations can access emotional reactions more readily than linguistic ones. The idea that imagery is well adapted for accessing emotion can be seen as parallel theoretically to the claims made by Kosslyn & Shepard that imagery gives access to spatial information and facilitates the manipulation of spatial information in a way that other representations do not. There would be no serious theoretical problems involved in conceding that imagery allows emotional information to be accessed more readily than do other representations. However, it would be more problematic to concede that imagery contained emotional information that was not, and could not be, contained in other forms of representation. For example, this would raise problems about how the information was stored when images were not in consciousness. It seems unnecessary to postulate that imagery contains unique emotional information in order to accommodate any special emotional properties it may have. There is another claim that has been made in the clinical literature that would have more farreaching theoretical implications if it were substantiated. This is that imagery is uniquely effective in recalling experiences from the pre-verbal stage of childhood (Kepecs, 1954). If correct, this would imply that it was easier to retrieve information in the same representational form as it was originally experienced. A descriptionalist theory of imagery that (1) regards imagery as an epiphenomenon of more basic cognitive processes, and (2) assumes that all non-experiential encodings are of the same general form, could not readily accommodate such a finding. Though no theoretical problem is raised by the idea that imagery facilitates the accessing of emotional information, there is a sense in which this may not be the appropriate way of handling the supposed emotive properties of imagery. The problem is that emotional information is not the same as emotion (Russell, 1987). How representations access emotion itself is less easily explained than how they access emotional knowledge. A parallel problem has arisen in associative network theories of relationships between mood and memory. Bower & Cohen (1982) were led to make a distinction in their associative network model between nodes for emotional feeling, for emotion
words, and for emotional concepts. Though Bower & Cohen include feeling nodes in their network, it is not clear that this is appropriate, i.e. that feelings are elicited by connections in the network in ways that are essentially similar to those by which knowledge is accessed (see Chapters 6 and 11). Page 177
Linguistic representations lend themselves to classification more readily than do images. The simplest level is the single word. A number of information processing tasks involving emotive words (e.g. colour naming) have been discussed in earlier chapters of this book. The success of these tasks is evidence that individual words can be potent emotional stimuli. As yet we know little about what kinds of word are best able to produce such effects, though Watts et al. (1986b) speculated that colour-naming effects may be strongest with concrete, highly imageable words. This would parallel word variables that are relevant to the strength of learning (e.g. Christian et al., 1978). The representational effectiveness of single words is also attested by the fact that therapeutic benefit can be derived from semantic desensitisation using individual words (Sergeant, 1965). There is an analogy between this procedure and the laboratory phenomenon of semantic satiation (Di Caprio, 1970) but it remains to be explored in detail. Though the effects that can be obtained with single words are striking, it is probably correct that the important linguistic determinants of specific behaviours (including emotional reactions) are sentences/facts rather than words/things (Rozeboom, 1972). Distinctions can be made between sentences that (a) describe particular stimuli or events, (b) contain general semantic knowledge and (c) provide guidance for actions. This classification roughly parallels that between episodic, semantic and procedural memory (Tulving, 1983). It is probably in the semantic and procedural domains that language has the advantage, as it is better able to represent general statements than images. In the clinical domain, negative cognitions in depression (e.g. Beck's well-known triad of negative views about the self, the world and the future) seem to represent general semantic propositions rather than specific episodes. An interesting issue here is the relationship between semantic and episodic representations. Either type of representation can be biased in emotional disorders, and clinical observations suggest that dissociation in either direction is possible. Thus a depressed patient might have a preponderance of negative specific memories of parenting, but have a general proposition in semantic memory that parenting was good. Equally, as we described in Chapter 6, there are patients who have negative general statements in semantic memory, but when pressed for detailed episodes have surprising difficulty in substantiating what they are based on. Perhaps the potentially opposing effects of mood-congruent retrieval (Teasdale, 1983b) and repression (Erdelyi & Goldberg, 1979) are partly responsible for these apparently assorted clinical phenomena.
Page 178
Though both imagery and language can be used to guide actions, the advantage of language seems to lie in the fact that it can more readily be used concurrently with action. The effects of imagery in rehearsal for action can be powerful, a fact that has been best demonstrated in sports psychology (Suinn, 1983). However, it is harder to use imagery than language concurrently with an activity; this gives imagery less flexibility than inner speech. As discussed in Chapter 3, anxiety and depression are associated with an excess of task-irrelevant worry and rumination. However, there have been few studies of whether they are also associated with a poverty of the kind of inner speech that might guide coping behaviour. It is relevant that Watts, Sharrock & Trezise (1986) found that spider phobics' descriptions of how to remove a spider from a bath were significantly less elaborated than those of non-phobics. There may be a general impoverishment of coping inner
speech in patients with emotional disorders. There has also been surprisingly little comparative evaluation of the effects of treatments based on linguistic and pictorial representations. Imagery enthusiasts have made claims for the therapeutic value of imagery-based techniques (e.g. Sheikh & Jordan, 1983). On the opposite side, it has been suggested that linguistic treatments involving representations have more widely generalised effects. Meichenbaum (1977, p. 114), in a seminal study, explored this issue by examining treatment effects in subjects who had two discrete phobias (snakes and rats). The imagery treatment was desensitisation and the verbal treatment was a stress inoculation technique focusing on anxious 'self-talk!. It seems that stress inoculation affected both phobias, whereas desensitisation affected only the phobia to which it was applied, Methodological questions can be raised about this study as the two treatments may have differed in their relative emphasis on stimulus and response processes, as well as in their use of verbal or visual representations, thus making interpretation difficult. A stress management technique using rehearsal in visual imagery would have been a more appropriate comparison with verbally based stress inoculation to answer the question being raised here of the relative generalisability of language- and imagery-based treatments. Nevertheless, it is plausible that language produces better generalisation of treatment effects than imagery. Characteristics of Emotional Imagery Clinical researchers have shown considerable interest in the vividness of emotional imagery. This no doubt partly reflects the long history of general research on imagery vividness (Marks, 1972), but also follows from a Page 179
more specific interest in the hypothesis that response to imagery-based treatments such as desensitisation is dependent on the vividness of the subject's imagery. Several studies directed at this hypothesis have used standard questionnaire measures of imagery vividness, though these scales have generally been unsuccessful in predicting response to desensitisation (e.g. McLemore, 1972; Dyckman & Cowan, 1978). Awareness of the limitations of these scales has led to interest in developing performance-based measures of imagery vividness that might relate better to desensitisation outcome (Rimm & Bottrell, 1969; Danaher & Thoresen, 1972; Rehm, 1973; McLemore, 1976; Hiscock, 1978). These studies have generally obtained weak and inconsistent relationships between objective and subjective measures of vividness. Apparently, no one has yet looked at whether these objective measures actually predict response to desensitisation, the purpose for which they were purportedly developed. It is not clear why performance tests of visuo-spatial imagery would be any more closely related to desensitisation than are general imagery measures. The basic mistake seems to lie in seeking to infer the vividness of emotional imagery from general imagery measures (whether questionnaire or performance) rather than looking specifically at the vividness of emotional imagery. Dyckman & Cowan (1978) obtained ratings of the vividness of desensitisation scenes and found that this was related to desensitisation outcome. This kind of direct examination of emotional imagery seems the appropriate strategy. Fuller reviews of this literature on vividness of imagery and desensitisation can be found in Strosahl & Ascough (1981) & Anderson (1981). Perhaps the most basic question about the vividness of emotional imagery is the nature of the relationship between such vividness and the degree of emotion associated with it. This may well be complex. Most image theorists have accepted uncritically that vividness is a unitary dimension, though Klinger (1978) has questioned this and suggested (on the basis of factor analyses of imagery ratings) that there are two main components which he designated 'sensory saturation' (i.e. richness and impact) and 'clarity' (i.e. precision). Watts, Sharrock & Trezise (1986), working with
the related components of 'awareness' and 'detail' in phobic imagery, found that awareness was positively related to phobic anxiety, though detail was not. Indeed there may be a negative relationship between anxiety and some aspects of vividness. Consistent with this, Mathews (1971) suggested that relaxation serves to increase the vividness of imagery in desensitisation, and Borkovec & Sides (1979) found that vividness increased during desensitisation. On the other hand, Marzillier, Carroll & Newland (1979) found no difference between phobics and non-phobics in the vividness of imagery. In a very different clinical field, Page 180
Parkes (1972) observed a tendency in bereaved people for the clarity of imagery for the deceased person to improve as preoccupation decreased. It has been suggested (Watts, Sharrock & Trezise, 1986) that the detail and precision of phobic imagery is particularly important in the effectiveness of imagery-based treatments. Watts (1974) attempted to manipulate the detail of imagery in desensitisation by either describing each scene in full each time it was imagined, or simply instructing the subject on the second and subsequent presentations to 'imagine the same scene again'. Detailed presentations resulted in more long-term (between-session) anxiety decrement. This was interpreted in terms of a dual-process habituation theory of desensitisation in which one of the component processes is relatively stimulus specific and depends on a clear cognitive representation of the stimulus (Watts, 1979). There have been no specific investigations of whether the sensory awareness component is related to treatment effectiveness, though there is no clear theoretical reason for predicting that there should be a relationship. Even for clarity of imagery, the theoretical question remains of whether this has a direct causal effect on the desensitisation process, or whether the causal properties are to be located in an underlying non-conscious cognitive encoding of which experiential detail is only an epiphenomenon. However, even if this latter is the case, increasing experiential detail may be an effective strategy for bringing about the desired changes in the non-conscious encodings. If images are generated from underlying cognitive representations they may show parallel characteristics. The lack of precision in emotional imagery may result from relatively undifferentiated cognitive representations. Landau (1980) found that dog phobics are able to list fewer instances of dogs than non-phobics, suggesting a less differentiated conceptual structure. However, this was also true of mammal instances, leaving it unclear how general were the cognitive differences between groups. As discussed in Chapter 3, obsessionals show a general tendency to use narrow categories in sorting tasks, but especially so with 'feared' material such as contamination or making serious mistakes (Persons & Foa, 1984). Similarly, studies of relationships between personal constructs have found that they are unusually highly correlated in patients with anorexia (Button, 1983a), depression (Sheehan, 1981), spider phobia (Watts & Sharrock, 1985b) and mixed neurotic problems (Ryle & Breen, 1972; Winter, 1983). MakhloufNorris, Jones & Norris (1970) also found this for obsessionals, but Millar (1980) failed to replicate this finding. Some of these studies have used constructs related to the disorders, others have used neutral constructs. Relationships between constructs are Page 181
especially strong when constructs related to the disorder are used (Winter, 1983). For a more detailed review of this literature see Button (1983b) Though sufficient positive findings have emerged to indicate some kind of association between emotional disorders and conceptual organisation, several problems of interpretation remain. It is not clear how far the findings are specific to emotional materials. It is also not clear whether the different methodologies (listing, sorting and repertory grid) provide indices of exactly the same cognitive phenomena. Further, any assumptions about relationships between these structural
phenomena and the properties of conscious representations such as images must be tentative, as they have not been studied directly. Images and Thoughts in Emotional Disorders Worry, Intrusive Thoughts and Ruminations There are a great many emotional conditions and disorders that are characterised by intrusive thoughts, and the analysis of these has been a major focus or research in recent years. A particular focus of interest has been the investigation of worry. One focus of the first phase of research on worry has been the development of questionnaires to measure worry and the identification of domains of worry, see, for example, Barlow (1988), Eysenck & van Berkum (1992), Meyer et al. (1990) and Tallis, Eysenck & Mathews (1992). However, in the present context, theoretical questions about the role and function of worry are more important. Borkovec has played a key role in the development of theoretically motivated research on worry. The clearest statement of his theoretical position is that of Borkovec, Shadick & Hopkins (1991), though this builds on an earlier formulation by Borkovec, Ketzsger & Pruzinsky (1986). One key idea that Borkovec has advanced is that worry inhibits emotional imagery; indeed he suggests that the reason why people worry is in part that it protects them against the emotional imagery that they would otherwise experience. This obviously assumes that worry, despite its apparent aversiveness, is in some ways less distressing than emotional imagery. The essential point here is that worry leads to less physiological activation than emotional imagery, and may even reduce it. What is the evidence for these ideas? Borkovec & Inz (1990) reported evidence that people experienced less imagery when they were worrying Page 182
than when they were in a state of relaxation, and this is consistent with the theory. However, it does not show that worry has any special imagery suppressant properties. Indeed, East & Watts (1994) found that thinking about any relatively neutral topic suppressed imagery just as well, or even slightly better. The general effect is probably just that any verbal-linguistic mentation suppresses imagery; there is nothing specific to worry here. Worry seems generally not to be associated with much physiological activation. Indeed, it has been clear in the test-anxiety literature for some time that worry and physiological arousal are relatively distinct components (Deffenbacher, 1980) and, in a more experimental approach, Borkovec et al. (1983) have confirmed that there is little more activation in worry than in rest periods, though York et al. (1987) found evidence of worry leading to physiological arousal. The evidence reviewed earlier in this chapterthat imagery leads to more accessing of emotion than comparable linguistic mentationis, of course, consistent with Borkovec's position. Despite the interest of Borkovec's somewhat counterintuitive ideas about the avoidance functions of worry, it does not claim to offer a complete account of the cognitive functions of worry. A theory focusing on some of these other aspects has been sketched out by Eysenck (1992) and Tallis, Eysenck & Mathews (1991). They suggest that worry serves the 'alarm' and 'prompt' functions of introducing threats into conscious awareness, enabling people to prepare for impending danger. Whatever the function of worry, there is no doubt that it is experienced as being aversive. The indications are that the uncontrollability of worry is the key factor heresee a review by Borkovec, Shadik & Hopkins (1991). For example, uncontrollability correlates with worry more strongly than does the subjective tension associated with a thought (Clark, 1986). Also, though thoughts may access emotion better than images, intrusive thoughts are more difficult to dismiss than
intrusive images (Parkinson & Rachman, 1981). Intrusive thoughts are a feature of a wide range of emotional disorders, and they are exacerbated by either anxiety or sadness. For example, Sutherland, Newman & Rachman (1982) tested and confirmed the hypothesis that depressed mood (experimentally induced in normal subjects) makes intrusive thoughts difficult to remove. There are also intrusive thoughts associated with anxiety that tend to be more emotionally intense than depressive ones (Clark & de Silva, 1985). However, among clinical disorders, it is generalised anxiety disorder (GAD) which bears a particularly close relationship to worry. GAD patients not only have a Page 183
high level of worry, but their worries are less controllable, less realistic and lend themselves less easily to corrective action than those of other people (Craske et al., 1989). The recent work of Nolen-Hoeksema has made an important contribution to understanding the role that rumination plays in depression. Some people in dysphoric moods are inclined to distract themselves, while others are inclined to ruminate. It is now clear that ruminators are likely to experience more prolonged periods of depressed mood (Nolen-Hoeksema, 1991), and the greater tendency to a ruminative style in women may contribute to their higher rates of depression (Nolen-Hoeksema, Morrow & Fredrickson, 1993). Rumination makes people less likely to engage in activities that could lift their mood. One reason why people may engage in rumination is that they believe it may increase their insight into their problems (Lyubomirsky & NolenHoeksema, 1993). In the final chapter, we consider how Johnson's multiple entry memory (MEM) model might account for such ruminative processes. We suggest that these represent nonmemorial elaboration and use subprocesses normally associated with creative problem solving. Emotional Processing We shall turn now to the role of thoughts and images in 'working through' the stressful impact of life-events such as bereavement. In this process, conscious thoughts and images occur frequently and involuntarily. The important theoretical question is how, if at all, this contributes to the process of 'working through' and to such changes in underlying knowledge structures as may accompany this. Freud's observations regarding repetition compulsion following trauma have been influential and will serve as a starting point. Horowitz (1983, p.131) summarises Freud's position as follows: A harrowing or frightening experience exceeded a person's state of preparedness and/or capacity to master the resulting stimulations and affects. A temporary protective mechanism shunted the experience out of awareness where it resided as a kind of undigested foreign body; the memory traces were still extremely vivid and the affects were still of potentially overwhelming intensity. At some later date, the 'repetition compulsion' asserted itselfthe person relived the experience repeatedly until it was mastereduntil associated feelings such as helplessness diminished. Until such mastery of affects, recall of the experience tended to evoke very vivid images. With mastery the memory traces were processed for storage in the usual way: they were stripped of sensory intensity and related to various schemata and concepts. Page 184
Such clinical observations have been supplemented by experimental studies. Horowitz has reported a series of studies comparing the effects of stressful and neutral films on subsequent intrusive representations and found that stressful films were followed both by more intrusive thoughts and more intrusive imagery (e.g. Horowitz, 1975). His interpretation of these findings hinges on the notion of 'completion' of information processing. Until this point is reached, material remains in what Horowitz calls 'active memory', and shows a tendency to surface repeatedly in conscious representations.
It is not clear how well this formulation can explain why stressful material is apparently more likely to surface in consciousness than other material in active memory. Another problem is whether the critical notion of completeness can be defined sufficiently independently for it to be of scientific value. If such an independent measure were available, it might be found that stressful material would be more likely to intrude than nonstressful materials that had both been processed only to a similarly 'incomplete' extent. Another problem is that it appears from Horowitz's experimental report (though his report is not wholly clear on this point) that stressful films led to increased intrusive representations of material that was unrelated to the film as well as of related material, though it is not clear how Horowitz's theory can predict this. Rachman (1980) has offered an integrative review of what he terms' emotional processing', i.e. the processing of stressful events. His position is similar to that of Horowitz at many points. Like Horowitz, he has a concept of incomplete or unsatisfactory emotional processing. Among his comprehensive list of signs of unsatisfactory processing are unpleasant intrusive thoughts. Intrusive imagery is not included, but would be compatible with Rachman's general position. Rachman's analysis concentrates on the factors that promote or impede emotional processing. Of particular interest is his proposal that calm rehearsal promotes emotional processing, whereas agitated rehearsal impedes it. This is an advance in specificity on Horowitz's position, which includes no hypothesis about different types of cognitive repetition. One of the puzzling features of memory phenomena in post-traumatic stress disorder is the range and variety of memory phenomena (Horowitz, 1992). Specifically, memory for traumatic events can be either enhanced or inhibited (Loftus & Kaufman, 1992). Highly detailed, intrusive memories of traumatic events are quite common. However, there are also many clinical reports of amnesic phenomena in post-traumatic stress disorder, and experimental reports of poor or overgeneral recall (McNally et al., 1994). Also, there are cases where the affect associated with the Page 185
memory is intense, and other cases where the affect appears to be inhibited. Memory phenomena in bereavement, which in some ways is like post-traumatic stress disorder, are equally diverse (Stroebe, Stroebe & Hausson, 1993). One of the current theoretical challenges in this area is to explain the range of memory phenomena associated with trauma. Intrusive cognitions can be found in a wide range of disorders apart from those associated with trauma, for example in the worry phenomena that characterises GAD. However, they may play a different functional role in trauma. As we have seen, the available evidence suggests that distraction is a more helpful strategy than rumination in depression, though distraction does not seem to be an effective way of coming to terms with trauma (Tait & Silver, 1989). Intrusive memories seem to be potentially more functional than in other cases such as worry in GAD. However, much seems to depend on the kind of accessing of traumatic memories that takes place. There is relevant evidence arising from the application to emotional processing of Borkovec's theoretical approach to worry to a situation in which subjects watched a stressful film. Borkovec has argued that worry inhibits negative emotion in the short term, but its long-term effects might be very different. Butler, Wells & Dewick (in press) found, as Borkovec would predict, that worry immediately after a stressful film lowered subjective anxiety in the short term compared to imagery or 'settle-down' conditions. However, three days later, there were more intrusive thoughts in those who had worried than in other groups of subjects. It thus seems that worry, after a stressful event that needs processing, may reduce activation in the short term, but at the longterm cost of interfering with emotional processing. The imagery condition apparently led to more affect in the short term, but there was evidence of better emotional processing in terms of reduced long-term intrusions. This suggests that individual differences in ability to tolerate affective
disturbance in the immediate aftermath of a stressor may have a disproportionate influence on the long-term tendency for worry about the stressor to intrude. There is also encouraging evidence for the health benefits that can be derived from working though traumatic memories, particularly from Pennebaker's programme of experimental research on getting people to write about troublesome memories (Harber & Pennebaker, 1992). In general, the indications are that (a) almost any emotional concerns can generate intrusive preoccupations, not just those relating to a previous stressful event; but that (b) the deliberate recollection of traumatic memories has a functional value in terms of creating the possibility of transforming them and rendering them less intrusive. Page 186
Even if this latter point is accepted, the question arises of why the working through cannot be carried on without any conscious representation. A specific possibility that might be considered (and could be tested experimentally) is that the repeated retrieval of conscious representations from long-term memory contributes importantly to their transformation. This may be more helpful to the completion of processing than either sustained conscious processing or sustained nonconscious processing. The Role of Representations in Psychological Treatments All psychological treatments make use of thoughts or images. Probably the most widely used treatment employing visual imagery and certainly the most thoroughly investigated is the technique of imaginal desensitisation developed by Wolpe (1958). A wide variety of other techniques making similar use of imagery have also been developed, including covert sensitisation, covert reinforcement, covert negative reinforcement, covert extinction, covert modelling and covert response cost (Upper & Cantela, 1979). Imagery has also been the basis of a variety of psychotherapeutic techniques (Singer, 1974). There is an equivalent family of techniques based on linguistic representations. Verbalisations are also, of course, the foundation of most psychotherapeutic techniques, including psychoanalysis. One of the first behavioural techniques to make use of verbal material was Homme's (1965) coverant control method. This was followed by Meichenbaum (1977) who developed a treatment for anxiety based on modifying 'self-talk', and Beck, who developed a related treatment for depression (Beck et al., 1979) and anxiety (Beck, Emery & Greenberg, 1985) based on modifying negative thoughts and assumptions. Indeed, there is a burgeoning family of clinical techniques making use of verbalisations. With all treatments that use symbolic representations, there is a question to be answered about how their effects transfer to situations in the real world. Clinical theories about how the effects of imaginal treatments such as desensitisation 'transfer' to external stimuli have been almost embarrassingly naive. The simplest view would accept the extreme pictorialist assumption that phobic imagery is an analogue of an external stimulus. It is simply assumed that modified reactions transfer from the image to the stimulus, rather in the same way as conditioned reactions to one stimulus transfers to other related stimuli along a stimulus generalisation gradient. The poverty of theorising about how transfer from representations to external stimuli occurs is so serious that few people have explicitly stated the assumptions involved. However, Cantela & Baron (1977) set out what Page 187
they regard as three key assumptions: (a) there is 'homogeneity' between overt and covert behaviours (i.e. that they share similar properties); (b) there is 'interaction' between overt and covert events such that one can influence the other; and (c) both are governed by the same laws of learning. Evidence can be adduced in partial support of these assumptions, but they remain
problematic (see Strosahl & Ascough, 1981). For example, it is true that the physiological reactions to imagery are similar to those to external stimuli, and that conditioning procedures have been successfully adapted to imagery-based exercises with acceptable therapeutic effects. However, this does not explain how the effects of imagery treatments on external stimuli come about. Another, somewhat more sophisticated suggestion that can be made is that imagery and perception involve common psychological processes, or rather that imagery represents the anticipatory phase of visual perception (e.g. Neisser, 1976). Modifying processes involving imagery would therefore require the modification of the section of the pathway through which external stimuli are processed. Imagery-based treatments might be able to modify the processing of external stimuli by operating on the common processes involved in both. A variant of this view would assume that, even though imagery processes themselves may not have a role in mediating reactions to external stimuli (Franks, 1974, pp. 23740), they may influence the longterm structures that modulate reactions to external stimuli. However, it is not clear whether or how imagery can affect subsequent stimulus processing in this way. It may be that imagery is a form of conscious representation that is derived from long-term memory, but is not itself involved in determining how external stimuli are processed. Another possibility is that the apparent transfer from imaginal exercises to real-life situations is mediated by expectations (e.g. Powell & Watts, 1973). A key question for such a theory to face is how imaginal exercises affect expectations about real-life situations. One possibility is that there is a simple failure of 'reality monitoring' (see Chapter 3). People often fail to distinguish in memory between what they have seen externally and what they have generated internally (Johnson, 1985). However, failures of memory for whether events actually occurred are altogether different from a failure to distinguish between imagination and reality at the time of occurrence. A repeated failure to do the latter would represent a psychotic state. There are other, perhaps more plausible, ways in which imaginal expectations might change expectations regarding real events. For example, for phobics to imagine a feared situation with only minimal anxiety may lead Page 188
them to believe that they could cope with the real situation. Beliefs may thus play an important part in mediating the effects of imaginal exercises. However, clinical observations suggest that beliefs alone may not be sufficient. Sometimes a phobic may believe he can do something but find that in practice his level of anxiety prohibits it. A fundamental issue for all theories of therapeutic transfer is the distinction between representations of the real world and representations of possible or imaginary worlds (e.g. Johnson-Laird, 1983a, p. 423). Such a distinction is crucial in psychological theory for a variety of purposes. Perceptual input seems to contact the models of the real world in a way that is different from whatever subsidiary contact it has with models of hypothetical worlds. Also, models of the real world have behavioural consequences in a way that models of hypothetical worlds do not. Consequently, the therapeutic benefits of treatments based on images or linguistic representations would be very limited if all they achieved was to give people a model of a hypothetical but unreal world (for example, a hypothetical world in which phobic objects could be approached without fear, or in which a depressed person was loved and esteemed by others). There are two possible alternative consequences of a phobic person entertaining representations of a world in which the phobic object can be approached without fear. One would be the creation of an additional model of a hypothetical world in which this was the case; the other would be the revision of the existing (phobic) model of the real world. The clinical results would be very different.
The question of whether a representation relates to real or imaginal worlds cannot be settled on the basis of whether or not it is true of the real world. Many depressed people have a distorted perception of the world which, though inaccurate, is clearly intended to be a representation of reality; they are not likely to confuse this and a model of a bleak but imaginary world. Though there are a number of ways in which models of real and imaginal worlds tend to differ, it is not clear that any provide a wholly reliable way of making the distinction. For example, it has been suggested in the literature on reality monitoring Johnson, 1985) that external events are associated with greater contextual information, though this is probably not invariably correct. It might not be true for an author who spent nearly all his time creating fictional characters. Johnson-Laird (1983, p. 423) suggests that people know the status of a model 'because they can remember what led to the construction of the model in the first place'. However, the fact that reality monitoring is very imperfect shows that this criterion is also likely to fail. Perhaps the essential property of a model of the real world is simply that it is intended to refer to the real world. There is a great deal of difference Page 189
between a phobic patient imagining an explicitly hypothetical world in which he had no phobic anxiety, and imagining himself functioning in the real world without phobic anxiety. Imaginary exercises divide according to whether they are intended to refer to hypothetical worlds or to the real world. It can be proposed that only exercises involving models intended to refer to the real world will affect subsequent functioning in it. Whether or not people take a model as being intended to refer to reality is not necessarily an allor-nothing matter. People are capable of giving varying degrees of credence to alternatives to their current model of reality. For example, someone might entertain conflicting judgements about the risks associated with a particular activity and be genuinely uncertain which was correct. The amount of credence given to the alternatives might vary over time depending on circumstances. This prompts a question about the impact of new information. If new information suggests that an existing model of reality is incorrect, one possible consequence would be to create a new alternative model and discard the old one. Another would be to retain both the old and the new models, but to assign to the new one a higher level of credence than the old one. Some clinicians have the impression that the latter is the more common situation. In therapy, old models seem to be retained alongside new models, but with less credence attached to them, rather than being discarded. As with most of the issues raised in this chapter, a great deal of additional empirical research is required before clear conclusions can be reached. However, there is one category of verbal representation, judgements of frequency and probability, which has been extensively researched, and it is to this literature that we now turn.
Page 190
Chapter 8 Judgement When the patient [a flying phobic] was not planning a flight in the predictable future, he would feel that the chances of the plane's crashing were one in a hundred thousand or one in a million. As soon as he decided to make a trip by air, his estimated probabilities of a crash would jump. As the time for the flight approached, the likelihood increased progressively. By the time the airplane took off, he would figure the chances as 50-50. If the trip was bumpy, the odds would switch over to 100 to 1 in favour of a crash. (Beck, 1976, p. 164)
Why should the anticipation of a flight increase the subjective probability of a crash? Are emotional states associated with systematic distortions in judgemental processes? The above quotation strongly suggests that they are. However, an issue arises here about whether the judgemental processes for emotional outcomes differ in any fundamental way from those involved in processing neutral material. It is by no means clear that biases in judgement are confined to people with emotional disorders, or even to those in highly emotional states. Systematic biases can be demonstrated in quite normal individuals whenever they make judgements concerning any uncertain event. To establish the extent that emotional states or disorders are associated with unusual degrees of judgemental bias requires us to examine our state of knowledge concerning errors or bias in the normal population. We therefore begin this chapter by discussing the evidence that even nonemotionally disturbed individuals are prone to errors associated with the use of simplifying heuristics, such as availability or representativeness. Next, we discuss evidence that emotional states do indeed affect judgements of past performance, future risk, and the degree of covariation between events. Other evidence will be cited to show that emotion produces similar effects when ambiguous events are interpreted, such that the perceived meaning is congruent with emotional state. Finally, the implication of this research for clinical practice is discussed, both in relation to errors of clinical judgement and the psychological treatment of emotional disorder. Page 191
At the most general level, errors in judgement are associated with the overuse of intuitive heuristic 'rules-of-thumb', and a corresponding failure to use logical or statistical decision methods. Kahneman, Slovic & Tversky (1982) discuss several commonly used heuristic devices, of which two of the most relevant for our purposes are availability and representativeness. Heuristics The Availability and Simulation Heuristics In using the availability heuristic people are influenced by the relative availability, or accessibility from memory, of events related to the judgement they are making. When judging the probability of an uncertain future event, for example, it may often be useful to base one's estimation on the ease with which similar events can be accessed from memory. Generally speaking, the ease of such access will be correlated with past event frequency, so that the use of this heuristic often leads to quite accurate judgements. However, there are many factors associated with accessibility other than objective frequency, including vividness or salience, which may render the use of the availability heuristic misleading. In one experiment subjects
were presented with a list of male and female names, in which either the males or the females were famous personalities, while the other sex names were those of unknown people. When subjects were later asked to estimate the number of men and women in the list, those who had seen the names of famous women but unknown men, tended to overestimate the number of women, while those whose list contained famous men and unknown women overestimated the number of men (Tversky & Kahneman, 1974). The simulation heuristic extends the concept of availability to those domains where an individual has little or no past experience to draw on. It refers to the tendency to base judgements on the ease with which a scenario of a situation may be simulated. Such 'ease of generation' may be assessed by examining the speed with which an individual can generate reasons why a particular event (such as being burgled) might happen, compared with the speed with which the person generates reasons why it would not happen. Examples of using this research paradigm will be given later, when describing mood and subjective risk (p. 201). Apparently, the more salient and memorable the exemplars within a category, the greater the estimate of their frequency. It also seems possible that the availability heuristic is involved when people overestimate the likelihood of unpleasant and frightening events occurring to them. For example, if we have just read an extremely lurid account in the newspaper about a victim of violent crime, the increased availability that results Page 192
may lead us to overestimate the true probability of being involved in such a crime ourselves. In systematic studies of this effect (Lichtenstein et al., 1978, Johnson & Tversky, 1983) the subjective probability of death from a number of causes was shown to be related to disproportionate exposure to lethal events via media descriptions, or to their memorability and imaginability. Tversky & Kahneman are not entirely explicit about the mechanisms thought to underlie availability; for example, whether availability refers to differential ease of generation in retrieval, and whether subjects actually engage in such retrieval at the time of making a judgement. Some evidence indicates the number of items which can be recalled when required does correlate with frequency estimates (Williams & Durso, 1986) consistent with the possibility that subjects actually generate examples in at least some tasks. However, since subjects can sometimes give judgements very rapidly, it would seem more likely that frequency information, sometimes in biased form, may often be encoded at the time of initial presentation. The Representativeness Heuristic This representativeness heuristic is said to be used whenever people base their judgements on the extent to which a specific event is seen as typical of a larger group of events. The heuristic can be particularly misleading when we ignore base-rate information as a consequence of its use. For example, suppose one has the information that an individual is relatively shy and withdrawn, is very fond of reading, and is good at keeping things neat and orderly. In deciding whether this individual is more likely to be a librarian than (say) a construction worker, many people would estimate the former as more likely, on the basis that the personality sketch given resembles their prototype of a representative librarian. However, use of representativeness in making this judgement has led to overlooking completely the fact that the frequency of librarians in the population (base rate) is very many times lower than that of construction workers, and the latter group must also contain a proportion of individuals who also fit the description. It is clearly desirable that clinicians and therapists should be aware of the biases and heuristics which their emotionally disturbed clients may use when judging situations as threatening or depressing. It is also worth remembering, as was indicated at the beginning of this chapter, that judgemental bias is not the prerogative of emotionally disturbed populations. Since we all use
heuristics in making judgements, clinicians are by Page 193
no means immune from similar errors. To illustrate this, suppose a particular disease has a prevalence rate of 1 in 1000. A screening test for this disease is 95% accurate, that is, the false positive rate is 5%. If one person obtained a positive result on the test, in the absence of any other symptoms or indications, what is the probability that they actually have the disease? Would your guess be 90, or 95%? If so, recall that in a random sample of 1000 people, only one will actually have the disease; while the 999 that do not have it will produce a 5% false positive rate. Since 5% of 999 is approximately 50, the chances are therefore only 1 in 50 that an isolated positive test indicates true presence of the disease! This fairly elementary application of base rates was sufficient to confuse the majority of students and staff at Harvard Medical School, many of whom gave the erroneous answer of 95% (Casscells, Schoenberger & Grayboys, 1978). Confirmation and Covariation Bias Problem-solving Tasks People often seem to cling to their initial beliefs, even when the evidence on which they were originally based has been totally discredited (Ross, Lepper & Hubbard, 1975). For example, subjects given the task of judging the validity of real versus faked suicide notes persisted in believing themselves to be accurate in these judgements, even after they had been informed that earlier confirmatory feedback given by the experimenter had been totally fabricated. It would appear that once formed, a hypothesis can lead people to focus on reasons why it might be true, and to disregard evidence of its falsity. Research on problem solving has also demonstrated that normal individuals show a persistent preference for gathering information which confirms their beliefs, rather than challenging them. When invited to test the validity of a rule, we typically proceed to do so by generating positive instances of that rule, and neglect to test negative instances (see Wason, 1971). For example, suppose that you are given the numbers 2, 4 and 6 and told that this sequence conforms to a specific rule. Your task is to establish the nature of this rule by generating more sequences, and receive feedback about whether each sample conforms to the rule or not. Only when you are confident that you know the rule should you announce your decision. In fact, the rule used by Wason was that of any series of ascending numbers; but subjects typically generated a few other Page 194
ascending series of three consecutive even numbers, and on having them confirmed as valid examples, then announced this as their answer. Judgements about People Confirmation biases are by no means restricted to abstract problems, but extend to judgements about other people. Consider the case of an individual who is told that a new colleague is much more friendly than appearances suggest. To test out this idea, the individual might strike up a conversation and probe for examples of friendliness, or for examples of unfriendliness, or both. In a direct test of the preferred strategy in such circumstances, subjects who were told to determine if another person was introverted (or extroverted) typically chose confirmatory rather than disconfirmatory questions from a prepared list (Snyder & Swann, 1978). Interestingly, blind judges who listened to the target person's answers to these questions and then judged their personality, tended to be swayed in the same direction, suggesting that confirmatory testing
elicits social behaviour consistent with the hypothesis being tested. Such results strongly suggest that, once established, judgemental biases about people are very hard to change. Judgements of Covariation The ability of animals and humans to detect that two events tend to co-occur is critically important to understanding the world we live in, and to anticipating what may happen in the future. We would be unlikely to survive long if we did not rapidly learn that the occurrence of certain stimuli signals a forthcoming aversive event, while others are associated with positive outcomes. After reviewing research on situations ranging from simple conditioning to making complex judgements about people, Alloy & Tabachnick (1984) concluded that current situational information frequently interacts with prior expectations to determine the covariation that is perceived. That is, a covariation consistent with prior expectations is typically perceived more readily than one which is inconsistent with expectation. In general, this interaction has adaptive consequences, because prior expectations are typically based on previous experience with similar situations, allowing more accurate assessment of appropriate responses. Sometimes, however, expectations may be based on sources that do not accurately match the current situation, and thus interfere with perception of the veridical covariation. Page 195
In a classic series of studies, Chapman (1967) and Chapman & Chapman (1967) showed that either semantic associations between stimuli, or their joint distinctiveness, can lead to the illusion that they co-occur, even when no such covariation exists. If subjects are shown lists of word-pairs, in which some pairings are semantically associated (e.g. table-chair) and others are not (e.g. table-grass, sheep-chair), then the frequency of the related pairings is seriously over estimated. Similarly, if most of the words are short and only a few are long, subjects overestimate how frequently the long words co-occur. Further studies showed how similar illusions may influence clinical judgements. Naive observers were given random combinations of drawings (of a man) and psychological symptoms (e.g. excessive suspiciousness) that were both attributed to the same patient. These subjects were subsequently asked to judge how frequently each type of symptom was associated with characteristic features in the drawings. Systematic relationships were consistently reported, despite the lack of any veridical relationship in the combinations they had seen (e.g. the symptom of excessive suspiciousness was typically linked with drawings of men with large eyes). Strikingly, these perceived relationships closely resembled those reported by clinicians who used the Draw-a-Person Test diagnostically. Finally, Chapman & Chapman found that, even without any exposure to sample drawings, the prior expectations of naive subjects about what covariations might occur, were very similar to those reported by others given prolonged exposure and offered a reward for being accurate. They conclude that perceived correlations among actually non-associated events may arise entirely on the basis of prior expectations, and can be remarkably impervious to corrective experience. Even a very cursory consideration of these findings suggests that illusory correlations have important (and potentially alarming) implications for social stereotyping and prejudice, as well as many of the judgements made routinely by clinical psychologists. Furthermore, as will be seen later, similar processes may operate so as to maintain biased judgements in emotional disorders. Mood State and Evaluative Judgements It seems fairly clear from the foregoing review of intuitive judgements, that judgemental heuristics are commonly used by normal individuals, and while useful in many circumstances, may sometimes lead to errors and the preservation of misleading beliefs. However, the cognitive
processes involved in such biases have not been clearly established. Candidates that have been proposed in the case of covariation judgements (see Alloy & Page 196
Tabachnick, 1984) include: (1) the initial selection and encoding of an instance or episode as relevant or not, (2) the interpretation of the selected instances as confirming or disconfirming, (3) the storage and subsequent selective recall of evidence, and (4) the combination of different sources of evidence into a composite judgement. Given that mood is likely to have effects at any or all of these stages, it would not be surprising to find emotional effects on judgement. For example, we have discussed in earlier chapters the evidence that mood has effects on the type of information that is available from memory, leading us to expect equivalent effects on judgement. Hence if a negative mood state were to increase the relative accessibility of unpleasant episodes in memory, then use of the availability heuristic would produce a consistent negative bias when judging the past frequency of negative events, or their future probability, or their co-occurrence ('It never rains but it pours'). A large number of experimental studies involving mood manipulations have demonstrated effects which are entirely consistent with this expectation, In an influential paper, Isen et al. (1978) suggested the existence of a cognitive loop connecting mood state, accessibility of positive or negative material in memory, and behavioural consequences. Positive-feeling states were induced by giving passers-by a free gift, and in a second (apparently unrelated) interview the same subjects were required to rate satisfaction with their own car and television set. Compared with controls, the subjects who had received free gifts rated themselves as significantly more satisfied with their possessions. Current mood state also influences normal subjects' ratings of their performance on a mental task. After induction of a sad mood prior to performance of a mental task, subjects were significantly less satisfied with their own performance, even though this had been standardised by providing predetermined feedback of success or failure (Wright & Mischel, 1982). Very similar results have been reported in a series of studies concerned with naturally occurring depression and the assessment of one's own performance in a variety of tasks (Blaney, 1986). Both depressed students (Nelson & Craighead, 1977) and depressed psychiatric out-patients (DeMonbreun & Craighead, 1977) underestimated the amount of positive feedback they had received during an earlier laboratory task. In other studies (Gotlib, 1983; Cane & Gotlib, 1985) similar effects were found when depressed patients rated standard evaluative feedback that they thought described their social performance, and when students directly rated their behaviour in simulated social situations. Even when confronted with identical neutral descriptions of their personality, depressed individuals rate such descriptions of their behaviour and themselves in a less positive way (Vestre & Caulfield, 1986). Page 197
This congruence of results arising from temporary mood manipulations, and more permanent states of depression, has been interpreted as implying that they arise from effects of depressed mood state on accessibility, possibly interacting with more enduring aspects of depressives' cognitive organisation (Teasdale & Spencer, 1984, Teasdale, 1988). However, very few studies have systematically investigated possible interactions between mood state and these more enduring individual differences (for an exception, see Teasdale & Dent, 1987). Equally, very few have considered whether the negative judgemental biases observed arise from depression or anxiety, or both. This last question certainly remains an open one, since at least one study in which both anxiety and depression were measured found that negative subjective evaluation of performance related more to trait anxiety than to depression (Zarantonello et al., 1984). One interesting possibility is that anxiety is associated with a bias to selectively anticipate negative
(threatening) events, while depression is also associated with a reduced anticipation for positive events (MacLeod, Rose & Williams, 1993). The 'How Do I Feel about It?' Heuristic Consistent with the earlier findings of Isen et al. (1978), Schwarz & Clore (1983) reported that mild happy or sad moods, varied by interviewing subjects under different conditions such as on sunny or rainy days, resulted in concordant changes in ratings of satisfaction with their life as a whole. The explanation for such effects that was offered above is based on the availability heuristic. A depressed mood will make negative material in memory more accessible, and via the availability heuristic, this will in turn bias one's judgements of present circumstances or past events. However, there was a novel manipulation involved in Schwarz & Clore's experiments, that calls this explanation into question. Half of their subjects were unobtrusively reminded of a probable cause for their happy or sad mood (e.g. a small uncomfortable room, bad weather, etc.), just prior to making their ratings. These subjects showed a much reduced mood congruent effect on their judgements of life satisfaction, which were not significantly different from those of controls. Why should this simple manipulation cause the mood effect to disappear? Schwarz & Clore (1989) argue that people use their current mood as a source of information in making evaluative judgements. That is, when asked a question about our feelings or attitudes towards something, it is as if we ask ourselves 'How do I feel about it?', and then use our current mood state as evidence for the answer. If this mood state is actually caused Page 198
by the thing being judged then use of this heuristic would be reasonable. However, if there are other unrelated causes for current mood, then it could be misleading. Reminding subjects that their mood may indeed have other causes, may thus serve to prevent the otherwise automatic use of mood-as-information, or may lead people to discount it subsequently as a valid source of evidence. Although it seems unlikely that such an explanation could apply to all moodcongruent cognitive biases, it is theoretically important because it suggests that simple availability explanations do not provide a complete account of the impact that mood states have on evaluative judgements. Emotion and Judgements of Contingency It has sometimes been noted that reports on frequency of past successes or failures in experimental tasks are provided more accurately by depressed than by non-depressed subjects (for example, Nelson & Craighead, 1977). More consistently, depressed individuals appear to be relatively pessimistic (but more accurate) than non-depressed individuals in their judgements of contingency and control. In a study by Alloy & Abramson (1979), depressed subjects proved very accurate at detecting a lack of contingency between their own button-pressing responses and a desired outcome. Non-depressed controls were both less accurate and less consistent in their judgements across situations in which outcomes were either desirable or otherwise. If the outcomes were desirable, normals tended to overestimate their control, but then underestimated their control if outcomes were undesirable. In another related experiment, normals consistently succumbed to a positive illusion of control over success regardless of exposure to induced helplessness manipulations; whereas depressed subjects were consistently more accurate in correctly judging non-contingency (Alloy & Abramson, 1982). This accuracy is apparently confined to observation of oneself, however, in that judgements of contingency made for others by depressed subjects tended to err in the direction of overestimating control (Martin, Abramson & Alloy, 1984). It would thus be misleading to characterise depressed people as invariably 'sadder but wiser' with the implication that their judgements are in some sense less biased than those of normals. Rather, it would appear that both depressed and non-depressed
people can make biased judgements of contingency and control but in circumstances where non-contingent success is being judged for oneself, the cognitive distortions shown by nondepressed people may be absent in depression. None the less, these findings may have implications for psychological treatments, since they suggest that therapists could try to induce the illusion of control over Page 199
positive outcomes. Even if normals are less accurate in this respect than are depressives, such a positive bias may be a valuable defence against becoming depressed in circumstances in which one cannot always have control over events. Covariation Bias and Emotional Events The illusory correlation effect described by Chapman & Chapman (1967) may sometimes lead phobics to overestimate the relationship between feared stimuli and aversive outcome. In a series of experiments, Tomarken, Cook & Mineka (1989) found that fearful subjects exposed to pictures of snakes or spiders that were followed one-third of the time by mild shocks, afterwards rated the probability of contingent shock at about 50%. Lower estimates were given for the probability of other outcomes after these slides (an aversive tone, or nothing) although in fact they occurred equally often. In addition to these overestimates of frequency, fearful subjects also rated the shocks after phobic stimuli as being more painful than those following neutral slides. In general, both frequency-and pain-rating biases seem to be more marked in high than low fearful subjects, and particularly so with 'biological' stimuli such as snakes or spiders, rather than with 'technological fears such as electric outlets (Mineka & Sutton, 1992). Subsequent research, however, suggests that both these differences may be a function of fear level. Subjects sensitised by news about recent stabbings showed similar effects with slides of knives (Sutton & Mineka, submitted), and the 'biological' status of stimuli does not seem to add significantly to the predictive power of fear ratings alone (Amin & Lovibond, submitted). After behavioural treatment, the strong overestimation of shock contingency shown by spider phobics was eliminated in one study (de-Jong et al., 1992), although not in another (de-Jong & Merckelbach, 1991). As with many other biases that are associated with emotional disorders, a critical question is raised: does covariation bias play any causal role in maintaining the disorder? As Mineka & Sutton (1992) point out, the form of the bias seems such as to confirm and strengthen a preexisting fear that aversive consequences will follow a phobic stimulus. This implies that selective learning takes place during the (actually random) slide/outcome pairings, perhaps due to preferential encoding of the instances when phobic stimuli are followed by shock. In contrast, it is possible that fearful subjects already have the strong expectation that phobic slides would be followed by shock, rather than learning anything during the experiment. In fact, a finding consistent with this possibility has been Page 200
reported by McNally & Heatherton (1993). After reading a description of the experiment and instructions used by Tomarken, Cook & Mineka (1989), subjects were asked to estimate the contingencies they would expect to experience. Results for high versus low snake fear subjects exactly matched those found by Tomarken and colleagues at the end of their experiment; although all subjects also had high expectancies for shock following electric outlets slides, unlike the findings of Mineka & Sutton. Similar experiments were reported by de-Jong, Merckelbach & Arntz (1995) who found higher estimated shock probabilities for spider slides, and that these persisted throughout the experiment in untreated high-fear subjects. It is therefore likely that perception of covariation was not created during the experiment, but already existed in the form of an expectation. According to correlational data from Tomarken,
Sutton & Mineka (1995), the persistence of this expectation is not due to the snake (or spider) and shock being perceived as 'belonging together', although it is associated with the perception that they are very similar in associated emotional reactions. In any event, the expectation can apparently be effectively disconfirmed in the case of electric outlet slides, or in low-fear subjects, but is less easily modified in the case of highly feared stimuli. The argument that covariation bias may serve to perpetuate some fears is thus weakened, but not disproved. An alternative hypothesis would be that a very strong emotionally driven expectation has already been acquired by highly fearful subjects, and this strength helps to protect it against the effects of disconfirmation. In some circumstances, particularly when outcome is more ambiguous or is avoided and thus unknown, such expectations could remain unchanged or even be strengthened. As was suggested by Hill, Lewicki & Neubauer (1991), if people have acquired an encoding rule, then ambiguous instances that do not actually disconfirm the rule may be treated as being consistent with expectation, and thus conceivably strengthen it. In this way, a more subtle version of the perpetuation hypothesis might be retained, and tested in future experimental work. Subjective Risk The Effect of Imagining on Outcome Probabilities In an extension of the research work concerned with judgements of past or present events, manipulations designed to alter the availability of specific future scenarios have also produced evidence of subjective probability changes. Suppose that you were asked to estimate the probability of a Government or Opposition victory in the next election. Having made Page 201
your prediction, you are then asked to imagine as vividly as possible a detailed scenario involving plausible events that culminate in a victory for your less expected outcome. Would this procedure lead you to revise your expectation, despite knowing that the exercise was a purely imaginary one? Based on research studies such as those by Carroll (1978), the answer is apparently yes. Subjects required to imagine either Jimmy Carter or Gerald Ford winning the 1976 Presidential election shifted their predictions so that these more closely matched the outcome that they had just imagined. In a related experiment, Gregory, Cialdini & Carpenter (1982) asked subjects to imagine themselves involved in a positive event (e.g. winning a contest) or negative event (e.g. being arrested for a crime they had not committed). Once again, systematic shifts of probability judgements occurred in which subjects moved towards increased likelihood of the event they had imagined. Both these studies were interpreted as being due to the imagined scenarios making related material more accessible in, memory. Alternative explanations in terms of demand seem unlikely. In a further experiment on imagined subscription to a cable TV service, subjects were more likely to actually subscribe to cable when offered it by company sales people (47% in the imagination condition versus 20% of the control group). This effect of imagined scenarios on later behaviour was subsequently replicated independently by Anderson (1983) who found that changes persisted over at least a three-day period. Under the pretext of asking subjects to plan a series of creative cartoons, they were asked to imagine themselves as the main character in scenarios which involved giving blood or joining a political action group. Three days later, subjects who had imagined donating blood were more likely to volunteer to do so than were subjects who had imagined the political scenario. It may be that similar processes are involved in psychological treatment methods that involve
imagining new behaviours. Systematic desensitisation, covert modelling and role-play rehearsal in imagination are all examples of treatments in which clients imagine themselves performing otherwise avoided behaviours, and which may exert their effects through enhanced availability of new behavioural scripts in memory. Mood and Subjective Risk Where mood has been experimentally manipulated, it appears to have surprisingly pervasive effects on estimates of risk across a wide range of future events, both positive and negative. In the previously cited study by Johnson & Tversky (1983), reading newspaper accounts describing the Page 202
death of an individual in detail apparently resulted in quite global increases of estimated risk across all causes of death, rather than local increases related to the similarity between the newspaper account and the causes of death to be rated. Kavanagh & Bower (1985) argue from similar results that the global increase results from activation spreading equally to all events linked to a particular emotional state, as predicted by the network theory of mood and memory. This does not explain, however, why events which are directly primed due to being semantically related to a mood-inducing event, as well as being linked via that mood state, do not show disproportionate increases. A possible explanation was provided by Constans & Mathews (1993), who found specific effects when only one mood-inducing event is used and then immediately rated, but global effects when many events were used, and then rated after a delay. It seems that specific priming effects do occur, but can be obscured by interference among items, when multiple mood-related events are presented. Global elevations in subjective risk, relative to normal controls, have also been obtained in various clinical populations: depressed and generally anxious patients (Butler & Mathews, 1983), and agoraphobics (McNally & Foa, 1987). In the first study, normal, anxious and depressed subjects rated the risk of a range of positive and negative events, both in relation to themselves and to others. No significant differences were found between the groups in the analysis of the ratings for positive events. On the other hand, ratings of negative events showed highly significant differences, with both anxious and depressed subjects rating negative events as more likely to happen than did normal controls. Furthermore, although normal individuals did not differentiate between probability ratings for themselves or for others, ratings of negative events made by both anxious and depressed subjects for themselves were significantly higher than ratings for the same items when made for another person. Similar findings were reported in the second study, but McNally & Foa also established that the elevation in subjective risk was no longer present in a sample of treated patients. An alternative to the spreading activation account suggests that subjects use their own mood state as a source of information when making judgements of risk (Schwarz & Clore, 1989). A particular mood state might then serve as evidence for a wide range of judgements, so long as the judged events were congruent with that mood. When feeling anxious, for example, we might rate all possible threats as more dangerous and more likely, because the anxiety is treated as evidence that the dangers are real. Thus, as with the experimental studies of normal subjects, the inflated risk estimates obtained from clinical groups could be Page 203
explained either by a 'priming by mood state, plus greater availability' account, or by the use of the 'How do I feel about it?' heuristic. Some recent evidence suggests that both types of process may occur, depending on conditions (Forgas, 1995). Consider first an experiment by MacLeod, Williams & Bekerian (1991), who required high or
low worriers to generate reasons for or against a specified negative event happening to them in the future. High worriers rated the negative events to be more probable, as expected. A posthoc median split by anxiety scores in the worry group produced some evidence that the more severely anxious subgroup were slower to generate an initial reason as to why the negative event would not happen, and they also rated this task as more difficult to do. However, if they eventually succeeded in generating such a reason, this had the effect of reducing the subjective probability of the negative event, consistent with the use of availability in judging risk. Similarly, MacLeod & Campbell (1992) found that speed of recalling specific emotional episodes was correlated with the subjective probability that something similar would happen again. Induction of a sad or happy mood resulted in faster recall of congruent memories, and this speeding effect of mood was correlated with a corresponding difference in probability ratings. What is it about generating a reason that something might or might not happen, or recalling an event, that influences judgement of future risk? More specifically, is it the quantity or salience of the information generated, or is it something about the process of generating it? Schwarz et al. (1991) provided some compelling evidence that the difficulty of generation or retrieval might be more important than the mere number of examples produced. Subjects required to generate six instances of their own assertive behaviour, which was easy to do, rated themselves as more assertive than subjects required to generate 12 instances, which was difficult. Thus, despite the greater number of accessed instances in the latter case, subjects presumably took their subjective difficulty in finding examples as evidence against their typicality. Perceived fluency of generation may thus be used as one source of evidence that influences judgements of probability. However, fluency is unlikely to be the only source of evidence. If we could express the 'How do I feel about it' heuristic in verbal form, it might be something like 'I must be in danger from whatever I am thinking about, because otherwise I would not feel this anxious'. Arntz, Rauner & van den Hout (1995) found strong support for this idea using scenarios that varied in the emotional reactions described as being experienced by the central character. Anxious patients, but not normal controls, rated the Page 204
danger in scenarios to be greater when the person involved was described as being anxious. This result is not easily explained by differential fluency of generation, but suggests instead that anxious people are more inclined to use feelings as a source of evidence, perhaps due to their greater salience in this group. This is not to imply that we are normally aware of such influences, whether due to mood state or to perceived fluency. For example, judgements can be influenced by familiarity, even in the absence of conscious recognition. Unknown names can be mistakenly judged as being those of famous people, if subjects have been briefly exposed to them within the same experiment, but are not being able to remember having seen them (Jacoby & Kelly, 1989). It seems that awareness of the origins of impressions is not necessary for judgements to be influenced by them. We suggest that multiple sources of evidence can be called upon in making judgements of risk, including the recall of prior judgements made, perceived fluency of generated reasons and instances, and current mood state. Which source predominates will presumably depend on its salience relative to other sources. In emotional disorder mood state, effects may be particularly strong and lead to global elevations in subjective risk, whether due to use of emotional state as evidence, or fluency of access to negative information. Equally, when a transient mood state has been induced by events, those feelings may also be globally influential, although they will be more easily discounted as evidence if attention is drawn to the events responsible. Alternatively, when information in memory has been primed by recent exposure, a related
judgement is more likely to be influenced by the feeling of familiarity or fluency with which relevant information is accessed, leading to more specific effects. Interaction of Mood with Trait Emotionality The above analysis suggests that the use of evidence such as mood state or fluency is sufficient to explain elevations in subjective risk, without needing to suppose an interaction with enduring cognitive characteristics. There is evidence, however, that subjects high in negative trait emotionality may show somewhat stronger, or more general effects. Butler & Mathews (1987) gave subjective probability questionnaires to students who were or were not expecting an important examination. As expected, state anxiety rose as the examination approached, and there was a related tendency to rate all negative events as being more likely. However, there was also an interaction between type of event and trait anxiety level. Individuals high in trait anxiety scores tended to predict that all positive Page 205
events were less likely to happen (to themselves) and all negative events were more likely. Low trait anxious students showed an increase in subjective risk of examination failure, but less evidence of global elevations in dangers unrelated to examinations. One theoretical framework within which these results can be understood supposes that trait emotionality differences reflect the extent and accessibility of mood-congruent information. In this view, high trait anxiety is associated with having information in memory encompassing a wider range of dangers than in low trait anxious individuals. Mathews, Richards & Eysenck (1989) and Eysenck (1992) have suggested that this information may be more extensive, and better integrated, in highly anxious individuals, and less associated with other more positive information. Thus, in episodes of worry, high trait anxious individuals may become aware of many possible threats in succession, without being able to access any positive alternatives. Similarly, in Teasdale's (1988) Differential Activation Hypothesis, individuals prone to depression are said to have more 'depressogenic' information in memory; that is, more negative information about themselves having severe long-term implications for self-worth. When in a depressed mood, for whatever reason, this information will become differentially available, and thus perpetuates depression longer than in those having less severely negative material in memory. Support for this hypothesis was provided by Teasdale & Dent (1987), who found that both self-rating and recall measures were more negative in previously depressed than in never-depressed subjects after the induction of a sad mood. Parallel results have been reported by Miranda & Persons (1988) and Miranda, Persons & Byers (1990) showing that endorsement of dysfunctional beliefs are mood-state dependent. Previously depressed subjects endorsed items on the Dysfunctional Attitude Scale to the extent that they were currently in a negative mood state. No such relationship existed for subjects who had not been depressed previously, who always endorsed fewer dysfunctional beliefs irrespective of current mood. In both these theoretical views, priming by mood state or related events is a necessary condition for the global elevations in subjective risk, and other judgemental biases, that have been documented earlier. The extent to which these biases occur, however, varies according to trait emotionality, or prior experience of emotional disorder. Emotional bias in judgement is thus best seen as the joint product of current state factors, and enduring emotionality, with mood or related events differentially eliciting larger biases in high negative-emotionality subjects (who are characterised by extensive congruent information in memory). Such Page 206
judgemental biases should remit on recovery, but may easily be reinstated in these vulnerable subjects in negative mood states. Interpretation of Ambiguous Stimuli Suppose you wake with a start in the middle of the night thinking you heard a noise, but all is quiet. What do you suppose woke you up? Clearly there are a number of possible interpretative schemata that one could use in understanding such ambiguous scenarios, or their equivalents in real life. We shall discuss the extent to which emotional disorders may be associated with systematic bias in the selection of different schemata in the following chapter. However, there is clearly considerable overlap between the concept of judgement and that of selection of schemata to use when arriving at an inference. In this chapter we have been concerned for the most part with reported subjective judgements of uncertain situations, rather than with comprehension and inference. However, this is necessarily an arbitrary division, and in some of the experimental paradigms that have been used it is not clear whether reports should be classed as subjective judgements, or as inferences which require selection among various possible schemata. In the remainder of this chapter we shall discuss some studies involving the interpretation of ambiguous words, while leaving experiments on inferences from more complex texts for the following chapter on schemata. Eysenck, MacLeod & Mathews (1987) required subjects differing in level of trait anxiety to write down homophones as they were read out in list form. Homophones consist of words that have more than one meaning associated with the same pronounced sound, and in this case they consisted of words where one of the meanings was neutral, while the other was mildly threatening (e.g. dye, die; pane, pain). In such homophones one or other of the meanings tends to be dominant, and this was balanced so that in half of the words the dominant meaning was threatening and in the other half it was non-threatening. Similarly, half of the threatening meanings were physical and half were social. Anxiety level of subjects did not affect results associated with these last two factors, but overall there was a highly significant correlation between trait anxiety score and the total number of threatening spellings chosen for the homophones. As expected, the higher the trait anxiety level the more threatening interpretations were selected. In an extension of this work Mathews, Richards & Eysenck (1989) used the same task with groups of currently anxious and recovered patients, Page 207
and normal controls, and with the addition of skin conductance measures. Anxious patients used more spellings corresponding to the threatening meaning than did normal subjects, with the recovered group being intermediate and not significantly different from either of the other two groups. Skin conductance reactions were greater for unambiguously threatening than for neutral words overall, with the ambiguous homophones being intermediate; but there were no differences involving groups in this respect. Thus, the association with anxiety with selection of threatening meaning was confirmed with a clinical population. It is possible that subjects were aware of both meanings and consciously selected one of them, although there were no indications from the skin conductance data that this was the case. A more rigorous demonstration of bias in the interpretation of single words was provided by Richards & French (1992). Having established norms for a set of homographs with both threatening and non-threatening meanings (e.g. stroke) they used them as primes in a lexical decision task, with associates corresponding to each meaning as targets (e.g. disease versus caress). Subjects with high trait anxiety scores were faster to endorse targets corresponding to threatening than to the non-threatening meaning, relative to low anxious subjects. However, this was only true
when there was a delay of 750 ms or more between presentation of the prime and target. With only 500 ms delay, there were no such differences, suggesting that in these anxious subjects, the interpretative bias may not be strongly automatic. None the less, it seems clear that selection of the threatening meaning of ambiguous words is a very rapid process, not requiring much in the way of reflective thought. These and other experimental paradigms involving text comprehension (discussed in Chapter 6) lead us to believe that anxious (and perhaps depressed) subjects make inferences at the time of encoding which are biased in a mood-congruent direction. The exact mechanisms responsible have been less researched than have other judgemental biases discussed in this chapter, but we would argue that the consequences are likely to be similar. Thus, emotional disorders seem to be associated with a range of judgemental biases: a tendency to be more sensitive to negative contingencies in their environment, or to perceive them even if they do not really exist; an inflated view of the risk of future negative events; and a tendency to view ambiguous stimuli as having a threatening meaning. If a similar bias were to operate in real life, the experience of an anxious or depressed person will be that of a relatively threatening and upsetting environment. Judgemental bias in emotional disorders thus takes a form that is capable, at least in principle, of supporting and maintaining negative mood state. Page 208
Clinical Judgements and De-Biasing The discussion earlier in this chapter of heuristics and biases common in normal individuals suggests that clinicians will not be immune from making similar errors. Of particular relevance to clinical judgements is the early work of the Chapmans (Chapman, 1967, Chapman & Chapman, 1967, 1969) and the later studies that have extended their conclusions. These researchers were originally interested in why clinicians persist in believing that certain features of projective test responses are diagnostic of emotional disorders, despite evidence that they were actually invalid. The Chapmans hypothesised that clinicians' prior beliefs about the aetiology of certain conditions led them to perceive an illusory association between specific reactions and diagnostic category. That this was indeed the case could be demonstrated by providing clinical judges with totally random pairings between test reactions and individual clients, and showing that an illusory correlation persisted despite the lack of any real association. It would seem that the occasional random pairings which apparently confirm a clinician's prior theory are particularly salient, and for this reason are noticed and retained, while disconfirming instances are ignored or forgotten. Reviews of this and subsequent work are provided by Ross (1977) and Turk & Salovey (1986). If the use of heuristics and consequent biases are so common, what corrective measures can be taken? Although the obvious answer is to warn people of their existence, abstract educational methods alone appear to have limited success in removing bias from intuitive judgements. Most discussions of methods which could be used to reduce bias suggest that, beyond educating judges about common sources or error, it is most helpful to provide corrective feedback which allows individuals to observe the nature of judgemental bias in their own responses. Specific errors, such as the confirmation bias, can be countered to some extent by practising corrective heuristics, such as listing reasons why one's initial solution might be wrong (Turk & Salovey, 1986). De-biasing research may eventually lead to helpful suggestions that can be used in cognitive or behavioural therapies, although relatively little work has been specifically directed at modifying the powerful judgemental biases encountered in emotional disorders. Consider, for example, the problem of an anxious and hypochondriacal patient who is convinced that palpitations are a
warning of imminent heart disease. Such a conviction is likely to be maintained by a series of heuristics and biases. Vivid and emotionally arousing scenarios of heart attacks, derived from the media, family or friends, may be a particularly compelling and available form of evidence. Page 209
Palpitations are a representative symptom of heart disease, although in using this heuristic the client is overlooking the high base rate of palpitations which occur in the absence of physical health problems. Once convinced that a threat of heart disease is present, the confirmation bias will ensure that attention is selectively drawn to corroborating evidence while disconfirming instances will be ignored. De-biasing attempts in such a client might be attempted in the context of cognitive therapy in the following ways. The high emotional salience of thoughts and images may be pointed out, while encouraging the client to see that they constitute ideas rather than real evidence. Attention may be drawn to the high rate of palpitations found in normals, and the client might be encouraged to gather information about the frequency of palpitations among healthy friends. Perhaps, given the documented power of pre-formed causal hypotheses, it may be more important, to provide an alternative causal model which the client can use in actively countering judgemental bias. By offering the client a model in which frightening scenarios of heart disease may cause autonomic symptoms which in turn serve to maintain anxiety, therapists may be able to help their clients to use judgemental heuristics in a positive rather than a negative way. In the clinical example cited at the beginning of this chapter we posed the question of whether emotional disorders are systematically associated with distortions in judgement, such as subjective probability. A great deal of evidence has been summarised to show that, although judgemental distortions are also common among non-emotionally disturbed populations, depressed and anxious individuals are indeed characterised by particular types of bias which distinguish them from normals. Hence emotionally disturbed people judge their performance more negatively, consider they have less control over success, rate the risk of unpleasant future events as more likely, and interpret ambiguous events in a more threatening manner. Many of these effects can be attributed to the use of current mood state as evidence, or to the effects of mood in making congruent events more available. Certainly, the characteristic distortions observed become more pronounced as mood becomes more negative, and tend to reverse on recovery. On the other hand, some research suggests that judgemental biases cannot be wholly attributed to short-term mood changes, and perhaps depend on previously acquired impressions and strategies. To the extent that this is true, the results of de-biasing research are likely to find useful applications in the psychological treatment of emotional disorders.
Page 210
Chapter 9 Schemata It would be impossibly inefficient if we had to understand each new event without reference to other similar events that have occurred in the past. The idea of schemata arises naturally from the intuition that regularities among similar events leads to their being connected in memory to form a memory structure, or schema, for that class of events. Thus, when encountering something new, but which resembles previous things now represented in memory, we can understand the present encounter by referring back to a summary of features common to previous similar encounters. Suppose you visit a building that you have never seen before, but has a large 'New Supermarket' sign above it. You can reasonably expect to see displays of goods laid out, know that you will need money to buy them, and should look for a checkout to take them to. As well as features common to previous encounters, there may also be some variable features, or some that are completely unexpected. Thus you know there will be a checkout assistant to scan the goods, although that person may be male or female, young or old, and so on. But if there seems to be no checkout at all in the new supermarket, you might be confused as to what to do, and may need to entertain new hypotheses not part of your current 'supermarket' schema (perhaps there is a new personal scanning system installed that you have to operate yourself?). The use of schemata as explanatory concepts has become common in cognitive psychology following Bartlett's (1932) classic demonstration of how a culturally alien folk-tale becomes changed in the re-telling so as to better fit with the listeners' own expectations. More recently, the term schema has been used with increasing frequency in social and clinical psychology, to explain anything from how we understand personality traits in others, to why some people are vulnerable to depression. In this chapter we shall first review the various uses of the term in cognitive psychology, before examining how schemata are claimed to guide the interpretation of ambiguity, and influence the encoding and retrieval of emotional information from memory. Page 211
Nature of Schemata Different researchers use the term schema with widely varying degrees of precision, and this has consequently led to considerable confusion and misunderstanding. We shall here consider several, increasingly specific, uses of the term. 1. At the most basic level all theorists agree that a schema is a stored body of knowledge which interacts with the encoding, comprehension and/or retrieval of new information within its domain, by guiding attention, expectancies, interpretation and memory search (see reviews by Graesser & Nakamura, 1982; and Alba & Hasher, 1983). While some may consider this definition sufficient, the majority of researchers would require that such memory structures fulfil certain additional criteria. 2. One of the more elementary criteria is that the schema should have a consistent internal structure, used as a template to organise new information. Such stimuli therefore come to be structured in a stereotypical manner. Thus, when subjects are presented with stories in which the positions of various statements have been rearranged, they tend to produce recall protocols with a more conventional organisation. For example, expected events in the description of
(say) a visit to the supermarket tend to be recalled in their appropriate temporal order, even if they had not been presented in that sequence, or even if a feature had been omitted in some descriptions (Bower, Black & Turner, 1979). 3. Another common criterion is that the knowledge contained within schematic structures should be generic in nature, constituting relatively abstract prototypical representations of environmental regularities. For example, we may have a schema representing our knowledge of birds, in which the main features correspond to features shared by all or most birds, such as being small, having feathers and being able to fly with wings. As a result, it may be easier to recognise a robin as a bird than an ostrich, because a robin is closer to this prototype. Thus a schema is often conceptualised as an organised summary of the attributes and relationships which typify certain classes or categories. Various researchers have produced evidence that we have generic schemata for such varied domains as motor skills (Chamberlin & Magill, 1992), routine social events (Trafimow & Wyer, 1993), body positions (Reed & Farah, 1995) and so on. Specific instances of such stimuli are then processed using the appropriate schematic prototype as a blueprint to impose structure, resolve ambiguity and provide supplementary information. The final representation in memory may therefore include both elements from the specific stimulus event and elements from its generic prototype. 4. It has sometimes also been suggested that before such representations can be considered schematic they must constitute a modular package of Page 212
such generic information, in the sense that activation of any part will tend to produce activation of the whole (Mandler, J., 1984). Suppose we are told: 'John sat down and looked at the menu.' A menu is probably a bound feature in most people's restaurant schema and, since the activation of this element will produce activation of the whole modular representation, we may immediately assume the presence of food, waiters and chefs; and anticipate more specific information concerning stereo-typed schema-bound actions such as ordering food, eating, paying the bill, etc. Reading-time studies do indeed indicate that activating a schema establishes the mental representation of unmentioned scenariobound information just as effectively as does the explicit introduction of this information, (e.g. Garrod & Sanford, 1982; Sanford, 1985). In summary, we have seen that while certain properties are necessary before any internal representation can be considered schematic, various additional criteria are required by most definitions of this concept. A schema certainly consists of a stored domain of knowledge which interacts with the processing of new information. However, certain definitions also require that this representation must have a consistent structure which exerts an organisational influence, must embody only generic prototypical information, and be modular in nature. Later in this chapter we review several different areas of psychology where schema-based models have been proposed, and consider which of these criteria are commonly met. The Functioning of Schemata Bottom-up and Top-down Interactions Most schema theorists assume that the comprehension of a stimulus input involves first identifying the relevant schema, and then applying it to direct the intake of information and organise its representation (cf. Norman & Bobrow, 1976). The first of these two processes is a 'bottom-up' or data-driven operation in which the cognitive system attempts to recognise the schema that is best able to accommodate the elements of the particular input. The second process is 'top-down' or conceptually driven, with the chosen schema being employed in the
control of attention, inferencing and the selective storage of particular details (cf. Graesser & Nakamura, 1982). Sanford & Garrod (1981) point out that stimulus input may often take the form of a fragmentary description of a situation about which we already have further knowledge. When this happens, these components of our existing internal representation will Page 213
become activated and, if they are a part of a modular schema, the entire structure will be activated in consequence. Selection of schemata by such 'partial matching' is a common device employed by schemabased AI systems. Of course, the system is not fool-proof, and errors will occur when such partial matching leads to the activation of an inappropriate schema. These errors will only be detected when some later information cannot be accommodated by the currently active schema which, in consequence, must be revised. Such errors are subjectively obvious in 'garden path' texts including the following examples from Sanford & Garrod (1980): (i) John was on his way to school. (ii) He was terribly worried about the maths lesson. (iii) He thought he might not be able to control the class again today. (iv) It was not a normal part of a janitor's duty. The schema activated by the first two sentences usually casts John as a pupil; sentence (iii) cannot be accommodated by this representation and a new schema is selected in which he is considered the teacher; while the final sentence indicates this also to be inappropriate and a further revision is required. In such instances the failure to accommodate the new information within the existing schema leads to comprehension difficulties, which are easily indexed by reading-time latencies. Once an appropriate schema is operative, it is assumed to guide processing in a conceptually driven fashion, and influence which aspects of the input receive most processing. Most attention is usually allocated to those elements which deviate from the schematic representation. However, when confronted with diverse information concerning unrelated topics, a schema can be employed to guide attention selectively towards schema congruent elements, elaborating vague information and filling in absent details. Ultimately an integrated representation will be stored which is heavily influenced by the active schema used during encoding, and this will lead to a range of memory biases. There is, however, dispute in the literature concerning the exact nature of such biases, which partly reflects underlying differences of opinion regarding the detailed operation of schemata. Graesser & Nakamura (1982) identify four alternative formulations based on different aspects of experimental phenomenon. Four Hypotheses (1) Filtering hypothesis. According to this account, information which is typical of the active schema becomes cohesively organised in the final Page 214
representation, whereas atypical information is only loosely associated. This predicts a recall advantage for schematically typical over atypical elements of the input. There is indeed empirical support for this prediction (e.g. Rothbart, 1981). However, such an effect may stem from a guessing bias, favouring schematically typical information, rather than representing a true memory effect. In fact, memory discrimination between stated and unstated information is
poorest for schematically typical elements (e.g. Bower, Black & Turner, 1979; Smith & Graesser, 1981). (2) Attention elaboration hypothesis. This suggests that, due to the disproportionate processing resources which schematically atypical elements attract during encoding, there will be a memory advantage for such information. Considerable experimental evidence is consistent with this alternative prediction (Graesser et al., 1980; Srull, 1981). Attempts to relate such superior recall for atypical elements directly to patterns of resource allocation suggest that the account may not be entirely adequate (Graesser, 1981), but a meta-analysis of some 60 studies does show a significant overall advantage for schema-inconsistent information, as predicted by the hypothesis (Rojahn & Pettigrew, 1992). (3) Partial copy model. Bower, Black & Turner (1979) proposed that, in addition to an episodic memory trace which is rapidly lost over a short time interval, a stereotyped event sequence is encoded by activating the corresponding elements in its generic schema. At short delays, explicitly stated elements are more highly activated than non-stated elements, which are only partially activated. However, such activation fades and, after a longer retention interval, it becomes impossible to distinguish stated from non-stated elements. Several predictions arising from this model were confirmed by Bower, Black & Turner (1979). For example, false-alarm rates were high for unstated, schematically typical actions, especially when several episodes requiring the same schema had been presented (permitting incremental activation of relevant but unstated actions). Also, as predicted, memory discrimination between stated and unstated actions was poorer for schematically typical than for atypical actions. Nevertheless, Graesser & Nakamura (1982) argue that the partial copy model is inadequate since, although the absence of memory discrimination for typical actions after a 30-minute retention interval indicates the loss of explicit propositions by this time, recognition memory for atypical actions persists even after three weeks (Graesser, 1981; Smith & Graesser, 1981). The generic schema should provide no basis for remembering such information. (4) Schema pointer + tag. According to Graesser's favoured account (Graesser, 1981; Graesser & Nakamura, 1982) the memory trace consists of Page 215
(i) a pointer to the relevant generic schema, and (ii) a set of 'tags' or additional information, for schema atypical or marginally atypical elements. The predictions arising from this account are similar to those produced by Bower, Black & Turner's (1979) partial copy model, but permit the continued retention of atypical information despite the complete absence of any ability to discriminate stated from unstated typical actions. To some extent these four formulations of schema functioning appear to stand in direct conflict. However, they are each specifically tailored to account for rather different aspects of the experimental literature. The first two models, for instance, are particularly concerned with accounting for differential recall of presented details, while the partial copy model was developed primarily to account for the erroneous 'recall' of missing information. Furthermore, some apparent contradictions appear to be terminological rather than conceptual in nature. For example, advocates of the filtering hypothesis predict and confirm that memory for typical information will be superior to that for atypical information, while the attention elaboration hypothesis leads to the supported predictions that memory for atypical information will be disproportionately good. However, each group of theorists conceptualise 'atypical' rather differently, to mean either information irrelevant to a particular schema (such as blowing one's nose in a restaurant), or incompatible with this schema (such as driving a tractor in a restaurant). In an investigation of
this issue, Davidson (1994) introduced different types of schema-atypical material in an attempt to separate out their effects on recall and recognition. Irrelevant elements were introduced into otherwise schematic accounts of routine social events (e.g. a visit to the cinema), which were described in a mundane, or rather more vivid, manner. The irrelevant elements were well remembered immediately, but tended to drop out later, especially if they were designed to be mundane rather than vivid. More strikingly, irrelevant events which interrupted or distracted from the normal flow of events in the schematic account, were much better remembered than schema-consistent information, even one week later. Thus, use of a schema can facilitate the processing of either schema-congruent or schema-incongruent elements of an input, depending upon the precise task conditions. Furthermore, it can facilitate memory for typical or for atypical information depending upon the definition of typicality and precise nature of the memory tests which are employed. Frames and Scripts Despite such complications, many have found schema theory useful to accommodate a wide variety of phenomena. Perhaps the most fully Page 216
developed formulations of schemata are expressed in the related theoretical constructs of 'frames' (Minsky, 1975) and 'scripts' (Schank & Abelson, 1975, 1977). Both frames and scripts represent highly structured modules of generic information. They are typically assumed to constitute organisational structures which serve to link in a co-ordinated manner 'slots' or 'terminals', capable of receiving particular details from the input. Each slot is designed to accept only one particular type of information, and may initially contain default values which, if no further details are supplied by the input, will be assumed valid. While 'frames' may embody a wide range of organisational relationships to arrange components slotsfrom simple spatial position to such complex rule-defined systems as scientific paradigms'scripts' represent a subset of frames involving temporal and causal relationships, as, for example, in the sequence of events representing a visit to a restaurant. Both theoretical concepts, however, have important aspects in common. Processing an input essentially involves assigning elements to appropriate slots in the relevant frame or script. The imposition of structure, the provision of unstated detail and the comprehension of inferences are all emergent properties of this representational system. There is certainly compelling evidence that identifying the appropriate generic script can impose a meaningful structure upon an input which may otherwise be extremely difficult to process. Consider, for example, the following passage, condensed from Bransford & Johnson (1971): The procedure is actually quite simple. First you arrange things into different groups. Of course, one pile may be sufficient depending on how much there is to do. After the procedure is completed one arranges the materials into different groups again. Then they can be put into their appropriate places. Eventually they will be used once more and the whole cycle will then have to be repeated.
Most readers find this text difficult to comprehend and to recall. However, if they are initially provided with the appropriate generic script by being given the title 'Washing Clothes', comprehension ratings and recall scores improve dramatically. Subsequent provision of alternative schemata can differently influence the availability of certain details, suggesting again that the frame can be employed as a retrieval device. For example, Anderson & Pichert (1978) presented subjects with a passage concerning a house which mentioned general layout, decoration, position of doors, open and closed windows, and so on. Some subjects were subsequently told to imagine they were house buyers, and others
burglars, and asked to recall everything they could about the house. Each group recalled significantly more of those details which could easily be accommodated by plausible slots in their Page 217
adopted schema, e.g. 'house buyers' recalled details of decor, while 'burglars' recalled details of access. Switching perspectives after initial recall led to further retrieval of much previously omitted information which could now be assimilated by the new schema. While clearly indicating that generic representations can affect comprehension and retrieval of new information, such studies do not directly address the hypothesis that those representations exert a structural influence. The best evidence that frames and scripts impose a particular structure upon new information is provided by those experiments which demonstrate that recall protocols tend to approximate a stereotypical organisation, even if this organisation is not explicitly represented in the initial stimulus. There is good evidence that clustering of propositions in the recall of text corresponds closely to this abstract theoretical schematic structure, even when such clustering is absent in the presentation order (e.g. Black & Bower, 1979). Using reading time as a measure of ease of comprehension has also produced evidence that familiar situations are represented as generic schemata with slots for standard elements. For example, Garrod & Sanford (1982) presented a short passage about a man being questioned, either explicitly stating the presence of a lawyer or not. Not surprisingly, a subsequent sentence which made a definite reference to 'the lawyer' was comprehended more rapidly when this character had already been mentioned. However, when the passage was introduced with the title 'in court', this effect disappeared entirely. Introducing the 'court' schema had thus apparently already activated a module that contained a 'lawyer slot'. Similarly, Anderson, Garrod & Sanford (1983) presented descriptions of standard situations (such as visiting a cinema), which then indicated a shift in either time or space, and finally made a reference to a schema element within the initial frame. The accessibility of the schema-bound element (as indexed by the comprehension latency for the final definite reference) was reduced, but only when the shift in time or space indicated that the original schema was no longer appropriate. While there is thus considerable converging evidence to demonstrate the existence of schemata, and the predictive power of frames and scripts, it is important to note that the evidence cited has dealt with relatively familiar and stereotyped situations. Quite frequently we have to try to understand events that we have never previously experienced, or at least not in that particular combination. To understand a novel situation we have to construct a new representation using elements of schematic knowledge that we already have: a mental model. In experimental studies of spatial variants of mental models, Bower & Morrow (1990) showed that providing a diagrammed layout of a house, and a description of a person walking through it, activated information about actions and Page 218
objects related to that description, even if they were not explicitly mentioned. For example, if they had just read about a route that implied walking through a specific room, readers were faster to answer questions about objects that had been depicted as in that room rather than elsewhere. Readers typically identify with the central character in such descriptions, and the introduction of information requiring the perspective of someone else in the narrative results in slowed reading times (O'Brien & Albrecht, 1992). Given a new situation, it appears that we use existing knowledge to construct a model having some, although not all, of the properties of schemata. That is, a mental model can be thought of as structured information in (working) memory, which actively guides understanding and interpretation, and which can influence how that situation is remembered later. However, models are not generic and modular in the
same way as frames or scripts, since they are constructed on line, are easily changed, and may never be used again. Repeated use of a model over related occasions may be necessary to develop a temporary model into something like an enduring frame or script. Schemata in Clinical Theory Many aspects of cognitive content in mood disorders suggest that emotional information is processed in an idiosyncratic or biased fashion. Observations of depressed individuals reveal that they tend to notice their failures rather than successes, infer the worst from situations which could be interpreted positively, and selectively recall unhappy memories from the past. By comparison with non-depressed controls, depressed subjects are more negative in their recall of performance feedback, evaluate themselves more negatively, and blame themselves for failure (Coyne & Gotlib, 1983; Blaney, 1986). Similarly, anxious individuals are selectively attentive to environmental cues related to threat, provided that these cues match current concerns, and are likely to perceive threatening meaning in ambiguous events (Mathews & MacLeod, 1994). Although these findings do not compel an explanation in terms of emotional schemata, such explanations have been frequently offered (e.g. Greenberg & Alloy, 1989; Ingram, Partridge, Scott & Bernet, 1994). The major contributor to the development of clinical theories involving the schema concept is Aaron Beck (see Beck, 1976; Beck et al., 1979; Beck, Emery & Greenberg, 1985). In his terminology, schemata are cognitive assemblies of structural elements, which are themselves further organised into larger constellations. When activated, a specific schema or constellation of schemata directly influences the content of a person's Page 219
perceptions, interpretations and memories; for example, by helping to select relevant details from the environment, or to recall relevant data. Individuals who develop specific emotional disorders are supposed to be characterised by the nature of those constellations of schemata concerned with interpreting emotional information. Thus depression is said to be associated with schemata concerned with loss, and with negative aspects of the self, the world and the future. Content concerned with personal vulnerability and danger activates the behavioural pattern of flight and the feeling of anxiety, and so on. It is clear from this account that Beck's use of the term schema corresponds to our first and most basic requirement of the concept; that of a stored body of knowledge which interacts with encoding, comprehension and retrieval of information. Presumably, it is supposed that individuals who are prone to develop depression or anxiety either have a more extensive database concerning loss or danger, etc. or have greater ease of access to such a database. A major problem with this view is that the relevant observations have been made on those who are currently in a depressed or anxious state, and there is little evidence that such individuals could have been distinguished from others not vulnerable to emotional disorders prior to its onset. The usual explanation offered is that the relevant schemata cannot be detected until activated by appropriate events. Thus Kovacs & Beck (1978) state: 'We postulate that the schemata that are active in depression are previously latent cognitive structures. They are reactivated when the patient is confronted with certain internal or external stimuli' (p. 529). A related suggestion is that vulnerability to depression or anxiety disorders depends on the type of information which is activated by a given mood state. If depressed mood activates highly aversive and deprecatory information, then the negative emotional state may persist and deepen rather than recover, as would be the case if little such information existed. This hypothesis predicts that vulnerability to emotional disorders arises more from the persistence of otherwise normal variations in mood than from initially different reactions to events (Teasdale, 1983, 1988). If it is indeed the
case that emotional schemata can remain latent but then be activated intact, at least a degree of modularity is being assumed. This assumption is not obviously required by any experimental data, and it seems equally possible to suppose that emotional information is stored in a more fragmented way, not necessarily leading to activation in unitary form. It is much less clear whether or not Beck views emotional schemata as containing generic representations. If so, then the theory would need to be more explicit concerning the type of representations involved, in order to become testable. Page 220
One might propose that depressed individuals have generic cognitive structures for situations such as social encounters that resemble those employed by non-depressed individuals, but that the frames or scripts involved have different default values. For example, when another person fails to provide definite signals of approval in a social encounter, the depressed person might retain a default value corresponding to 'dislikes me' in the relevant slot. Alternatively, depression might affect the choice of frame that is appropriate to any particular situation. If there are different frames for accepting and rejecting encounters, then depression might be associated with a bias towards selecting the rejection frame. Since no systematic research has been carried out on these issues they can only be the subject of speculation. In fact, most research on emotional schemata in depression has been concerned with content as revealed by selective recall. Specifically, it has been proposed that depression is associated with a particularly negative self-schema, and that it is this structure which explains the increased probability of recall for negative information about oneself. To evaluate this proposal we must first consider the nature of self-knowledge, and the extent to which this knowledge should be considered as a schematic structure. Self-Knowledge and the 'Self-Schema' The term 'self-schema' has commonly been used to refer collectively to the knowledge about oneself stored in long-term memory. We shall first examine evidence about the nature of selfknowledge, and then return to the question of whether self-knowledge may be considered to constitute a 'self-schema'. According to Markus (1977), cognitive structures concerning the self develop out of individuals' attempts to explain their own behaviour, and are used subsequently in encoding or reporting information about oneself. To demonstrate that selfknowledge operates in this way, Markus proposes that within any designated self-descriptive domain, one should (a) make judgements of applicability to oneself relatively rapidly; (b) be able to retrieve consistent behavioural evidence; (c) find it relatively easy to predict one's own behaviour; and (d) resist counter-schematic information. To provide evidence relevant to these propositions, Markus showed that individuals who consistently rated themselves as independent, individualistic and a leader (or as dependent, conforming and a follower) were more rapid in arriving at a decision as to whether words related to such traits were selfdescriptive than were other subjects who had less decided views (described as 'aschematics'). The same subjects also recalled more behavioural evidence, and predicted more future behaviour, that was consistent with adjectives related Page 221
to their own self-descriptions. Finally, when subjects were given feedback on their supposed 'suggestibility' that was incongruent with these self-descriptions, those who had described themselves as either independent or dependent rated the feedback as less accurate than did the so-called aschematic subjects. Thus, on each of the four counts proposed by Marcus as necessary evidence to support the idea that a consistent body of self-knowledge is used to describe one's own behaviour, confirmatory data were reported. More recently, however, Markus & Nurius (1986) and others (e.g. Linville, 1985; Strauman, 1992; Niedenthal, Setterlund & Wherry, 1992) have introduced the idea of multiple selves or
self-guides. As argued by Markus & Nurius, people can report distinct conceptions of what they might become in the future, as well as how they actually see themselves now, so that it makes little sense to think of a single self-schema. Neidenthal and colleagues (1992) assessed the complexity of self-knowledge in terms of the number and type of independent roles that are seen as important to the individual (e.g. as parent, spouse, worker, etc.), and confirmed earlier findings by Linville (1985) that complexity is associated with reduced vulnerability to emotional upsets, presumably because self-esteem is not uniquely determined by any one role. Complexity of possible selves also mediated the emotional reaction to feedback about future goals, in a similar way. The idea of self-guides is a related one, but specifically distinguishes between actual self, ideal self, and the self as one ought to be (e.g. Strauman, 1992), with the latter two acting as guides for desired changes in oneself. Words related to these self-guides were relatively efficient cues for recalling emotional memories from childhood, suggesting that they may be based on earlier experiences. Furthermore, discrepancies between actual and ideal self were associated with depression, and between actual and ought self with anxiety. Possible selves or self-guides thus appear to have some validity as predictors of emotional vulnerability. Rogers, Kuiper & Kirker (1977) were the first to show that self-knowledge appears to act as a particularly powerful mnemonic, using a depth-of-processing paradigm. Judgements of whether adjectives either described oneself, meant the same as another word (semantic), rhymed with another word (phonemic) or were presented in large or small letters, were compared for their capacity to facilitate incidental recall. The probability of recall following self-judgements was twice as great as following semantic judgements, which were in turn more effective than either phonemic or size judgements. In discussing these results, Rogers and co-workers state: 'In order for self-reference to be such a useful encoding process, the self must be a uniform, well-structured concept. During the Page 222
recall phase of the study, subjects probably used the self as a retrieval cue. In order for this to be functional the self must be a consistent and uniform schema.' The superiority for selfreferenced material is now a well-established finding, which has been replicated by many other researchers (e.g. Bower & Gilligan, 1979; Lord, 1980; Klein & Kihlstrom, 1986; Greenwald & Banaji, 1989; Ingram, Partridge, Scott & Bernet, 1994). None the less, it does not seem essential to invoke such concepts as a consistent and uniform self-schema in order to understand the self-referent recall data. The more extensive the information base within which new input can be encoded, the more associations are possible and thus the easier it will be to retrieve that input. This expectation arises because connections between old and new information may help in the search process. For example, in a network model of memory, search is said to occur via spreading activation from startingpoints such as the context within which the to-be-recalled word was presented. If activation spreads simultaneously from such contextual cues, and from activation of information about the self, then intersection and consequent word retrieval will occur more rapidly. Klein & Kihlstrom (1986), while replicating the self-reference effect, made clear how the superior recall of self-related material can be explained without appealing to the concept of a uniform and consistent self-schema. Essentially, they have shown that self-reference is no more effective than other forms of semantic encoding, provided that the categorical organisation of the material is strictly controlled. Hence, if encoding involves category questions such as 'Is this an external body part: skin?', then recall is just as good as following questions like 'Can you think of an accident involving your: skin?' Furthermore, self-referent encoding may result in significantly poorer recall if it does not encourage categorisation as self-descriptive vs non-self-descriptive, as in questions such as 'I prefer to keep my . . . short:
hair?' Thus it would appear that in earlier experiments, self-referent encoding has been confounded with the organisation of material into categories, and that it may be this organisation which is responsible for recall superiority. Consequently, the evidence on selfreferent recall does not necessarily tell us much about the structure of self-knowledge, as opposed to its use as a category in encoding or retrieval. In free recall paradigms, a search process involving the activation of self-descriptors may be an easy and effective one, while equivalent strategies could not be used to search for words which happen to have similar meanings, or which rhyme with each other, or were written in capital letters. On the other hand, if simple categories are made explicit by the encoding instructions, then the organisation which results will aid Page 223
retrieval to the same extent as does self-reference. Thus it would seem that while the effects of self-encoding and autobiographical recall might reveal something about the extent or type of self-knowledge, they do necessarily require use of the self-schema concept. For example, Bower & Gilligan (1979) tested recall of phrases such as 'a cheerful mood' or 'a broken bone' after these phrases had been encoded using judgements of whether they had occurred to oneself, or using other semantic and graphic judgements. Once again the self-reference condition was markedly superior compared with most other judgements, but encoding with reference to very familiar other people (such as one's mother) led to levels of recall similar to those achieved with self-encoding. Furthermore, although memory for self-descriptive adjectives and autobiographical events have been used interchangeably as evidence for a self-schema, it seems that the benefit from each is relatively independent (Klein, Loftus & Burton, 1989), indicating that they do not derive from a single structure in memory. These results suggest that there is nothing unique about the structure of self-knowledge, and that it cannot be adequately described in terms of a single self-schema. The more extensive the knowledge base about the person or topic, the easier it will be to find points of contact between this knowledge and the new information to be remembered (cf. Greenwald & Banaji, 1989). The existing evidence can thus be interpreted as strong support for the idea that an extensive and well elaborated knowledge base about oneself exists in memory, and that this knowledge base can be used in encoding by categorisation, and as an aid to subsequent retrieval. However, it does not show that all self-information is structured as a 'self-schema' having a consistent internal structure, nor that it has generic content, or is typically activated as a modular unit. Depression and Self-Schema Theory Early experiments by Davis (1979), modelled on that of Rogers, Kuiper & Kirker (1977), failed to show superiority for self-referred recall in depressive subjects, and this result was taken to throw doubt on the existence of a well-differentiated self-knowledge structure in depression. In subsequent work (e.g. Derry & Kuiper, 1981) it became clear that demonstration of a self-reference recall advantage in depressive subjects depends crucially on the use of negative mood-congruent words. Comparisons of positive and negative words showed that clinical depressives have a self-reference advantage for negative (but not positive) words, while with non-depressed subjects the self-encoded positive words are recalled better. This distinction between the two groups persists when self-encoding Page 224
is compared with judgements concerning other people (Bradley & Mathews, 1983).
Clinically depressed individuals recalled more negative words that had been encoded in relation to themselves, but more positive words encoded in relation to others, while nondepressed controls recalled more positive words both for themselves and others. This and other studies of depressed subjects (Kuiper & MacDonald, 1982: Bargh & Tota, 1988) have shown that, consistent with the criteria advanced earlier by Marcus (1977), latency to arrive at negative self-judgements during the encoding task is shorter for depressed subjects, while non-depressed subjects are quicker to give positive self-judgements. Furthermore, Bargh & Tota (1988) found evidence that these negative self-judgements in depression were relatively automated, in the sense that they were not slowed by a mental load, unlike other judgements. Results such as these cannot be attributed to a general bias in depression which favours the endorsement of all negative judgements, since both the recall and the latency data show differences between self- and other person judgements. At the same time, it remains unclear whether or not such data constitute strong support (as they are often taken to be) for the self-schema concept. While they certainly could arise from the operation of enduring and unitary cognitive structures, they might also be attributable to selection processes or strategies employed by different individuals. Perhaps depressed individuals have easier access to negative self-relevant information because it is already activated, or because it has been activated frequently enough in the past so as to become chronically accessible. That is, rather than depressed individuals necessarily having different cognitive structures than do non-depressed people, depressed mood may prime the selection of negative information in memory (e.g. Bower, 1981), or such selection may be well practised and semi-automated. Reasoning that existing methods cannot discriminate between the memory structure and accessibility hypotheses, Segal and colleagues (Segal, Hood, Shaw & Higgins, 1988; Segal & Vella, 1990; Segal et al., 1995) have utilised a priming method to assess the association among self-relevant concepts. Subjects first read a prime, followed by a related or unrelated target word to be colour named, and the differential interference with colour naming used as an index of prime-target association. It is important to note, however, that in this method, 'related' refers to whether or not prime and target are both self-descriptive, as judged by the subject, rather than normatively associated in the sense of having similar meaning. Both normal and depressed subjects showed greater interference when prime and target were related (Segal & Vella, 1990), although this Page 225
was often greater in depressed subjects (Segal et al. 1988), and particularly so for negative self-descriptive targets (Segal et al., 1995). Overall, despite some inconsistent results, the data suggest that negative self-descriptors are more effective primes for related targets in depressed than in normal subjects. The data thus suggest the existence of more structured negative self-knowledge in depressed than in nondepressed people. However, because the words were idiosyncratically chosen on the basis of self-descriptions, direct comparison across groups becomes problematic. To illustrate this problem, let us suppose that a depressed subject selects hopeless, useless and helpless as highly self-descriptive, while a normal selects aggressive, clumsy and forgetful. In this example, the depressed subject's selections are more stereotyped, due to their greater availability in memory, and the semantic associations among them are stronger as a result. Unless the inherent semantic links among the selected words is balanced across groups, the greater priming in depressives is difficult to interpret (e.g. hopeless might prime helpless equally in all subjects). If so, then the results might show only that the available selfknowledge in depressed subjects is confined to a small set of highly negative and interrelated concepts, without more definite conclusions being possible.
On the other hand, if the results discussed above are interpreted as indicating the existence of a well-integrated negative self-schema in depression, it might be expected that such a cognitive structure should persist in reliable form. Evidence on this issue is mixed, but overall it suggests a considerable degree of change over time. In a study of normal students, Mathews & Bradley (1983) found that the extent of self-related recall bias varied considerably across two occasions of testing that were separated by several months. In a similar but more extensive study, Hammen, Dyck & Miklowitz (1986) found that the extent of reported past depression contributed to current recall bias in depressed students, but that when their mood improved, all evidence of enhanced recall for negative words disappeared. Parallel results were obtained by Ross (1989). In recovered patients the evidence is also mixed. Bradley & Mathews (1988) found that fully recovered depressives no longer showed any negative self-related bias in recall, and in fact recalled more negative words that had been encoded in relation to other people. Rather different results, however, were obtained by Teasdale & Dent (1987), who compared previously depressed patients with never-depressed controls. The recovered group still showed a more general negative bias in recall, particularly after a sad mood had been induced. While thus indicating that retrieval operations may not be fully normalised following recovery, it is clear that the negative recall bias for self-encoded material in depression is Page 226
not very stable over time, and that the earlier idea that it reflects the operation of a unitary and consistent self-schema is not supported. One way of accommodating these results is to suggest that we have many possible selfschemata or self-guides (cf. Strauman, 1992; Niedenthal et al., 1992) some of which remain latent until activated by events or mood states. Alternatively, it is equally plausible to argue that events or emotional states serve to facilitate the recall of congruent self-related material, and this mechanism maintains depressed mood by selectively activating unpleasant episodic and semantic information in memory (Teasdale, 1983, 1988). Perhaps the most satisfying way of integrating both of these ideas with the present data is to assume that individuals differ in the extent and type of information about themselves in memory (without postulating a unitary and consistent self-schema) and that different types of self-knowledge are activated and used in different circumstances or mood states. If so, then the recall of self-encoded material may be a useful index of clinical change, but not necessarily of vulnerability to future depression. In contrast to the fairly consistent findings on the relationship between recall of self-encoded material and depression, similar experiments with anxious subjects have proved much less conclusive. Although test anxiety has been found to be associated with a variable recall deficit (cf. Mueller, 1980), self-encoded emotionally threatening material is not always recalled better than is neutral information by clinically anxious subjects compared with controls (Mogg, Mathews & Weinman, 1987; but see McNally, Foa & Donnell, 1989, for positive findings in panic disorder). After reviewing all studies of anxiety and memory, Eysenck & Mogg (1992) concluded that evidence of memory bias was often absent in clinical anxiety, perhaps because any schematic effects were countered by intentional avoidance of elaboration. If anxious subjects do indeed impose strategic control over elaborative encoding, this poses additional problems for using memory tasks such as recall as a test of the schema concept. An alternative method, discussed below, is to use incomplete or ambiguous text as a means of probing the schemata currently being used to comprehend
events. As will be seen later, this method does provide evidence for biases in comprehension, consistent with the operation of emotional schemata. Organisation of Self-Knowledge After reviewing the data concerned with the idea of a self-schemata, we have concluded that, although self-knowledge emerges as a very extensive and well-elaborated information base, we are not forced by this evidence to think of it as being structured as a consistent, generic and Page 227
modular schema. Rather, we would propose that self-knowledge can be selected and variously structured at different times, depending on current events, situations and moods. It remains quite possible, however, that individuals have distinctive modes of organising their self-knowledge, so that patterns do recur over time, and that such recurring patterns can influence the likelihood of emotional reactions such as depression. Conceivably, there are many 'selves' depending on current role, circumstances or mood (Markus & Nurius, 1986). As a result, we may encode information about ourselves quite differently in different situations, such as when we are at home or at work, encounter success or failure, and thus may be using quite different schemata or frames. If at work I am required to teach students, then presumably there will be slots in my work schema for lectures, students and so forth; while at home there will be other slots for recreation and housework tasks. Similarly, it is not impossible that other aspects of self-knowledge, such as personal attributes, are also differentially available in different emotional contexts. It does not seem inconceivable that one such 'self-schema' may be active at any one time, and then be replaced by another 'self-schema' at another time, depending on circumstances. In this sense the self-schema could be used to refer to the particular frame that we are currently using to extract information from the general database. This conception is related to the idea of mental models (Bower & Morrow, 1990), and would have no difficulty in accounting for apparent inconstancy in the structure and content of 'self-schemata' over time. Changes in negative recall bias with recovery from depression do not have to be attributed to the 'latent' status of a negative self-schema, but rather to a change in the choice of frames used to encode or retrieve information about oneself. Future research could thus be directed towards investigating the factors that influence frame selection, rather than the content of any particular schema. Schemata and Ambiguity In the previous chapter on judgement, we presented evidence showing that ambiguous words were often interpreted by anxious subjects as having the more threatening of two possible meanings. We now return to this general topic, but in the context of methods used to demonstrate the operation of schemata in text comprehension. In earlier self-report studies, emotional disorders have been found to be associated with an increased tendency to interpret descriptions of ambiguous events in a Page 228
negative manner. For example, Krantz & Hammen (1979) offered a series of social scenarios (e.g. your friends are invited to a party, but you are not) to depressed or nondepressed subjects, and provided them with a list of possible attributions. As expected, the depressed subjects tended to select the items containing negative attributions (e.g. you are not wanted).
In a related study by Butler & Mathews (1983), both anxious and depressed patients read a series of ambiguous scenarios (Suppose you wake with a start in the middle of the night thinking you heard a noise, but all is quiet. What do you suppose woke you up?) and then provided an open-ended response, before being asked to place possible explanatory alternatives in the order in which they would be most likely to come to mind. Threatening explanations (e.g. there is a burglar in your house) were ranked as more likely to come to mind by both anxious and depressed subjects than they were by normal controls. Related studies by McNally & Foa (1987) with agoraphobics, and both Clark et al. (1988) and Harvey et al. (1993) with panic disorder patients, revealed similar results. To overcome the objection that these studies were too dependent on self-report, Eysenck et al. (1991) used a recognition test for disambiguated forms of ambiguous sentences with threatening and non-threatening meanings, that had been heard previously (e.g. The doctor examined little Emma's growth). Anxious patients endorsed relatively more of the threatening disambiguations (e.g. The doctor looked at little Emma's cancer) than did control subjects, consistent with the hypothesis that they had encoded the ambiguous versions as having a threatening meaning. In a second experiment, a signal detection analysis confirmed that the difference between groups was due to sensitivity rather than response bias. None the less, these studies did not establish that only one meaning had been encoded, since subjects could have been aware of both interpretations and then chosen to endorse one of them at the time of testing. Strictly speaking, therefore, all of the above results could be attributed to intentional processes at the time of report, rather than an automatic schema-driven interpretation. In experimental studies of text comprehension a number of methods have been developed to establish the extent to which normal readers make inferences about meaning 'on-line', or as they read. Consider, for example, the following text (adapted from Graesser, Singer & Trabasso, 1994): 'The driver saw the child at the edge of the road. He braked hard, but it was too late.' Although the surface meaning of the text can be comprehended without making any inferences, readers may assume that the driver braked suddenly because the child ran out into the road, causing an accident, although these details are nowhere specifically stated. To assess Page 229
whether or not such inferences had been made, researchers may require readers to make a speeded lexical decision, such as whether 'accident' is a real word or not. Alternatively, reading time for continuations that are consistent or inconsistent with a possible inference can be used. For example, the continuation 'The driver felt very proud' might be understood more slowly than 'The driver felt very guilty'. In just such an experiment, Gernsbacher et al. (1992) found that readers were slower to read descriptions of incongruent than of congruent emotional states, and concluded that inferences are typically made about the feelings likely to be experienced by central characters, even when these are unstated. Using similar material to that reported by Eysenck et al. (1991), MacLeod & Cohen (1993) employed a reading time measure for threatening or non-threatening continuations of ambiguous sentences, to determine if the former results could be attributed to on-line inferences. To provide a baseline measure of responses to be expected following each type of interpretation, some conditions included a title that suggested the appropriate schema. Thus the ambiguous sentence 'They completed the service by filling in the hole' could be headed 'Burial' or 'Renovate', or by neither. Using differences in reading time for threatening or non-threatening continuations (e.g. The funeral/repairs finished much sooner
than expected), MacLeod & Cohen found less speeding for threatening continuations in anxious subjects when a benign title was presented, relative to the other conditions. Although this finding suggested that anxious subjects had arrived at a threatening interpretation, and non-anxious subjects at a benign interpretation, even in the absence of a title, the raw data presented a more complex pattern. Longest reading times in anxious subjects were for threatening continuations following the benign title (and vice-versa for non-anxious controls). Thus, initially cueing a non-threatening schema seemed to create particular problems for anxious subjects when a threatening inference is required later, and this was the main factor responsible for the pattern of difference scores observed. Fewer problems arose in the pattern of results found by Calvo, Eysenck & Estevaz (1994), using a speeded lexical decision probe following an ambiguous paragraph. Test anxious subjects under performance stress were faster to endorse a word matching a threatening than a non-threatening continuation, and slower to reject a non-word that resembled such a threatening word. Further work by Calvo and colleagues has confirmed the reliability of these observations, as have similar experiments by Hirsch (1995) using longer texts that included ambiguous sentences describing reactions to job interviews. Anxious subjects were consistently faster to endorse valid words matching threatening versus non-threatening interpretations, but only Page 230
when they were placed in the text at an ambiguous point where such inferences were possible. In summary, these experiments have provided objective confirmation that anxious (and perhaps depressed) subjects make mood-congruent inferences on-line, without any demand on them other than to read and understand ambiguous descriptions. In their investigations of inferences during reading, McKoon & Ratcliffe (1992) have concluded that inferences are rarely made unless the relevant information is easily accessible, or is required for text to be locally coherent. Emotional inferences may constitute a major exception to this rule, or perhaps emotionally relevant information is more easily accessible than most. In any event, comprehension measures may be superior in some respects to conventional memory paradigms for studying emotional schemata: they do not seem to depend on intentional encoding and retrieval strategies, and they allow the flexible use of naturalistic and complex text. The nature of on line measures is such as to tap the moment-to-moment construction of schematic models. Rather than encouraging the assumption that schemata are static immutable structures, such as a self-schema, they provide a means of assessing how the construction of schematic models differ among emotionally disordered groups, when encountering relevant events.
Page 231
Chapter 10 Nonconscious Processing This book has raised many issues concerning the relation between conscious and nonconscious processing. It is now time to address these directly. As discussed in Chapter 7, much theorising about the function of nonconscious processes assumes a simple isomorphic mapping with processes which are consciously accessible. However, if the relationship is more complex, it is difficult to assess whether, in changing conscious cognitive appraisals during therapy, one potentially is modifying an underlying nonconscious bias in information processing. A further reason for examining the way in which conscious and nonconscious mental processes relate to one another is to prepare the ground for building a framework within which the results of the experiments on emotional disorders can be integrated. We shall leave such attempts at integration until the final chapter, where we shall also discuss the role of underlying processing biases in the maintenance and remediation of emotional disturbance. First, however, we shall review some of the most relevant issues concerning the relationship between conscious and nonconscious processing. Historical Overview The conceptual distinction between conscious and nonconscious mental processes enjoyed a long and relatively uncontroversial history before the advent of experimental psychology in the nineteenth century. Pascal anticipated the psychoanalytical concept of the unconscious by three centuries when he wrote, 'the heart has its reasons, which reason knows not'. Similarly, the seventeenth-century English Platonist John Norris provided a prophetic description of a more cognitive account when he made the claim, 'there are infinitely more ideas impressed on our minds than we can possibly attend to or perceive'. Indeed, Whyte (1978) reviews excerpts from the writings of over 50 philosophers, between AD 130 and 1800, all of which demonstrate an appreciation of both conscious and nonconscious mental operations. The nineteenth-century experimenPage 232
tal psychologists, such as James and Ebbinghaus, were severely restricted by their exclusion of nonconscious processes from scientific investigation. Although Freud attempted to reintroduce a more detailed account of the unconscious, defining it both in terms of a hypothetical unique content (instincts, motivation and repressed memories) and distinctive processes (pleasure principle and defence mechanisms), the psychoanalytical approach proved too unreliable to qualify among experimental psychologists as a formal scientific method. The thorny issue of consciousness was subsequently simply set aside by the arrival of behaviourism in the mid-1920s (Watson, 1925), and for the next 35 years scientific progress within psychology was restricted to the rigorous study of how environmental factors influence and control behaviour. Ultimately, however, theoretical psychologists found the constraints of behaviourism too restrictive. In the late 1950s, by borrowing concepts from the newly established science of information technology, psychologists became capable of formulating cognitive models with sufficient precision to enable highly detailed and specific predictions concerning the effects of controlled manipulations upon behavioural measures such as reaction times, or forced choice discriminations. In this way mentalism was reintroduced to psychology without any reduction of scientific rigour, and the information-processing paradigm was established. As this approach developed, so the distinction between conscious and nonconscious processes gradually re-emerged. Early resistance to reinstate the concept of the unconscious has left a legacy of terminological
confusion, with labels such as 'pre-conscious', 'pre-attentive', 'pre-cognitive', and 'implicit' often being used in preference to the term 'unconscious' with its controversial Freudian connotations. Over the past two decades, there has been a major revival of interest in issues concerning the distinction between conscious and nonconscious information processing within the discipline of psychology (cf. Dennett, 1991; Kihlstrom, 1993; Block, 1995). Experimental psychologists have asked important questions concerning the extent of nonconscious processing. How fully is information processed in the absence of consciousness? Can our behaviour be influenced by past memories that we are unable to consciously retrieve? Can we learn rules that enable us to control our environment without ever becoming conscious of these rules? Do we comprehend the meaning of stimuli which never reach our awareness? Are the rules of nonconscious processing different from those of conscious processing? How does information enter consciousness, and why? Chapter 1 indicated how the traditional commitment to behaviourism shown by clinical psychologists has been weakened recently by an increased appreciation of how cognitive concepts may illuminate various Page 233
psychological disorders. In attempting to determine how patients construe their world, therefore, clinical psychologists have found themselves confronting similar questions concerning consciousness to those which have interested the experimentalists. Can nonconscious processes, such as biased assumptions and inferencing, affect our conscious appraisal of situations? Can such nonconscious operations be made available to consciousness and, if so, how can this best be achieved? Do certain psychiatric or neurological disorders reflect disruptions in the transfer of information from nonconscious levels into awareness? Given the theoretical and applied importance of such questions, it is not surprising that they have attracted a great deal of recent research attention. Evidence concerning the nature of nonconscious processing is now available from several areas of research. In evaluating this evidence, we consider it helpful to relate it to two alternative forms of postulate regarding the 'psychological unconscious' introduced by Shevrin & Dickman (1980). The weak form of this postulate states simply that nonconscious processes exist and actively affect conscious processes. The strong form adds that these nonconscious processes follow different laws of organisation from those which govern conscious processing. Having established the reality of the weak form, and argued that the balance of evidence supports the strong form of this postulate, we shall discuss the nature of the relationship between conscious and nonconscious aspects of the cognitive system, before briefly considering the clinical implications of this dichotomy. Evidence for Nonconscious Processing Verbal Report of Higher Order Cognitive Operations Few would argue that low-level processes, such as feature extraction, need be available to consciousness. However, it often is assumed that the higher order cognitive operations underlying choice, evaluation and judgement are represented within awareness and thus can be addressed through introspection. Frequently, clinical psychologists will ask patients questions which require them to report verbally on such higher order processes. What is the main thing which upsets you about your marriage? Why does this aspect of the situation exacerbate your anxiety? Asking such questions presupposes that higher order cognitive operations are mediated by, or at least are accessible to, conscious thought. However, many theorists (e.g. Miller, 1962; Neisser, 1967; J. Mandler, 1975; Nisbett & Wilson, 1977) have argued that such operations actually Page 234
proceed nonconsciously. In a classic article, Nisbett & Wilson (1977) provide an extensive review
of evidence that supports the existence of nonconscious higher order processing by demonstrating that participants often are entirely inaccurate when asked to report verbally on such cognitive processing. A few examples from the many offered by these writers should illustrate this point. Many of the studies reviewed by Nisbett & Wilson originally were designed to test predictions derived from either dissonance theory or attribution theory. The central idea behind dissonance theory is that behaviour which is intrinsically undesirable will, when performed for inadequate extrinsic reasons, be construed as more attractive than when it is performed for adequate extrinsic reasons. Thus, for example, participants receiving shock tend to experience this as more aversive when they are given adequate justification for the necessity of the shock procedure by the experimenter (Zimbardo et al., 1969). Closely related to dissonance theory is attribution theory, which postulates that a participant's attitudinal, emotional and behavioural responses to any event or experience will be influenced by the particular attributions which that participant makes concerning this event or experience. For example, Storms & Nisbett (1970) treated two groups of insomniacs with a placebo pill. One group of participants was told that this pill would produce increased heart rate, alertness and other symptoms of insomnia. The other group of participants was told that the pill would reduce arousal, lower heart rate, and generally facilitate sleep. As anticipated, following consumption of the pill, the former group's insomnia improved 28% whereas the latter group took 42% longer to get to sleep. The most common explanation for this effect is that when arousal symptoms were easily attributable to the pill they would have been interpreted as less threatening, making them easier to ignore and attenuating their disruptive influence on sleep. Nisbett & Wilson's central claim, when reviewing such areas of research, is that participants typically do not mention the influence of dissonance or attribution when asked to report verbally on the factors which contribute to observed behavioural changes. Indeed, they tend to produce plausible explanations which reveal no awareness of the critical influence exerted by the experimental manipulations. For example, the insomniac participants explained their improvement by reporting that previously worrying problems had been resolved, making them feel more relaxed. All denied having thought about the pills at all. An equivalent inability to verbally report any awareness of influential factors affecting a wide range of problem solving, appraisal and choice behaviours also is conveyed by Nisbett & Wilson's review. For example, when asked to assess the relative quality of four identical pairs of nylon stockings, Nisbett & Wilson's participants tended Page 235
strongly to select the stockings on the right-most side of the display, but all explained their choices by reference to the relative quality of the articles and denied that position had affected their decisions in any way. Nisbett & Wilson argue from these and similar kinds of data that conscious awareness is limited to the products of mental processes, and that these processes themselves are beyond the reach of introspection. Furthermore, they suggest that verbal reports are based not upon true introspection, but upon culturally supplied, a priori, causal theories about the likely cognitive operations involved. Such verbal reports will be accurate only to the extent that the actual processes which mediate a particular decision or behaviour might coincidentally correspond to those implicated by such a priori theories. According to this view, the anxious patient may willingly report, for example, that he or she becomes tense in social situations because he or she anticipates receiving criticism within such settings. However, this may reveal only the patient's a priori theory, rather than the actual cognitive basis of this emotional response to social stimuli. The many examples illustrating inaccurate verbal reporting of higher order cognitive processes described by Nisbett & Wilson provide evidence supporting at least the weak form of Shevrin & Dickson's postulate. That is, they suggest that nonconscious processing may indeed take place, and may exert an active influence on conscious processes such as judgement and decision making.
Given the importance of this conclusion, it perhaps is not surprising that, while Nisbett & Wilson's review received a great deal of attention and has served to catalyse intensive further research on this topic, it also has attracted a range of criticisms. For example, researchers such as Smith & Miller (1978) and White (1980) questioned whether the inaccuracy of verbal reports, especially those taken subsequent to the occurrence of a cognitive process, could serve adequately to exclude the possibility that this process was never available to consciousness. Specifically, they pointed out that, in all of the studies reviewed by Nisbett & Wilson, because the verbal reports temporally followed the occurrence of the cognition, it was possible that the inaccuracy of these reports may have reflected memory failures. In fact, Nisbett & Wilson's discussion acknowledges this possibility when they tell us: at this (immediate) point . . . subjects have some chance of accurately reporting that a particular stimulus was influential . . . at some later point, the existence of the stimulus may be forgotten, or become less available, and thus there would be little chance that it could be correctly identified as influential.
Additionally, it should be noted that participants' limited abilities to provide accurate verbal reports may stem not from the absence of any Page 236
conscious experience of the cognitive process, but rather may reflect a difficulty in translating such a conscious experience into a verbal code. This possibility can be illustrated with reference to everyday experience. For example, we all would admit to consciously experiencing, say, the taste of an apple. However, it is unlikely that we could fully convey the conscious experience of this taste verbally to another individual who has never eaten an apple. This failure does not imply that the taste was only experienced nonconsciously but, rather, that certain conscious experiences cannot be fully reported verbally. Thus failure to report verbally the nature of a process can never be sufficient to demonstrate a lack of any conscious awareness of that process. Nisbett & Wilson's critics also questioned the logical viability of the assertion that only the products of processing are available to consciousness, while the processing itself remains nonconscious (Smith & Miller, 1978; White, 1980). As no satisfactory definitions of mental product and mental process presently exist, there are no criteria whereby a mental event can be defined as one or the other. We run the risk of circularity if we decide to use consciousness as the criterion for making this distinction calling everything that gets into consciousness product and everything else process as no experimental predictions can be derived from such a viewpoint. Therefore, while Nisbett & Wilson's review suggests that nonconscious processing may occur, it is less informative about the particular classes of cognitive operations which most plausibly can proceed without awareness. Is it the case, for example, that people tend to retrieve, without awareness, memories of past experiences, which exert a nonconscious influence on their ongoing behaviours? Or can people actually acquire, without awareness, knowledge of complex regularities within their environments, enabling them to regulate their behaviour by following rules which they are not conscious of having learned? Or is it possible that people actually perceive stimuli, and process them to a deep conceptual level that influences their conscious experiences and behaviours, without ever experiencing any awareness that such perception has occurred? In summary, early demonstrations of participants' inability to accurately describe verbally the factors that influence their decision making are consistent with the weak form of Shevrin & Dickman's (1980) postulate, according to which nonconscious processes do occur, and actively affect conscious processing. Such studies, however, have provided little evidence for the stronger claim that such nonconscious processes follow different laws of organisation. Also, there are interpretive problems with these data, as inaccuracy of verbal reports may reflect memory failure or difficulty in conveying the nature of certain conscious mental events through the use of a verbal code. Finally, this approach offers little
Page 237
opportunity to delineate with precision the particular classes of cognitive operations that may proceed nonconsciously. Subsequent research on nonconscious processing more often has focused specifically upon the study of some distinctive class of cognitive operations. This has resulted in the development of extensive and largely independent literatures concerning nonconscious memory, nonconscious learning, and nonconscious perception. Although some methodological criticisms still remain within each of these research domains, the following brief reviews of these literatures offer substantial support not only for the weak form, but also for the strong form of Shevrin & Dickman's postulate. Non-conscious Memory Few would dispute the observation that our thoughts, feelings, decisions and behaviours often are influenced profoundly by our memories of past experiences. Perhaps it is because of this that the study of memory has remained a central preoccupation for experimental psychologists since the early days of our discipline. However, throughout the century that elapsed between Ebbinghaus's (1885) seminal efforts to decouple the memory system from other cognitive subsystems in order to study its operation in the most abstract form, and Baddeley's (1986) focus upon the important ways in which memory interacts with these other cognitive subsystems, memory research was almost exclusively concerned with the study of conscious recollection. Only across the past 15 years or so has there been a dramatic growth of interest in the hypothesis that memory also may function without awareness to retrieve past information which, without entering consciousness, can exert an effect upon our subjective experience and actions. This hypothetical form of retention and retrieval has popularly been termed 'implicit memory' (cf. Schacfer, 1987) and a tremendous amount of empirical support for its existence now resides within a large and diverse body of experimental literature (cf. Graf, 1994; Roediger & McDermott, 1993). Conscious recollection typically is revealed by accuracy on recall or recognition procedures, which require participants intentionally to retrieve and reflect upon past memories. In contrast, nonconscious memory is revealed when performance on a task is influenced by previous exposure to critical information, without this effect being mediated by the conscious recollection of this information (Graf & Schacter, 1985). It has, indeed, been found that prior exposure to stimulus information (usually, but not always, word sets) can significantly influence performance on a wide range of tasks which require the subsequent processing of these Page 238
same stimuli, but which do not require conscious recollection of their initial exposure. For example, prior word exposure decreases lexical decision latencies for repeated words (e.g. Scarborough, Gerard & Cortese, 1979), increases the accuracy with which such words can be identified under short or degraded exposure conditions (e.g., Jacoby & Dallas, 1981), increases the likelihood that incomplete word stems or fragments will be completed to yield those particular words (e.g. Graf, Mandler & Haden, 1982), sharpens their apparent perceptual contrast (e.g. Merikle & Reingold, 1991) and decreases the perceived volume of white noise within which these words are embedded (e.g. Jacoby et al., 1988). These findings, and other effects of a similar nature, are commonly assumed to reflect the operation of nonconscious memory. However, while such demonstrations do indicate that memory has impacted upon task performance, they do not constitute strong evidence that this influence has been independent from conscious recollection. Although lexical decisions, word stem completion, and the perceptual processing of word stimuli do not require conscious recollection of the initial exposure, it nevertheless remains possible that such conscious recollection may occur, and serve to mediate the effects of prior
exposure. Better evidence for nonconscious memory would be provided, therefore, by studies which reveal a strong effect of prior exposure, while also demonstrating that conscious recollection of this exposure is poor. Consequently, a number of experiments have simultaneously employed both 'direct' measures of memory, such as recall or recognition, to assess conscious retrieval, as well as 'indirect' measures of memory, such as word fragment completion, capable of revealing the impact of nonconscious retrieval. The rational underpinning these studies is that nonconscious memory retrieval can be assumed whenever the indirect measure shows a greater sensitivity to prior exposure than is shown by the direct measure (cf. Reingold & Merikle, 1988, 1990) Not uncommonly, researchers have indeed found that their indirect memory test has proved more sensitive to prior exposure than their direct memory test, suggesting that conscious recollection is unlikely to account adequately for the observed memory effects (e.g. Kunst-Wilson & Zajonc, 1980; Graf & Mandler, 1984; Mandler, Nakamura & van Zandt, 1987; Merikle & Reingold, 1991). For example, Kunst-Wilson & Zajonc found that participants initially exposed to 10 irregular geometric shapes tended subsequently to show an above-chance aesthetic preference for the exposed shapes relative to previously unseen shapes. While this preference judgement task (indirect memory test) led to participants selecting the previously exposed shape on 60% of the trials, a recognition memory task (direct memory test) led to participants selecting the previously Page 239
exposed shape on only 48% of trials. Thus, it appears unlikely that conscious recollection could account for the preference judgements, which Kunst-Wilson & Zajonc therefore attribute to the influence of nonconscious memory. A similar conclusion, supporting the existence of nonconscious memory, was drawn by Tulving, Schacter & Stark (1982) when they compared the effects of prior word exposure on a subsequent word fragment completion task (indirect memory test) and a cued recall task (direct memory test), and found that the significant impact of this prior exposure on the former task was stochastically independent from its impact on the latter task. Nevertheless, one might harbour reservations concerning the assumption that any comparison of performance on tests designed to measure conscious retrieval, and those designed to measure nonconscious memory, ever can yield entirely conclusive evidence in support of the latter hypothetical process. As Jacoby (1991) has argued, even if conscious and nonconscious memory do both exist, it is highly unlikely that any memory test will be 'process pure', in the sense that it will be influenced by only one of these two forms of memory. Not only is it possible that performance on indirect memory tests may be facilitated by conscious recollection (as, for example, when the conscious retrieval of a previously exposed word results in that word being used in a stem completion task) but also it is possible that performance on direct memory tests may be facilitated by nonconscious recollection (as, for example, when participants' guesses in a recall or recognition task are biased by nonconscious memory in ways that increase accuracy). Consistent with this position, Perruchet & Baveux (1989) have shown that performance on many conventional indirect tasks (such as word stem completion) is highly correlated with performance on conventional direct memory tasks (such as free recall), leading these researchers to conclude that such tasks are not process pure. Given the likelihood that most memory tasks could be partially sensitive to both conscious and nonconscious memory, it becomes difficult to interpret with confidence the theoretical implications of observed differences in the degrees to which any two tests are affected by prior stimulus exposure. Fortunately, however, there are other experimental approaches which offer strong converging support for the existence of nonconscious memory. One of the most ingenious, and powerful, methodological techniques developed to discriminate the operations of conscious and nonconscious memory is the process dissociation procedure introduced by Jacoby (cf. Jacoby, 1991; Jacoby & Kelley, 1992). Recognising that conscious and
nonconscious memory could both facilitate performance on conventional direct and indirect memory tasks, Jacoby adopted the approach of placing Page 240
the influence of conscious and nonconscious memory in direct opposition to one another. Within Jacoby's tasks, the nonconscious influence of memory for prior exposure would automatically produce an effect in one direction, whereas task instructions were designed to ensure that the conscious recollection of this prior exposure would elicit its effect in the opposite direction. For example, Jacoby, Woloshyn & Kelley (1991) exposed participants to a list of non-famous names, before requiring these same individuals to make fame judgements about each member of a subsequent set which, in addition to famous names, included non-famous names which had, or which had not, been presented in the initial list. It was reasoned that nonconscious memory would result in a sense of familiarity that would lead participants to falsely class as famous a disproportionate number of the previously exposed non-famous names. Of course, the possibility was recognised that conscious memory also could contribute to such a pattern of results if, for instance, the recollection that a name had been presented twice led participants to believe it more likely that this must refer to a famous individual. Therefore, half of the Jacoby and colleagues' participants were explicitly informed, before the judgement task, that all names presented in the previous list were non-famous, and were directly instructed to therefore classify as non-famous any names which they could recollect having seen within that list. Thus, while nonconscious memory would increase false fame judgements for previously exposed names, conscious memory of such previous exposure would decrease false fame judgements for those same names. The fact that Jacoby and colleagues observed a pronounced elevation of false fame judgement for those names given prior exposure led them to conclude that some memory of exposure, not accessible to consciousness, must have influenced performance. In its most refined form, Jacoby's process dissociation approach involves the comparison of performance under exclusion task instructions (designed to elicit opposing effects of conscious and nonconscious memory) and inclusion task instructions (designed such that conscious and nonconscious memory will produce effects in the same direction). A simple but powerful statistical equation is applied to these data to provide precise independent estimates of conscious and nonconscious memory. The growing popularity of this approach not only has led to many convincing demonstrations to support the existence of nonconscious memory, but also has enabled the more precise investigation of the factors that influence both conscious and nonconscious retrieval processes (cf. Jacoby, Ste-Marie & Toth, 1993; Toth, Reingold & Jacoby, 1994). Another impressive line of evidence against the proposal that undetected conscious memory might adequately account for effects that have been Page 241
attributed to the operation of a nonconscious memory system is provided by studies that have contrasted conscious and nonconscious memory within clinical amnesic patients. Such patients are classified on the basis of profound impairments in recollective experience, resulting from neurological conditions such as Korsakoff's syndrome, encephalitis, vascular disorders, head injury, or hypoxic episodes. Despite the fact that amnesic patients often show virtually no accuracy on tests requiring conscious access to memory, such as recall and recognition, they frequently have been found to display intact levels of nonconscious memory (cf. Schacter, McAndrews & Moscovitch, 1988). Preserved nonconscious memory in amnesics has now been observed on a wide variety of indirect memory tasks, including word fragment completion (e.g. Graf & Schacter, 1985), lexical decision (Moscovitch, 1982), perceptual identification (e.g. Jacoby & Dallas, 1981) and homophone interpretation (Jacoby & Witherspoon, 1982). Normal levels of nonconscious memory in amnesics also has been observed in studies that have employed Jacoby's process dissociation approach to more precisely assess nonconscious retrieval (e.g. Squire & McKee,
1992). Indeed, when reviewing this literature, Shinamura (1989,1993) has reported several dozen studies that have convincingly demonstrated this pattern of findings. The fact that tasks designed to assess nonconscious retrieval yield strong evidence of normal functioning within a population characterised primarily by the absence of the ability to consciously retrieve information from memory, renders it implausible to argue that performance on these tasks may actually reflect their sensitivity to conscious recollection. Instead, such findings appear to require the conclusion that memory retrieval can take place, in ways that exert an influence upon behaviour, without this retrieval being associated with any conscious recollective experience. This appraisal of recent research on memory provides strong support for Shevrin & Dickman's weak postulate that nonconscious processes exist and actively affect conscious processes. Furthermore, the observation that nonconscious memory is more robust to neurological insult than is the case for conscious memory, suggests possible support for Shevrin & Dickman's strong postulate that these nonconscious processes may follow different laws of organisation from those which govern conscious processing. Additional support for this strong postulate is provided by a wealth of studies that have demonstrated other qualitative dissociations between conscious and nonconscious memory functioning. For example, it appears that nonconscious memory may involve a more literal representation of the original stimulus than is the case for conscious memory. The former type of memory, but not the latter, is often substantially impaired by superficial alterations to stimuli between Page 242
presentation and test, such as changes to modality (e.g. Graf, Shimamura & Squire, 1985), or from pictorial to verbal representations (e.g. Weldon & Roediger, 1987), or manipulations of voice and intonation (e.g. Church & Schacter, 1994). Also, while conscious memory can be demonstrated for any type of stimulus materials, nonconscious memory often is not found for non-words (e.g. Rueckl, 1990; Musen & Squire, 1991) or for pictorial representations of impossible objects (e.g. Schacter, Cooper & Delaney, 1990; though see McKoon & Ratcliffe, 1995). This has led a number of researchers to suggest that, while conscious memory can result in the creation of new mental representations, nonconscious memory may operate primarily by changing the state of activation within existing representations (cf. Dorfman, 1994). Another difference between conscious and nonconscious memory is that the latter appears to be generally resistant to a variety of strategic influences which powerfully affect the former. For example, the adoption of encoding strategies which serve to increase depth of processing strongly facilitates conscious memory, but has no impact upon measures of nonconscious memory (Graf & Schacter, 1989; Roediger, Weldon, Staedler & Riegier, 1992). Likewise, encoding manipulations that influence participants' intention to learn powerfully affect conscious memory without having any effect on measures of nonconscious memory (Bowers & Schacter, 1990; Roediger et al., 1992). Post-encoding instructions to forget, or remember, previously exposed stimulus items also systematically modify performance on tests of conscious memory, without exerting any influence on tests of nonconscious memory (Paller, 1990; Basden, Basden & Gargano, 1993). In general, therefore, nonconscious memorial processes appears to follows laws of organisation that are less amenable to intentional modification than are those which govern conscious memorial processing. Additionally, it would appear that nonconscious memory is resistant to the disruptive influence commonly produced on tests of conscious memory by secondary tasks that consume cognitive capacity (Jacoby, Woloshyn & Kelley, 1989; Parkin, Reid & Russo, 1990). For example, using a version of the false fame experiment described earlier, Jacoby and colleagues required half their participants to complete a demanding simultaneous task, involving detection of three consecutive odd numbers within a stream of continuous digits, while being exposed to the initial list of names.
The addition of this capacity-consuming task increased the false fame effect, suggesting that it served to impair conscious memory for the initial list (whom participants had been told were not famous), without any equivalent decrease in the degree to which exposure produced the feelings of familiarity assumed to underpin false fame judgements. In a Page 243
more sophisticated partial replication of this experiment, Jennings & Jacoby (1993) employed the process dissociation approach to calculate independent indices of conscious and nonconscious memory, and found that the addition of a secondary task substantially lowered the conscious memory index without producing any decline in the index of nonconscious memory. Such findings suggest that conscious memory is sustained by cognitive processes which rely upon limited capacity resources, whereas nonconscious memory proceeds in a manner that makes no capacity demands upon the cognitive system. In summary, therefore, there appears to be compelling evidence to conclude not only that nonconscious memory processes do indeed occur, and exert an influence on conscious experience and behaviour, but also that nonconscious memory follows qualitatively different laws from those which govern conscious memorial processes. The picture that emerges is of a nonconscious memory system which, unlike conscious memory, operates in a manner that is largely free from capacity constraints, but which, relative to conscious memory, is inflexible and unresponsive to strategic manipulations. Thus, within the domain of memory, good support can be found for Shevrin & Dickman's strong postulate concerning the psychological unconscious. In all of the studies reviewed in this section, the information which participants appear able to remember without awareness has been represented in consciousness at some point. Thus, on the basis of these data alone, it would be possible to maintain that nonconscious processing might be restricted to retrieval operations. However, recent research within the domain of learning suggests that this conclusion would be unwarranted. It is to this research that we now turn our attention. Nonconscious Learning Consider the situation in which an individual is exposed to a task environment within which certain regularities reflect the operation of complex or highly abstract rules. In these circumstances, it is not unusual to find that task performance develops in ways that reveal the acquisition of knowledge concerning these rules, despite the apparent absence of any conscious awareness that anything has been learned. This phenomenon, which has been popularly termed 'implicit learning' (Reber, 1965), has now been demonstrated on a wide variety of experimental paradigms (cf. Reber, 1993; Berry, 1994; Seger, 1994). Page 244
Some of the earliest systematic studies of nonconscious learning made extensive use of an artificial grammar task introduced by Reber and his colleagues (cf. Reber, 1989). This procedure involves first exposing participants to a large number of letter strings each of which, unknown to the participant, has been generated by the application of a synthetic grammar (specifying a finite set of legitimate transitions between letters). Subsequent to this exposure, participants are informed of the fact that the strings reflected the application of a grammar, then are required to decide whether new letter strings do, or do not, represent legitimate applications of this same grammar. Reber's basic finding is that participants show an above chance ability to accurately classify the grammatical status of these new strings, despite the fact that they are quite unable to accurately report the rules of the grammar (e.g. Reber & Allen, 1978; Reber et al., 1980). Reber concludes that participants nonconsciously abstract the rules governing the structure encountered within the task environment. His claim that participants acquire abstract rule knowledge is further supported by his finding that, even when all the original letters in the grammar are substituted by
new letters, participants still demonstrate an above-chance ability to discriminate grammatical from non-grammatical strings (Reber, 1969). These effects appear to be robust, and have been replicated by other authors using a range of procedural variations (e.g. Mathews et al., 1989; Dienes, Broadbent & Berry, 1991; Brooks & Vokey, 1991). Using a rather different sequence learning task, Lewicki similarly has drawn the conclusion that people can nonconsciously learn and make use of rules governing environmental regularities (Lewicki, Czyzewska & Hoffman, 1987; Lewicki, Hill & Bizet, 1988). Lewicki measured participants' detection latencies for target stimuli which appeared in differing screen locations on a VDU. Unknown to participants, the particular location within which the target appeared on certain trials was governed by a complex algorithm that took account of target locations on the preceding six trials. Not only were participants unable to verbally report this algorithm when interrogated following the study, but an additional group of participants, informed of the algorithm's presence before being exposed to the sequences and offered a high financial inducement for learning it, also was incapable of subsequently reporting the sequencing rule. Nevertheless, participants showed substantial speeding on the target detection task as exposure to the sequence progressed. More importantly, when the sequencing algorithm was changed unexpectedly, following an extended period of such exposure, detection latencies for the previously predictable target increased sharply. Lewicki's conclusion from this observation was that the participants must have learned to predict the Page 245
location of the critical targets by nonconsciously abstracting the rule that governed the regularity within the presentation sequences. Substantial support for the existence of nonconscious learning also has been forthcoming from Berry & Broadbent's extensive work on the control of complex systems (Berry & Broadbent, 1984, 1988; Broadbent, 1989). In this research program, participants were exposed to computercontrolled simulations of complex systems such as a sugar production plant, or a social interaction, and were required to control these systems by establishing and maintaining some target state. Participants could influence the system only by inputting the values of certain variables, and the output of the simulation reflected the operation of a complex algorithm performed upon these input values. Berry & Broadbent report that, with substantial practice, participants do learn to control the system effectively, as is evidenced by their success in attaining and maintaining the desired goal states, but remain quite unable to verbally report knowledge of the rules that they must have learned to enable this accomplishment. Such findings strongly suggest that people can acquire and make use of the critical rule knowledge without this learning being mediated by, or represented within, conscious experience. Clearly, the above learning studies provide considerable support for at least the weak version of the Shevrin & Dickman (1980) postulate, by indicating that nonconscious learning processes exist and actively influence performance. Nevertheless, despite this accumulation of evidence, some researchers have expressed residual doubts over the veracity of nonconscious learning (cf. Shanks & St John, 1994). The major criticism most often levelled against demonstrations such as those reviewed above is simply that they may not have adequately assessed conscious knowledge, thus leaving open the possibility that learning in all cases was actually mediated by awareness. There are two slightly different versions of this criticism. One version involves the proposal that, when asked to verbally report their knowledge of the relevant rules, participants may adopt high response criteria that lead them to omit reports of accurate conscious hypotheses of which they are uncertain. This possibility has led some researchers, such as Brody (1989), to suggest that forced choice discrimination tasks, requiring participants to select the accurate rule description from among two or more alternatives, may be more appropriate than free verbal report for the assessment of conscious rule knowledge.
The second version of the criticism levelled against these studies concerns the possibility that the scoring methods applied to verbal reports may underestimate relevant conscious knowledge. Specifically, it has been argued that participants may consciously learn rules which differ from Page 246
those the experimenter has built into the task, but which nevertheless lead to above-chance (though imperfect) performance. Verbal reports of such 'correlated' rules will be scored as incorrect by the experimenter, leading erroneously to the conclusion that nonconscious knowledge of the true rules, rather than conscious knowledge of correlated rules, mediated performance accuracy. A number of observations lend some support to this possibility. For example, Perruchet, Gallego & Savy (1990), using a version of Lewicki's sequence learning task, drew the conclusion that participants' performance could be accounted for by conscious knowledge concerning only differing frequencies of targets within each location. Likewise, a number of researchers investigating Reber's artificial grammar task have observed that, despite their inability to report the actual rules of the grammar, participants sometimes report conscious knowledge of certain legitimate letter pairs or triplets, which could enable above-chance performance when classifying the grammatical status of new strings (Dulany, Carlson, & Dewey, 1984; Perruchet & Pacteau, 1990; Perruchet, Gallego, & Pacteau, 1992). Such expressions of scepticism concerning the veracity of nonconscious learning have contributed greatly to the development of methodological rigour. However, it has proved difficult to sustain the argument that the development of skilled performance on artificial grammar tasks, sequencing tasks and tasks concerning the control of complex systems, might actually reflect the results of conscious learning. Several classes of evidence militate strongly against this position. First, it often has been shown that specific explanations offered as alternatives to the nonconscious learning account ultimately prove incapable of accommodating the data. For example, Gomez & Schvaneveldt (1994) directly assessed the plausibility of the proposal that artificial grammar task performance might reflect conscious knowledge of legitimate letter pairs, by directly training some participants to identify such legitimate letter pairs. They found that participants who were provided with conscious knowledge of legitimate letter pairs performed in quite different ways from that of participants trained only through exposure to the letter strings generated by the grammar. Specifically, they showed less ability to subsequently discriminate grammatical from ungrammatical strings, and failed to display transfer to a new letter set, as was demonstrated by the group trained only through exposure. Thus Gomez & Schvaneveldt conclude that conscious knowledge of letter pair associations simply cannot account for artificial grammar learning. A second class of evidence, which also serves to undermine the proposal that undetected conscious learning might account for performance on tasks intended to assess nonconscious learning, comes from studies that Page 247
have examined the performance of individuals with neurological or psychiatric conditions known to seriously impair conscious learning. Typically, such cognitively impaired individuals have been found to exhibit no learning deficit on those particular tasks which normal participants can learn without apparent awareness. Abrams & Reber (1988), for example, found that psychiatric patients suffering from a variety of psychotic conditions which grossly interfered with their general problem-solving abilities displayed equivalent learning performance on an artificial grammar task to that shown by normal control participants. Indeed, using this same type of task, Knowlton and her colleagues have found that even patients suffering profound amnesia (resulting from conditions such as Korsakoff's syndrome, thalamic infarction or penetrative brain injury) demonstrate equivalent learning on artificial grammar tasks to that displayed by control
participants (Knowlton, Ramus & Squire, 1992; Knowlton & Squire, 1994). Likewise, it has been found that amnesics show unimpaired learning on complex control tasks similar to those employed by Broadbent (Squire & Frambach, 1990). Also, on sequence learning tasks similar to those used by Lewicki to demonstrate nonconscious learning, normal levels of learning have been observed in amnesics (Nissen, Willingham & Hartman, 1989), and in patients suffering from Altzheimer's disease (Knopman & Nissen, 1987). This common finding, that conditions which grossly impair consciously mediated learning do not affect performance on those tasks developed to assess nonconscious learning, refutes the claim that all learning observed on those tasks actually takes place consciously. Furthermore, it provides support not only for the weak form, but also for the strong form, of Shevrin & Dickman's postulate. Specifically, the demonstration that nonconscious learning is not impaired by factors which compromise conscious learning, suggests that those nonconscious processes may follow different laws of organisation from those which govern conscious processing. Considerable additional evidence of other qualitative dissociations between conscious and nonconscious learning can be found elsewhere in the literature, serving to further support Shevrin & Dickman's strong postulate, and to discredit the notion that supposed demonstrations of nonconscious learning represent only failures to adequately assess conscious learning. One such dissociation concerns the relationship between performance accuracy and confidence observed within these two forms of learning. Under normal learning conditions, which permit participants to explicitly report what they have learned, accuracy and confidence are strongly correlated across trials. Thus, those trials on which participants express a higher level of confidence in their decision tend to be associated with Page 248
higher levels of accuracy. In contrast, recent research has shown that this correlation between confidence and accuracy can be eliminated by inducing learning in a manner that renders it inaccessible to verbal report, for example by using artificial grammar tasks, sequence learning tasks, or tasks requiring the control of complex systems (Chan, 1992; Dienes et al., 1995). The elimination of this association has been interpreted as further evidence that these learning conditions prevent the development of 'metaknowledge'. That is, the presence of significant rule learning in the absence of any correlation between confidence and accuracy enables the conclusion that participants have come to know the relevant rule, without having come to know that they know this rule. Indeed, because it circumvents the potential criticisms that have been directed against other criteria used to define the presence of nonconscious learning, Chan has argued that presence of a zero correlation between accuracy and confidence should serve as the universal criterion for determining the occurrence of learning without awareness. A different type of dissociation that appears to distinguish conscious and nonconscious learning concerns the degree to which each type of learning varies as a function of individual differences. There is considerable evidence to support Reber's contention that, being an evolutionarily primitive form of knowledge acquisition, nonconscious learning should show reduced variance across participants, and heightened independence from common dimensions of individual difference. For example, age differences have no appreciable impact upon nonconscious learning ability. Children as young as 4 years old demonstrate equivalent learning to that demonstrated by 8 and 14 year olds on a modified form of Reber's artificial grammar task (Roter, 1985), while very elderly individuals show no impairment relative to young adults on sequence learning tasks of the type developed by Lewicki (Howard & Howard, 1992). Furthermore, despite having a very powerful impact upon many forms of conscious knowledge acquisition, individual differences in intelligence show no association with nonconscious learning performance. Reber, Walkenfeld & Hernstadt (1991) contrasted performance on the standard artificial grammar task against performance on a letter series-completion problem which also required learning of the rules
governing the structure of letter strings, but which strongly encouraged the development of conscious knowledge through the application of deliberate code-breaking strategies. Although exposure to both tasks produced substantial learning, far more pronounced individual differences in the level of such learning were observed on the conscious learning task than on the nonconscious learning task. Furthermore, on the conscious learning task these individual differences in learning performance were correlated strongly with Page 249
individual differences in IQ scores, while no such correlation was evidenced in the nonconscious learning task. Yet another dissociation between conscious and nonconscious learning concerns the degree to which each type of knowledge acquisition is vulnerable to interference from secondary tasks that consume cognitive capacity. While conscious learning tends to be severely impaired when available cognitive capacity is reduced by the addition of such a secondary, this is not the case for nonconscious learning. For example, using a sequence learning task which resulted in the acquisition of rule knowledge that participants could not verbally report, Cohen, Ivry & Keele (1990) found that the level of this learning was quite unaffected by the difficulty of a concurrent secondary task. This contrasts with the finding of Nissen & Bullemer (1987) who, using a sequence learning task which did result in consciously accessible rule knowledge, observed substantial learning decrements under a concurrent task condition. Hayes & Broadbent (1988) observed a similar dissociation within a single experiment, in which participants were to learn the same rule to control a computer simulation under one of two different conditions. In one condition, the relationship between participant input and computer output was rendered salient by having this rule operate upon the participant's most recent input, and this resulted in conscious learning that was available to verbal report. In the other condition, the same rule was rendered less salient by having it operate upon the participant's earlier input and, although this still resulted in learning, this knowledge appeared to be acquired nonconsciously as it could not be reported verbally. After an initial period of exposure to the simulation, Hayes & Broadbent changed the nature of the critical rule without warning, and assessed how many trials it took participants to regain control. For some participants, but not for others, this simulation task was performed in the presence of a secondary, capacity-consuming task involving letter generation. Hayes & Broadbent found that, for those participants performing the version of the simulation which resulted in conscious learning, the presence of the secondary task substantially impaired their ability to regain control of the system following this rule change. In contrast, those participants performing the version of the simulation that produced nonconscious learning recovered control of the system at least as fast in the presence of the secondary task as when no such secondary cognitive load was present. Such demonstrations strongly suggest that nonconscious learning proceeds in a manner that makes minimal demand on cognitive resources, while conscious learning is dependent upon the availability of central resources within the limited capacity system that mediates performance of other intentional cognitive operations. Page 250
In summary, therefore, it can be seen that a great deal of evidence not only supports the hypothesis that the nonconscious learning of complex environmental regularities can affect conscious experience and overt behaviour, but also suggests strongly that the rules governing nonconscious learning follow different laws of organisation from those which govern conscious learning. Thus the study of learning, like the study of memory, provides clear support for Shevrin and Dickman's strong postulate concerning the psychological unconscious. In both of these areas of research, the
actual stimuli which the cognitive system operates upon are, at some point in time, represented within consciousness (though, in the case of implicit learning, the rule is not so represented). Research in the area of perception, however, suggests that stimulus information which never enters consciousness can also be extensively processed, and can exert a powerful influence on subjective and behavioural measures. This literature now will be reviewed. Nonconscious Perception Early research on attentional control yielded findings suggesting that information presented outside awareness might nevertheless be processed nonconsciously. Typically, in these attentional studies, participants were required to ignore one source of information while attending to another. For example, in the dichotic listening procedure introduced by Cherry (1953), participants were instructed to attend towards and repeat aloud (or 'shadow') a message presented to one ear, while ignoring a message simultaneously presented to the other ear. Despite participants' common claims that such unattended messages are not consciously perceived, there is evidence that they do undergo semantic analysis. In a seminal dichotic listening study, for example, Treisman (1960) presented different prose to each ear and instructed participants to shadow one particular car. Unexpectedly, during the task, she switched the messages at the two ears. When this occurred participants typically switched to shadowing the previously unattended channel, indicated that this unattended channel must have received sufficient semantic analysis for participants to know that it provided a meaningful continuation for the message they had been shadowing originally. A similar conclusion is required to account for many other findings that have been obtained using divided attention tasks, such as the demonstration that interpretation of an attended message can be influenced by the semantic content of an ignored message (e.g. Lewis, 1970; Treisman, Squire & Green, 1974; Page 251
Underwood, 1976; Philpott & Wilding, 1979). For example, in a dichotic listening study, Mackay (1973) observed that participants were likely to interpret the shadowed sentence 'she sat by the bank' in quite different ways when the words 'river' or 'money' occurred simultaneously in the other ear. Indeed there is evidence that people not only semantically process unattended information, but also categorise it, relate it to past learning experiences, and respond with appropriate emotions. A number of studies (e.g. Corteen & Wood, 1972; Corteen & Dunn, 1974; von Wright, Anderson & Stenman, 1975; Forster & Govier, 1978; Govier & Pitts, 1982) have conditioned a GSR response to a certain class of stimulus word, such as city names, before presenting these words to the unattended ear within a dichotic listening task. Despite participants' claims to be unaware of the stimuli presented to this ear, they have been found to show GSR responses not only to the originally shocked words, but also to synonyms and to new words from the same semantic category. Evidence to suggest that individuals nonconsciously perceive, and semantically process, information presented outside the focus of attention also has been provided by studies on visual information processing. A number of researchers have reported that stimuli presented within the parafoveal region of the visual field (which extends between 1 and 5 degrees of visual angle from fixation point) cannot be accurately identified, but do undergo semantic processing. In a conceptual replication of Mackay's (1973) dichotic listening study, Bradshaw (1974) found that the meaning individuals assigned to a centrally fixated homograph was influenced by the semantic content of simultaneous parafoveally presented words that participants could not report perceiving. Other researchers have demonstrated that the processing of centrally fixated words is facilitated by the unreportable presence of semantically related words within the parafoveal region (Underwood, Whitefield & Winfield, 1982), and have observed that Stroop interference can be
produced by having participants colour name centrally fixated patches while incongruent colour names, which participants cannot report, are presented parafoveally (Kahneman & Chajczyk, 1983). Despite this volume of evidence, some researchers have severely criticised the capacity of attentional studies to permit strong conclusions concerning the occurrence of nonconscious perception (e.g. Merikle, 1982; Holender, 1986). Such critics point out that attentional paradigms require a high level of participant compliance to ensure that information remains unattended, and suggest that awareness of 'unattended' information may be more common than is reported. Momentary awareness of such information may occur due to attentional switching, and may be underestimated by the retrospective appraisal of conscious experience usually Page 252
employed in attentional experiments. Some attempts to check for momentary awareness by unexpectedly halting the experiment and immediately asking participants to report the unattended information, have provided no support for this argument (e.g. Bargh, 1982). However, using this same procedure, Glucksberg & Cohen (1970) found good evidence of memory for unattended material at delays of 5 seconds, and Klapp & Lee (1974) argue that such memory can be detected at delays of up to 10 seconds. When the probability of momentary awareness is reduced, then evidence for the semantic processing of unattended information is more difficult to obtain. Newstead & Dennis (1979) modified Mackay's (1973) procedure to make attentional switching less likely, and found it impossible to replicate Mackay's results, leading them to suggest that momentary awareness, arising through attentional switching, may have mediated the original effect. A more rigorous test for the occurrence of perception without awareness is provided by those studies which have precluded awareness by manipulating presentation procedures in ways that have prevented participants from consciously apprehending stimulus information even when this is given full attention. This hypothesis that faint or brief (subliminal) stimulation might influence conscious processing, while giving rise to no intervening perceptual experience, has been addressed frequently throughout the history of experimental psychology (Sidis, 1898; Stroh, Shaw & Washbourne, 1908; Coover, 1917; Williams, 1938; Spence & Holland, 1962; Gordon, 1967; Dixon, 1968, 1971, 1981; Marcel & Patterson, 1978; Marcel, 1983a, 1983b; Bornstein & Pittman, 1992). Many studies have reported evidence that, when stimuli are rendered undetectable by reducing their energy content, either through the faintness or brevity of presentation, they nevertheless may undergo cognitive processing. In one of the earliest such experiments Dunlap (1900) found that the Muller-Lyer illusion still occurs even when the 'arrows' which produce the illusion are too faint to be perceived consciously. Although Dunlap's results have not replicated consistently (Titchener & Pyle, 1907; Trimble & Eriksen, 1966), there now exist numerous studies which have demonstrated that the structural properties of undetectably brief tachistoscopic displays can affect conscious experience (Bressler, 1931; Smith & Henriksson, 1955; Farne, 1963; Worthington, 1964; Walker & Myer, 1978; Gellatly, 1980; Kunst-Wilson & Zajonc, 1980; Bornstein, 1989). There also is compelling evidence that undetectable stimuli can be processed not only structurally, but also semantically. For example, the semantic content of a subliminally presented message has been shown to affect conscious perception of expressionless faces (Smith, Spence & Klein, 1959; Allison, 1963; Somekh & Wilding, 1973; Henley, 1975; Page 253
Sackeim, Packer & Gur, 1977). Such faces are described as significantly more pleasant when their presentation is accompanied by the subliminal word 'happy' rather than 'angry'. It also has been reported that measures of imagery (Henley & Dixon, 1974; Mykel & Daves, 1979), projective test
responses (Goldstein & Barthol, 1960), dream content (Fisher, 1960), social appraisal (Bargh et al., 1986; Erdley & D'Agostino, 1988), and affective experience (Kemp-Wheeler & Hill, 1987) can be influenced by the semantics of stimuli rendered undetectable as a result of their brevity or low intensity. A somewhat different method of eliminating stimulus awareness involves the use of the 'backward pattern masking' procedure (cf. Neisser, 1963, 1967). Although this technique also involves brief presentations, the duration of these exposures is sufficient to permit conscious perception. However, the subsequent presentation of a second pattern-stimulus serves to 'mask' or obscure conscious experience of the first stimulus. The effectiveness of this pattern-masking procedure in eliminating conscious awareness of the preceding stimulus increases as the onset asynchrony between the stimulus and the mask decreases, with maximum effectiveness being achieved at stimulus onset asynchronies of between 20 and 100 ms (Kolers, 1962; Fehrer & Smith, 1962). The backward-masking phenomenon appears to operate centrally, as is evidenced by the fact that it will occur when the initial stimulus and the mask are each presented to separate eyes (Turvey, 1973). Furthermore, an extensive body of literature now suggests that this form of masking can eliminate awareness of a stimulus without preventing its semantic processing (e.g. Marcel, Katz & Smith, 1974; Marcel, 1980, 1983a; Allport, 1977). In an influential early study using this procedure, Marcel (1983a) presented participants with a stimulus field which either contained a single word or was blank. The effectiveness of the subsequent pattern mask was gradually enhanced by decreasing the asynchrony between the onset of the word and the mask stimuli. Participants were required to make three judgements: (a) Presence: did any word appear? (b) Graphic: which of two subsequently presented words had the more similar structural appearance to the masked word? (c) Meaning: which of two subsequently presented words had the more similar meaning to the masked word? The participants were instructed to guess an answer for (b) and (c), even if they did not believe that any word actually had appeared. As the asynchrony between stimulus and masking fields was reduced, participants first lost their ability to report accurately the presence or absence of a Page 254
word. While such simple detection judgements were at chance level, however, they could still accurately judge the graphic structure and the meaning of the word, though such judgements typically were experienced subjectively as guesses. A further decrease in the word-mask onset asynchrony reduced graphic judgements to chance level without eliminating accuracy on the semantic judgement task. This important finding, that forced choice discriminations concerning the identity of backward-masked words can remain accurate at stimulus onset asynchronies which eliminate accuracy on concurrent forced choice discriminations concerning the presence or absence of these words, has been replicated by subsequent research (e.g. Merikle & Reingold, 1990). Clearly, it represents powerful evidence that perceptual processing of stimuli need not be mediated by conscious awareness of their presence. A diverse range of measures sensitive to semantic processing have provided additional support for the claim that the elimination of stimulus awareness through backward masking does not prevent the occurrence of semantic processing (e.g. Marcel, 1983a; Hines et al., 1986). Thus, for example, Marcel (1983a) has reported finding that normal Stroop interference is obtained when participants name the colour of stimulus patches, within which incongruent colour names are presented under backward-masking conditions which render these words undetectable. The normal semantic
priming effect commonly observed on the lexical decision task, which involves a reduction in lexical decision latencies for target words preceded by semantically related prime words (Schvaneveldt & Meyer, 1973), also has been observed using backward-masked prime words that participants are unable to detect (e.g. Fowler, Wolford, Slade & Tassinary, 1981). Likewise, participants are speeded in making judgements about the affective valence of target words when these are preceded by undetectable masked words sharing the same emotional tone (Greenwald, Klinger & Liu, 1989), and the responses given by participants on a word stem completion task are biased following exposure to masked word stimuli that participants cannot report (Forster et al., 1990). Such findings have led many to conclude that the backward-masking paradigm provides strong evidence in support of nonconscious perception. Nevertheless, other researchers have argued that such apparent evidence in favour of nonconscious perception simply reflects inadequacies in the assessment of conscious perceptual experience (e.g. Merikle, 1982; Holender, 1986). Some of these criticisms concern methodological issues. Holender, for example, has expressed concern over the manner in which the stimulus mask onset asynchronies, deemed appropriate for the prevention of stimulus awareness, are often calculated for each participant within an initial block of calibration trials. Holender points out that if the Page 255
eyes are not fully dark-adapted during those early calibrating trials, then stimuli encountered later in the experimental session may come to be consciously perceived as dark-adaptation develops. Other criticisms, however, border more on the philosophical than the methodological, and concern the types of criteria which different researchers consider adequate to define the absence of awareness. Holender's position, which is perhaps the most extreme, is that evidence of ability to accurately perform any discrimination concerning any aspect of a stimulus should be taken as evidence of stimulus awareness, rather than evidence of perception without awareness. Thus Holender demands, as minimal evidence for nonconscious perception, the demonstration that semantic processing of a stimulus has taken place without the presence of accompanying evidence that any discriminations are possible concerning this stimulus. Not surprisingly, many have criticised this position, arguing that it virtually defines nonconscious perception out of existence, by classifying all types of evidence that possibly could demonstrate the occurrence of semantic processing as evidence of awareness. Preference has been shown for the more moderate position developed by Cheesman & Merikle (1985, 1986), who propose the introduction of a distinction between the subjective perception threshold, below which individuals experience no phenomenal awareness of stimuli, and the objective perceptual threshold, below which there is no evidence of any discriminative ability whatsoever. Cheesman & Merikle hold that no semantic processing can occur below the objective threshold, but suggest that satisfactory evidence for nonconscious perception can be provided by the clear demonstration that this objective threshold is lower than the subjective threshold. Although some researchers recently have claimed to have obtained evidence indicating semantic processing below the objective threshold, as defined by Cheesman & Merikle (e.g. Kemp-Wheeler & Hill, 1988; Snodgrass, Shevrin & Kopka, 1993; van Selst & Merikle, 1993), most have focused upon refining methodologies to exclude the possibility that all observed semantic processing of stimulus materials might take place above the subjective threshold. A powerful source of evidence against the claim that stimulus processing always is mediated by conscious perception, even though this may go unreported, is provided by experiments on individuals who, as a result of biological factors, show an absence of the relevant class of conscious perceptual experience. Sometimes, the elimination of such perceptual experience has resulted from brain damage, and Farah (1994) has reviewed extensive evidence indicating that perceptual processing often appears to proceed nonconsciously even when perceptual experience
has been destroyed through such neurological insult. For example, Page 256
despite suffering total phenomenal blindness within large regions of their visual field, usually due to the destruction of tissue within the visual cortex, many neurological patients continue to show the capacity accurately to make discriminative judgements concerning visual stimuli presented within these regions, a phenomenon that has been labelled 'blindsight' (Weiskrantz, 1986). Although such patients typically fail to reveal even the most restricted or rudimentary visual capacity under sophisticated physiological and ophthalmic assessment, and though they insist that their responses in the experimental tasks represent only random guesses, these patients often retain an ability to accurately 'guess' the location, orientation, motional direction, shape and size of stimuli, when they are required to respond on forced-choice judgement tasks (cf. Milner, 1992; Cowey & Stoerig, 1992; Weiskrantz, 1993). In other studies of a biological nature, the elimination of conscious perceptual experience has been induced temporarily within normal people through the administration of a general anaesthetic. These experiments also have yielded substantial evidence that nonconscious perceptual processing continues to occur under such conditions (cf. Andrade, 1995). When unconscious surgical patients have been presented auditorily with stimulus words, they later show an increased tendency to complete word frames to generate these items (Block et al., 1987) and to produce these items on subsequent category exemplar generation tasks (Roorda-Hrdlickova et al., 1990; Jelicic et al., 1992a) despite being entirely unable to discriminate presented from nonpresented words on forced-choice recognition tests. The presentation of names to these unconscious patients has also been shown to give rise subsequently to the false fame effect described earlier in this chapter (Jelicic et al., 1992b). Evidence from several studies that have presented full sentences to anaesthetised patients indicates that even stimulus materials with this level of complexity can undergo extensive nonconscious processing involving full syntactical and semantic analysis (e.g. Humphreys, Asbury & Millar, 1990; Jelicic et al., 1992b). The demonstrated presence of perceptual processing within individuals who have lost the capacity for conscious perceptual experience as a result of biological factors militates strongly against the argument that all perceptual processing takes place within awareness. Additional evidence against this position is also provided by those recent studies of perceptual processing within normal intact individuals which have employed more refined experimental methodologies to enable the precise separation of conscious and nonconscious contributants to perception. It has become common, for example, for researchers carefully to compare the relative magnitudes of 'indirect' effects, plausibly attributPage 257
able to nonconscious perception, and 'direct' effects, which clearly permit a contribution from conscious perception, in order more rigorously to assess the plausibility of the suggestion that conscious perceptual experience might actually underlie the supposedly nonconscious perceptual processing. Often such comparisons have yielded results that support the strong criteria for nonconscious processing provided by Reingold & Merikle (1988), by demonstrating effects of greater magnitude on those tests that plausibly assess nonconscious perception than is observed on those tests that clearly permit conscious perception to play a role (e.g. Bonanno & Stillings, 1986; Bornstein & D'Agostino, 1992). For example, some researchers who have manipulated exposure conditions to permit or restrict perceptual awareness have observed a strengthening of certain perceptual processing effects as the likelihood of awareness is reduced. Jacoby & Whitehouse (1989) found that, when they immediately preceded an item in a recognition memory test by the masked exposure of that same item (which participants could not report perceiving), then this served to elevate probability that
participants would make a false recognition response on this item, presumably because it is experienced as more familiar. However, no equivalent effects were observed by Jacoby & Whitehouse when these same primes were exposed for long enough to permit their conscious perception, making it improbable that the false recognition effects observed with masked presentations could have been mediated by conscious perception of the masked words. Similarly, when assessing the facilitation of lexical decision responses produced by the immediately prior exposure of related words under backward-masking conditions, Dagenbach, Carr & Wilhelmsen (1989) found that reductions in the prime-mask onset asynchronies which reduced ability verbally to report prime identity served simultaneously to increase the magnitude of the observed priming effects. Likewise, when measuring the degree to which masked primes produced facilitation on an association judgement task, Klinger & Greenwald (1995) found that the largest priming effects were observed in those participants who showed the least ability to discriminate prime identity. Such findings have led some researchers to go beyond the conservative conclusion that some small amount of perceptual processing may occur nonconsciously, and to argue instead that the perceptual processing taking place outside awareness may be more extensive than that which is accompanied by awareness. According to this position, consciousness may serve more to inhibit and constrain perceptual processing than to augment and enrich it. Page 258
The range and the sophistication of methodologies for studying nonconscious perception continue to increase. Some researchers have made more extensive use of advanced statistical methods. Thus, for example, Greenwald, Klinger & Schuh (1995) recently employed a statistical procedure involving the regression of 'direct' perceptual effects onto 'indirect' perceptual effects, to yield residuals that they present as evidence for nonconscious perception. Other researchers, such as Debner & Jacoby (1994), have developed more powerful experimental tasks employing Jacoby's process dissociation procedure, and have cleanly separated the influences of conscious and nonconscious perceptual processing in ways that yield unambiguous evidence for the occurrence of the latter. In the face of this large body of accumulated research, it now is abundantly clear that extensive perceptual processing does indeed occur without conscious awareness, and can exert an influence on experience and behaviour. Thus, as has been observed in research on memory and learning, the study of perception provides ample support for Shevrin & Dickman's weak postulate, that nonconscious processes exist and actively affect conscious processes. There also are indications within this literature to suggest that conscious and nonconscious perceptual processes may each be governed by different laws of organisation, as is proposed by the strong version of Shevrin & Dickman's postulate. Marcel (1980) has provided findings relevant to this issue, using a masked-priming paradigm. On the critical trials within this study, participants made lexical decisions on a target word, which followed two sequentially presented prime words. The prime word immediately preceding the target was a polysemous word, with one of its two meanings potentially sharing a semantic relationship with the target word. The first prime word, which preceded this polysemous prime, was related semantically to only one meaning of the polysemous word. Thus participants were sequentially presented with word sequences such as HAND-PALM-WRIST or HAND-PALMTREE, in both cases being required to make a speeded lexical decision response concerning the final letter string. Without pattern masking, when the word PALM was preceded by HAND it facilitated lexical decisions for WRIST but not for TREE. In contrast, under such an exposure condition, when the initial word was changed to an associate of the other meaning of PALM, such as OIL, then PALM facilitated lexical decisions for TREE but not for WRIST. Thus, when conscious perception of the polysemous word was permitted, only one of its meanings was
activated, and the selection of this meaning was constrained by prior context. However, when the word PALM was pattern masked in a manner that eliminated conscious perception of this prime, then its presence facilitated lexical decisions equally Page 259
for both WRIST and TREE, regardless of whether the initial prime was HAND or OIL. Thus, Marcel's data suggest that both representations of the polysemous prime words are simultaneously activated when these ambiguous stimuli are perceived nonconsciously, though only one of those representations remains active when conscious perception of the stimuli takes place. Marcel (1983b) argues from such data that nonconscious perceptual processing 'automatically redescribe sensory data into every representational form and to the highest level of description available to the organism'. According to Marcel, nonconscious perception is unconstrained by capacity limitations, and permits multiple representations of stimuli to be processed in parallel. In contrast, he argues, that within our limited capacity, serially organised, conscious perceptual system, inhibitory processes operate to suppress perceptual processing in ways that preserve activation only of those particular representations selected as relevant. Marcel's suggestion that consciousness serves to inhibit rather than to enrich perception is consistent with evidence and conclusions that since have appeared elsewhere within this literature, some of which already has been reviewed (e.g. Jacoby & Whitehouse, 1989; Dagenbach, Carr & Wilhelmsen, 1989; Klinger & Greenwald, 1995). Additionally, evidence has accumulated to support Marcel's proposal that nonconscious perception proceeds in a manner which, unlike conscious perception, is unconstrained by capacity limitations. Perhaps the most sophisticated demonstration of this apparent dissociation concerning the capacity requirements that govern conscious and nonconscious perception has been that recently provided by Debner & Jacoby (1994). To distinguish and independently assess conscious and nonconscious perception, these researchers employed a formal process dissociation procedure paralleling the method employed previously by Jacoby to discriminate the influence of conscious and nonconscious memory (e.g. Jacoby, 1991). The degree to which conscious and nonconscious perception of briefly exposed primes served to influence a target word stem completion task, was revealed by contrasting task performance under two critical conditions. In the inclusion condition, participants were instructed to complete each stem to make the masked word that had just been briefly exposed, while in the exclusion condition they were instructed to complete each stem to make any word other than the masked word that had just been briefly exposed. The comparison of stem completions under inclusion and exclusion conditions enabled precise calculation of independent indices to represent the separate contributions of conscious and nonconscious perceptual processes to task performance. The significant values obtained on both of those indices revealed that Page 260
conscious and nonconscious perception of primes both occurred to some extent across the duration of the experiment. Of particular relevance to our current considerations, Debner & Jacoby required that, on half of the trials, the experimental task be performed under divided attention conditions, thereby reducing the extent to which limited capacity central resources could be devoted to the perceptual processing of the prime stimuli. This manipulation served to reduce substantially the value of the index representing the extent of conscious perceptual processing, without modifying in any way the value of the index representing the extent of nonconscious perceptual processing. Thus, consistent with Marcel's early claim, it does indeed appear that conscious perception reflects the operation of a limited capacity system, while nonconscious perception proceeds in a manner that is unaffected by capacity limitations.
In conclusion, therefore, the evidence concerning perception, like that concerning learning and memory, is consistent with Shevrin & Dickman's (1980) strong postulate concerning the psychological unconscious, which holds that fundamentally different rules govern conscious and nonconscious mental operations. Not only is it clear that extensive higher order information processing takes place without awareness, including the perception of new stimulus information, the retrieval of past memories, and the learning of complex rule knowledge concerning environmental regularities. Also, it appears that nonconscious cognition is insensitive to strategic manipulation, and involves extensive parallel processing, taking place in a manner that is unconstrained by apparent capacity limitations. In contrast, the weight of evidence suggests that conscious cognition, while being more flexible in the sense that it is open to strategic modification, may restrict and inhibit certain aspects of information processing, perhaps as a result of the capacity limitations which appear characteristic of those cognitive processes that operate consciously. Six decades ago our psychoanalytically oriented predecessors wrestled with the problem of formulating a credible account of the unconscious. Paradoxically, perhaps, having gathered such convincing evidence in recent years to support the existence of extensive and elaborate nonconscious information processing, contemporary psychologists now are faced with precisely the reverse problem. A major challenge confronting modern psychology is the need to develop an adequate account of the nature and function of consciousness, which is capable of explaining both how and why particular aspects of information processing results in a conscious experience. The next section examines some of the alternative models that have been developed in response to this challenge. Page 261
Accounting for Consciousness While the evidence reviewed so far within this chapter strongly suggests that a substantial amount of processing may proceed nonconsciously, it nevertheless remains clear that we all do experience conscious thoughts. Clinical psychologists often are faced with the need to evaluate the significance of such conscious experiences, and may perhaps attempt to modify certain clinical disorders through techniques aimed at directly changing such patterns of thinking. Consequently, it is important to know which aspects of information processing correspond to conscious experience, and to understand how conscious representation may be achieved. Perhaps it is not surprising, therefore, that many researchers recently have turned their attention directly towards the challenge of accounting for consciousness (e.g. Baars, 1988; Marcel & Bisiach, 1988; Dennett, 1991; Searle, 1992; Natsoulas, 1994; Block, 1995). One common proposal has been that, despite the extensive nonconscious information processing that takes place, certain types of cognitive operation only can occur in the presence of accompanying awareness. Often, for example, it is argued that consciousness is required to permit performance of novel tasks, especially those tasks that are neither stimulus driven nor biologically prepared (e.g. Umilta, 1988). In a more specific version of this position, G. Mandler (1975) claims that relational mental operations such as comparison, serial ordering and conceptual grouping, cannot occur without awareness, because the simultaneous juxtaposition of two or more 'mental contents' can only take place within consciousness. Through its unique capacity to sustain relational processing, Mandler proposes that consciousness permits the choice and selection of action systems on the basis of a consideration afforded their likely relative merit. It is this capacity, Mandler suggests, that enables long-range plans to be interrogated and modified, and novel processing strategies to be formulated. Such a position is not dissimilar to that put forward by Baars (1988), who construes consciousness as necessary to permit adaptation to novel inputs,
to make possible the 'debugging' of faulty processes and, most generally, to enable the establishment and maintenance of a dominant goal hierarchy through prioritisation of options and executive decision making. Other theorists have echoed and extended such claims, often adding the proposal that consciousness is essential to support 'metacognition', which involves the acquisition of knowledge about one's cognitive processes themselves (Marcel, 1988). Without the metacognitive ability provided by consciousness, it has been argued, it would be impossible to run the internal simulations of cognition that not only inform our personal decision making, but that also Page 262
underpin our capacity to understand and appreciate other individuals' mental states (Leslie, 1987; Oatley, 1988). Researchers have attempted to model consciousness in rather different ways. One approach has been to construe consciousness as the experiential correlate of a particular state that can apply to cognitive structures. Adopting this approach, G. Mandler (1975) has argued that cognitive structures may, in certain circumstances, become conscious though, when they do not, they continue to function nonconsciously. According to this position, certain operations of the types described earlier, such as choice, reappraisal and comparison, can only be performed by cognitive structures in the conscious state. Mandler attributes the limited capacity of consciousness to the constraint that only a restricted number of structures may be maintained in this state at any point in time. However, he provides little indication of the factors that will determine which processing structures become conscious, attributing this to specific 'organism-environment interactions', and the precise mechanisms through which such selection occurs remains unclear in Mandler's model. A somewhat similar, but rather more detailed, position has been developed by Shallice (1972, 1978) who also proposes that processing structures vary in terms of their activation, and argues that it is this energy level which determines whether a processing structure attains conscious status. Specifically, Shallice suggests that the structures capable of becoming conscious are a set of action systems, and he distinguishes these from perceptual, motivational and effector systems. According to this account, the brain contains a vast number of such action systems, each of which monitors its input in search of a pre-set target configuration. Detection of this target configuration triggers the operation of the action system and results in an output either to another action system or to effector units. The speed and accuracy with which an action system operates depends on its level of activation which can vary continuously, though simple or highly learned action systems require only minimum activation for normal functioning. Level of activation will be influenced by the inputs that every action system receives from perceptual and motivational systems. However, one important feature of Shallice's cybernetic model is that each action system also shares inhibitory connections with every other action system, with the amount of inhibition increasing monotonically with elevations in the activation of these other systems. In consequence, only one action system can become strongly active, or dominant, at any given time. It is this dominant action system (or, more precisely, the inputs to this system) which correspond to the contents of consciousness within Shallice's theoretical account. Though only one system can become dominant, and hence conscious at any Page 263
time, simple action systems and highly learned action systems that are not dominant still will continue to function at low levels of activity, and it is the operation of these nondominant action systems that corresponds to nonconscious processing. Because accounts like those proposed by G. Mandler (1975) and Shallice (1972, 1978) associate consciousness only with the state of a processing structure, they do not predict that conscious and nonconscious processes will follow different laws of organisation. Within Shallice's model, for
example, the operating rules for any action system remain unchanged whether it becomes dominant or not; only its speed and accuracy are affected by its attaining conscious status. Theorists wishing to accommodate observed qualitative dissociations' between the rules of conscious and nonconscious processing have been inclined to view these as different components of the cognitive systems, rather than as different states that can obtain within the same component of this system. Perhaps the simplest such approach is that which sometimes has resulted from a stages-of-processing framework, and has involved labelling one particular, usually late, stage of processing as 'conscious identification' or 'phenomenal awareness'. Erdelyi (1974), for example, locates such a stage 'in or near the short-term storage system, beyond the encoding system but prior to long-term memory'. Indeed, there was considerable consensus for such a view among theorists at one time (e.g. Sperling, 1967; Haber & Hershenson, 1973). The basic idea is that information flows through a linear series of processing stages, often incorporating feedback mechanisms reflecting the effects of attention or motivation which may modify signal strength. Any information that reaches the stage of conscious identification becomes represented in awareness, prior to proceeding to the final stages which involve highlevel analysis and storage in long-term memory (LTM). Whether or not information enters consciousness, therefore, depends upon the likelihood that it maintains sufficient signal strength to reach this processing stage. One advantage of the stages approach is that, because each stage can involve the operation of unique processes, it is straightforward to model qualitative differences in the rules that govern conscious and nonconscious processing. However, simple linear stage models have been rather discredited by a range of evidence. Indeed, some of the most compelling evidence against this account comes from demonstrations, many of which have been reviewed earlier in this chapter, that stimulus information can be processed to a high level, and may enter LTM, without ever passing through a stage where it becomes represented within awareness. Consequently, stage accounts of consciousness have largely been supplanted by alternative conceptions which construe conscious experience Page 264
as corresponding to the functioning of some particular level within a hierarchically arranged cognitive system. In more recent developments of Shallice's model, for example, he has added a higher control level, labelled the Supervisory Attentional System (SAS), which makes a powerful input to the selection of the dominant action system (Norman & Shallice, 1986; Shallice, 1988, 1994; Shallice & Burgess, 1993). Without suggesting a perfect correspondence between the SAS and conscious experience, Shallice now posits that the functioning of the SAS will be incorporated within such experience. A stronger position has been adopted by JohnsonLaird (1988), who also construes the cognitive system as a collection of semi-autonomous processing modules, operating in parallel but organised within a tightly structured control hierarchy. Although lower levels within this hierarchy usually proceed without higher level control, especially when performing well-learned or biologically prepared cognitive operations, higher levels within the hierarchy can intervene to influence processing at lower levels when necessary, for example to overcome unusual problems that might arise or to modify processing to deal with novel variants of previously encountered situations. In order to enable such control, Johnson-Laird points out that each level within this hierarchy must contain a representational model of its subordinate levels, which can be employed to run simulations of functioning within those lower levels. The highest level within the cognitive hierarchy, therefore, must contain a comprehensive model of the whole system. Indeed, JohnsonLaird proposes that, at this highest level, the system must have a recursive model of itself. This means that it must contain not only a representation of itself as a system, but also a representation of itself as a system that contains a representation of itself, and even a
representation of itself as a system that knows it contains a representation of itself (and so on, until the system's limited representational capacity is exhausted). Conscious experience, according to Johnson-Laird, corresponds to the operation of this highest level system, and reflects the experiential consequences of performing simulations using recursive models of the self, as is necessary to enable effective control of the cognitive system. Falling between stage theories and level theories are other accounts which also postulate a direct correspondence between consciousness and some discrete component within the cognitive system, but which locate this component neither at the top of a control hierarchy nor within a compulsory sequence of operations that inevitably precedes long-term storage. Such a view is taken by Schacter (1989), who proposes the existence of a specific Conscious Awareness System (CAS) that performs no executive functions itself, and serves only as a gateway to the executive system. Schacter, like Johnson-Laird, holds that most information processing is Page 265
performed by semi-autonomous cognitive modules, which operate without awareness. The output from these modules may provide inputs to other modules; which influences their functioning (and so affecting experience and behaviour) and may result in long-term memory representations, without any of this being accompanied by conscious experience. The likelihood of this is greatest for modules that perform well-learned cognitive activities, especially of a procedural nature. According to Schacter, however, some modules have outputs that enter the CAS, and conscious experience is the global awareness of the composite output from these processing modules. Schacter conceives of the executive system as functionally independent of, but receiving input from, the CAS. He recruits evidence for this partial dissociation between consciousness and executive control from a variety of neurological findings, which also lead him to localise the CAS within a particular posterior region of the cortex critically involving the inferior parietal lobes. Damage to this region can result in dramatic disorders of awareness, such as anosognosia (denying illness), hemianopia, aphasia or unilateral neglect, without producing any general disruption of higher order executive functioning (cf. McGlynn & Schacter, 1989). A somewhat different account, developed by Baars (1988), also construes consciousness as a specific cognitive subsystem which is neither an inevitable stage in a linear-processing sequence nor a high-level executive control mechanism. In keeping with the views of both Johnson Laird and Schacter, Baars conceives of the cognitive system as a distributed collection of autonomous processing modules, but unlike either of these two theorists Baars sees no need to postulate the presence of a central executive. Rather, he suggests that control is decentralised, emerging as a consequence of the local 'dialogues' that take place among the processing modules. Within such a system, Baars conceptualises consciousness as the global workspace to which all processing modules have access. According to this model, consciousness serves the function of a 'bulletin board', to which are posted processing tasks that cannot be handled locally by one or more local expert modules. The whole constellation of processing modules then works co-operatively on problems submitted to this workspace, in order to develop appropriate processing algorithms for their execution. Once such algorithms are established, it then becomes possible for more local expert modules to handle similar processing tasks without further submission of these to the bulletin board. Thus, within Baars model, consciousness embodies no intrinsic executive function. However, without the ability to transfer information into consciousness, in order to enable modules to co-operate in the development of new processing algorithms, the system proposed by Page 266
Baars would be incapable of performing a great many cognitive tasks, especially those novel to the system.
From this brief review, it can be seen that many models incorporate the proposal that information can be transferred from nonconscious processes into conscious awareness. Few of these models, however, provide a detailed account of how such transfer might actually take place. In contrast, Marcel (1983b) has focused directly upon this particular issue, arguing that conscious experience requires a constructive act, whereby distinctive processes operate to retrieve and synthesise the output of nonconscious systems, thereby transforming this information into a qualitatively different representational code. Marcel conceptualises nonconscious processing as extensive and elaborate, involving the analysis, transformation and redescription of stimulus information into the highest level of every possible representational form available to the organism. He considers such processing not to be bound by capacity, and whenever there is ambiguity of any type all possible parsings and interpretations will be carried out and represented at this nonconscious level. According to this account, therefore, nonconscious analysis leads to multiple representations at every level and in every available code, and such representations are not segmented in any wayfor example, into events, objects or episodes. Phenomenal experience, Marcel proposes, results from the imposition of a particular segmentation, structure and interpretation on the records of this nonconscious processing. To clarify how this occurs, Marcel invokes the concept of 'perceptual hypotheses' which reside in LTM and mediate access to consciousness. Like Minsky's (1975) concept of 'frames', or Rumelhart's (1978) notion of 'schemata', perceptual hypotheses are stereotypical representations corresponding to complex structural descriptions of world knowledge. Together they form an exhaustive set of canonical representations for all things we are capable of consciously perceiving. A perceptual hypothesis is conceived of as a structural description which specifies its criterial features, non-criterial features, and their relationship. The values of those features, however, are left open to particular instantiations which vary from occasion to occasion. A fuller discussion of similar hypothetical structures can be found in Chapter 9 of this text. According to Marcel's formulation, perceptual hypotheses are activated automatically if any of their criterial features can be detected among the immediate records of nonconscious processing. In practice, multiple hypotheses will be simultaneously activated and tested in parallel against the records of nonconscious processing. Which hypothesis will be selected depends partly on factors such as expectancy and frequency. However, the chosen hypothesis usually will be that which accounts for most data in the records, and Page 267
does so at the highest possible level of description. At any moment, what we are aware of is that hypothesis best instantiated by the fit of data in the relevant records. Thus, in Marcel's model, data are recovered from the record of nonconscious processing, and a subset of these data is synthesised into a unitary model by being fitted into the highest level perceptual hypothesis which can accommodate the greatest proportion of the record. We then become conscious of this instantiated hypothesis. By this account, consciousness depends upon two independent factors. First, nonconscious operations must leave adequate records for a sufficient period to enable hypothesis-matching to take place. Reducing stimulus intensity or duration (e.g. Henley, 1975; Sackeim, Packer & Gur, 1977) may result in inadequate records, whereas backward masking (e.g. Allport, 1977; Marcel, 1980, 1983a) may over-write such records, and therefore curtail their duration below that is necessary for hypothesis matching. Second, the range of perceptual hypotheses actually stored in LTM will impose limits on the potential scope of awareness. Conscious representations of events or episodes can only be accurate to the extent that appropriate hypotheses exist in LTM. It is interesting to note that many of the studies cited in Nisbett & Wilson's (1977) review involved specifically those situations where no schemata or hypotheses would have been likely to exist in LTM that could accurately have accounted for the records of the nonconscious processing which likely took
place. Introspection may have been inaccurate in these tasks, not because the records of the nonconscious operations involved were unrecoverable, but rather because they could not be synthesised into an accurate pre-existing schema. Nisbett & Wilson argue that participants supply introspective reports 'not by consulting a memory of the mediating process, but by applying a priori causal theories'. However, Marcel's model permits the existence of both a priori hypotheses and access to records of nonconscious processing. The a priori causal theories apparent in the introspective reports of Nisbett & Wilson's participants may, in fact, correspond to the highest level existing perceptual hypothesis which could accommodate the greatest amount of data from the recovered records. It is evident that our understanding of the relationship between conscious and nonconscious processing has benefited greatly in recent years from the careful consideration afforded this topic by many astute cognitive researchers. Such theoretical progress bears upon a range of questions with possibly substantial clinical importance. Can patients be trained consciously to access cognitive processes previously beyond the reach of introspection? Could idiosyncrasies in nonconscious processing underlie certain clinical conditions? Can cognitive manipulations which Page 268
are themselves not accessible to conscious awareness affect nonconscious processes in ways that have therapeutic benefits? In the final section of this chapter we briefly examine consciousness from a clinical perspective, and illustrate how the reviewed research can aid our ability to understand, and further investigate, matters of considerable clinical significance. Clinical Implications A central goal of cognitive therapy involves enabling patients to access their 'automatic thinking' in order to recognise its erroneous nature and substitute more rational, strategically controlled lines of processing where appropriate. Beck et al. (1985) describe automatic thoughts as interpretations, inferences or predictions which occur so rapidly that the patient is often aware only of the emotion they generate. The implication, therefore, is that the interpretations or predictions themselves are often not spontaneously reportable. However, Beck does not clearly address the important issue of why such automatic thoughts may be difficult to report. One possibility is that they may occur with only momentary awareness, perhaps because they are not fully attended to, and therefore produce a relatively weak memory trace. Since such thoughts could otherwise be qualitatively similar to phenomenal mental events, they would be amenable to conscious representation. If so, then straightforward introspection could be a useful technique to access automatic thoughts and, indeed, most therapists do rely largely upon this approach. Patients are encouraged to search for answers to questions like: 'What did you think your wife was really saying when she told you to wash uphow did you interpret it deep down, automatically?', or 'You felt panicky before the speech; were you perhaps having the automatic thought that you would mess it up?' If the difficulty in reporting automatic thoughts does in fact stem from memory failure, then another obvious implication is that introspective access will be most successful if attempted immediately, while the thought is taking place, as indexed by the occurrence of an emotional response. An alternative possibility, however, is that a thought may be automatic not because it has only briefly been represented consciously and hence is poorly remembered, but because it actually has occurred completely outside of awareness. If so, then theorists would disagree over the issue of whether or not such a thought could ever be introspectively accessed. Nisbett & Wilson possibly would argue that the product of automatic thinking may be accessed but not the underlying processes generating Page 269
this product. While, as we have already argued, the distinction between product and process may be difficult to maintain, the implication is that conscious access may be limited to certain aspects of automatic thinking. For those theorists who conceptualise consciousness as a state imposed on processing structures, which nevertheless function in a qualitatively similar manner whether conscious or not, then processes which normally occur automatically may possibly be rendered conscious. For example, Shallice may argue that certain automatic thoughts occur outside awareness in particular situations because other cognitive action systems are more dominant at this time. An automatic thought could, therefore, become conscious if the relevant action system underlying this thought were to become dominant. This rationale may suggest certain techniques to enhance access to such automatic thinking. For example, during an episode of emotional distress, a patient may access automatic thoughts more easily by suppressing competing action systems through cessation of irrelevant activity and temporary abandonment of concurrent goals. In this way the possibility of the appropriate action system, underlying the automatic thought, becoming dominant, and therefore consciously represented, would be increased. While other theoretical positions, such as the CAS account developed by Schacter, also may permit the possibility of initially automatic thoughts ultimately attaining representation within consciousness, this is not the case for all accounts of the association between conscious and nonconscious processing. As we have seen, many theorists consider nonconscious processes to differ qualitatively from those which can be represented consciously (e.g. Marcel, 1983b; Johnson-Laird, 1988), with a common view being that the former operate in parallel and are unconstrained by capacity while the latter are capacity limited and serial in nature. For such theorists, if automatic thoughts represent nonconscious processes, then they can never truly be represented consciously. At best a therapist could hope only to encourage a conscious experience which may give only some indication of what might be occurring nonconsciously. For example if, as Johnson-Laird's model proposes, conscious experience primarily represents a high-level simulation of the types of processing which would be possible at lower levels, then the clinician might best hope only that nonconscious thoughts will be accurately simulated within the modelling carried out in consciousness (a position similar to that taken by Nisbett & Wilson). Likewise, Marcel's conception of consciousness as instantiated perceptual hypotheses, constructed by synthesising specific records left by nonconscious processing, implies that nonconscious processes cannot themselves be directly experienced. According to this position, nonconscious Page 270
processes can only be reflected within conscious experience to the extent that the records of such nonconscious processes instantiate available perceptual hypotheses. Thus, to identify consciously an automatic inference, prediction or interpretation, Marcel's model requires that one must have an appropriate 'blueprint' for such a mental event stored in LTM. This suggests that one productive use of therapy time may be to supply patients with, and encourage them to actively speculate about, the hypothetical automatic thoughts which feasibly could have triggered critical negative emotions in the past. Though Marcel's account does not predict that this will permit accurate retrospective access to these actual thoughts, it does suggest that this procedure may generate the perceptual hypotheses necessary consciously to apprehend those same thoughts should they occur subsequently. Cognitive research on nonconscious processing, therefore, may help to identify the procedures most likely to enable conscious experiences that correspond to automatic thoughts. It also is possible that this research may contribute to the construction of explanations to account for
clinical pathology. Within this chapter, we have presented strong evidence to support the operation of nonconscious memory, nonconscious perception and nonconscious learning. Consider how each of these processes may contribute to the development of psychopathology. The idea that memories which patients cannot consciously recollect might play a functional role in mediating emotional pathology is not new. Indeed, such an assumption was central to even the earliest psycho-dynamic models of neuroses (e.g. Freud, 1901). However, the lack of any direct evidence to sustain the hypothesis that memory could operate nonconsciously led to this view being either neglected or rejected by mainstream psychology for many decades. This situation has been dramatically reversed in recent years, following the convincing laboratory demonstrations of nonconscious memory provided by experimental psychologists such as Schacter (1987). Since then, a great many clinical psychologists have become firmly convinced that the influence of nonconscious memory, for early episodes of childhood abuse, may account for a constellation of psychopathological symptoms observed in adult patients, including obsessive-compulsive disorder (Lipinski & Pope, 1994), phobia (Cowen, 1983), depression (Alpert, 1994) and multiple personality (Merskey, 1995). In consequence, therapies designed to enable conscious retrieval of these supposed memories, in order to eliminate their pervasive nonconscious influences, have been developed and widely promoted (e.g. Siegel & Romig, 1990; Roland, 1993). Page 271
For a wide variety of reasons, such claims have been hotly disputed, and it would be fair to say that the 'repressed memory debate' currently one of the most contentious issues within clinical psychology (cf. Bekerian & Goodrich, 1995; Whitfield, 1995; Zaragoza & Mitchell, 1995). It lies beyond the scope of the present text to provide an adequate critical review of this debate, which extends beyond psychological theory to raise many sensitive social and legal issues (cf. Quirk & DePrince, 1995; Sales, Shuman & O'Connor, 1994). Nevertheless, it perhaps is not inappropriate for us to point out that some revival of interest in the possibility that nonconscious memories of trauma may exert maladaptive effects on psychological functioning is fully warranted by our recently acquired knowledge that nonconscious memories can systematically influence our interpretations of ambiguity, our affective appraisals of emotional stimuli, and our overt behaviours. Indeed, it is a striking indication of how profoundly clinical conceptions of memory have changed in recent years that the contemporary repressed memory debate involves virtually no dispute concerning the potential impact of nonconscious memories. Rather, this debate focuses primarily upon doubts concerning the validity of the conscious 'memories' supposedly recovered in later life, often through therapy (cf. Belli & Loftus, 1994; Loftus & Yapko, 1995; Ware, 1995). It also seems highly probable that nonconscious perception may make a functional contribution to clinical pathology. For example, given the facts that clinically anxious individuals tend to selectively encode threatening stimuli they cannot consciously perceive (e.g. MacLeod & Hagan, 1992; MacLeod & Rutherford, 1992), and that nonconscious perception of threatening stimuli can elicit state anxiety elevations (Kemp-Wheeler & Hill, 1987), it seems reasonable to postulate that elevated state anxiety levels in anxiety disordered patients may result from the nonconscious perception of threat. Recently, Ohman and his colleagues have reported a series of experiments which add empirical weight to this proposal. In some of these studies, Ohman has demonstrated that when skin conductance responses (SCRs) were conditioned to certain stimuli by initially pairing these with shock, subsequent masked exposures to these stimuli served to elicit SCRs in the absence of stimulus awareness (e.g. Ohman, Dimberg & Esteves, 1989, Soares & Ohman, 1993a, 1993b; Ohman & Soares, 1993). Interestingly, although equivalent SCRs could initially be conditioned to a variety of stimuli, it was only for those particular stimuli commonly represented within phobic disorders (such as spiders, snakes or angry faces) that SCRs were preserved under the masked exposure conditions.
Such findings have led Ohman to propose that phobic responses may be directly initiated by the nonconscious perceptual analysis of feared stimuli. He has provided further support for this proposal by demonstrating Page 272
that, relative to normal controls, snake-fearful individuals show elevated SCRs to masked snake slides, and spider-fearful individuals show elevated SCRs to masked spider slides, despite all individuals' inability to correctly identify slide content (Ohman & Soares, 1994). In addition to elevated SCRs, the fearful individuals also reported higher levels of arousal and aversion in response to the masked presentations of their feared stimuli. It is reasonable to conclude, therefore, that individual differences in emotional experience associated with psychopathology can arise as a result of perceptual processing that occurs outside awareness. As yet, the possible contribution of nonconscious learning to the development of psychopathology has received little direct consideration, though this represents a promising area for future research. Of course, behavioural theorists have long assumed that the learning of conditioned associations, which may contribute to pathological symptomatology, can occur without awareness. However, given recent demonstrations that nonconscious learning can also result in the acquisition of complex rule knowledge, which individuals can use without awareness to predict the occurrence of future events and to exert control over complex systems, it is tempting to speculate that this form of nonconscious processing may make a more extensive contribution to the production of psychological dysfunction. For example, it seems possible that the idiosyncratic predictions about the short- and long-term future which appear to underlie certain fear responses in anxiety disorders, and which plausibly contribute to the development of 'hopelessness' in depression, may result from the application of rules nonconsciously abstracted from exposure to previous environmental regularities. Likewise, patients whose dysfunctions plausibly stem from the use of maladaptive control strategies to regulate their social interactions, their marital relationships, or their levels of gambling, drinking or eating, may unwittingly be applying rules acquired nonconsciously within some earlier context, which can neither be accessed nor modified through introspective analysis. Despite the speculative nature of such suggestions, they highlight the potential enrichment to our understanding of psychopathology that one might reasonably expect to result from the fuller consideration of nonconscious learning processes. There is some evidence to suggest that the occurrence of nonconscious processing may not only serve to illuminate possible explanations for psychopathology, but might also be exploited to provide direct therapeutic benefit. A number of studies have attempted to exploit the treatment potential of nonconscious processing, often proposing that therapeutic information implanted without permitting awareness will evade intentional rejection, and hence may be capable of producing beneficial effects beyond those that could be expected from conscious Page 273
appraisal. For example, Lee and his colleagues (Tyrer, Horn & Lee, 1978; Lee & Tyrer, 1980) have reported successfully desensitising agoraphobic patients by presenting films of the phobic situation at levels of illumination which did not permit conscious perception. Using this procedure, fear and avoidance behaviour were reduced in the subliminal exposure group at least to the same degree as in a supraliminal exposure group, and both those groups improved to a greater degree than controls who experienced the same procedure with no film in the projector. The subliminal procedure, however, was experienced as significantly less stressful than either the supraliminal or even the control conditions, and produced lower subjective ratings of muscle tension, sweating, shaking or difficulty breathing. This pattern of results has been replicated by Lee, Tyrer & Horn (1983). In this later study a hybrid exposure condition,
involving the gradual incrementation of initially subliminal presentations to a supraliminal level of illumination as treatment progressed, was found even more effective than either purely subliminal or purely supraliminal conditions in treating phobic responses. While these studies have tended to employ small samples, with only five or six individuals in some groups, such results are certainly encouraging. They clearly indicate the clinical potential of exploiting procedures which may directly influence nonconscious processing without either permitting rejection or producing the discomfort which can accompany the conscious representation of aversive stimuli. A rather different approach to the therapeutic use of nonconscious processing has been introduced by Silverman. This rests upon a more controversial theoretical framework, but has generated an extensive series of studies which also largely support the clinical utility of subliminally presenting information. One problem with evaluating this research is that its experimental predictions are sometimes based upon psychoanalytical hypotheses which are not always made fully explicit. Nevertheless, Silverman and his associates present considerable evidence to suggest that short statements given at very brief exposure durations can serve useful diagnostic and therapeutic functions. The suggestion that such presentations can serve a diagnostic function is based upon the psycho dynamic supposition that certain psychiatric disorders arise as a result of underlying conflicts involving aggressive and libidinal wishes. Silverman argues that statements capable of activating such conflicts, such as 'Destroy mother', when presented subliminally will increase any Psychopathology which is itself a consequence of those conflicts. Silverman reports that different types of stimulus material related to dynamic conflicts produce quite specific effects upon particular classes of psychopathology, when presented subliminally (4 ms exposure durations) Page 274
but not when presented supraliminally (Silverman, Bronstein & Mendelsohn, 1976). The effects of such subliminally presented statements have been offered as support for various psychoanalytical hypotheses, such as the role of oral aggression in schizophrenia (Silverman & Spiro, 1968), repressed aggression in depression (Miller, 1973), anal conflict in stuttering (Silverman et al., 1972) and incestuous wishes in homosexuality (Silverman et al., 1973). Silverman has also extended this approach to the treatment of psychopathology, by presenting subliminal messages aimed at gratifying early symbiotic wishes which, according to the psychoanalytical literature, can underlie adult emotional distress. The most common therapeutic phrase employed in this research has been 'Mommy and I are one', which has been reported to be effective in reducing a remarkable range of psychopathology (cf. Kaye, 1975; Silverman, 1976, 1980, 1983, 1984). Four millisecond exposures to this message have been found to facilitate normal desensitisation procedures (Silverman, Frank & Dachinger, 1974), to decrease thought disorder in schizophrenics (Silverman & Candell, 1970) and to reduce anxiety in male homosexuals (Silverman et al., 1973). Objective measures of improvement, such as weight loss in obese individuals (Silverman et al., 1978b) or improved high-school grades in students (Arian, 1979) also have been reported using this subliminal psychodynamic activation (SPA) procedure. Related research suggests that mild manifestations of psychopathology in the normal population also may be manipulated by the subliminal presentation of messages designed to influence unconscious motives. Again, the hypothesised nature of unconscious motives, and hence the content of the subliminal statements employed, usually have been derived from a psychoanalytical framework. Hence it has been suggested that ability to perform well in competitive situations will be impaired by unconscious conflict over Oedipal wishes, and thus the influence of subliminal messages related to such conflict on competitive performance has been examined. For example, Silverman et al. (1978b) report four studies in which dart-
throwing ability was greatly impaired by 4 ms exposures to the phrase 'Beating dad is wrong' and improved by similar exposures to 'Beating dad is OK'. Such effects were not, however, found when the exposures were made supraliminally. Similar effects of such subliminal messages on dart-throwing performance have been replicated by Lonski & Palumbo (1978). The validity of the SPA paradigm remains a controversial issue (cf. Balay & Shevrin, 1988). Replication attempts have not always been successful (e.g. Oliver & Burkham, 1982; Haspel & Harris, 1982; Porterfield & Page 275
Golding, 1985; Kothera, Fudin & Nicastro, 1990; Borgeld, 1990; Holmes, 1991). However, Silverman (1985) attributes many such failures to procedural inadequacies, and emphasises that he and his colleagues have obtained significant effects in approximately 80% of their 60 or so studies. A sufficient number of independent replications have been reported to demand that Silverman's results be taken seriously (e.g. Parker, 1977; Sackeim, 1977; Talbot, Duberstein & Scott, 1991; Samide, 1991; Gustafson & Kallmen, 1991; Hudesman, Page & Rautianen, 1992). Indeed, in a large scale meta-analysis of 56 SPA studies, Hardaway (1990) concludes that subliminal exposure to 'symbiotic-like stimuli' results in a reliable enhancement of adaptive behaviour. In a more recent review, even out-spoken critics Fudin & Benjamin (1992) class 38 existing experimental studies as 'clearly supportive' of the SPA approach. It is easy to understand some resistance among many theoreticians to embrace a therapeutic paradigm which rests upon so many controversial assumptions. The entire psychoanalytic framework, from which the specific experimental stimuli are derived, is itself largely unvalidated. Furthermore, the experimental paradigm requires the assumptions not only that subliminal words produce semantic activation, but that phrases can be syntactically parsed and encoded as relatively complex messages without awareness. At least, until recently, there has been little direct experimental evidence available to support this assumption (though see Masling et al., 1991). Nevertheless, the experimental procedures often resist trivial criticism, frequently employing double-blind methodologies and commonly including appropriate checks for awareness. Therefore, this area of research adds further weight to the proposal that the direct manipulation of nonconscious processes may potentially exert a therapeutic influence in cases of psychopathology. This chapter has provided some indication of the exciting progress that experimental psychologists have recently made in their endeavours to understand the nature of, and the distinctions between, conscious and nonconscious processes. Novel paradigms have been developed, and increasingly more powerful models have been generated. It seems highly likely that, by drawing upon this research, clinical psychologists now may enhance their ability to assess, treat and explain a wider variety of psychopathology. What the field lacks at present, however, is an integrated model within which the different types of cognitive bias found in emotional disorders can be understood with respect to the roles of conscious and nonconscious information processing. In the next, and final, chapter we attempt to effect such an integration.
Page 276
Chapter 11 Theoretical Overview At the start of our book, we set as our aim to review how the information processing paradigm has been applied to emotional disorders, and to consider theoretical and clinical implications of this body of work. The first chapter included a description of the relevant clinical phenomenology, how emotionally disturbed people are preoccupied with upsetting events. Was it that they experience more negative events, or do they notice them more than other events in their environment? Was it that, for them, the effect of even a mild stressor was exaggerated? Or could it be that the impact was initially the same, but by ruminating about and elaborating the event or its interpretations, emotional disturbance is increased. Until the increased interest in this field in the 1980s, there had been a dearth of methods for investigating these phenomena, and a lack of precision in clinical theorising about cognitive aspects of emotional disorders. In the first edition of this book, we reviewed a number of experiments on the effects of emotion on general cognitive tasks, as well as the way emotion may bias attention, memory and judgements. We reviewed evidence on the dissociations between conscious and nonconscious processing, and suggested that inconsistencies in the pattern of results might be due to dissociations between different processing demands made by tasks that assessed different aspects or stages of information processing. This final chapter reexamines the heuristic model that we proposed at that time, and how further experiments have supported the model, or have suggested that further development of it is necessary. Specificity in Processing Bias In our 1988 review, a number of experimental paradigms had been used to examine differences between emotionally disturbed and control subjects in the processing of emotional materials: dichotic listening, lexical decision, the Stroop test, incidental recall of self-referent adjectives, intentional recall of personality trait adjectives, recall of stories with mixed affective context, cued autobiographical memory and probability judgements. There was a Page 277
great deal of evidence that, in emotional disorders, attention is biased, memory is biased and judgements are biased. It was possible to find biases in pre-attentive mechanisms as well as in the contents of consciousness. At this global level, this was consistent with the predictions of the two main frameworks within which these paradigms had been used: network theory and schema theory. However, both these frameworks were relatively general in their predictions. They suggested that cognitive processing was biased in emotional disorders, but not that different emotions may have different effects or that any one emotion may have distinct effects on different aspects of processing. However, our review of the experimental evidence showed that bias favouring mood-congruent stimuli was not shown by all mood-disordered individuals on every paradigm. Many studies had shown attentional bias in anxious patients and high trait anxious normals, but had not found consistent results on attentional tasks in depression. In contrast to the equivocal results of experiments looking for attentional bias in depression, our 1988 review found a considerable number of studies reporting a memory bias in depressed subjects. Similar memory biases had been reported in agoraphobics (Nunn, Stevenson & Whalan, 1984), but in addition to these findings not being replicated (Pickles & van den Broeck, 1988) findings with other highly anxious subjects had
been negative despite the use of self-referent threat adjectives in a paradigm which exactly paralleled that used with depressives (Mogg, Mathews & Weinman, 1987). In the light of these findings we suggested an integrative model to account for why there should be a different pattern of results in different mood states, using the differences between anxiety and depression as our guide. In the next section, we revisit this model, and then examine further experiments which allow us to examine how closely the data match its predictions. The 1988 Integrative Model The model took as its starting point the data from information-processing studies that supported Freud's claim that unconscious processes moderate our conscious experience. Furthermore, the processes which operate nonconsciously appeared to follow different rules from those which are available to introspection. This conclusion corresponds to the distinction drawn by Schneider & Shiffrin (1977) between automatic and strategic processes (see Chapter 2). Automatic processes operate without awareness, are rapid, unconstrained by capacity and occur in parallel. Strategic processes on the other hand are capacity limited, relatively slow and usually serial in nature. Such a distinction is found in many areas of clinical Page 278
and experimental cognitive psychology, often arising out of studies which examine very different aspects of cognitive function. It is similar to the distinction made by Hasher & Zacks (1979) between automatic and controlled processing and that made by Baddeley (1982) between an initial passive aspect of memory and a strategic awareness-based recollection procedure, the latter being a problem-solving process which requires strategic deployment of mnemonic cues. Similarly, in their theory of emotion, Leventhal & Scherer (1987) distinguished between an automatic sensorimotor level, a schematic level and a controlled conceptual level (see Chapter 1). The notion that there may be more than one form of process which acts on more than one form of representation was, we suggested, assumed by many cognitive psychologists, but it was only relatively recently that this notion had been elaborated on the basis of detailed experimental results. We argued that both encoding and retrieval involve a passive, automatic aspect and an active, strategic aspect. Second, we suggested that a bias in one aspect need not entail a bias in the other. Third, we showed how the discrepant results were explicable if different moods had their effects on different aspects of processing. Priming vs Elaboration The 1988 model was derived from the observations made in the perceptual memory experiments of Jacoby & Witherspoon (1982) and Graf & Mandler (1984). As described in Chapter 5, a central purpose of their experiments was to understand how a person could show evidence of having been exposed to a stimulus without necessarily being able to remember having been so exposed. In perceptual memory paradigms, a study phase (in which, for example, a list of words is shown to subjects) is followed by a test phase involving such tasks as identification of briefly presented words (e.g. 35 ms) or completion of word stems (for . . .: forest, forbid, forget, etc.). These test tasks show evidence that subjects' responses are influenced by the material to which they were exposed in the study phase. However, the extent of this influence is independent of the subject's ability to explicitly recognise or recall the items from the study phase. Indeed, since amnesic patients show equivalent bias on such indirect memory tasks without being able to recall anything about the study phase, such effects are clearly independent of 'aware' forms of remembering (Jacoby & Witherspoon, 1982). Many experiments seeking to distinguish implicit from explicit memory found that the extent of perceptual memory bias was independent of the
Page 279
depth to which an item was processed. Graf & Mandler (1984) compared structural and semantic encoding in the study phase. They found that this manipulation had no effect on the extent to which word-stem completion was biased by the items which had been studied. Despite this there was the predicted large enhancement of recognition and recall for the items encoded semantically. They accounted for these results by proposing a distinction between two processes which operate upon mental representations: integration (priming) and elaboration. Priming is automatic, occurring because the processing of a stimulus (e,g. a word) involves automatic activation of the multiple components involved in the representation of that stimulus. The result is a strengthening of the internal organisation of the representation, making the word more 'accessible'. That is, the word will come to mind more readily when only some of its features or components (e.g. initial letters) are presented. Elaboration, a more strategic process, consists of the activation of a representation in relation to other associated representations to form new relationships between them and to activate old relationships. The result of the spread of activation to associates produced by this process of elaboration is to make the word more 'retrievable' because such elaboration generates new and reinstates old paths for retrieving the word. In summary, processing a word results in activation which unifies or primes the representation of the word. (We used the concept of 'priming' rather than that of 'integration' because of the latter's ambiguity.) Such priming 'raises the probability that the word will be produced (or seen, or heard) when only some of its components are presented.' By contrast, 'elaborative processing is required to perceive relations among a set of previously unrelated words and to relate the occurrence of a word to its context' (Graf & Mandler, 1984, p. 554). These suggestions encouraged us to look again at the information requirements of the tasks which have been used to assess emotional disturbance. Biased Encoding The level to which the representation of a word or concept has been primed is assessed by tasks such as perceptual threshold. We suggested that these were similar to the type of attentional tasks used to study anxiety. This opened up the possibility that biases such as those shown by anxious patients on the visual dot-probe experiment (see Chapter 4) reflected processes similar to those which mediated automatic priming. The extent of elaboration of a word or concept in relation to associated words or concepts is assessed by tasks such as depth of processing. We suggested that Page 280
these were similar to the types of task used to study memory in depression. This opened up the possibility that biases such as those shown on recall (in neutral mood) of negative material encoded in depressed mood (see Chapter 6) reflected the extent to which that material has been elaborated at encoding. The important implication about the dissociation between pre-attentive priming and strategic elaboration was that bias in priming could exist without there being any bias in elaboration. This dissociation was not readily predicted by either schema or network theories. We assumed that at the pre-attentive stage of processing there existed a decision mechanism capable of assessing the affective valence (e.g. threat value) of an item (see Figure 6). On the basis of this decision, priorities for subsequent processing are determined; resources are oriented towards or away from the source of the stimulus. This account is similar to Neisser's (1976) characterisation of perception as a cyclical process. The first stage involves the passive intake of partial information from the environment which is then mapped onto internal representations or schemata. These schemata accommodate both the information and direct processing resources during the next intake cycle towards particular elements of the stimulus array. It is at this point of the cycle, where processing resources are allocated for subsequent cycles, that decisions may be made on the basis
of affective salience. Non-disturbed subjects tend to orient processing resources away from the location of a stimulus which has been judged to be threatening (see Figure 6). This may be protective, limiting increases in anxiety by excluding such minor threatening stimuli from the cognitive system at an early stage. By contrast, it appears that anxious patients shift processing resources towards the location of a threatening item. We suggested that it was this switching of processing resources which MacLeod, Mathews & Tata (1986) were assessing in their visual probe task. It also explained the poorer performance found on a secondary reaction-time task when threatening material was played to anxious subjects in a non-shadowed message in a dichotic listening paradigm (Mathews & MacLeod, 1986). In this experiment the fact that subjects were not consciously aware of the threatening material in the non-shadowed message did not prevent it from depleting the resources available for the reaction-time task. The model suggested that once a negative item had received priority in the allocation of resources, it would increase the extent to which an item is primed ('integrated' in Graf & Mandler's sense). That is, increased allocation of pre-attentive resources towards an item was functionally equivalent to multiple exposure to that item. Irrespective of the degree of further processing of the item at the elaboration stage, the representation Page 281
Figure 6 Williams et al. (1988) model representing how state and trait mood (e.g. (a) anxiety, (b) depression) may affect resource allocation at priming and elaboration stages. (ADM = affective decision mechanism; RAM = resource allocation mechanism)
of that item will be more likely to be produced (or heard, or seen) when only some of its components are presented. Furthermore, negative items may appear to come into consciousness of their own accord due to the operation of previously primed items acting to disambiguate otherwise neutral stimuli in a negative way. The person need not be aware that the original items have been encoded; the phenomenon is independent of a person's ability to recognise those items to which he or she has been exposed. In summary, the visual probe and dichotic listening experiments of MacLeod, Mathews & Tata (1986) and Mathews & MacLeod (1986) were explained in terms of these pre-attentive activation processes. We assumed that high trait anxious individuals have a
tendency at this pre-attentive stage to orient resources towards the location of threat. The greater the evidence that a stimulus has been disambiguated as threatening, the greater the tendency to orient towards it for these subjects. Similarly, at the strategic stage of encoding (see Figure 6), relations between associates of the disambiguated item are activated (Graf & Page 282
Mandler, 1984). We assumed that, as such elaboration of an item begins, a further decision mechanism assesses the affective valence of the item. As at the pre-attentive stage, resource allocation is contingent upon the results of this decision. If additional resources are allocated, the stimulus will be more elaboratively encoded. These elaborations are encoded with the item, and are able to act as mnemonic cues to aid later recognition and recall, even if such recall is performed in neutral mood. We assumed that enhanced incidental recall following self-referent encoding of negative material in depressives (see Chapter 6) arises because of the extra resources deployed in the recruitment of mnemonic encoding cues. In some types of emotional disturbance (e.g. spider phobia) the memory data were explained by assuming, that there were strong tendencies to direct processing resources away from valent (phobic) material at the elaboration stage. In this case, fewer mnemonic cues would be stored with an item rendering it less retrievable. Different mood states may therefore have different effects in such tasks, independent of the effect they have had on pre-attentive priming, by varying the type and extent of elaborative encoding that takes place. That is, they may differentially affect the degree to which mood-congruent mnemonic cues are recruited and encoded with an item. Biased Retrieval Retrieval from long-term memory also has a passive automatic aspect and a strategic 'recollection' aspect (Baddeley, 1982). We suggested that here, too, some emotional states affect the passive aspect and others the strategic aspect. The passive aspects of retrieval may determine which memories 'come to mind' apparently unprompted. The strategic aspect of retrieval determines how successfully one can make an active search of memory. More highly primed material would come to mind more readily when only partial cues were available. This may, for example, explain the finding that highly anxious people are more likely to give a negatively biased spelling of a homophone (e.g. weak, week; ail, ale; pale, pail). Although alternative explanations were availablee.g. that such highly anxious subjects consciously retrieve both spellings and then consciously select the more threatening interpretationthe notion that these biases arise independently of awareness was more consistent with the evidence of Jacoby & Witherspoon (1982). They used a similar paradigm and found that amnesics' spelling of a homophone could be biased by prior exposure to the non-dominant meaning, despite the fact that the patients could not recall having undergone the study phase. It suggested that the Page 283
anxious person may, either by virtue of mood or by virtue of frequent rumination on threatening themes, have relatively frequently activated representations which bias towards threat on these interpretive tasks. Such biases may exist without the anxious patient being able to recall the episodes which have primed the threat interpretations. We suggested that depression affected the active strategic element of memory retrieval. In particular, depression may enhance the deployment of mood-congruent cues at retrieval, enhancing the recall of negative material even if that material has been encoded in neutral mood (Teasdale & Russell, 1983). By contrast, if the material to be recalled involves an element of distaste or threat (as it may do in phobic or anxious patients) then mood-congruent recall may be inhibited. That is, there may be, for these patients on these materials, a reduction in the extent to
which mood-congruent mnemonic cues are generated at retrieval, whether or not such cues have been generated at encoding. Has the Subsequent Data Matched the Model's Predictions? The aim of the model was to provide a framework within which it would be possible to see how dissociations between different emotions arise. It accounted for the pattern of findings by distinguishing between the passive, automatic aspect of encoding and retrieval, and the more active, effortful aspects of encoding and retrieval. In looking at the results of experiments conducted since, we can ask, first, whether this distinction between priming and elaboration has proved helpful, in general. Second, we can ask the more specific question whether anxiety and depression do differ in the specific way predicted, that is, in the terms used by Graf & Mandler (1984) and coined by us in 1988, whether anxiety does indeed make certain items more accessible, whereas depression makes certain items more retrievable. Attentional Bias In our 1988 review, a considerable number of studies had found attentional biases in highly anxious subjects. The additional data reviewed in this volume in Chapter 4 shows that such attentional biases have continued to be found, using the Stroop task, visual dot-probe task, visual search tasks and dichotic listening tasks. Although we suggested that further research was needed to determine the extent to which failure to find Page 284
effects was due to differences in tasks or materials, such attentional biases appeared then to be less associated with depressed mood. Although colour naming in the Stroop task has been found to be retarded for depressed content words (Gotlib & Cane, 1987), such effects may not occur under all conditions, requiring priming by self-referent material related to the colour word (Segal et al., 1995). Other perceptual effects remain elusive. Following an earlier study by Gerrig & Bower (1982), who had failed to find that experimentally induced depressed mood affected perceptual thresholds, a similar experiment by Small & Robins (1988) found reduced thresholds for both negative and positive words, suggesting priming of the words' semantics ('cold' priming) rather than of their emotional implications ('hot' priming). Further, early studies which failed to find an effect of depressed mood on lexical decision for negative words (Martin & Clark, 1983; MacLeod et al., 1987) have been followed up by more recent study using depressed patients, which also found no effect of depression on lexical decision (Matthews & Southall, 1991). These negative results using lexical decision must, however, be set against the positive results using a similar priming paradigm, but which asks the participant to judge whether targets are positive or negative words. Such affective decisions are significantly faster if the target is preceded by an affectively related prime (Hermans, De Houwer & Eelen, 1994) and following a mood induction procedure (MIP, Hermans, De Houwer & Eelen, in press), a result which offers general support to Williams et al's (1988) assumptions about the presence of an affective decision mechanism (ADM) at an early stage of processing. The question remains how much such mechanisms influence the subsequent processing of emotional material in different emotional disorders. Gotlib, McLachlan & Katz (1988) reported a difference between depressed and non-depressed subjects in their perceptual fluency (a judgement about which of two colour patches appeared first on the screen, each patch occurring in the same location at which a negative or control word had been briefly presented). They reported that non-depressed subjects oriented attention selectively towards positive stimuli. However, they did not measure anxiety, so it was not possible to judge which mood was responsible for the effects. A procedural replication of Gotlib and colleagues,
experiments was conducted by Mogg et al. (1991b) using measures of depression, and of state and trait anxiety. This study found that the bias was wholly due to state anxiety, rather than depression. Only one study had examined both anxious and depressed subjects on the same perceptual task (the visual dot-probe experiment of MacLeod, Mathews & Tata, 1986). There was no evidence of biased processing by Page 285
depressives. Further attempts to find similar biases using such tasks have also produced equivocal results (Hill & Dutton, 1989). Indirect Tests of Memory In Chapter 6, we reviewed several studies that had examined implicit (indirect) memory in anxiety. Of four studies that used subjects varying in trait anxiety, three (MacLeod, 1990 using homophone spelling; Richards & French, 1991, using fragment completion; Bradley, Mogg & Williams, 1994, using lexical decision following repetition priming) found evidence of memory on these indirect tests, though the lexical decision paradigm result may have been due to the presence of depression. One (Nugent & Mineka, 1994, using word stem completion) found no such effect. There have been three studies examining indirect tests of memory in GAD patients. Two of these (MacLeod, 1990, using a perceptual threshold task; and Mathews et al., 1989, using a word stem completion task following imagination instructions in the study phase) found a significant bias effect, and one (Mathews et al., 1995, using word stem completion following orthographic encoding in the study phase) did not. Further experiments have examined indirect tests of memory in other clinical states. Two studies looked at panic disorder patients, and both (Cloitre et al., 1994, using word completion, and Amir et al, submitted, using judgement of loudness of noise in which previously studied words were embedded) found significant bias. (A third study (Cloitre & Liebowitz, 1991) that found biased perceptual memory (speeded recognition of panic words previously studied using surface or deep encoding) will be considered later, since it was not a test of implicit memory.) In summary, out of nine studies using indirect tests of memory in anxious subjects or patients, seven have found significant bias towards negative material. While this is encouraging for Williams and co-workers' model, such findings are not universal, with some investigators reporting difficulty in replicating the implicit memory bias with high trait subjects (Richards & French, personal communication, 1994), so the discrepancies may still need to be accounted for. Second, better implicit memory performance can sometimes be traced to the use of explicit memory strategies, and not all studies report the correlations between the bias for the two tests. Fewer studies have examined similar test performance in depression. Although two studies (Hertel & Hardin, 1990, Elliott & Greene, 1992) Page 286
examined implicit memory in experimentally induced and clinical depression, respectively, their purpose was not to examine bias in memory, but general deficits for neutral materials, so their relevance to the present discussion is only tangential. Only two studies have looked for memory bias in depression using indirect tests of memory for negative words (that is, for words that normally produce a bias in explicit memory). Both studies (Denny & Hunt, 1992, using word fragment completion, and Watkins et al., 1992, using word stem completion) found no evidence of implicit memory bias. In summary, no study has yet found mood congruent bias in implicit memory in depression.
Direct Tests of Memory Williams et al. (1988) had reviewed evidence to suggest that explicit memory bias was more consistently found in depression than in anxiety. More recent tests on depression have confirmed that memory bias, or overgeneral memory on an autobiographical memory task, is a highly consistent finding. Memory bias emerges from two studies using mood induction (Bradley et al., 1993; Nasby, 1994) and four studies using depressed patients (Greenberg & Beck, 1989, though see below for shortcomings of this study; Denny & Hunt, 1992, Watkins et al., 1992; Fromholt, Larsen & Larsen, 1995). Five studies of depressed patients have examined autobiographical memory and found such patients to be overgeneral in recall (Williams & Scott, 1988; Puffet et al., 1991; Brittlebank et al., 1993; Kuyken & Brewin, 1995; Kuyken & Dalgleish, 1995). To our knowledge, no study has yet been published which has failed to find effects of depression on explicit memory. By contrast, the results on explicit recall in anxious subjects is more difficult to characterise. Those studies that have examined attentional bias using the emotional Stroop task have continued to show that colour-naming interference for negative material does not result in enhanced memory for that material. This dissociation has been shown in high trait anxious subjects (Richards & Millwood, 1989), in GAD patients, (Mogg et al, 1989), in panic disorder patients (Ehlers et al, 1988) and in snake-avoidant subjects (Mathews & Sebastian, 1993). Since two of the studies (Richards & Millwood, 1989, and Mathews & Sebastian, 1993) found that all subjects showed better memory for the negative material, the lack of memory bias could not have been due to the insensitivity of the memory tests. In addition to these memory tests given after a Stroop task, 12 studies have focused only on memory in anxiety. Six of these examined explicit Page 287
memory in high trait anxious or high N subjects. Two failed to find an effect (Nugent & Mineka, 1994; Richards & French, 1991), one found an effect (Ingram et al., 1987) and three found effects which appeared to depend partly or wholly on interactions with depression (Bradley et al., 1993; Bradley & Mogg, 1994; Bradley, Mogg & Williams, 1994). Six studies have examined explicit recall in GAD patients. Two have found a memory bias (Greenberg & Beck, 1989, though this study has been criticised for not controlling for 'yes' rated adjectives; and Mogg & Mathews, 1990, which they interpret as due to response bias since they found an increase in negative intrusions in the anxious patients). Three have found no such bias (Mathews et al., 1989, using cued recall; Mogg et al., 1992 and MacLeod, 1990, using recognition tasks). Finally, one study (Otto et al., 1994) found a bias towards memory for threat, but only in those GAD subjects with left hemisphere preference on a laterality task. In summary, there have been 16 studies examining explicit recall in trait anxiety or GAD, including the four Stroop studies. Only five of these have found an explicit memory bias, and three of these five provide relatively weak evidence (the first because of response bias, the second because of failure to control for 'yes' encodings, and the third because bias was found only after subjects' laterality was taken into account). Similarly, weak evidence comes from the study of memory for phobic materials in spideravoidant subjects. Whereas Rusted & Dighton (1991) found enhanced recall for material related (by inference) to the presence of spiders in a story, Watts & Dalgleish (1991), in two experiments, found reduced recall for words relating directly to spiders compared with control words. Delaying the free recall for 24 hours made no differencethere was still no enhancement of recall of phobic materials.
In 1988, we pointed out how anxiety and depression appeared to have different effects on different aspects of information processing. Anxiety seemed to affect automatic priming, and depression to affect strategic elaboration. We can now see more clearly what the answer to the two questions posed at the start of this chapter are: whether, in general, we still need to suppose that biases in the automatic and strategic aspects of processing are independent from each other; and whether the specific dissociations proposed in 1988 have been replicated. Sufficient experiments have found bias in some aspects of processing in the absence of bias in others to suggest that, in general, some such dissociation has to be assumed. Further, the facts that (a) there is as yet no evidence of biased implicit memory in depression, yet eight out of ten studies on anxious subjects (high trait, GAD, PTSD or panic patients) show such a bias; Page 288
(b) all published studies appear to find explicit memory biases in depression, yet only a third of the studies on trait anxiety or GAD find explicit memory bias; (c) attentional bias on tasks such as the emotional Stroop and visual dot-probe task is a consistent finding in anxiety, but more difficult to find in depression; suggest that the Williams et al. model continues to be supported. If this were all the evidence, there would be little motivation for re-examining the model at all. But there are other findings to be accounted for. Findings Inconsistent with the Model Despite the balance of evidence reviewed above, we are concerned that some well-controlled studies (e.g. Nugent & Mineka, 1994; Mathews et al., 1995) have failed to find implicit memory biases in trait anxiety and GAD patients. We are also aware that other recent work (for example, in Richards' laboratory) have looked for evidence of implicit memory in trait anxiety and have not found it (Richards, personal communication, 1994). In addition, more consistent evidence of explicit memory effects in anxiety emerges from two sources: first, from research using subjects other than high trait anxious, phobic or GAD patients; and second, from work using autobiographical recall rather than recall of previously presented words. The research on patients other than GAD has mostly examined panic disorder patients. Of six studies focusing on memory only, five have found explicit memory bias towards panic words (McNally, Foa & Donnell, 1989; Cloitre & Liebowitz, 1991; Becker, Rinck & Margraf, 1994; Cloitre et al., 1994; the fifth being that of Otto et al. (1994), who found a bias for those patients showing laterality scores favouring left-hemisphere processing.) Only one study )that of Amir et al., in press, using a recognition task) failed to find a bias. Finally, a study of socially anxious subjects (Claeys, 1989) also found recall bias for unlikeable trait adjectives. The research using alternatives to the word list has focused on autobiographical memory paradigms. Each has used a cue word technique (similar to Williams & Broadbent 1986a) and has found evidence of mood-congruent bias. Richards & Whittaker (1990) using an anxiety induction procedure, found reduced latency for retrieval of threat memories. Mayo (1989) and Burke & Matthews (1992) described similar results for high trait anxious and GAD patients, respectively. Neither study found clear evidence of overgeneral memories, however. By contrast, autobiographical Page 289
memory studies in PTSD patients (McNally, Litz & Prassas, 1994; McNally et al., 1995) have found both memory bias and overgenerality in memories, similar to the pattern shown in
parasuicide patients by Williams & Broadbent (1986a) and Williams & Dritschel (1988a). These results cannot be explained within an unmodified version of Williams et al.'s (1988) model. Although we said at that time that it was unlikely that anxiety would never be found to be associated with explicit memory deficits, and depression with attentional deficits, there is sufficient evidence now for us to need to try to see if the discrepancies fit into any sort of pattern. How can such discrepancies be explained? We shall first examine possible reasons why direct and indirect tests of memory sometimes do not show bias towards threat words in anxiety, then move to consider the possibility that the distinction between implicit and explicit masks a more important difference between perceptual and conceptual processes at encoding and retrieval. Finally we shall briefly consider the relevance of two new multiple subsystems approaches to cognition and emotion, those of MEM (Johnson & Hirst, 1993) and ICS (Teasdale & Barnard, 1993). In particular, we shall suggest that although the discrepancies in the effects of emotion on biasing information-processing can be partly explained by accidental features of experimental procedure, other discrepancies require substantive changes to our understanding of the information-processing mechanisms involved in emotional processing. An example of a procedural issue is the way indirect tests of memory require subjects to respond in the test phase with the item that they would normally avoid. The substantive change we shall suggest is to follow Johnson's MEM theory to distinguish different types of elaboration subprocesses, one for memorial and the other for non-memorial elaboration. Explaining Inconsistencies in the Data Explicit Memory Results One source of inconsistency in the results on explicit memory in anxiety may be that the extent of bias depends on the way anxiety interacts with other moods such as depression. We have seen that those experiments which have examined both anxiety and depression together have often found that both are involved in predicting extent of memory bias. Thus high N subjects tended to show a bias against recall of negative material after neutral MIP, but showed a memory bias for such material after a depression MIP (Bradley et al., 1993). A similar conclusion follows from the study of Bradley & Mogg (1994) in which the level of N predicted Page 290
recall bias in high BDI scorers, but not in low BDI scorers. Finally, Richards & Whittaker (1990), Burke & Mathews (1992) and Eysenck & Byrne (1994) each found that level of depression was involved with at least some aspects of retrieval bias on explicit tests, particularly a reduced retrieval of positive or neutral material. Since many of these experiments involve a situation in which an individual is free to recall either negative or positive material from memory, faster recall of one type of information is likely to have a detrimental effect on the recall of other types of information, depending in subtle ways on how much mnemonic support the experimenter or subject can provide at retrieval. Second, recent research has suggested that there are likely to be individual differences between subjects in neurological characteristics which might mediate their ability to use such mnemonics effectively in these tasks. Otto et al. (in press) tested 12 GAD patients, 12 panic disorder patients and 12 controls on a task of hemispheric laterality. Rhyming words were presented simultaneously to subjects in a dichotic listening task, and subjects reported which word they heard, to yield a laterality score. In the memory experiment, subjects were presented with 24 panic words (e.g. suffocate, collapse), 24 threat words (cancer, immature) and 24 neutral words (carpet, surplus). The test task involved using word stems as cues for recall. The main analysis of cued recall showed no overall bias in memory in GAD or panic patients. However, multiple regression showed that the greater the left-hemisphere advantage in patients, the greater the bias
towards recall of threat material, and the greater the left-hemisphere advantage in controls, the greater the bias away from recall of threat material. Since laterality was associated with extent of bias, rather than with overall memory performance, these results cannot be explained simply in terms of some subjects' greater ability in elaborating mnemonic material. Rather, they suggest individual differences in the extent to which subjects can tolerate allocating resources towards threatening material at encoding, at search, or at response. Further research will be necessary to see if this interesting result replicates, and to examine the aspect of information processing that mediates the effect. Both the involvement of other moods, and the mediating influence of these neurological characteristics need to be taken into account in future research. What effect do these variables have? We suggest that when a mood or neurological characteristic prevents explicit memory bias, it does so by preventing the material from being sufficiently elaborated at encoding or retrieval. There are three main ways in which this may occur. First, the material may be categorised too broadly (i.e. the words 'cobweb, hairy', etc. being all encoded as 'spider words'), causing a cue overload phenomenon (see Chapter 6). Consistent with this is the finding that there are sometimes Page 291
intrusions into recall protocols by words that are related to the TBR material, but were not seen in the study phase (Watts & Dalgleish, 1991; Mogg, Mathews & Eysenck, 1992). Second, the mood or neurological differences may increase the aversion to rehearsing or elaborating the material to be learned or retrieved. Any factor which reduces such elaboration will reduce explicit memory for that material. Third, subjects in certain moods or with certain neurological characteristics may divide their attention, for example, by elaborating the wrong material. Jacoby (1991) has shown that dividing attention at encoding reduces explicit recall. Could this apply to some anxious patients too? For example, worriers may elaborate the implications of the word 'cancer' rather than the word itself. So, for example, they may think of a person they knew who died of cancer, then of hospitals in general, then of their partner's recent operation, then of the tests their friend had for a stomach ulcer. The result will be reduced recall for the original word list, since attention was divided during encoding. The subjects are elaborating something, but not the TBR material itself. Such tendencies might be explored in the future using a false recognition task, in which the previously presented material is given at the test phase along with associates of the material. Further, we should be able to mimic such effects by using divided attention tasks at encoding. Implicit Memory Biases in Anxiety We turn, now, to the problem that some studies have had in finding implicit memory biases in anxiety. This inconsistency presents a puzzle in its own right, irrespective of whether there is found a dissociation with other tasks. This is because, where no evidence of implicit memory bias in anxiety is found, the result appears to suggest that words associated with anxiety are not more perceptually fluent than neutral words. Compare this pattern of results with those of attentional bias tasks such as the emotional Stroop and dichotic listening tasks (Chapters 4 and 5). Whichever way one explains such attentional bias, the phenomenon appears robust, and can be found across a large range of different types of psychopathology (see Table 1, Chapter 4). Yet both types of task might be argued to be assessing 'processing fluency' or 'activation level' or degree of priming/integration. Let us compare what is required of subjects in a typical emotional Stroop experiment with that required of them in a typical implicit memory experiment. Concurrent vs Delay First, in the Stroop task, subjects are given the threat stimulus, and its impact is assessed
concurrently. In implicit memory paradigms, the torea Page 292
word is presented in the study phase, and its impact assessed after a delay. There are two aspects of this difference that may be significant. First, the dependence of the implicit memory task on prior presentation means that variation in the mode or context of presentation may affect performance in the test phase in addition to any variation at the test phase itself. This refers not only to the difference between conceptual and perceptual encoding (Roediger & McDermott 1992, and see later), but also to the differences in study context. Jacoby (1983) found that individual words presented in the study phase without a context (e.g. xxx-cold, rather than hotcold) produced greater priming in a perceptual identification task. Similarly, a word presented in isolation, or out of context in a text (or a low-frequency word) will show more priming on a word completion test than high-frequency words or words presented in a meaningful context (MacLeod, 1989). The reduced priming due to context in these experiments appear to reflect the fact that less perceptual evidence is needed by the subject to instantiate the representation of the word. So in the case of 'hot-cold', the perceptual evidence required to activate the representation of the word 'cold' is small (e.g. 'co..'). When there is no context, the word requires there to be more integration or perceptual evidence, and later perceptual tests are more likely to show priming has taken place. No experiments on emotional bias have yet taken such variables into account in their experiments. However, these results clearly have implications for studies on priming for emotional words. These results illustrate an important principle: that perceptual implicit test tasks may reflect the incremental priming due to the study phase. Prior rumination on fear themes may mean that, when a word is presented during study, less perceptual evidence is required to instantiate the word. These tendencies may sometimes work against the possibility of showing implicit perceptual memory effects. Second, the dependence on prior presentation in a study phase allows opponent processes to operate (see Chapter 5). We have suggested that such processes might operate to suppress the availability of units denoting threat at the point at which activation of such units are needed for construction of mnemonic cues for later recollection. However, the time course of opponent processes is, at present, unknown, though the logic of our argument suggests that opponent processes result in the inhibition (or normalisation) of activation levels of threat units for at least long enough to affect explicit recall. Only further research will allow us to be more precise on this point. Primary vs Secondary Task Effects The Stroop examines processing bias by studying its effect on another task (colour naming). In the most commonly used 'implicit' tasksword Page 293
stem and word fragment completionthe subject has to complete the task by responding with the threat word itself. This may be a critical feature. It is possible that the unit representing the threat word remains more 'fluent', but that the response requirement introduces dysfluency. Should subjects have an aversion to responding with threat words, this may offset any advantage the subjects may have had by virtue of the increased accessibility of the word. The Effect of Current Stress Some experiments on the Stroop have shown that colour naming is most likely to occur when a concern is activated or primed (e.g. when an important exam is approaching, in which case high trait anxious subjects show more interference, or when a subject has been induced to fail on an
experimental task, in which case all subjects tend to show interference for achievement words). It is possible that bias on implicit memory tests may also show up more clearly under conditions of high stress or concern. Nugent & Mineka (1994), in discussing the fact that their study had not replicated that of Richards & French, point out that Richards & French's subjects had higher levels of state anxiety and lower levels of trait anxiety than subjects in their own study. Paradigms That May Help We need a task in which greater fluency of processing can be shown up by its effect on a secondary task, rather than demands a response from the subject which the subject may wish to avoid. Jacoby's noise judgement task is one such paradigm. It has already proved useful in assessing processing fluency of threat sentences in panic disorder patients (Amir, et al., in press). In this paradigm, words or sentences are played against the background of white noise (around 6065 dB). Subjects judge how loud the noise was for each stimulus. Jacoby and colleagues have found that words that have been heard before in the study phase are more likely to have their background noise judged as quieter in the test phase compared to newly presented words (Jacoby, Lindsay & Toth, 1992). They have also found that the noise that is background to emotional sentences is judged quieter. If all words were read in the test phase of such an experiment, then only in the conceptual encoding condition (does it describe you?) should there be priming which can be picked up by a reduction in loudness judgement at test. If perceptual encoding were used (is there an 'e' in it?) there should be little effect on later noise judgement. If a visual analogue of the task were to be developed (judging the darkness or contrast of a screen in which is embedded threat or non-threat Page 294
words), a fully cross-modal experiment is then possible. If auditory and visual presentation are used at both study and test, then conceptual encoding should transfer across modalities, whereas perceptual encoding should be shown up only on the modality in which encoding took place. If such a paradigm were used, then both conceptual and perceptual test tasks would need to be usedeach used with explicit versus implicit instructions. Perceptual versus Conceptual Encoding In discussing the null results of their experiment Nugent & Mineka (1994) raised the possibility that difficulties in finding consistent patterns of implicit and explicit memory biases in anxiety may be due to the fact that experiments have not controlled for the type of processing they demand of subjects at study and test. Roediger & McDermott (1992) had alluded to the same confound in their discussion of implicit and explicit memory biases in depression. Perceptual priming is shown where a word fragment (e.g. -h--t-h) is completed after the subject has studied the item (Cheetah) in the same modality. The cue specifies that it is the perceptual form of the item which requires to be matched. This matching may be implicit (the first thing that comes to mind) or explicit (use the cue to deliberately retrieve a previously studied item). By contrast, conceptual priming would be shown where a subject (having seen the word 'cheetah' at study, comes up with this item when asked to generate exemplars of a category (animals), or is asked a general knowledge question such as 'What is the fastest animal on earth?' Once again, these cues may be part of an implicit or explicit test. Dissociations can be found between perceptual and conceptual processing which, it has been argued, wholly account for the dissociations previously attributed to implicit versus explicit memory. Thus, reading words at study leads to higher performance on perceptually cued implicit memory than generating the same word from a conceptual cue at study. By contrast, such a 'generate' condition at study leads to enhanced conceptual implicit memory. Second, depth of
processing affects conceptual priming, not perceptual priming. Third, differences in the way the word is typed (upper or lower case) at study or test affects perceptual, but not conceptual priming. Fourth, conceptual priming is higher for categorically organised than for unorganised lists. Finally, there are modality effects in perceptual priming which are not shown in conceptual priming. Of perhaps the greatest interest is the fact that amnesics show evidence of intact conceptual priming (Shinamura, 1986; Tulving, Hayman Page 295
& MacDonald, 1991). According to the pure processing view, memory tests benefit to the extent that type of processing permitted at study overlaps with type of processing required for the performance of the test. Previous studies in cognition and emotion tend to have confounded perceptual (or bottomup/data-driven) processing with conceptual (top-down/concept-driven) processing. They have used explicit conceptual tests (free recall) and perceptual implicit tests (word or fragment completion). According to Roediger & McDermott (1992) the lack of implicit memory effects in depression (Denny & Hunt, 1992; Watkins et al., 1992) might be due to the fact that, using perceptual tasks (word fragment or stem completion at test) they were unlikely to pick up differences in type of material (positive or negative: a conceptual difference) despite this variable affecting explicit performance (which was conceptual). However, this does not account for the studies which have found perceptual implicit memory in GAD, trait anxiety, panic disorder or PTSD, all of which have used word stems or fragments at test following conceptual encoding (see earlier). Ironically, one of the studies which failed to find implicit perceptual memory (Mathews et al., 1995) was the only study to use perceptual (orthographic) encoding. According to Roediger & McDermott's pure processing view, this study should have been a paradigm example of 'transfer appropriate processing'. This is not to say that these distinctions are not important. Eysenck & Byrne (1994) and Cloitre & Liebowitz (1991) contrasted data-driven and concept-driven encoding and retrieval in trait anxious subjects and panic patients, respectively. Both found evidence of a match between encoding and retrieval processing demands. For example, Cloitre & Liebowitz (1991) found explicit conceptual bias (free recall) in panic patients following semantic encoding. However, the amount of perceptual explicit retrieval (a speeded recognition task) did not vary as a function of encoding: both conceptual and perceptual encoding were equally likely to produce later perceptual explicit bias in panic patients. It would be possible to argue that, in this case, subjects must have been doing some conceptual encoding in the (ostensibly) perceptual study phase, thus preserving the integrity of the pure 'transfer appropriate processing' view. The difficulty is that, without independent evidence of which processing operations are being used, it will always be possible to make such post hoc analyses. Furthermore, there is experimental evidence that a pure processing view is insufficient to account for the data. For example, consider an experiment by Challis et al., (1993). One hundred and twenty target words (e.g. molasses) were presented to subjects auditorily or visually during the study phase; then they were given a recognition test (auditorily or visually). In a third phase, they were Page 296
required to do either an explicit or implicit memory test, and in each case there was a perceptual cue (a word fragment such as m---ss-s) or a conceptual cue (the question: 'What is the syrup drained from raw sugar?'). In the explicit condition, they were told that the cues were to help them retrieve items from the study phase. In the implicit condition they were simply told to say the first thing that came to mind. Results showed that the modality of study and recognition phases (visual vs auditory) affected perceptually cued retrieval (both explicit and implicit) but
had no effect on conceptual (explicit and implicit) retrieval. This is consistent with a pure processing view. However, results also showed that, regardless of the type of retrieval cue (perceptual fragment or conceptual question) the degree of dependency between recognition and explicit retrieval was greater than that between recognition and implicit retrieval. There are processes involved in explicit retrieval over and above the matching of processing requirements between study and test. The authors suggest the need to combine a processing view with a multiple memory systems view, in which explicit retrieval taps episodic memory traces, and implicit memory is driven by the semantic memory system. Such distinctions between memory systems and not just memory processes is also advocated by Schacter (1993) on the basis of neuropsychological data that show dissociations in the regions of brain mediating different memory subsystems. In the next section, we consider two recent subsystems views of information processing. Before doing so, we need to consider what implications the distinctions made so far have on the work we have been reviewing. We have seen that a 'transfer appropriate processing' view, though it might account for some failures to find implicit or explicit memory effects in some studies, cannot account for the positive results that have been found. In particular we may need to assume that explicit recollection (whether perceptual or conceptual processing is involved) demands the use of special resources or processing modules. Roediger & McDermott (1992) suggest that, because depressive bias was found on the explicit test, this was due to the conceptual nature of the test. On the basis of their transfer appropriate processing view they predict that if a conceptual implicit task was used, then depressed people would be biased on this too, despite it not demanding strategic recollection. This is a testable prediction. If depressed people are found to be biased on explicit, but not implicit conceptual recall (following conceptual encoding in both cases) this would strongly suggest that the bias in depression occurs in the use of conscious recollection strategies. Such an emphasis on conscious reflection processes forms a central element in the multiple-entry memory (MEM) system of Marcia Johnson, and it is to this model that we now turn. Page 297
Multi-Level Models: Can They Account for Differences between Emotional States? We saw in Chapter 6 how, within single-level models such as associative network theory, it is difficult to understand how one emotion, such as sadness/depression, could be associated with a congruent bias in a particular encoding operation, while another, such as fear/anxiety, was not. However, if emotions are associated with different functions, together with facilitation of encoding processes appropriate to those functions, then it would not be surprising if such inconsistencies occurred. In this case, the level or encoding process most characteristic of that emotion, would be where congruent bias would be most consistently observed. The experimental findings so far thus seem easier to account for in terms of models that posit different representational levels or subsystems, than within a single system using a common code. The most obvious fit with the data just discussed is the MEM model described by Johnson & Multhaup (1992) and Johnson & Hirst (1993). We give a general description of the model first, then say how it has been seen as relevant to emotion. MEM (Johnson & Hirst, 1993) The aim of MEM is to provide an appropriate vocabulary to guide and constrain discussion of memory tasks and processes. It hypothesises 16 basic processes configured into four subsystems. The two perceptual systems (P1 and P2) are lower level modules designed to extract perceptual information from the environment. A person is usually unaware of perceptual information
involved in the associations learned by P1 processes (for example, cues in a speech signal specifying a particular vowel), while the P2 subsystem involves learning about the perceptual world as it is perceived phenomenally. The two reflective subsystems (R1 and R2) have the capability of executive control (R2) and supervision and monitoring (R1). They enable the subject to go beyond the immediate consequences of perception, to manipulate information and anticipate events, and to imagine possible outcomes of alternative behaviour plans. The 16 processes involved within the perceptual and reflective subsystems are further organised into themes and are shown in Table 2. Each theme represents a hypothesised continuity through evolution, the higher functions building on mechanisms of the lower. Thus, in Table 2, 'rehearsing' is hypothesised to have grown out of lower 'identifying' mechanisms. The model implies that reflective mechanisms are involved in source judgements necessary for episodic memory. That is, if, during Page 298 Table 2 Perceptual and reflective subsystems and their themes in a multiple-entry memory system (after Johnson & Hirst, 1993) Theme
Perceptual
Reflective
P1
P2
R1
R2
'Identifying' and maintaining active the objects of perception and thought
Resolving stimulus configurations
Identifying objects
Refreshing information so it remains active
Rehearsing with aim of increasing probability of later recall
'Going back' to earlier objects of perception or thought
Locating stimuli
Placing Reactivating objects in spatial information that has relationship to each other dropped out of consciousness
Retrieving strategic re of cues in active recollection
'Relating' across time and/or events
Extracting invariants from perceptual arrays
Structuring or abstracting patterns across temporal extensions
Noting relations
Discovering strategic search for new links
'Introducing change' in activation pattern to system
Tracking stimuli
Examining or perceptually investigating stimuli
Shifting attention to more Initiating useful stimuli search for different cues Page 299
encoding, a person does not note the context, the record of that context will not be available at recall. Such problems in reflective monitoring of contexts is assumed to mediate memory problems in ageing, amnesia, as well as the confabulation of some brain-damaged groups and delusional patients. The model further assumes no distinction between attention and memory. Attention involves 'activities with memorial consequences' being . . . a shorthand term for engaging in the component processes' shown in Table 2 (Johnson & Hirst, 1993, p. 257). MEM and Cognitive Bias: Memorial vs Non-Memorial Elaboration Within MEM, fear can be represented at all levels, starting with the most basic perceptual processing; whereas remorse or disappointment (perhaps the closest emotion that is considered to depression) are predominantly represented at higher reflective levels (Johnson & Multhaup, 1992). It would seem to follow directly that selective perceptual processing would occur in fear and anxiety states to a greater extent than in depression. Reflective levels can also be involved in some anxiety statesfor example, where worry is a prominent featurebut even if such strategic operations were to be inhibited, the basic perceptual schemas and processes could remain active.
This might have two consequences. First, the emotion experienced might change to a felt sense of danger without any reasons being obvious to the anxious individual. Second, distinguishing between different domains within the reflective subsystem allows the possibility that processes may be selectively inhibited. So, if only those reflective processes involved in 'identifying' (refreshing and rehearsing) and 'going back' (reactivating and retrieving) were inhibited, then the reflective processes which would remain active would be those responsible for noting, discovering, shifting and initiating. This is significant, for these are just the constellation of processes that Johnson & Hirst (1993) suggest are involved in the normally 'creative organisational activities found, for example, in problem solving or brain-storming' (p. 270). That such non-memorial elaboration may proceed without involving processes which would result in increased retrievability is a powerful aspect of the MEM model. It is consistent with the explanation we suggested earlier for the inconsistency in explicit recall in some forms of anxiety. We suggested that some anxious patients engage in elaboration which is irrelevant to the TBR material. The MEM account adds to this explanation. It is not only the case that such 'worry' divides attention, Page 300
filling a 'limited capacity channel' with material which is irrelevant to the TBR material. It suggests that such non-memorial elaboration is subserved by a system that would normally have the positive outcome of creative problem solving, a system that takes some fragments of mnemonic material and elaborates it in new ways, not bound by the mnemonic material itself. The similarity of 'worry' to an unresolved problem-solving process is at the centre of many theories of worry (e.g. Borkovec, 1994; Davey, 1994). In contrast, it is the fact that depression involves the memorial processes (refreshing, rehearsing, reactivating, retrieving), that explains the more robust elaborative memory biases seen in this condition. We suggest, therefore, that the model we proposed in 1988 be modified to distinguish between memorial and non-memorial elaboration processes. (see Table 3). Non-memorial elaboration we take to include noting and discovering which emerge from the 'relating' perceptual processes of extracting and structuring; and the themes of shifting and initiating which emerges from the 'introducing change' perceptual processes of tracking and examining. We have already noted the involvement of these subprocesses in creative problem solving, and the way in which the same processes might subserve worry. Making this distinction allows us to predict that such worry/ruminative elaboration may not result in increased memory, unless, for some reason, other subprocesses involving reactivation, retrieval and rehearsal are involved. Clearly, such subprocesses subserving memorial and non-memorial elaboration might in some conditions, such as panic disorder, social phobia or PTSD, co-occur. In others, such as spider phobia or GAD, they may tend to inhibit each other. In a third case, the presence of a mood such as depression, itself normally involving memorial elaboration, might be sufficient to produce explicit memory bias. We suggest that future research on explicit retrieval emotion takes account of these subtypes of reflective elaborative processes. This needs to be done explicitly, rather than in a post hoc fashion. Clearly, there is the danger that any inconsistent result might be explained after the fact by assuming that non-memorial elaboration has taken place. We therefore need independent evidence for such processes. Such evidence could be provided by the use of thought-sampling techniques. Following a study phase in which negative material is presented for later recall, a filler task between study and test phases might be interrupted at random for thought-sampling. The resulting samples could be independently rated for the conceptual 'distance' between the thought content and the TBR material. We predict that some anxious patients (e.g. GAD) would have ratings showing their rehearsal processes had travelled a greater distance from the target material than would be the case with depressed patients. Within groups,
Page 301 Table 3 Memorial and non-memorial elaborative processes in reflective subsystems (the non-memorial elaboration processes, that normally subserve creative problem solving, become active in anxious patients and worriers, reducing memorial elaboration which would normally assist encoding and retrieval mnemonics) Process
R1
R2
Memorial elaboration 'Identifying' 'Going back'
Refreshing Reactivating
Rehearsing Retrieving
Non-memorial elaboration 'Relating' 'Introducing change'
Noting Shifting
Discovering Initiating
there should be an association between the degree of non-target material in the rehearsal pool and later performance on a direct test of memory. MEM and OvergeneraI Recall An advantage of MEM is that it can add to our account of overgeneral memory phenomena. In MEM, there is no separate store for autobiographical memory; semantic and episodic memories are not different subsystems. Rather, the senses of 'remembering' and 'knowing' reflect the attribution made on the basis of the subjective quality of the memory experience. (Note that the term 'attribution' is being used in a different sense from that usually used in experimental abnormal psychology. Here it refers to the attribution of a percept to different sources (e.g. 'I recall seeing it', versus, 'I imagined seeing it').) According to MEM, the reactivation subprocess is critically involved in the maintenance, over time, of contextual or perceptual detail. It is disruption of this subprocess in amnesia which gives rise to the lack of specificity in their recollections. In MEM, people do not either remember autobiographically or non-autobiographically, but rather experience, while remembering (and depending on the integrity of the reactivation subprocess) different degrees of specificity, clarity, confidence in the veridicality of the image, etc. In addition, specificity of memory arises from the ability, at encoding, to tie an episode to other episodes occurring at the time. (Amnesics cannot retain these other episodes to help such specific encoding, so specificity of memory is damaged on this second count too.) According to this view, overgeneral memory in suicidal, depressed and PTSD patients Page 302
might arise either from impairment in reflective processes at encoding preventing the encoding of other events that are occurring at the same time, or might arise from impairment in the reactivation subprocess. Such an account is consistent with the mnemonic interlock model proposed by Williams (1996), adding to it the notion that the generation of 'intermediate descriptions' into which the subjects eventually become enmeshed when attempting to retrieve a specific memory may be the result of the other reflective processing cycles: noting, shifting, discovering and initiating. Once again, we see the value of a model which includes reflective processes which do not necessarily help memory to function efficiently. So far we have described how discrepancies might be accounted for by confounds between the match in types of processing which tasks demand at encoding and retrieval. Where this does not account for the data, it is possible that subtle differences in the way subjects are encoding the material, how much items are pre-primed (and thus less susceptible to perceptual priming), and the requirements of the task to give the feared item as response, each contribute to the outcome.
Further, we have identified differences between memorial and non-memorial elaboration in MEM and suggested that this distinction be incorporated into the affective decision-resource allocation model we proposed in 1988. Although such additional considerations may explain the experimental data reported, they do not appear to explain the difference between cold and hot cognition. For the most comprehensive discussion of this issue we need to turn, finally, to the interacting cognitive subsystem framework of Teasdale & Barnard (1993). In describing the model, we draw upon the summary of it by Williams (1994). Interacting Cognitive Subsystems (ICS) Teasdale & Barnard (1993) propose a number of different subsystems to capture the clinical phenomena of hot cognition in depression: three sensory/proprioceptive subsystems (acoustic, visual and body state), two intermediate structural description subsystems (the morphonolexical and object) two meaning subsystems (the propositional and implicational ) and two effector subsystems (the articulatory and limb ). Each subsystem uses a different 'code', and each has its own memory storage system. The important aspect of the model is that emotion results only from the activity in the highest level 'implicational' code. The function of the implicational code is to extract general, schematic patterns/regularities from the information encoded by other subsystems. These other subsystems, including the proprioceptive and sensory subsystems, may feed Page 303
directly to the implicational subsystem (that is, by-passing the propositional code) thus affecting emotion directly. In this way, emotion escapes from the need to be represented within a propositional network. The model captures the difference between hot and cold cognition by locating them within separate subsystems. Cold processing is likely to be propositional. Hot processing involves the implicational subsystem. ICS suggests that negative depressive thinking reflects changes in schematic models encoding the interrelationships between constructs (such as the relationship between social disapproval and personal worth) rather than changes in accessibility of individual negative constructs. This predicts that depressed patients can be made to use more positive individual words if the whole meaning of the sentence thereby becomes dysfunctional. In an experiment reported in the book (p. 190) a depressed patient completed the sentence: 'Always to put others' interests before your own is a recipe for __________' with the word 'success'. When less depressed three months later, the same patient completed the same sentence with the negative word 'disaster' thus making the entire sentence less dysfunctional (though perhaps still a little catastrophic!). Perhaps the most radical aspect of ICS is the way it conceives emotion. Emotional subjective feelings with implicit information content (such as, hopelessness following repeated attempts to succeed at a task), 'mark the processing of related Implicational code patterns by the COPY process of the Implicational subsystem. Holistic ''senses" or feelings of this type are the phenomenal experience corresponding to activity in the Implicational subsystem, just as phenomenal experience of pitch or timbre mark activity in the Acoustic subsystem' (p. 84). This clearly marks their departure from more conventional cognitive models of emotion. Depression is associated with ruminative response style caused by the reciprocal interaction of the implicational and propositional subsystems. These may become interlocked when propositions such as 'I am a failure' contribute to the recreation of schematic models like 'me as a worthless, hopeless, culpable person' similar to that which produced the proposition 'I am a failure' in the first place. It is this interlock phenomenon that has been influential in guiding explanations of generic memory in depression (see Chapter 6).
The ICS framework is explicitly designed to account for developmental aspect of psychopathology in a way that MEM is not. Consider the following observations by Bettelheim (1987). From the outset, the interplay between parents and child is continuous . . . incidents which adults consider trifles are immensely significant to a child, providing the parental signals which guide the child's awakening to the world. Page 304 A parents words and gestures, tone of voice and facial expression, can suddenly throw a different light on things . . . giving the cues upon which (the child) bases his view of himself and of his world . . . this is the child's guide to the meaning of what is happening. (Bettelheim, 1987, pp. 467, 48)
Clearly, these observations relate to a great deal of implicit learning that takes place in the development of the child. Much of it is conceptual rather than perceptual, but all of these cues are integrated into meaningful chunks that go beyond perceptions or words. It is this 'going beyond' which the implicational subsystem of Teasdale & Barnard seeks to capture. ICS locates the predisposition to depression in the way biologically preprimed tendencies (for example, reaction to loss with distress and 'searching' followed by inactivity) combine with regularly experienced environmental features (e.g. threatened abandonment by a parent). The result is the encoding of a prototypical set of features of the situation which will elicit emotionsan implicational pattern. In this respect, the pattern behaves like the 'schema' in Beck's model of emotion. However, the matching between original and current context is much more generic in ICS. According to ICS, in adulthood, the implicational pattern may be activated by the presence of any stimuli which has been related in the past to the schematic model. Such a stimulus may arise from sensory, propositional or bodily sensation, or any of the other subsystems. Since the context may affect the implicational system directly, without necessarily affecting the propositional subsystem, the ICS framework provides a potentially powerful way of explaining some aspects of the onset, maintenance, and vulnerability to emotional disorders. How might the model account for differences between anxiety and depression? Within ICS it is possible that the implicational schemas activated in anxiety states were predominantly derived from patterns of perceptual and body-state information, and are associated with perceptual vigilance as a result. In contrast, the implicational schemas involved in depression may have been derived from predominantly propositional information, and are thus associated with elaborative processing of internally generated negative memories. Comparing the Models This raises the question of whether there are any real differences between these two multi-level models. Are they saying the same thing in different words, or, if not, can they generate different predictions that could be experimentally tested? Clearly there are obvious similarities. Both have Page 305
higher conceptual levels, or subsystems, that are involved when we talk or think about emotions, but do not in themselves represent emotional feelings. This feature allows us to make sense of the distinction between experiencing and talking about feelings. In MEM there is also provision for downward influence over some emotions at this level. For example, a schematically driven emotion (at R1) might be reduced by consideration of new evidence at the executive level (R2), as for example in cognitive therapy for depression. Similarly, in ICS, propositional input may initiate or modify the activity of the emotional schema operating at the implicational level. In some anxious patients, for example, just talking about symptoms and disease can precipitate a panic.
Perhaps the main contrast between models is whether all emotions are seen as being represented at the same level as in ICS, or whether, as in MEM, different emotions can occur at different levels. The latter way of thinking allows the possibility that contradictory emotions may coexist, by virtue of being at different levels. Johnson & Multhaup (1992), for example, suggest that a dog phobic might fear a guide dog (represented at P2), while also admiring the help given by the dog to a blind person (represented at R1 or R2). Awareness of contradictory emotions might lead to a resolution in favour of one or the other, but the experience we describe as 'mixed feelings' or ambivalence suggests that this does not always happen. In ICS, all emotions are represented in the implicational subsystem, and thus only one emotional schema will be dominant at any one time, although this schema may give way rapidly to another. Different emotions may derive from information from different sources (e.g. various other subsystems that use other codes) but will all operate within the implicational subsystem using the same code. Distinguishing between these possibilities in experiments seems very difficult. Because implicational schemata can potentially integrate across information from all other subsystems, it is possible that some emotional schemas are more derived from perceptual information, and others more from conceptual information. This makes it difficult to distinguish ICS from a model in which the emotional information is actually retained in different codes. It is not obvious how one could distinguish experimentally between the two models. Predictions about the different processing operations that are associated with fear-anxiety versus sadnessdepression seem to arise more naturally within MEM, although the flexibility of ICS allows these results to be accommodated without too much difficulty. Future research might profitably be directed to further specifying the apparent cognitive Page 306
differences among emotions. If the encoding processes associated with specific emotional disorders are indeed distinct, then this should constrain the assumptions that can be made within very flexible models such as ICS. Similar constraints arise from evidence about the emotional effects associated with loss of cognitive function. Rather than searching for a test that will allow us to choose decisively between models, we need to find out how helpful each proves to be in generating new predictions. For example, ICS is able to account for the mood-incongruency results of Parrott & Sabini (1990) by making the assumption (among others) that mood states are sometimes derived and maintained by external factors (sunny days) rather than by depressive schematic thinking derived from reciprocal implicational-propositional links. This is one of a number of domains in which ICS generates new predictions. The Parrott & Sabini work involved autobiographical memory, but since the ICS explanation for mood-congruent encoding and retrieval of selfdescriptive adjectives also depends on their linkage with implicational schematic models, biases in recall of such material should also depend on whether the mood is being maintained by external context on the one hand or by implicational to propositional transformations on the other. We already know that such biases depend on the extent to which the material itself requires elaborative processing to encode, such as being processed in relation to the self, but we can now see to what extent this effect interacts with the method by which mood is being maintained. We suggest that it follows from ICS that people who are feeling low because of the weather should not preferentially encode or retrieve negative words. The Importance of Processing Priority Neither of the multi-level models we have considered appears to account for why, when no
competition for resources is involved (for example, on a standard lexical decision task), it is difficult to find evidence for the perceptual priming we have suggested, occurs. Despite this, we have very good evidence that anxious individuals do selectively attend to threatening words, when they are presented in circumstances that allow the selective process to be revealed. In several experiments (see Chapter 4), we have shown that anxious subjects (whether clinically anxious or high trait anxious normals) are faster to detect neutral probes in the same location as threatening words, and show greater interference effects when the threatening words are presented as distracters. Furthermore, although studies have not found any speeding in lexical decisions for single mood-congruent words in anxious subjects, they have found speeding when Page 307
two competing words are presented simultaneously (see Chapter 4). In this task, subjects are told that either one or two stimuli will be presented, and that they should press a 'yes' key as fast as possible if either of them is a word, or press 'no' if both are non-word letter strings. In two experiments, anxious patients, but not controls, were faster to report that a word was present when it had a threatening rather than neutral meaning, but only when presented in the dualstimulus condition. These results suggest that anxious individuals do indeed selectively process threatening stimuli in preference to competing information. However, it does not seem that this selective process leads to any advantage when making either lexical or affective decisions for single words (Mathews & Milroy, 1994). This suggests the hypothesis that anxiety is associated with a tendency to give priority to processing threatening stimuli, but that the encoding involved is predominantly perceptual rather than conceptual in nature. One reason is that the results indicate that threatening stimuli are identified as such early on, yet the form of this identification is apparently not helpful in facilitating conscious affective decisions (Mathews & Milroy, 1994). Second, interference due to threatening distracters in colour-naming tasks does not seem to be associated with the ability to report or recall them later. Even when the threatening words are presented rapidly and then barkward masked to prevent awareness, selective emotional interference can still be observed in colour-naming (see Chapter 4). Third, even when evidence of selective recall is lacking, selective priming effects can be observed on implicit tasks such as stem completion or tachistoscopic identification. Selective perceptual encoding of threatening information is proposed as the central cognitive characteristic of phobias and anxiety states. This is not to say that other semantic or conceptual processing may not be involved, either in initiating or maintaining anxiety. For example, Priestley (1995) has recently found conceptual implicit priming bias for definitions of homonyms (e.g. flash, mug) in anxious, but not depressed, patients. However, we suggest that the arousal of fear or anxiety is associated with a characteristic mode of processing that gives priority to perceptual encoding. Both perceptual and conceptual encoding and retrieval effects in anxiety are most likely to be shown when the task requires decision between competing inputs or outputs. Trait and State Effects Most of the research described in this book has been aimed at finding differences in information processing between emotionally disturbed Page 308
patients and non-disturbed controls or between different types of emotional disturbance. From these findings it is difficult to determine whether the cognitive biases represent responses to particular events in a person's life, represent a particular cognitive style or trait disposition, or
arise from transient affective disturbance. For example, Lloyd & Lishman (1975) demonstrated that the greater the level of depression in depressed patients, the more biased the latency of recall of hedonically toned autobiographical memories. From this study it was not clear to what extent this phenomenon was due to the fact that the more depressed people may have suffered more negative events in the past. However, Teasdale & Fogarty (1979) showed that transient manipulation of mood in normal volunteers was sufficient to bring about a similar bias in recall latencies. Similarly, it is difficult to determine whether the fact that anxious patients tend to overestimate the probability of unpleasant things happening to them is dependent on trait or state mood. The research on exam anxiety reported in Chapters 4 and 5 found that as the exam approaches, all subjects become anxious, but that high-trait anxious subjects become more oriented towards threat on attentional tasks and more pessimistic about a wider range of concerns. This suggests an interaction between transient state mood and trait predisposition. Let us consider how each of these in turn may be represented in terms of the model. In the 1988 model transient mood shifts were assumed to affect cognitive processes by modifying the outcome of the decision on the affective salience of an item at both the preattentive and elaboration phases. We suggested that transient mood adjusted the decision mechanism in a certain direction, similar to the effect of presenting the subject with a more negative item. The greater the affective disturbance, the greater the output of the decision mechanisms to those other processes which were contingent upon them; orientation towards or away from the source of threat at the pre-attentive phase (see Figure 6a); facilitated or inhibited search for associative mnemonics at the elaboration phase (see Figure 6b). However, following our discussion of Cohen and colleagues' PDP model in Chapter 5, it becomes clear that we no longer need to assume the existence of such an affective decision mechanism. We need only assume that units be associatively 'tagged' with a threat value or loss value. Such tagging might generally reflect prior learning, though some may be biologically pre-wired, such as the threat value of large looming objects. Increase in state mood can then be represented as neuromodulatory control, with different neurotransmitters (e.g. noradrenaline or seratonin) or combinations of neurotransmitters affecting all units having a certain tag for threat, loss or anger association. Page 309
The outcome is the same as that envisaged by Williams et al. (1988), with the unit representing the activated percept (at the earliest stage) or concept (at a later stage) increasing in activation and affecting the resource allocation mechanism (the Task Demand Unit, in PDP terms). As discussed in Chapter 5, in the light of increased activation from a threat stimulus, the Task Demand Unit has to increase its activation in order to remain 'on task', but this can occur only up to a point. The switching of processing resources towards threat may be delayed but, if so, it can switch suddenly with catastrophic results. Trait predisposition may be represented as a permanent tendency to react to the increased activation of input units representing emotion by switching task demands either towards or away from the source of the emotion. Some individuals are prone to direct more processing resources towards an item if it is judged to be negative at the pre-attentive stage. For these individuals, threatening items will tend to be in a more activated state. Other individuals are prone to direct processing resources away from an item which is judged to be negative at this stage. The item will be less well primed. In particular, we suggest that high trait anxiety reflects a permanent tendency to react to increased activation of threat units at the pre-attentive stage by switching resources towards the source of threat.
Independently of these predispositions, individuals may also differ in their tendencies to direct processing resources towards or away from an item if it is judged to be negative at the elaboration stage. The result will be differences in the extent to which a negative item is elaborated. In particular we suggest that depression-prone individuals have a permanent tendency to react to increased activation of threat concepts at the elaboration stage by directing more resources to a negative item. For these individuals, negative concepts will always tend to be more elaboratively encoded. In terms of the model, the high trait anxious or depression prone person may show little evidence of psychopathology when transient emotional disturbance is low. As state mood increases, however, the predisposition to allocate processing resources towards or away from negative stimuli at the pre-attentive or elaboration stages becomes more evident. In addition to increasing the likelihood that such people will exhibit greater psychopathology in response to smaller degrees of affective disturbance, the effect of such tendencies would be to prolong any episode of emotional disturbance. The recovery process is likely to be retarded if any small increase in affective disturbance produces a greater amount of biased processing. Such a prolongation of the emotional disturbance has been Page 310
found for people currently depressed who have great cognitive disturbance, assessed by selfreport measures and for depressed patients who have remitted but who continue to have more dysfunctional attitudes (see Williams, 1992, ch. 10 for review). Other Implications Phenomenology of Anxiety and Depression The model is based largely on studies with word stimuli. The question arises whether it is generalisable. Especially, can one talk about 'priming' and 'elaboration' outside the context of individual words? Some interesting work by Thorpe (1994) suggests that effects similar to those seen with word stimuli (with spider phobics) can be found with videos of spiders and with real spiders. In one experiment, she was able to demonstrate no selective recognition advantage for spider phobics using short video clips of spiders. She also carried out a real-life visual dot-probe experiment, with subjects having to react as fast as possible to lights which flashed near the place where a spider was located. Spider-avoidant subjects oriented towards the threat location, as evidenced by their reduced latencies to respond to those locations. In addition to this further experimental work, clinical observation would suggest that these notions are generalisable to the world outside the laboratory. The evidence reviewed in this book suggests that anxiety may arise from the interaction of stimuli and the pre-attentive mechanisms which operate upon them. Attention will be more likely to switch to a threatening location, threatening items become more activated and more likely to prime subsequently ambiguous stimuli. The result is that they too will be disambiguated as negative. The significance of this becomes clear when it is realised how many stimuli in the environment are potentially mildly threatening. 'On the road, oncoming vehicles are potentially lethal, in dark streets passers by are potential assailants, and . . . minor somatic sensations are potentially symptoms of some serious malady' (MacLeod, Mathews & Tata, 1986, p. 18). The model emphasises how, by the time a person becomes aware of the stimulus, it may already have been disambiguated as negative. For example, an anxious patient may not first feel a chest pain and subsequently interpret it as a heart attack. He or she may experience the pain as a heart attack.
If depression is less associated with biased pre-attentive priming, but rather more with bias in the active strategic aspect of encoding, it may explain why depressed patients attribute the source of their negative Page 311
mood to themselves and attribute the causes of events that happen to them in a self schematic way. This contrasts with anxious and phobic patients who tend to attribute the source of their negative mood to specifiable bodily or external sources (e.g. chest pain or supermarkets). This way of characterising the nature of emotional disturbance may offer clinicians a framework which may help them notice aspects of their patients' problems they had not noticed before. The clinician may now not simply wish to take account of level of anxiety or depression or the situations in which these occur, but also how specific are the stimuli which the person notices, and whether it relates to external or bodily stimuli or to self-schemata. They might wish to assess to what extent the mood is being maintained by enhanced attention to negative items or to disambiguation (in a negative way) of neutral or ambiguous items. Alternatively, the mood may be being maintained by enhanced elaboration of negative items when encoded, or the biased selection of cues to recall events from the past at retrieval. The more exactly such assessments coincide with the processes which underlie dysfunction in emotional disturbance, the more helpful they are likely to be in generating specific remedial strategies. Cognitive Impairments on Neutral Tasks Deficits in performing neutral tasks can be understood in terms of the switching of resources (pre-attentive or elaborative) towards mood-congruent items and away from mood-neutral or incongruent material. In anxious patients reaction time to a neutral stimulus is impaired in both the dichotic listening paradigm and the visual dot-probe task (see Chapter 4) when negative material is presented in another message or location. These deficiencies in performance in anxiety may arise from the pre-attentive switching of attention away from a task in hand to other mood-relevant current concerns. Such switching is likely to occur even when neutral or ambiguous stimuli occur in the environment of the individual if such material is easily disambiguated as negative. He or she need not be aware of their existence for them to disrupt performance. In depressed patients, memory for neutral material is often impaired. Chapter 3 reviewed evidence suggesting that such memory impairment was not due to cautious response criteria. Neither do depressive memory deficits occur because depressed patients do not bother to make the effort to recall material they have learned. It appears that such patients are unable to structure material at encoding. This can be seen as representing a deficiency at the elaboration stage of encoding. The affective decision Page 312
(how negative is this item?) which is made on the to-be-remembered material at the elaboration stage determines the extent to which material will be rehearsed, structured and elaborated. Neutral material will not receive priority in the allocation of resources. Dissociation between Judgement and Memory A further implication of the contention that emotional disorders may affect different aspects of processing is that it becomes easier to understand why some aspects of conscious experience do not correlate with some aspects of underlying cognitive biases. For example, the availability heuristic described by Tversky & Kahneman (1982) has often been adduced as an explanation of how memory (the ease with which relevant instances come to mind) may affect judgements of frequency and probability. However, research (see Chapter 8) has suggested that judgements
may not correlate with the speed of recollection of relevant instances from the past. These findings imply that modifying speed of recollection may have little effect on judgements of future probability. The model allows us to see, however, that a link between memory and judgements is not precluded by these results. Speed of recollection reflects the active, strategic aspect of memory (generating and refining the cues, searching for relevant episodes, checking their appropriateness, etc.). As such it may be relatively independent of the more automatic form of memory demonstrated in perceptual memory experiments discussed above. Judgements may be a function of the extent to which the memory system has been primed by differential exposure to salient instances. It is possible that such priming might be more readily assessed by word completion or tachistoscopic recognition tasks than by strategic recall tasks, and that these would be found to relate more closely to judgements of frequency or probability. Implications for Processes Underlying Remediation The relationship between judgements and memory discussed in the previous section represents only part of a much needed re-examination of the relation between underlying biases and conscious experience. Some of these issues have been discussed in the previous chapter in relation to Nisbett & Wilson's contention that only the product but not the processes of cognitive bias can be introspectively accessed. It was suggested that even if nonconscious and conscious processes represent fundamentally different processes, the patient may, over time, acquire 'blueprints' for Page 313
becoming more aware of the way perceptual hypotheses are contributing to current mental events. But even if a person became more aware of errors and biases, the question still remains whether any such conscious knowledge could itself modify the biases themselves. There is very little evidence on whether cognitive and behavioural therapies do in fact modify underlying processing biases (though see Williams, 1992, ch. 11, for a discussion of these issues). If they do so, it is because they provide opportunities for emotional processing to take place which may previously have been impeded. First, treatments which involve confrontation with situations the patients previously found difficult may overcome the tendencies for patients cognitively to avoid stimuli they find threatening. This possibility arises from the findings that the relationship between increased imageability of feared stimuli and relaxation during the course of therapy is reciprocal. Watts (1974) found that increased detail in desensitisation produced longer between-session reductions in anxiety, and more recent work has confirmed the extent to which phobic patients are impoverished in the amount of detail they encode about their feared stimuli (Watts, Trezise & Sharrock, 1986). It is an interesting outcome of this research that such impoverished detail of an image may also be associated in these patients with impoverished detail about what to do and how to cope with the feared subject (Watts, Sharrock & Trezise, 1986), reminiscent of the effect that imagining an outcome has on behavioural scripts (see Chapter 8). This implies that encouraging the patient to encode previously avoided stimuli in a detailed way will facilitate extinction and allow the development of adequate behavioural scripts for how to cope in previously feared situations. The more recent work reviewed in this chapter suggests other ways in which emotional processing may be facilitated. The work by MacLeod (1989) showed that the extent to which a word is perceptually primed is dependent on whether it is presented alongside other associated words. This suggests one way of viewing the distinction between guided imagery (the strategic bringing to mind of traumatic imagery) and flashbacks (the unbidden coming to mind of perceptual fragments of
traumatic imagery). Could it be that strategic elaboration reduces a stimulus's power to prime perceptually? Alternatively, given the modification of the model to distinguish between memorial and non-memorial elaboration, it is possible that forcing deliberate retrieval inhibits those reflective (non-memorial) processes that would otherwise remain in problem-solving mode, trying to solve the unsolvable problems raised by prior trauma. Page 314
Second, if the affective disturbance associated with exposure to a stimulus can be reduced, in terms of our model this will reduce the amount of neuromodulatory control. At the preattentive stage this will reduce attentional deployment towards other sources of threat for anxious patients. At the elaboration stage this will reduce the spread to associated negative concepts for depressed patients. Third, the fact that most treatments involve explicit discussion of specific positive and negative aspects of the patient's past and current life may alter the specificity of these events in memory, providing a richer set of cues that the patients may use between sessions for recollecting aspects of their past. Johnson & Hirst's (1993) MEM model has shown how instructions on how to do better source monitoring by inducing subjects to adopt stringent criteria about what constitutes a memory for visually presented material can be used to teach people to distinguish between real and imagined events. This is clearly relevant to the problems emotionally disturbed people have. We saw in Chapter 6 how non-specificity in autobiographical memory inhibits problem solving, and retards recovery from depression. As a result of this work, therapists are able to be vigilant for signs of such non-specificity, and encourage patients to be more specific with the aim of aiding the recoding of previous memories, facilitating current problem solving, and undermining the tendency to encode ongoing events in schematic terms. Concluding Remarks Part of the attraction of specifying a model which distinguishes between initial priming and subsequent elaboration is that it maps more readily onto the nature of anxiety and depression. We do not believe that the association of emotions with particular cognitive processes is arbitrary: rather, we assume that it arises naturally from the biological and social functions that are served by different emotions, as suggested in the model put forward by Oatley & Johnson-Laird (1987). The primary function served by fear is to avoid danger, so we would expect the processing involved to relate to perceptual vigilance. Anxiety is a multi-component system which, in normal living, helps an individual to anticipate and avoid danger. It makes sense that it should include, as one of its components, a system which reacts rapidly to even a partial representation of a possibly threatening stimulus. The organism may need to take quick avoidant action, so there is little reason (at that time) to recruit a further system which elaborates the stimulus, and might only interfere with the necessary action. Page 315
However, there are likely to be differences between different types of anxious patient in the nature of the bias shown. For example, for some patients (especially those characterised by general worry) such elaboration of a threat stimulus may take place over a longer time-interval. It will be necessary to ensure that the entire time-course of possible elaboration bias in anxiety is assessed before one will be able to obtain a full picture of the nature of the cognitive psychopathology for these patients. Perhaps just as important is the distinction between those patients whose anxiety is an object or feeling which may occur at unpredictable times (for example, suddenly seeing a spider, or suddenly feeling a pain) and patients (such as agoraphobics) who know which situations are likely to cause anxiety and how to avoid them. In
other words, there may be important differences in cognitive processing between those anxious patients who can provide themselves with adequate safety signals, and those who cannot. Depression is also a multi-component system, part of the function of which in day-to-day living may be to help come to terms with loss. This is also consistent with the model of Oatley & Johnson-Laird, who suggest that sadness and depression involve the need to revise a major lifeplan or goal that has failed. If so, then the conceptual processing of internally generated material related to failure or loss may be more relevant to this function than perceptual vigilance. In its more severe form, depression loses its adaptive significance and becomes characterised by rumination over past failures and disappointments. In neither mild nor severe depression, however, does the problem which causes or exacerbates the depression require very quick avoidant or escape action. It is not surprising then to find that depressives do not in general show enhanced 'pick-up' of negative information. Of course, depressives may very quickly interpret an ambiguous situation as negative, but this may result from biased elaboration, rather than enhanced priming. The significance of this distinction is that the biased elaboration in depression is memorial, that is, it encodes the event in a form that can be strategically retrieved later, the enhanced priming and much of the non-memorial elaboration by anxious subjects does not enhance later retrieval. If depression is, in some form, a response to an event which requires longer term problem-solving behaviour (rather than immediate avoidance or escape), then it can be seen why it is adaptive for it to involve a system that allows strategic access to previous events. In our 1988 review, we were careful to caution against attributing dissociations between attention and memory bias simply to differences between anxiety and depression. This remains true for differences between perceptual and conceptual encoding, or between perceptual and reflective processing. There remain too many differences between different studies Page 316
in paradigms, materials, task demands, diagnoses, severity of mood disturbanceto make sweeping conclusions. However, on the basis of current evidence we can be more confident than we were able to be then, that not all emotional states affect cognitive processing in the same way. The usefulness of exploiting a simple distinction between anxiety and depression is to help make clear what dissociations may exist between different emotions and between different aspects of cognitive processing. Having derived a model which allows such distinctions to be made, we need to assert again that the picture is likely to be complex. For one thing, we have seen that anxiety and depression often co-occur. Although we should expect that anxious patients will be especially characterised by overvigilance for threat cues on the basis of which immediate action might be necessary, some depressed patients might also show similar overvigilance. In the same way, some anxious patients will show bias in the way they recollect personal events. What the distinction between priming and reflective elaboration (and those between perceptual and conceptual processing) offers is a framework within which such phenomena can be understood whether they occur in patients with a diagnosis of anxiety, depression or any other emotional disorder. It helps us to understand why a bias in one aspect of processing does not necessitate biases in other aspects. It suggests that we view the switch towards the pick up of perceptual information about threatening or negative material and its subsequent reflective elaboration as distinct aspects of emotional disturbance needing distinct assessment and remediation. In 1988 we were able to suggest that the field needed to move away from general statements about the effects of mood on cognitive processing. Our present review has confirmed the wisdom of so doing, and has made further important distinctions, such as that between memorial and non-memorial elaboration. There remains the need to specify which mood affects what sub-processes to produce which bias.
When clinical psychologists assess patients' problems they do not simply make diagnoses, but also carry out a functional analysis. This includes a description of those aspects of behaviour the patient wishes to decrease, and those aspects the patient wishes to increase. It includes a detailed analysis of the activities, places, people and times when the symptoms are most apparent and when they are least apparent. It includes a detailed analysis of the different components of the mood disturbance itself: behaviour (e.g. what objects or situations are avoided?); somatic symptoms (e.g. heart rate increase, reduced weight); and experiential symptoms (e.g. thoughts and images). The work reviewed in this book has shown that continued progress is being made in understanding how perception, attention and memory contribute to these phenomena.
Page 317
References Abelson, R. P. (1981) Psychological status of a script concept. Amer. Psychol., 36, 715729. Abrams, M. & Reber, A. S. (1988). Implicit learning: robustness in the face of psychiatric disorders. J. Psycholing. Res., 17, 425439. Abramson, L. T., Seligman, M. E. P. & Teasdale, J. D. (1978) Learned helplessness in humans; critique and reformulation. J. Abnorm. Psychol., 87, 4974. Adamowicz, A. (1989). The role of anticipation in discourse. Polish Psychol. Bull., 20, 153160. Adams, M. J. & Collins, A. A. (1979) A schematic-theoretic view of reading. In R. O. Freedle (ed.), New Directions in Discourse Processing, Vol. II. Norwood, N.J.; Ablex. Adler, A. (1927) Practise and Theory of Individual Psychology. New York: Harcourt, Brace and World. Adler, A. (1944) Disintegration and restoration of optic recognition in visual agnosia. Arch. Neurol. Psychiat., 51, 243259. Alba, J. W. & Hasher, L. (1983) Is memory schematic? Psychol. Bull., 93, 203231. Alho, K., Woods, D.L. & Algazi, A. (1994). Processing of auditory stimuli during auditory and visual attention revealed by event-related potentials. Psychophysiology, 31, 469479. Allison, J. (1963) Cognitive structure and receptivity to low intensity stimulation. J. Abnorm. Soc. Psychol., 67, 132138. Alloy, L. B. & Abramson, L. Y. (1979) Judgement of contingency in depressed and nondepressed students: sadder but wiser? J. Exp. Psychol., 108, 441485. Alloy, L. B. & Abramson, L. Y. (1982) Learned helplessness, depression, and the illusion of control. J. Pers. Soc. Psychol., 42, 11141126. Alloy, L. B. & Tabachnick, N. (1984) Assessment of covariation by humans and animals: the joint influence of prior expectations and current situational information. Psychol. Review., 91, 112149. Allport, D. A. (1980) Attention and performances. In G. Claxton (ed.), Cognitive Psychology: New Directions. London: Routledge and Kegan Paul. Allport, T. A. (1977) On knowing the meaning of words we are unable to report: the effects of visual masking. In S. Dornic (ed.), Attention and Performance, Vol. VI. New York and London: Academic Press.
Alpert, J. L. (1994). Analytic reconstruction in the treatment of an incest survivor. Psychoanal. Rev., 81, 217235. Amin J. & Lovibond, P. F. (submitted) Covariation bias and expectancy bias for biological and technological fear-relevant stimuli. Amir, N., McNally, R. J., Riemann, B. C. & Clements, C. (in press). Implicit memory bias for threat in panic disorder. Behav. Res. Ther. Page 318
Anderson, A., Garrod, S. C. & Sanford, A. J. (1983) The accessibility of pronominal antecedents as a function of episode shifts in narrative text. Quart, J. Exp. Psychol., 35A, 427440. Anderson, C. A. (1983) Imagination and expectation: the effect of imagining behavioural scripts on personal intentions. J. Pers. Soc. Psychol., 45, 293305. Anderson, J. & Bower, G. H. (1973) Human Associative Memory. Washington, D.C.: Winston. Anderson, J.R., Bjork R.A. & Bjork, E.L. (1994). Remembering can cause forgetting: retrieval dynamics in long-term memory. J. Exp. Psychol: Learn. Mem. Cog., 20, 10631087. Anderson, M. (1992). Intelligence and Development: A Cognitive Theory. Oxford, UK: Blackwell Press. Anderson, M. P. (1981) Assessment of imaginal processes: Approaches and issues. In T. V. Merluzzi, C. R. Glass & M. Genest (eds), Cognitive Assessment. New York: Guilford Press. Anderson, M. C. & Spellman, B. A. (1995). On the status of inhibitory mechanisms in cognition: Memory retrieval as a model case. Psychol. Rev., 102, 68100. Anderson, R. C. & Pichert, J. W. (1978) Recall of previously unrecallable information following a shift in perspectives. J. Verb. Learn. Verb. Behav., 17, 112. Anderson, R. C., Pichert, J. W., Goetz, E. T., Schallert, D. L., Stevens, K. V. & Trollip, S. R. (1976) Instantiation of general terms. J. Verb. Learn. Verb. Behav., 15, 667679. Andrade, J. (1995) Learning during anaesthesis: a review. Br. J. Psychol., 86, 479506. Andreasen, N. C. (1976) Do depressed patients show thought disorder? J. Nerv. Ment. Dis., 163, 186192. Arian, S. (1979) The effect of subliminal symbiotic stimuli in Hebrew on academic performance of Israeli high school students. Unpublished PhD thesis, New York University. Arntz, A., Rauner, M. & van den Hout, M. (1995) If I feel anxious there must be danger. Exconsequentia reasoning in inferring danger in anxiety disorder. Behav. Res. Ther., 33, 917925. Ausubel, D. P. (1968) Educational Psychology: A Cognitive View. New York: Holt, Rinehart and Winston. Baars, B. J. (1988) A Cognitive Theory of Consciousness. Cambridge: University Press. Baddeley, A. D. (1966) The capacity for generating information by randomisation. Quart. J. Exp. Psychol., 18, 119129. Baddeley, A. D. (1982) Domains of recollection. Psychol. Rev., 89, 708729.
Baddeley, A. D. (1986) Working Memory, Oxford: Clarendon Press. Baddeley, A. D. & Hitch, G. (1974) Working memory. In G. H. Bower (ed.), The Psychology of Learning and Motivation, Vol. 8. London: Academic Press. Baddeley, A. D. & Liberman, K. (1980) Spatial working memory. In R. Nickerson (ed.), Attention and Performance. Vol. VIII. Hillsdale, N. J.: Lawrence Erlbaum. Baddeley, A. D. & Weiskrantz, L. (eds) (1993) Attention: Selection, Awareness, and Control. A Tribute to Donald Broadbent, Oxford: Clarendon Press. Baddeley, A. D. & Wilson, B. (1986) Amnesia, autobiographical memory, and confabulation. In D. C. Rubin (ed.), Autobiographical Memory (pp. 225252). Cambridge, Cambridge University Press. Bainbridge, L. (1993) Types of hierarchies imply types of model. Ergonom., 36, 13991412. Page 319
Baker, L. L. & Jessup, B. A. (1980) The psychophysiology of affective verbal and visual information processing in dysphoria. Cog. Ther. Res., 4, 13148. Balay, J. & Shevrin, H. (1988) The subliminal psychodynamic activation method: A critical review. Amer. Psychol., 43, 161174. Balint, R. (1909) Seelenlanhnung des 'Schauens' optische ataxie raumliche storung der aufnerksankeit. Monatssch. Psychiat. Neurolog., 25, 5771. Balota, D. A. (1983) Automatic semantic activation and episodic memory encoding. J. Verb. Learn. Verb. Behav., 22, 88104. Bandura, A. (1977) Self-efficacy: towards a unifying theory of behavioural change. Psychol. Rev., 84, 191215. Banquet, J.P., Ouardirhi, S.E., Spinakis, A., Smith, M.J. & Gunther, W. (1994) Automatic vs controlled processing in variable temporal context and stimulus-response mapping. In D. S. Levine & M. Aparicio (eds), Neural Networks for Knowledge Representation and Inference pp. 447469. Hillsdale, NJ: Erlbaum. Bara, B. G. (1983) Modification of knowledge by memory process. In M. A. Reda & M. J. Mahoney (eds), Cognitive Psychotherapies. Cambridge, Mass.: Ballinger. Bargh, J. A. (1982) Attention and automaticity in the processing of self-relevant information. J. Pers. Soc, Psychol., 43, 425436. Bargh, J.A., Bond, R.A., Lombardi, W.J. & Tota, M.E. (1986) The additive nature of chronic and temporary sources of construct accessibility. J. Pers. Soc. Psychol., 50, 869878. Bargh, J. A. & Pietromonaco, P, (1982) Automatic information processing and social perception: the influence of trait information presented outside of conscious awareness on impression formation. J. Pers. Soc. Psychol., 43, 437449. Bargh, J. A. & Tota, M. E. (1988) Context-dependent automatic processing in depression: accessibility of negative constructs with regard to self but not others. J. Pers. Soc. Psychol., 54, 925939. Barlow, D. H. (1988) Anxiety and its Disorders: The Nature and Treatment of Anxiety and
Panic. New York: Guilford Press. Baron, R. & Treiman, R. (1980) Some problems in the study of differences in cognitive processes. Mem. Cog., 8, 313321. Bartlett, F. C. (1932) Remembering. Cambridge: Cambridge University Press. Bartlett, F. C. (1941) Fatigue following highly skilled work. Proc. Roy. Soc. Lond. B, 131, 247257. Basden, B., Basden, D. & Gargano, G. (1993) Directed forgetting in implicit and explicit memory tests: some tests of the retrieval inhibition explanation. J. Exp. Psychol.: Learn. Mem. Cog., 19, 603616. Beck, A. T. (1976) Cognitive Therapy and the Emotional Disorders. New York: International Universities Press. Beck, A. T. & Clark, D. A. (1988) Anxiety and depression: an information processing perspective. Anx. Res., 1, 2336. Beck, A. T., Emery, G. & Greenberg, R. L. (1985) Anxiety Disorders and Phobias: A Cognitive Perspective. New York: Basic Books. Beck, A. T., Rush, A. J., Shaw, B. F. & Emery G. (1979) Cognitive Therapy of Depression, New York: Guilford Press. Becker, C. A. & Killion, T. H. (1977) Interaction of visual and cognitive effects in word recognition. J. Exp. Psychol.: Hum. Percept. Perf., 3, 389401. Becker, R., Rinck, M. & Margraf, J. (1994) Memory bias in panic disorder. J. Abnorm. Psychol., 103, 396399. Page 320
Bekerian, D. A. & Goodrich, S. J. (1995) Telling the truth in the recovered memory debate. Consc. Cog., 4, 120124. Belleza, F. S. & Bower, G. H. (1981) The representation and processing characteristics of scripts. Bull. Psychonom. Soc., 18, 1, 4. Belli, R. F. & Loftus, E. F. (1994) Recovered memories of childhood abuse: A source monitoring perspective. In S.J. Lynn & J.W. Rhue (eds), Dissociation: Clinical and Theoretical Perspectives, (pp. 415433). New York: Guilford Press. Bentall, R. P. & Kaney, S. (1989) Content-specific information processing and persecutory delusions: An investigation using the emotional Stroop test Br. J. Med. Psychol., 62, 355364. Bentall, R. P. & Thompson, M. (1990) Emotional Stroop performance and the manic defence. Br. J. Clin. Psychol., 29, 235237. Berry, D. C. (1994) Implicit learning: twenty-five years on. A tutorial. In C. Umilta & M. Moscovitch (eds), Attention and Performance, Vol. XV: Conscious and Nonconscious Information Processing, (pp. 755782) Cambridge, Mass.: Bradford. Berry, D. C. & M. Broadbent, D. E. (1984) On the relationship between task performance and associated verbalizable knowledge. Quart. J. Exp. Psychol., 39A, 585609. Berry, D. C. & Broadbent, D. E. (1988) Interactive tasks and the implicit-explicit distinction.
Br. J. Psychol., 79, 251272. Bettelheim, B. (1987) A Good-Enough Parent. London: Thames and Hudson. Biederman, I. (1982) On the semantics of a glance at a scence. In N. Kubovy & J. R. Pomerantz (eds), Perceptual Organization. Hillsdale, N.J.: Lawrence Erlbaum. Black, J. & Bower, G. (1979) Episodes as narrative chunks in memory. J. Verb. Learn. Verb. Behav., 18, 309310. Blackburn, I. M. (1985) Mental and psychomotor speed in depression and mania. Br. J. Psychiat., 126, 329335. Blaney, P. H. (1986) Affect and memory: a review. Psychol. Bull., 99, 229246. Blankstein, K. R., Flett, G.L. & Johnston, M. E. (1992) Depression, problem-solving ability and problem-solving appraisals. J. Clin. Psychol., 48, 749459. Block, N. (1981) Imagery. Cambridge, Mass.: MIT Press. Block, N. (1995) On a confusion about a function of consciousness. Behav. Br. Sci., 18, 227287. Block, R. I., Ghoneim, M. M., Sum Ping, S.T. & Ali, M. A. (1987) Human learning during general anaesthesia and surgery. Br. J. Anaesth., 66, 170178. Bolles, R. C. & Bailey, D. E. (1956) Importance of object recognition in size constancy. J. Exp. Psychol., 51, 222225. Bonanno, G. A. & Stillings, N. A. (1986) Preference, familiarity, and recognition after repeated brief exposures to random geometric shapes. Amer. J. Psychol., 99, 403415. Bootzin, R. R. & Natsoulas, T. (1965) Evidence for perceptual defence uncontaminated by response bias. J. Pers. Soc. Psychol., 1, 461468. Bootzin, R. R. & Stephens, M. W. (1967) Individual differences and perceptual defence in the absence of response bias. J. Pers. Psychol., 6, 408412. Borgeld, T. W. (1990) The Effects of reference to mother during early childhood using the subliminal psychodynamic activation method. Diss Abs. Int., 50, 3684B. Borkovec, T. D. (1994) The nature, function and origins of worry. In G. L. Davey & F. Tallis (eds) Worrying: Perspectives on Theory, Assessment and Treatment (pp. 534). Chichester: John Wiley. Page 321
Borkovec, T. D. & Inz, J. (1990) The effect of worry in generalized anxiety disorder: a predominance of thought activity. Behav. Res. Ther., 28, 153158. Borkovec, T. D., Ketzger, R. L. & Pruzinsky, T. (1986) Anxiety, worry, and the self. In L.M. Hartman & K.R. Blankstein (eds) Perception of Self in Emotional Disorder and Psychotherapy. New York: Plenum. Borkovec, T. D., Robinson, E., Pruzinsky, T. & DePree, J. A. (1983) Preliminary exploration of worry: Some characteristics and processes. Behav. Res. Ther., 21, 916. Borkovec, T. D., Shadick, R. & Hopkins, M. (1991) The nature of normal and pathological
worry. In R. Rapee & D. H. Barlow (eds), Chronic Anxiety and Generalized Anxiety Disorders. New York: Guilford Press. Borkovec, T. D. & Sides, J. K. (1979) The contribution of relaxation and expectancy to fear reduction via graded imaginal exposure to feared stimuli. Behav. Res. Ther., 17, 529540. Bornstein, R. F. (1989) Subliminal techniques as propaganda tools: review and critique. J. Mind Behav., 10, 231262. Bornstein, R. F. & D'Agostino, P.A. (1992) Stimulus recognition and the mere exposure effect. J. Pers. Soc. Psychol., 63, 545552. Bornstein, R. F. & Pittman, T. S. (1992) Perception Without Awareness: Cognitive, Clinical & Social Perspectives. New York: Guilford. Bower, G. H. (1981) Mood and memory. Amer. Psychol., 36, 129148. Bower, G. H. (1983) Affect and cognition. Phil. Trans. Roy. Soc., London, B302, 387402. Bower, G. H. (1985) A review of research on mood and memory. Paper presented at Symposium on Affect and Cognition, Cognitive Psychology Section, British Psychological Society, Oxford, September 1985. Bower, G. H. (1987) Commentary on mood and memory, Behav. Res. Ther., 25, 443455. Bower, G. H., Black, J. B. & Turner, T. J. (1979) Scripts in memory for text. Cog. Psychol., 11, 177220. Bower, G. H. & Cohen, P. R. (1982) Emotional influences in memory and thinking: data and theory. In S. Fiske & M. Clark (eds), Affect and Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Bower, G. H. & Gilligan, S. G. (1979) Remembering information related to oneself. J. Res. Perso., 13, 420432. Bower, G. H., Gilligan, S. G. & Monteiro, K. P. (1981) Selectivity of learning caused by affective states. J. Exp. Psychol.: Gen., 110, 451473. Bower, G. H. & Mayer, J. D. (1985) Failure to replicate mood-dependent retrieval. Bull. Psychonom. Soc., 23, 3942. Bower, G. H., Monteiro, K. P. & Gilligan, S. G. (1978) Emotional mood as a context for learning and recall. J. Verb. Learn. Verb. Behav., 17, 573587, Bower, G. H. & Morrow, D. G. (1990) Mental models in narrative comprehension. Science, 247, 4448. Bowers, J. S. & Schacter, D. L. (1990) Implicit memory and test awareness. J. Exp. Psychol.: Learn. Mem. Cog., 16, 404416. Bradley, B. & Mathews, A. (1983) Negative self-schemata in clinical depression. Br. J. Clin. Psychol., 22, 173182. Bradley, B. P. & Mogg, K. (1994) Mood and personality in recall of positive and negative information. Behav. Res. Ther., 32, 137141. Bradley, B. P., Mogg, K., Galbraith, M. & Perrett, A. (1993) Negative recall bias and neuroticism : state vs trait effects. Behav. Res. Ther., 31, 125127.
Page 322
Bradley, B. P., Mogg, K. & Williams, R. (1994) Implicit and explicit memory for emotional information in non-clinical subjects. Behav. Res. Ther., 32, 6578. Bradshaw, J.L. (1974) Peripherally presented and unreportable words may bias the perceived meaning of a centrally fixated homograph. J. Exp. Psychol., 103, 12001202. Braff, D. L. & Beck, A. T. (1974) Thinking disorder in depression. Arch. Gen. Psychiat., 31, 456460. Brand, N. & Jolles, J. (1987) Information processing in depression and anxiety. Psychol. Med., 17, 145153. Bransford, J. D. (1979) Human cognition: Learning, Understanding and Remembering. Belmont, California: Wadsworth. Bransford, J. D. & Johnson, M. K. (1971) Semantic prerequisites for comprehending prose. Paper presented at Eastern Verbal Investigators League Meetings, October 1971. Bransford, J. D. & MacCarrell, N. S. (1975) A sketch of a cognitive approach to comprehension. In E. B. Weimer & D. F. Palermo (eds), Cognition and the Symbolic Processes. Hillsdale, N. J.: Lawrence Erlbaum. Breslow, R., Kocsis, J. & Belkin, B. (1981) Contribution of the depressive perspective to memory function in depression. Amer. J. Psychiat., 138, 227230. Bressler, J. (1931) Illusion in the case of subliminal visual stimulation. J. Gen. Psychol., 5, 244251. Brewer, M. (1988) A dual process model of impression formation. In T. K. Srull & R. S. Wyer (eds), A Dual Process Model of Impression Formation: Advances in Social Cognition, Vol. 1, (pp. 111118) Hillsdale, NJ: Erlbaum. Brewer, W. F. & Treyens, J. C. (1981) Role of schemata in memory for places. Cog. Psychol., 13, 207230. Brewin, C. R. (1985) Depression and causal attributions: what is their relation? Psychol. Bull., 98, 297309. Brewin, C. R., Andrews, B. & Gotlib, I. H. (1993) Psychopathology and early experience: a reappraisal of retrospective reports. Psychol. Bull., 113, 8298. Brittlebank, A. D., Scott, J., Williams, J. M. G. & Ferrier, I. N. (1993) Autobiographical memory in depression; state or trait marker? Br. J. Psychiat., 162, 118121. Broadbent D. E. (1958) Perception and Communication, New York: Pergamon Press. Broadbent D. E. (1971) Decision and Stress. London: Academic Press. Broadbent D. E. (1977a) The hidden pre-attentive processes. Amer. Psychol., 32, 109118. Broadbent, D. E. (1977b) Levels hierarchies, and locus of control. Quart. J. Exp. Psychol., 29, 181201.
Broadbent, D. E. (1982) Task combination and selective intake of information. Acta. Psychologia., 50, 253290. Broadbent, D. E. (1984) The Maltese Cross: a new simplistic model for memory. Behav. Brain Sci., 7, 5594. Broadbent, D. E. (1985) A question of levels: comment on McClelland & Rumelhart. J. Exp. Psychol.: Gen., 114, 189192. Broadbent, D. E. (1989) Lasting representations and temporary processes. In H. L. Roediger III & F. I. M. Craik (eds), Varieties of Memory and Consciousness: Essays in Honour of Endel Tulving, (pp. 211228). Hillsdale, NJ: Erlbaum. Broadbent, D. E. & Broadbent, M. (1988) Anxiety and attentional bias: state and trait. Cog. Emot., 2, 165183. Page 323
Broadbent, D. E. & Gregory, M. (1967) Perception of emotionally toned words. Nature, 215, 581584. Broadbent, D. E., Broadbent, M. H. P. & Jones, J. L. (1986) Performance correlates of selfreported cognitive failure and obsessionality. Br. J. Clin. Psychol., 25, 285299. Broadbent, D. E., Fitzgerald, P. & Broadbent, M. H. T. (1986) Implicit and explicit knowledge in the control of complex systems. Br. J. Psychol., 77, 3350. Brody, N. (1989) Unconscious learning of rules: comment on Reber's analyis of implicit learning. J. Exp. Psychol.: Gen., 118, 236238. Brooks, L. R. & Vokey, J. R. (1991) Abstract analogies and abstracted grammars: comments on Reber (1989) and Mathews et al. (1989) J. Exp. Psychol.: Gen., 120, 316323. Brown, T. (1820) Lectures on the Philosophy of the Human Mind. Edinburgh: Trait, Longman. Brown, W. P. (1961) Conceptions of perceptual defence. Br. J. Psychol. Mono. Supps., 35. Bruch, M. A., Juster, H. R. & Kaflowitz, N. G. (1983) Relationships of cognitive components of text anxiety to test performance: implications for assessment and treatment. J. Couns. Psychol., 30, 527536. Bruch, M. A., Kaflowitz, N. G. & Keuthe, M. (1986) Beliefs and the subjective meaning of thoughts: analysis of the role of self-statements in academic test performance. Cog. Ther. Res., 10, 5169. Bruner, J. S. & Postman, L. (1947a) Emotional selectivity in perception and reaction. J. Person., 16, 6977. Bruner, J. S. & Postman, L. (1947b) Tension and tension-release as organising factors in perception. J. Person., 15, 300308. Bruner, J. S., Postman, L. & Rodrigues, J. (1951) Expectation and the perception of color. Amer. J. Psychol., 64, 216227. Bundesen, C. (1990) A theory of visual attention. Psychol. Rev., 97, 523547.
Burgess, I. S., Jones, L. N., Robertson, S. A., Radcliffe, W. N., Emerson, E., Lawler, P. & Crow, T. J. (1981) The degree of control exerted by phobic and non-phobic verbal stimuli over the recognition behaviour of phobic and non-phobic subjects. Behav. Res. Ther., 19, 223234. Burke, A., Heuer, F. & Reisberg (1992) Remembering emotional events. Mem. Cog., 20, 277290. Burke, M. & Mathews, A. (1992) Autobiographical memory and clinical anxiety. Cog. Emot., 6, 2335. Buschke, H. & Schaier, A. (1979) Memory units, ideas, and propositions in semantic remembering. J. Verb. Learn. Verb. Behav., 18, 549563. Butler, G. & Mathews, A. (1983) Cognitive processes in anxiety. Adv. Behav. Ther., 5, 5162. Butler, G. & Mathews, A. (1987) Anticipatory anxiety and risk perception. Cog. Ther. Res., 91, 551565. Butler, G., Wells, A. & Dewick, H. (in press) Differential effects of worry and imagery after exposure to a stressful stimulus. Behav. Cog. Psychother. Butterfield, E. C. & Belmont, J. M. (1971) Relations of storage and retrieval strategies as short-term memory processes. J. Exp. Psychol., 89, 319328. Button, E. J. (1983a) Construing the anorexic. In J. R. Adams-Webber and J. C. Mancuso (eds), Applications of Personal Construct Theory. London: Academic Press. Page 324
Button, E. (1983b) Personal construct theory and psychological well-being. Br. J. Med. Psychol., 56, 323327. Byrne, D. (1961) The repression-sensitisation skill: rationale, reliability, and validity. J. Personal., 29, 334349. Byrne, D. (1964) Repression-sensitisation as a dimension of personality. In B. A. Maher (ed.), Progress in Experimental Personal Research. New York: Academic Press. Byrne D. G. (1975a) A psychophysiological distinction between different types of depressive states. Aust. & N.Z.J. Psychiat., 8, 261267. Byrne, D. G. (1975b) Note on decision time/movement, time relationships in normal and depressed subject. Percept. Mot. Skills, 41, 907910. Byrne, D. G. (1976a) Choice reaction times in depressive states. Br. J. Soc. Clin. Psychol., 15, 149156. Byrne D. G. (1976b) Vigilance on arousal in depressive states. Br. J. Soc. Clin. Psychol., 15, 267274. Byrne, D. G. (1977) Affect and vigilance performance in depressive illness. J. Psychiat. Res., 13, 185191. Caine, E.D., Yerevanian, B.I. & Bamford, K.A. (1984) Cognitive function and the dexamethasone suppression test in depression. Amer. J. Psychiat., 141, 116118. Calev, A. (1984) Recall and recognition in chronic nondemented schizophrenics: use of matched task. J. Abnorm. Psychol., 93, 172177.
Calev, A. & Erwin, P. G. (1985) Recall and recognition in depressives: use of matched task. Br. J. Clin. Psychol., 24, 127128. Calev, A., Nigal, D. & Chazran, S. (1989) Retrieval from semantic memory using meaningful and meaningless constructs by depressed, stable bipolar and manic patients. Br. J. Clin. Psychol., 28, 6773. Calvo, M. G., Eysenck, M. W. and Estevaz, A. (1994) Ego-threat interpretive bias in test anxiety: on-line inferences. Cog. Emot., 2, 127146. Campbell, D. (1957) A study of some sensory-motor functions in psychiatric patients. Unpublished PhD thesis, University of London. Cane, D. B. & Gotlib, I. H. (1985) Depression and the effects of positive and negative feedback on expectations, evaluations and performance. Cog. Ther. Res., 9, 145160. Cantela, J. R. & Baron, M. G. (1977) Covert conditioning: a theoretical analysis. Behav. Mod., 1, 351368. Carpenter, P., Miyake, A. & Just, M. A. (1994) Language comprehension: sentence and discourse processing Ann. Rev. Psychol., 46, 91100. Carrier, M. L. & Pashler, H. (1995) Attentional limits in memory retrieval. J. Exp. Psychol. Learn. Mem. Cog., 21, 13391348. Carroll, J. M., Thomas, J. C. & Malhotra, A. (1980) Presentation and representation in design problem solving. Br. J. Psychol., 71, 143153. Carroll, J. S. (1978) The effect of imagining an event on expectations for the event: An interpretation in terms of the availability heuristic. J. Exp. Soc. Psychol., 14, 8896. Caspar, F., Rothenfluh, T. & Segal, Z. (1992) The appeal of connectionism for clinical psychology. Clin. Psychol. Rev., 12, 719762. Casscells, W., Schoenberger, A. & Grayboys, T. (1978) Interpretations by physicians of clinical laboratory results. New Engl. J. Med., 299, 9991000. Cassens, G., Wolfe, L. & Zola, M. (1990) The neuropsychology of depressions. J. Neuropsychiat. Clin. Neurosci., 2, 202213. Page 325
Cassiday, K. L., McNally R. J. & Zeitlin, S. B. (1992) Cognitive processing of trauma cues in rape victims with post-traumatic stress disorder. Cog. Ther. Res., 16, 283295 Cave, K. R. & Wolfe, J. M. (1990) modeling the role of parallel processing in visual search. Cog. Psychol., 22, 225271. Cermak, L. M. & Craig, F. I. M. (1979) Levels of Processing in Human Memory. Hillsdale, N.J.: Lawrence Erlbaum. Challis, B. H., Chin, C. Y., Kerr, S. A., Law, J., Schneider, L., Yonelinas, A. & Tulving, E. (1993) Perceptual and conceptual cueing in implicit and explicit retrieval. Mem., 1, 127151. Chamberlin, C. J. & Magill, R. A. (1992) The memory representaion of motor skills: a test of schema theory. J. Mot. Behav. 24, 309319.
Chambers, D. & Reisberg, D. (1985) Can mental images be ambiguous? J. Exp. Psychol. Hum. Percept. Perf., 11, 317328. Chan, C. (1992) Implicit cognitive processes: Theoretical issues and applications in computer systems design. Unpublished doctoral dissertation. University of Oxford. Channon, S., Baker, J. E. & Robertson, M. M. of (1993) Effects of structure and clustering on recall and recognition memory in clinical depression. J. Abnorm. Psychol., 102, 323326. Channon, S. & Hayward, A. (1990) The effect of short-term fasting on processing of food cues in normal subjects. Int. J. Eat. Dis., 9, 447452. Channon, S. M., Hemsley, D. & de Silva, P. (1988) Selective processing of food words in anorexia nervosa. Br. J. Clin. Psychol., 27, 259260. Chapman, C. R. & Feather, B. W. (1972) Modification of perception by classical conditioning procedures J. Exp. of Psychol., 93, 338342. Chapman, L. J. (1967) Illusory correlation in observation reports. J. Verb. Learn. Verb. Behav., 6, 151155. Chapman, L. J. & Chapman, J. P. (1967) Genesis of popular but erroneous psychodiagnostic observation J. Abnorm. Psychol., 72, 193204. Chapman, L. J. & Chapman, J. P. (1969) Illusory correlation as an obstacle to the use of valid psychodiagnostic signs. J. Abnorm. Psychol., 74, 271280. Chapman, L. J. & Chapman, J. P. (1973) Disordered thought in Schizophrenia. New York: Appleton-Century-Crofts. Charniak, E. (1972) Towards a model of children's story comprehension. Technical Report 266, Artificial Intelligence Laboratory, Massachusetts Institute of Technology. Chase, C. & Tallal, P. (1990) A developmental, interactive activation model of the word superiority effect. J. Exp. Ch. Psychol., 49, 448487. Cheal, M. & Lyon, D. R. (1992) Attention in visual search: Multiple search classes. Percept. Psychophys., 52, 113138. Cheesman, J. & Merickle, P. A. (1985) Word recognition and consciousness. In D. Besner, T. G. Waller & G. E. Mackinnon (eds), Reading Research: Advance in Theory and Practice (pp. 311352). Orlando, FL: Academic Press. Cheesman, J. & Merickle, P. A. (1986) Distinguishing conscious from unconscious perceptual process. Canad. J. Psychol., 40, 343367. Cherry E. C. (1953) Some experiments on the recognition of speech, with one and with two ears. J. Acoustic. Soc. Amer., 25, 975979. Chi, M. T. H. (1978) Knowledge structures and memory development. In R. S. Siegler (ed.), Children's Thinking: What develops? Hillside, N.J.: Lawrence Erlbaum. Page 326
Chmiel, N., Totterdell, P., & Folkard, S. (1995) On adaptive control, sleep loss, and fatigue. Appl. Cog. Psychol., 9, 3953. Christian, J., Bickley, W., Tarka, M. & Clayton, K. T. (1978) Measures of free recall of 900
English nouns: Correlations with imagery, concreteness, meaningfulness and frequency. Mem. Cog., 6, 379390. Christianson, S. A. & Loftus, E. F. (1987) Memory for traumatic events. Appl. Cog. Psychol., 1, 225239. Christianson, S. A. & Loftus, E. F. (1991) Remembering emotional events: The fate of detailed information. Cog. Emot., 5, 81108. Christianson, S. A., Loftus, E. F., Hoffman, H. & Loftus, G. R. (1991) Eye fixations and memory for emotional events. J. Exp. Psych: Learn. Mem. Cog., 17, 693701. Christianson, S. A. & Safer, M. A. (1996) Emotional events and emotions in autobiographical memories. In D. C. Rubin (ed.) Remembering our Past: Studies in Autobiographical Memory. Cambridge: Cambridge University Press. Church, B.A. & Schacter, D.L. (1994) Perceptual specificity of auditory priming: implicit memory for voice intonation and fundamental frequency. J. Exp. Psychol.: Learn. Mem. Cog., 20, 521533. Cirilo, R. K. (1981) Referential coherence and text structure in story comprehension. J. Verb. Learn. Verb. Behav., 2, 358367. Claeys, W. (1989) Social anxiety, evaluative threat and incidental recall of trait words. Anx. Res., 2, 2743. Clark, A. (1989) Microcognition: Philosophy, Cognitive Science, and Parallel Distributed Processing. Cambridge, Mass.: MIT Press. Clark, D.A. (1986) Factors influencing the retrieval and control of negative cognitions. Behav. Res. Ther., 24, 151159. Clark, D. A. & de Silva, P. (1985) The nature of depressive and anxious intrusive thoughts: distinct or uniform phenomena. Behav. Res. Ther., 23, 383393. Clark, D. M. (1988) A cognitive model of panic attacks. In S. Rachman & J. D. Maser (eds), Panic: Psychological Perspectives. (pp. 7189). Hillsdale, N.J.: Erlbaum. Clark, D. M. & Teasdale, J. D. (1982) Diurnal variations in clinical depression and accessibility of memories of positive and negative experiences. J. Abnorm. Psychol., 91, 8795. Clark, D. M. & Teasdale, J. D. (1985) Constraints of the effects of mood on memory. J. Pers. Soc. Psychol., 48, 15981608. Clark, D.C., Clayton, P.J., Andraesen, N.C. et al. (1985) Intellectual functioning and abstraction ability in major affective disorders. Comp. Psychiat., 26, 313325. Clark, D. M., Salkovskis, P. M., Gelder, M., Koehler, C. K., Anastasides, P., Hackmann, A., Middleton, H. & Jeavons, A. (1988) Tests of a cognitive theory of panic. In I. Hand & H. Wittehan (eds), Panic and Phobias. Berlin: Springer-Verlag. Clark, H. H. & Chase, W. G. (1972) On the process of comparing sentences against pictures. Cog. Psychol., 3, 452517. Clark, H. H. & Marshall, C. R. (1981) Definite reference and mutual knowledge. In A. K. Joshi, B. L. Webber & I. A. Sag (eds), Elements of Discourse Understanding. Cambridge:
Cambridge University Press. Cloitre, M. & Liebowitz, M. R. (1991) Memory bias in panic disorders: an investigation of this cognitive avoidance hypothesis. Cog. Ther. Res., 15, 371386. Cloitre, M. Shear, M. K., Cancienne, J. & Zeitlin, S. B. (1994) Implicit and explicit memory for catastrophic associations to bodily sensations words in panic. Cog. Ther. Res., 18, 225240. Page 327
Cohen, J. D., Dunbar, K. & McClelland, J. L. (1990) On the control of automatic processes: a parallel distributed processing account of the Stroop effect. Psychol. Rev., 97, 332361. Cohen, J. D. & Servan-Schreiber, D. (1989) A parallel distributed processing approach to behaviour and biology in schizophrenia. Technical Report AIP-100. Dept. of Psychology, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213. Cohen, N. J., Ivry, R. & Keele, S. W. (1990) Attention and structure in sequence learning. J. Exp. Psychol.: Learn. Mem. Cog., 16,1730. Cohen, R. M., Weingartner, H., Smallberg, S. A., Pickar, D. & Murphy, D. L. (1982) Effort and cognition in depression. Arch. Gen. Psychiat., 39, 593597. Colby, C. A. & Gotlib, I. H. (1988) Memory deficits in depression. Cog. Ther. Res., 12, 611627. Colby, C. L. (1991) The neuroanatomy and neurophysiology of attention. J. Ch. Neurol., 6, 90118. Conway, A. R. A. & Engle, R. W. (1994) Working memory and retrieval: A resource dependent inhibition model. J. Exp. Psychol.: Gen., 123, 354373. Cooper, A. & Marshall, P. H. (1985) Spatial location judgements as a function of intention to learn and mood state: an evaluation of an alleged automatic encoding operation. Amer. J. Psychol., 98, 261269. Cooper, M. J., Anastasiades, P. & Fairburn, C. G. (1992) Selective processing of eating-, shape-, and weight-related words in persons with bulimia nervosa. J. Abnorm. Psychol., 101, 352355. Cooper, M. J. & Fairburn, C. G. (1992) Selective processing of eating, weight and shape related words in patients with eating disorders and dieters. Br. J. Clin. Psychol., 31, 363365. Cooper, R. & Franks, B. (1993) Interruptibility as a constraint on hybrid systems. Minds Machines, 3, 7396. Coover, J. E. (1917) Experiments in Psychical Research at Leland Stanford Junior University, California. Stanford, California: Stanford University Press. Cornell, D. G., Suarez, R. & Berent, S. (1984) Psychomotor retardation in melancholic and nonmelancholic depression: cognitive and motor components. J. Abnorm. Psychol., 93, 150157. Corteen, R. S. & Dunn, D. (1974) Shock associated words in a non-attended message: a test for momentary awareness. J. Exp. Psychol., 102, 11431144. Corteen, R. S. & Wood, B. (1972) Autonomic responses to shock associated words. J. Exp.
Psychol., 9, 308313. Courrieu, P. (1994) Connectionism and symbolic functions. Psychol. Française, 39, 231236. Covington, M. V., Omelich, C. L. & Schwarzer, R. (1986) Anxiety, aspirations and selfconcept in the achievement process: a longitudinal model with latent variables. Mot. Emot., 10, 7188. Cowen, L. W. (1983) A fear of thunderstorms: a case study. Aust. J. Clin. Hypnother. Hypnosis, 4, 7982. Cowey, A. & Stoerig, P. (1992) The neurobiology of blindsight. Trends in Neurosci., 29, 580. Cox, W. M. & Klinger, E. (1988) A motivational model of alcohol use. J. Abnorm. Psychol., 97, 168180. Coyne, J. C. & Gotlib, I. (1986) Studying the role of cognition in depression: well-trodden paths and cul-de-sacs. Cog. Ther. Res., 10, 695705. Page 328
Coyne, J. C. & Gotlib, I. (1983) The role of cognition in depression: a critical appraisal. Psychol. Bull., 94, 472505. Craig, M. J., Humphreys, M. S., Rocklin, T. & Revelle, W. (1979) Impulsivity, neuroticism and caffeine: Do they have additive effects on arousal? J. Res. Person., 13, 404419. Craik, F. I. M. & Lockhart, R. S. (1972) Levels of processing: a framework for memory research. J. Verb. Learn. Verb. Behav., 11, 671684. Craik, K. J. W. (1966) The Nature of Psychology. Cambridge: Cambridge University Press. Craske, M. G., Rapee, R. M., Jackel, L. & Barlow, D. H. (1989) Qualitative dimensions of worry in DSM-III-R generalized anxiety disorder subjects and non-anxious controls. Behav. Res. Ther., 27, 397402. Cromwell, R. L. & Levenkron, J. C. (1984) Psychological care of acute coronary patients. In A. Steptoe and A. Mathews (eds), Health Care and Human Behaviour. London: Academic Press. Cronholm, B. & Ottosson, J.-O. (1961) The experience of memory function after electroconvulsive therapy. Br. J. Psychiat., 109, 251258. Crossman, E. R. S. W. (1953) Entropy and choice-time: the effect of frequency on balance on choice response. J. Exp. Psychol., 5, 4151. Cuthbert, B. D., Melamed, B. G., Cook, E. W. III, McNeil, D. W. & Lang, P. J. (1987) Phobic imagery and the assessment of anxiety disorders. Psychophysiol., 24, 585. Dagenbach, D., Carr, T. H. & Wilhelmsen, A. L. (1989) Task-induced strategies and nearthreshold priming: conscious influences on unconscious perception. J. Mem. Lang., 28, 412443. Dalgleish, T. (1995) Performance on the emotional Stroop task in groups of anxious, experts and control subjects: a comparison of computer and card presentation formats. Cog. Emot., 9, 341362 Dalgleish, T. & Watts, F. N. (1990) Biases of attention and memory in disorders of anxiety and
depression. Clin. Psychol Rev., 10, 589604. Danaher, B. & Thoresen, C. (1972) Imagery assessment of self-report and behavior measures. Behav. Res. Ther., 10, 131138. Danion, J. M., Willard-Schroeder, D., Zimmermann, M. A., Grange, D., Sclienger, J.L. & Singer, L. (1991) Explicit memory and repetition priming in depression: preliminary findings. Arch. Gen. Psychiat, 48, 707711. Dannenbaum, S. E., Parkinson, S. R. & Inman, V. W. (1988) Short-term forgetting: comparisons between patients with dementia of the Alzheimer type, depressed, and the normal elderly. Cog. Neuropsychol., 5, 213233. Dark, V. J., Vochatzer, K. G. & Van Voorhis, B. A. (1996) Semantic and spatial components of selective attention. J. Exp. Psychol.: Human Percept. Perf., 22, 6381. Darke, S. (1988) Anxiety and working memory capacity. Cog. Emot., 2, 145154. Davey, G. L. (1994) Pathological worrying as exacerbated problem solving. In G. L. Davey & F. Tallis (eds) Worrying: Perspectives on Theory. Assessment and Treatment. (pp. 3560) Chichester: John Wiley. Davidson, D. (1994) Recognition and recall of irrelevant and interruptive atypical actions in script-based stories. J. Mem. Lang., 33, 757775. Davis, H. (1979) Self-reference and the encoding of personal information in depression. Cog. Res. Ther., 4, 97110. Dawes, R. M. (1964) Cognitive distortion. Psychol. Rep., 14, 443459. Dawkins, K. & Furnham, A. (1989) The color naming of emotional words. Br. J. Psychol., 80, 383389. Page 329
de-Jong, P. & Merckelbach, H. (1991) Covariation bias and electrodermal responding in spider phobics before and after behavioural treatment. Behav. Res. Ther., 29, 307314. de-Jong, P. J., Merckelbach H. & Arntz, A. (1995) Covariation bias in phobic women: the relationship between a priori expectancy, on-line expectancy, autonomic responding, and a posteriori contingency judgment. J. Abnorm. Psychol., 104, 5562. de-Jong, P. J., Merckelbach, H., Arntz, A. and Nijman, J. (1992) Covariation detection in treated and untreated spider phobics. J. Abnorm. Psychol., 101, 724727. De Jong, R. (1993) Multiple bottlenecks in overlapping task performance. J. Exp. Psychol.: Human Percept. Perf., 19, 965980. Debner, J. A. & Jacoby, L. L. (1994) Unconscious perception: attention, awareness and control. J. Exp. Psychol.: Learn Mem. Cog., 20, 304317. Deffenbacher, J. L. (1980) Worry and emotionality in test anxiety. In J. G. Sarason (ed.), Test Anxiety, Theory, Research and Application. Hillsdale, N.J.: Lawrence Erlbaum. DeMarco, S. & Harrell, R. M. (1995), Perception of word junctures by children. Percept. Mot. Skills, 80, 10751082. DeMonbreun, B. G. & Craighead, W. E. (1977) Selective recall of positive and neutral
feedback. Cog. Ther. Res., 1, 311329. Dennett, D.C. (1991) Consciousness Explained. Boston: Little Brown Books. Denny, E. B. & Hunt, R. R. (1992) Affective valence and memory in depression: dissociation of recall and fragment completion. J. Abnorm. Psychol., 101, 575580. Den Uyl, M. & Van Oostendorp, H. (1980) The use of scripts in text comprehension. Poetics, 9, 275294. Derry. P. A. & Kuiper, N. A. (1981) Schematic processing and self-reference in clinical depression. J. Abnorm. Psychol., 90, 286297. Di Caprio, N. S. (1970) Essentials of verbal satiation therapy: a learning-theory-based behaviour therapy. J. Couns. Psychol., 17, 419424. Dienes, Z., Altmann, G. T. M., Kwan, L. & Goodge, A. (1995) Unconscious knowledge of artificial grammars is applied strategically. J. Exp. Psychol.: Learn. Mem. Cog., 21, 13221338. Dienes, Z., Broadbent, D. & Berry, D. (1991) Implicit and explicit knowledge bases in artificial grammar learning. J. Exp. Psychol.: Learn. Mem. Cog., 17, 875887. Dillon, R. F. (ed.) (1985) Individual Differences in Cognition, Vol. 2. New York: Academic Press. Dixon, N. F. (1968) Perception without awareness. A reply to K. M. Banreti-Fuchs. Acta Psychol., 28, 171180. Dixon, N. F. (1971) Subliminal Perception: the Nature of a Controversy. London: McGrawHill. Dixon, N. F. (1981) Preconscious Processing. London: John Wiley. Dixon, W. A., Heppner, P. P., Burnett, J. W. et al. (1993) Distinguishing among antecedents, concomitants, and consequences of problem-solving appraisal and depressive symptoms. J. Couns. Psychol., 40, 357364. Dobson, K. S. & Shaw, B. F. (1987) Specificity and stability of self-referent encoding in clinical depression. J. Abnorm. Psychol., 96, 3440. Doctor, R. M. & Altman, F. (1969) Worry and emotionality as components of test anxiety, with replication and further data. Psychol. Rep., 24, 563568. Donnelly, E. F., Waldman, I. N., Murphy, D. L., Wyatt, R. J. & Goodwin, F. K. (1980) Primary affective disorder; thought disorder in depression. J. Abnorm. Psychol., 89, 315319. Page 330
Dooling, D. J. & Lachman, R. (1971) Effects of comprehension on retention of prose. J. Exp. Psychol., 88, 216222. Dooling, D. J. & Mullet, R. (1973) Locus of thematic effects in retention of prose. J. Exp. Psychol., 97, 404406. Dorfman, D. D. (1967) Recognition of taboo words as a function of a priori probability. J. Pers. Soc. Psychol., 7, 110. Dorfman, D. D., Grosberg, J. M. & Kroecker, L. (1965) Recognition of taboo stimuli as a
function of exposure time. J. Pers. Soc. Psychol., 2, 552562. Dorfman, J. (1994) Sublexical components in implicit memory for novel words. J. Exp. Psychol.: Learn. Mem. Cog., 20, 11081125. Dornic, S. (1977) Mental Load, Effort and Individual Differences. Reports of the Department of Psychology University of Stockholm, No. 509. Dreyfus, H. L. (1979) What Computers Can't Do: A Critique of Artificial Reason, 2nd edition. New York: Harper and Row. Dulany, D.E., Carlson, R. & Dewey, G. (1984) A case of syntactic learning and judgement: How concrete and how abstract? J. Exp. Psychol.: Gen., 113, 541555. Dulany, E. (1957) Avoidance learning of perceptual defence and vigilance. J. Abnorm. Soc. Psychol., 55, 333338. Dunbar, D. C. & Lishman, W. A. (1984) Depression, recognition-memory, and hedonic tone: a signal detection analysis. Br. J. Psychiat., 144, 376382. Duncan, J. (1980) The locus of interference in the perception of simultaneous stimuli. Psychol. Rev., 87, 272300. Duncan, J. & Humphreys, G.W. (1989) Visual search and stimulus similarity. Psychol. Rev., 96, 433458. Duncker, K. (1939) The influence of past experience upon perception of properties. Amer. J. Psychol., 52, 255265. Dunlap, K. (1900) The effect of imperceptual shadows on judgement of distance. Psychol. Rev., 7, 435453. Dutta, A., Schweickert, R., Choi, S. & Proctor, R.W. (1995) Cross-task cross talk in memory and perception. Acta Psychologica, 90, 4962. Dutta, S. & Kanungo, R. N. (1975) Affect and Memory: A Reformulation. Oxford: Pergamon. Dyckman, J. & Cowan, P. (1978) Imaging vividness and the outcome of in vivo and imagined scene desensitisation. J. Consult. Clin. Psychol., 46, 11551156. East, M. P. & Watts, F. N. (1994) Worry and the suppression of imagery. Behav. Res. Ther., 32, 851855. Ebbinghaus, H. (1885) Uber das Gedachtnis. Leipzig: Duncker and Humbolt. Edmundson, E. D. & Nelson, D. L. (1976) Anxiety, imagery and sensory interference. Bull. Psychonom, Soc., 8, 319322. Ehlers, A., Margraf, J., Davies, S. & Roth, W. T. (1988) Selective processing of threat cues in subjects with panic attack. Cog. Emot., 2, 201219. Elliott, C. L. & Greene, R. L. (1992) Clinical depression and implicit memory. J. Abnorm. Psychol., 101, 572574. Ellis, H. C. & Ashbrook, P. W. (1988) Resource allocation model of the effects of depressed mood states on memory. In K. Fiedler and J. Forgas (eds), Affect, Cognition and Social Behaviour. Toronto: Hogrefe.
Ellis, H. C. & Ashbrook, P. W. (1989) The 'state' of mood and memory research: A selective review. Special Issue: Mood and memory: Theory, research and applications. J. Soc. Behav. Pers., 4, 121. Page 331
Ellis, H. C., Thomas, R. L., McFarland, A. D. & Lane, J. W. (1985) Emotional mood states and retrieval in episodic memory. J. Exp. Psychol.: Learn, Mem, Cog., 11, 363370. Ellis, H., Thomas, R. L. & Rodrigez, I. A. (1984) Emotional mood states and memory: elaborative encoding, semantic processing and cognitive effort. J. Exp. Psychol.: Learn. Mem. Cog., 69, 237243. Ellis, H. C. & Ashbrook, P. W. (1988) Resource allocation model of the effects of depressed mood states on memory. In K. Fiedler & J. Forgas (eds), Affects, Cognition and Social Behaviour. Toronto: Hogrefe. Ellis, H. C., Varner, C. J., Becker, A. S. &. Ottaway, S. A. (in press) Emotion and prior knowledge in memory and comprehension of ambiguous sentences. Cog. Emot. Elman, J. L. (1995) Language as a dynamic system. In R. F. Port & T. V. Gelder (eds), Mind as Motion: Explorations in the Dynamics of Cognition. Cambridge, Mass.: MIT Press. Enns, J. T. (1990) Three-dimensional features that pop out in visual search. In D. Brogan (Ed.), Visual Search. London, UK: Taylor and Francis. Enns, J. T. & Resnick, R. A. (1990) Influence of scene-based properties of visual search. Science, 247, 721723. Enright, S. J. & Beech, A. R. (1990) Obsessional states: Anxiety disorders of schizo-types? An information processing and personality assessment. Psychol. Med., 20, 621627. Enright, S. J. & Beech, A. R. (1993a) Reduced cognitive inhibition in obsessive compulsive disorder. Br. J. Clin. Psychol., 32, 6774. Enright, S. J. & Beech, A. R. (1993b) Further evidence of reduced cognitive inhibition in obsessive-compulsive disorder. Pers. Indiv. Diff., 14, 387395. Erdelyi, M. (1985) Psychoanalysis: Freud's Cognitive Psychology. New York: W. H. Freeman. Erdelyi, M. H. (1974) A new look at the new look: perceptual defence and vigilance. Psychol. Rev., 81(1), 125. Erdelyi, M. H. & Goldberg, B. (1979) Let's not sweep repression under the rug: toward a cognitive psychology of repression. In J. F. Kihlstrom and F. J. Evans (eds), Functional Disorders of Memory. Hillsdale, N.J.: Lawrence Erlbaum. Erdley, C. A. & D'Agostino, P.R. (1988) Cognitive and affective components of automatic priming effects. J. Pers. Soc. Psychol., 54, 741747. Ericsson, K. A. & Simon, H. A. (1984) Protocol Analysis: Verbal Reports on Data. Cambridge, Mass.: MIT Press. Eriksen, C. W. (1963) Perception and personality. In N.J. Wepman and R. W. Heine (eds), Concepts of Personality. Chicago: Aldine. Eriksen, C. W. (1990) Attentional search of the visual field. In D. Brogan (ed.), Visual Search
(pp. 221240) London: Taylor and Francis. Eriksen, C. W. & Schultz, D. W. (1979) Information processing in visual search: A continuous flow conception and experimental results. Percept. Psychophys., 25, 249263. Eskridge, T. C. (1994) A hybrid model of continuous analogical reasoning. In K. J. Holyoak & J. A. Barnden (eds), Analogical Connections: Advances in Connectionist and Neural Computation Theory, Vol. 2 (pp. 207246). Norwood, NJ: Ablex. Estes, W.K. (1991) Cognitive architectures from the standpoint of an experimental psychologist. Ann. Rev. Psychol, 42, 128. Page 332
Evans, J., Williams, J. M. G., O'Loughlin, S. & Howells, K. (1992) Autobiographical memory and problem solving strategies of individuals who parasuicide. Psychol. Med., 22, 399405. Eysenck, M. W. (1979a) Anxiety, learning and memory: a reconceptualisation. J. Res. Pers. 13, 363385. Eysenck, M. W. (1979b) Depth, elaboration, and distinctiveness. In L. S. Cermak & F. I. M. Craik (eds), Levels of Processing in Human Memory. Hillsdale, N.J.: Lawrence Erlbaum. Eysenck, M. J. (1982) Attention and Arousal: Cognition and Performance. Berlin: SpringerVerlag. Eysenck, M. W. (1988) Anxiety and attention. Anxiety Res., 1, 915. Eysenck, M.W. (1992) Anxiety: The Cognitive Perspective. Hove, N.J.: Erlbaum. Eysenck, M. & Byrne, A. (1994) Implicit memory bias, explicit memory bias and anxiety. Cog. Emot., 8, 415432. Eysenck, M.W. & Calvo, M.G. (1992) Anxiety and performance: The processing efficiency theory. Cog. Emot., 6, 409434. Eysenck, M. W., MacLeod, C. & Mathews, A. (1987) Cognitive functioning and anxiety. Psychol. Res., 49, 189195. Eysenck, M. W. & Mogg, K. (1992) Clinical anxiety, trait anxiety, and memory bias. In S.-A. Christianson (ed.), The Handbook of Emotion and Memory: Research and Theory. (pp. 429450). Hillsdale, N.J.: Erlbaum. Eysenck, M. W., Mogg, K., May, J., Richards, A. & Mathews, A. (1991) Bias in interpretation of ambiguous sentences related to threat in anxiety. J. Abnorm. Psychol., 100, 144150. Eysenck, M. W. & van Berkum, J. (1992) Trait anxiety, defensiveness, and the structure of worry. Pers. Indiv. Diff., 13, 12851290. Farah, M. J. (1994) Visual perception and visual awareness after braindamage: a tutorial overview. In C. Umilta and M. Moscovitch (eds), Attention and Performance, Vol. XV: Conscious and Nonconscious Information Processing (pp 3776). Cambridge, Mass.: Bradford Press. Farne, M. (1963) Effects of so-called subliminal stimuli upon visual perception. Psychol. Res. Bull. 3, 173182. Faust, M., Kravetz, S. & Babkoff, H. (1993) Hemisphericity and top-down processing of
language. Brain and Lang., 44, 118. Fehrer, E. & Smith, E. (1962) Effects of luminance ratio on masking. Percep. Mot. Skills. 14, 243253. Fennell, M. J. V., Teasdale, J. D., Jones, S. & Damle, A. (1987) Distraction in neurotic and endogenous depression: an investigation of negative thinking in major depressive disorders. Psychol. Med., 17, 441452. Fielder, K., Semin, G. R. & Bolten, S. (1989) Language use and reification of social information: top-down and bottom-up processing in person cognition. Europ. J. Social Psychol., 19, 271295. Finger, R. & Galassi, J. P. (1977) Effects of modifying cognitive versus emotionality, responses in the treatment of text anxiety. J. Consult. Clin. Psychol., 45, 280287. Fischer, S. C., Hickey, D. T., Pellegrino, J. W. & Law, D. J. (1994) Strategic processing in dynamic spatial reasoning tasks. Learn. Indiv. Diff., 6, 65105. Fischhoff, B. & Beyth, R. (1975) I knew it would happen: remembered probabilities of oncefuture things. Organ. Behav. Hum. Perf., 13, 116. Page 333
Fischhoff, B., Slovic, P. & Lichtenstein, S. (1977) Knowing with certainty: the appropriateness of extreme confidence. J. Exp. Psychol.: Hum. Percept. Perf., 3, 552564. Fisher, C. (1960) Subliminal and supraliminal influences on dreams. Amer. J. Psychiat., 116, 10091017. Fitzgerald, J. M., Slade, S. & Lawrence, R. H. (1988) Memory availability and judged frequency of affect. Cog. Ther. Res., 12, 379390. Foa, E. B., Feske, U., Murdock, T. B., Kozak, M. J. & McCarthy, P. R. (1991) Processing of threat-related information in rape victims. J. Abnorm. Psychol., 100, 156162. Foa, E. G. & Kozak, M. J. (1986) Emotional processing and fear: exposure to corrective information. Psychol. Bull., 99, 2035. Foa, E. B. & McNally, R. J. (1986) Sensitivity to feared stimuli in obsessive-compulsives: a dichotic listening analysis. Cog. Ther. Res., 10, 477486. Foa, E. B., Ilai, D., McCarthy, P. R., Shoyer, B. & Murdock, T. (1993) Information processing in obsessive-compulsive disorder. Cog. Ther. Res., 17, 173189 Foa, E. B., McNally, R. J. & Murdock, I. B. (1989) Anxious mood and memory, Behav. Res. Ther., 27, 141147. Fodor, J. A. (1983) The Modularity of Mind. Cambridge, Mass.: MIT Press. Fodor, J. A. & Pylyshyn, Z. W. (1988) Connectionism and cognitive architecture: a critical analysis. Cognition, 28, 371. Fogarty, S. J. & Hemsley, D. R. (1983) Depression and the accessibility of memories: a longitudinal study. Br. J. Psychiat., 142, 610617. Forgas, J. P. (1995) Mood and judgment: the Affect Infusion Model (AIM). Psychol. Bull., 117, 3966.
Forster, K., Booker, J., Schacter, D. L. & Davis, C. (1990) Masked repetition priming: lexical activation or novel memory trace? Bull. Psychonom. Soc., 28, 341345, Forster, P. M. & Govier, E. (1978) Discrimination without awareness. Quart. J. Exp. Psychol., 30, 282292. Foulds, G. A. (1952) Temperamental differences in maze performance II: The effect of distraction and of electroconvulsive therapy on psychomotor retardation. Br. J. Psychiat., 43, 3341. Fowler, C.A., Wolford, G., Slade, R. & Tassinary, L. (1981) Lexical access with and without awareness. J. Exp. Psychol.: Gen., 110, 115125. Fox, E. (1993) Attentional bias in anxiety: selective or not? Behav. Res. Ther., 31, 487493. Fox, E. (1994) Attentional bias in anxiety: a defective inhibition hypothesis. Cog. Emot., 8, 165195. Fox, E. (1995) Negative priming from ignored distractors in visual selection: a review. Psychon. Bull. Rev. 2, 145173. Franks, J. J. (1974) Towards understanding understanding. In W. B. Weimer and D. S. Palermo (eds), Cognition and the Symbolic Processes. Hillsdale, N.J.: Lawrence Erlbaum. Fransson, A. (1977) On qualitative differences in learning IV: Effects of intrinsic motivation and extrinsic test anxiety on process and outcome. Br. J. Educ. Psychol., 47, 244257. Frederiksen, C. H. (1975) Acquisition of semantic information from discourse: effects of repeated exposures. J. Verb. Learn. Verb. Behav., 14, 158169. Freeman, J. T. (1954) Set or perceptual defence? J. Exp. Psychol., 48, 283288. Freud, S. (1901) The Psychopathology of Everyday Life. London: Hogarth. Page 334
Friedman, A. (1979) Framing pictures: the role of knowledge in automatised encoding and memory for gist. J. Exp. Psychol.: Gen., 108, 316355. Friedman, A. B., Polson, M. C., Defoe, C. G. & Gaskill, S. J. (1982) Dividing attention within and between hemispheres: testing a multiple resources approach to limitedcapacity information processing. J. Exp. Psychol.: Hum. Percept. Perf., 8, 625650. Frisby, J. P. (1980) Seeing, Illusion, Brain and Mind. Oxford: Oxford University Press. Frith, C. D., Stevens, N., Johnstone, E. C., Deakin, P. L., Lawler, P. & Crowe, T. J. (1983) Effects of ECT and depression on various aspects of memory. Br. J. Psychiat., 142, 610617. Fromholt, P., Larsen, P. & Larsen, 5. (1995) Effect of late-onset depression and recovery on autobiographical memory. J. Gerontol: Psychol. Sci. Soc. Sci., 50, 7481. Fudin, R. & Benjamin, C. (1992) Subliminal psychodynamic activation: updates comprehensive list of experimental results and comments on previous lists. Percept. Mot.
Skills, 72, 959977. Galambos, J. A. (1982) Normative studies of six characteristics of our knowledge of common activities. Cog, Sci. Tech. Rep., 14, New Haven: Yale University. Galambos, J. A. & Rips, L. J. (1982) Memory for routines. J. Verb. Learn. Verb. Behav., 21, 260281. Galassi, J. P., Frierson, H. T., Jr & Sharver, R. (1981a) The behavior of high, moderate, and low test anxious students during an actual test situation. J. Consult. Clin. Psychol., 49, 5162. Galassi, J. P., Frierson, H. T., Jr & Sharver, R. (1981b) Concurrent versus retrospective assessment in text anxiety research. J. Consult. Clin. Psychol., 49, 614615. Galassi, J. P., Frierson, H. T., Jr & Siegel, R. G. (1984) Cognitions, test anxiety, and test performance: a closer look J. Consult. Clin. Psychol., 51, 292293. Ganzer, V. J. (1968) Effect of awareness presence and test anxiety on learning and, retention in a serial learning situation. J. Pers. Soc. Psychol., 8, 194199. Garnham, A. (1982) On-line Construction of Representations of the Content of Texts. Bloomington, Indiana: University Linguistics Club. Garnham, A. (1983) What's wrong with story grammars? Cognition, 15, 145154. Garnham, A., Oakhill, J. V. & Johnson-Laird, P. N. (1982) Referential continuity and the coherence of discourse. Cognition, 11, 2946. Garrod, S. C. & Sanford, A. J. (1982) Bridging inferences and the extended domain of reference. In J. Long and A. Baddeley (eds), Attention and Performance, Vol. IX, Hillsdale, N.J.: Lawrence Erlbaum. Garrod, S. C. & Sanford, A. J. (1983) Topic-dependent effects in language processing. In G. B. Flores D'Arcais and J. Jarvella (eds), Processes of Language Understanding. Chichester: Wiley. Gass, C. S. & Russell, E. W. (1986) Differential impact of brain damage and depression on memory test performance. J. Consult. Clin. Psychol., 54, 261263. Geary, D. C., Frensch, P. A. & Wiley, J. G. (1993) Simple and complex mental subtraction: strategy choice and speed-of-processing differences in younger and older adults. Psychol. Ageing, 8, 242256. Gellatly, A. R. H. (1980) Perception of an illusory triangle with masked inducing figure. Perception, 9, 599602. Geller, V. & Shaver, P. (1976) Cognitive consequences of self-awareness. J. Exp. Soc. Psychol., 12, 99108.
Page 335
Gentner, D. R. (1976) The structure and recall of narrative prose. J. Verb. Learn. Verb. Behav., 15, 411418. Gerlsma, C., Kramer, J. J. A. M., Scholing, A. & Emmerlkamp, P. M. G. (1994) The influence of mood on memories of parental rearing practices. Br. J. Clin. Psychol., 33, 159172. Gernsbacher, M. A. & Faust, M. E. (1991) The mechanism of suppression: a component of general comprehension skill. J. Exp. Psychol.: Learn., Mem. Cog., 17, 245262. Gernsbacher, M. A., Goldsmith, H.Hill & Robertson, R.R. (1992) Do readers mentally represent character's emotional states? Cog. Emot., 6, 89111. Gerrig, R. J. & Bower, G. H. (1982) Emotional influences on word recognition. Bull. Psychonom. Soc., 19, 197200. Gilligan, S. G. & Bower, G. H. (1984) Cognitive consequences of emotional arousal. In C. Izard, J. Kagan & R. Zajonc (eds), Emotions, Cognitions and Behaviour. New York: Cambridge University Press. Glaser, M. O. & Glaser, W. R. (1982) Time course analysis of the Stroop phenomenon. J. Exp. Psychol.: Hum. Percept. Perf., 8, 875894. Glass, R. M., Uhlenhuth, E. H., Hastel, F. W., Matuzas, W. & Fischman, M. W. (1981) Cognitive dsyfunction and imipramine in outpatient depressives. Arch. Gen. Psychiat., 38, 10481051. Glucksberg, S. & Cohen, N. J. (1970) Memory for non-attended auditory material. Cog. Psychol., 1, 149156. Godden, D. R. & Baddeley, A. D. (1975) When does context influence recognition memory? Br. J. Psychol., 71, 99104. Gogel, W. C. & Newton, R. E. (1969) Perception of off-sized objects. Percept. Psychophys., 5, 79. Going, M. & Read, J. D. (1974) Effect of uniqueness, sex of subject, and sex of photograph on facial recognition. Percept. Mot. Skills., 39, 109110. Goldiamond, I. (1958) Indicators of perception: I. Subliminal perception, subception, unconscious perception: an analysis in terms of psychophysical indicator methodology. Psychol. Bull., 55, 373411. Goldiamond, I. (1962) Perception. In A. J. Bachrach (ed.), Experimental Foundations of Clinical Psychology., New York: Basic Books. Goldin, S. E. (1978) Memory for the ordinary: typicality effects in chess memory. J. Exp. Psychol.: Hum. Learn. Mem., 4, 605616. Goldstein, A. J. & Chance, J. E. (1985) Effects of training of Japanese face recognition: reduction of the other-race effect. Bull. Psychonom. Soc., 23, (3), 211214.
Goldstein, M. J. & Barthol, R. P. (1960) Fantasy responses to subliminal stimuli. J. Abnorm. Soc. Psychol., 68, 2226. Golinkoff, M. & Sweeney, J.A. (1989) Cognitive impairments in depression. J. Affect. Disord., 17, 105112. Golombok, S., Starron, A., Bonn, J. & Mogg, K. (1991) The effects of diazepam on anxiety related cognition. Cog. Ther. Res., 15, 459467. Gomez, R. L. & Schvaneveldt, R. W. (1994) What is learned from artificial grammars? Transfer tests of simple associations. J. Exp. Psychol.: Learn. Mem. Cog., 20 396410. Goodman, G. S. (1980) Picture memory: how the action schema affects retention. Cog. Psychol., 12, 473495. Goodwin, A. M. & Williams, J. M. G. (1982) Mood induction research its implications for clinical depression. Behav. Res. Ther., 20, 373382. Page 336
Gordon, G. (1967) Semantic determination by subliminal verbal stimuli: a quantitative approach. PhD thesis, University of London. Gordon, P. K. (1985) Allocation of attention in obsessional disorder. Br. J. Clin. Psychol., 24, 101107. Gorham, D. R. (1956) A proverbs test for clinical and experimental user. Psychol. Rep., 2, 112. Gotlib, I. H. (1983) Perception and recall of interpersonal feedback: negative bias in depression. Cog. Ther. Res., 7, (5), 399412. Gotlib, I. H. & Asarnow, R. F. (1979) Interpersonal and impersonal problem-solving in mildly and clinically depressed university students. J. Consult. Clin. Psychol., 47, 8595. Gotlib, I. H. & Cane, D. B. (1987) Construct accessibility and clinical depression: a longitudinal investigation. J. Abnorm. Psychol., 96, 199204. Gotlib, I. H. & Hammen, C.L. (1992) Psychological Aspects of Depression: Towards a Cognitive-Interpersonal Integration. Chichester: John Wiley. Gotlib, I. H. & McCann, C. D. (1984) Construct accessibility and depression: an examination of cognitive and affective factors. J. Pers. Soc. Psychol., 47, 427439. Gotlib, I.H., McLachlan, A.L. & Katz, A.N. (1988) Biases in visual attention in depressed and nondepressed individuals. Cog. Emot., 2, 185200. Govier, E. & Pitts, M. (1982) The contextual disambiguation of a polysemous word in an unattended message. Br. J. Psychol., 73, 537545. Graesser, A. C. (1981) Prose Comprehension Beyond the Word. New York: Springer-Verlag. Graesser, A. C. & Nakamura, G. V. (1982) The impact of a schema on comprehension and memory. Psychol. Learn. Motiv., 16, 59109. Graesser, A. C., Singer, M. & Trabasso, T. (1994) Constructing inferences during narrative text comprehension. Psychol. Rev., 101, 371395.
Graesser, A. C., Woll, S. B., Kowalski, D. J. & Smith, D. A. (1980) Memory for typical and atypical actions in scripted activities. J. Exp. Psychol.: Hum. Learn. Mem., 6, 503513. Graf, P. & Mandler, G. (1984) Activation makes words more accessible, but not necessarily more retrievable. J. Verb. Learn. Verb. Behav., 23, 553568. Graf, P. & Schacter, D. L. (1985) Implicit and explicit memory for new associations in normal and amnesic subjects. J. Exp. Psychol.: Learn. Mem. Cog., 11, 501518. Graf, P. & Schacter, D. L. (1989) Unitization and grouping mediate dissociations in memory for new associations. J. Exp. Psychol.: Learn. Mem. Cog., 15, 930940. Graf, P. (1994) Explicit and implicit memory: a decade of research. In C. Umilta and M. Moscovitch (eds), Attention and Performance, Vol. XV: Conscious and Nonconscious Information Processing (pp. 681696) Cambridge, Mass.: Bradford. Graf, P. & Masson, M. E. J. (1993) Implicit Memory: New Directions in Cognition, Development and Neuropsychology. Hillsdale, NJ: Erlbaum. Graf, P., Mandler, G. & Haden, P. E. (1982) Simulating amnesic symptoms in normal subjects. Science, 218, 12431244. Graf, P., Shimamura, A. P. & Squire, L. R. (1985) Priming across modalities and priming across category levels: extending the domains of preserved functioning in amnesia. J. Exp. Psychol.: Learn. Mem. Cog., 11, 385395. Granholm, E., Asarnow, R. F. & Marder, S. R. (1991) Controlled information processing resources and the development of automatic detection responses in schizophrenia. J. Abnorm. Psychol., 100, 2230. Gray, J. A. (1990) Brain systems that mediate both emotion and cognition. Cog. Emot., 4, 269288. Page 337
Grayson, J. B., Foa, E. B. & Steketee, G. (1982) Habituation during exposure treatment: distraction vs. attention focusing. Behav. Res. Ther., 20, 323328. Greenberg, L. & Safran, J. D. (1984a) Integrating affect and cognition: a perspective on the process of therapeutic change. Cog. Ther. Res., 8, 559578. Greenberg, L. & Safran, J. D. (1984b) Hot cognition-emotion coming in from the cold: a reply to Rachman. Cog. Ther. Res., 8, 591598. Greenberg, M. S. & Alloy, L. B. (1989) Depression vs anxiety: processing of self-and otherreferent information. Cog. Emot. 3, 207223. Greenberg, M. S. & Beck. A. T. (1989) Depression vs anxiety: a test of the content-specificity hypothesis. J. Abnorm. Psychol., 98, 913. Greenwald, A. G. & Banaji, M. R. (1989) The self as a memory system: powerful, but ordinary. J. Pers. Soc. Psychol., 57, 4154. Greenwald, A. G., Klinger, M. R. & Liu, T.J. (1989) Unconscious processing of dichoptically masked words. Mem. Cog., 17, 3547. Greenwald, A. G., Klinger, M. R. & Schuh, E. S. (1995) Activation by marginally perceptible
('subliminal') stimuli: dissociation of unconscious from conscious cognition. J. Exp. Psychol.: Gen., 124, 2242. Gregory, W. L., Cialdini, R. B. & Carpenter, K. M. (1982) Self-relevant scenarios as mediators of likelihood estimates and compliance: does imagining make it so? J. Pers. Soc. Psychol., 43(1), 8999. Griffin, J. A., Dember, W. N. & Warm, J.S. (1986) Effects of depression on expectancy in sustained attention. Mot. Emot., 10, 195205. Griggs, S. S. & Green, D. W. (1983) How to make a good cup of tea: exploring the scripts of thought-disordered and non-thought-disordered patients. Br. J. Med. Psychol., 56, 125133. Grosz, B. (1977) The representation and use of focus in a system for understanding dialogues. In Proceedings of the 5th International Joint Conference on Artificial Intelligence. Cambridge, Mass.: MIT. Grosz, B. (1978) Focusing in Dialogue. Tindap-II. Theoretical Issues in Natural Language Processing, ACM and ACL. New York: Academic Press. Grove, W. M. & Andreasen, N. C. (1985) Language and thinking in psychosis: is there an input abnormality? Arch. Gen. Psychiat., 42, 2632. Gustafson, R. & Kallmen, H. (1991) Subliminal psychodynamic activation: an experiment controlling for major possible confounding influences outlines by Fudin. Percept. Mot. Skills, 73, 163171. Guttentag, R.E. & Lange, G. (1994) Motivational influences on children's strategic remembering. Learn. Indiv. Diff., 6, 309330. Haber, R. N. & Hershenson, M. (1973) The Psychology of Visual Perception. London: Holt, Rinehart and Winston. Hadley, R. F. (1993) Connectionism, explicit rules and symbolic manipulation. Minds Machines., 3, 183200. Hamilton, V. (1983) The Cognitive Structures and Processes of Human Motivation and Personality. Chichester: John Wiley. Hammen, C., Dyck, D. G. & Miklowitz, D. J. (1986) Stability and severity parameters of depressive self-schema responding. J. Soc. Clin. Psychol., 4, 2345. Hammen, C., Marks, T. Mayol, A. & de Mayo, R. (1985) Depressive self-schemas, life stress and vulnerability to depression, J. Abnorm, Psychol., 94, 308319. Harber, K. D. & Pennebaker, J. W. (1992) Overcoming traumatic memories. In S-A. Christianson (ed.), The Handbook of Emotion and Memory Research and Theory. Hillsdale, N.J.: Lawrence Erlbaum Associates. Page 338
Hardaway, R. A. (1990) Subliminally activated symbiotic fantasies: facts and artifacts. Psychol. Rev., 107, 177195. Hare, R. D. & Blevings, G. (1975) Defensive responses to phobic stimuli. Biol. Psychol., 3, 113.
Harper, R. S. (1953) The perceptual modification of colored figures. Amer. J. Psychol., 66, 8689. Harvey, J. M., Richards, J. C., Dziadosz, T. & Swindell, A. (1993) Misinterpretation of ambiguous stimuli in panic disorder. Cog. Ther. Res. 17, 235248. Hasher, L. & Zacks, R. T. (1979) Automatic and effortful processes in memory. J. Exp. Psychol.: Gen., 108, 356388. Hasher, L., Rose, K. C., Zacks, R. T., Sanft, H. & Doren, B. (1985) Mood, recall and selectivity effects in normal college students. J. Exp. Psychol.: Gen., 114, 104118. Haspel, K. C. & Harris, R. S. (1982) Effect of tachistoscopic stimulation to subconscious Oedipal wishes on competitive performance: a failure to replicate. J. Abnorm. Psychol., 91, 437444. Hayes, N. & Broadbent, D. E. (1988) Two modes of learning for interactive tasks. Cog., 28, 249276. Healy, D. & Williams, J. M. G. (1988) Dysrhythmia, dysphasia and depression: the interaction of learned helplessness and circadian dysrhythmia in the pathogenesis of depression. Psychol. Bull., 103, 163178 Hemsley, D. R. & Zawada, S. L. (1976) Filtering and the cognitive deficit in schizophrenia. Br. J. Psychiat., 128, 456461. Henderson, L. (1977) Word recognition. In N. S. Sutherland (ed.), Tutorial Essays in Experimental Psychology, Vol. 1. Hillsdale, N.J.: Lawrence Erlbaum. Henley, S. H. A. (1975) Cross-modal effects of subliminal verbal stimuli. Scand. J. Psychol., 16, 3036. Henley, S. H. A. & Dixon, N. F. (1974) Laterality differences in the effects of incidental stimuli upon evoked imagery. Br. J. Psychol., 65(4), 529536. Henry, G. M., Weingartner, H. & Murphy, D. L. (1973) Influence of affective states and psychoactive drugs on verbal learning and memory. Amer. J. Psychiat., 130, 966971. Hermans, D., De Houwer, J. & Eelen, P. (1994) The affective priming effect: automatic activation of evaluative information in memory. Cog. Emot., 8, 515533. Hermans, D., De Houwer, J. & Eelen, P. (in press) Evaluative decision latencies mediated by induced affective states. Behav. Res. Ther. Hertel, P. & Hardin, T. S. (1990) Remembering with and without awareness in a depressed mood: evidence of deficits in initiative. J. Exp. Psychol.: Gen., 119, 4559. Hertel, P. T. & Rude, S. S. (1991a) Recalling in a state of natural or experimental depression. Cog. Ther. Res., 15, 103127. Hertel, P. T. & Rude, S. S. (1991b) Depressive deficits in memory: Focusing attention improves subsequent recall. J. Exp. Psychol.: Gen., 120, 301309. Heuer, F. & Reisberg, D. (1990) Vivid memories of emotional events: The accuracy of remembered minutiae. Mem. Cog., 18, 496506. Hicks, W. E. (1952) On the rate of gain of information. Quart. J. Exp. Psychol., 4, 1126.
Hill, A. B. & Dutton, F. (1989) Depression and selective attention to self-esteem threatening words. Pers. Indiv. Diff, 10, 915919. Hill, A. B. & Knowles, T. H. (1991) Pers. Indiv. Diffs., 12, 481485. Page 339
Hill, T., Lewicki, P. & Neubauer, R. M. (1991) The development of depressive encoding dispositions: a case of self-perpetuation of biases. J. Exp. Soc. Psychol., 27, 392409. Hillyard, S. A. & Kutas, M. (1983) Electrophysiology of cognitive processing. Ann. Rev. Psychol., 34, 3361. Hines, D., Czerwinski, M., Sawyers, P.K. & Dwyer, M. (1986) Automatic priming effects: effect of category levels and word association level. J. Exp. Psychol.: Hum. Percept. Perf. 12, 370379. Hinton, G. E., McClelland, J. L. & Rumelhart, D. E. (1986) Distributed representations. In J. L. McClelland and D. E. Rumelhart (eds), Parallel Distributed Processing: Explorations in the Microstructure of Cognitions, Vol. 1: Foundations. Cambridge, Mass.: Bradford Books. Hirsch, C. (1995) Anxiety and cognitive schemata. Unpublished PhD Thesis, Univ. of Cambridge. Hirst, W. (1986) Aspects of divided and selective attention. In J. Le-Doux and W. Hurst (eds), Mind and Brain. New York: Cambridge University Press. Hirst, W. & Kalmer, D. (1987) Characterising attentional resources. J. Exp. Psychol.: Gen., 116, 6881. Hiscock, M. (1978) Imagery assessment through self-report: What do imagery questionnaires really measure?. J. Consult. Clin. Psychol., 46, 223231. Hobbes, T. (1651) Leviathan. London: Andrew Crooke. Hochberg, J. E. (1978) Perception. Engelwood Cliffs, N.J.: Prentice Hall. Hockey, G. R. J. (1986) A state control theory of adaptation to stress and individual differences in stress management. In A. W. K. Gailland & M. G. H. Coles (eds), Energetics and Human Information Processing. Dordrecht: Nijhoff. Holender, D. (1986) Semantic activation without conscious identification. Behav. Brain Sci., 9(1), 1166. Holmes, D. A. (1991) Effects of subliminal psychodynamic activation. Diss. Abs. Int., 52, 2801B. Holmes, D. S., (1974) Investigations of repression: differential recall of material experimentally or naturally associated with ego threat. Psychol. Bull., 81, 632653. Homme, L. E. (1965) Control of coverants, the operants of the mind. Psychol. Rec., 15, 501511. Hope. D. A., Rapee, R. M., Heimberg, R. G. & Dombeck, M. J. (1990) Representations of the self in social phobia: vulnerability to social threat. Cog. Ther. Res. 14, 477485. Horowitz, M. J. (1975) Intrusive and repetitive thoughts after experimental stress. Arch. Gen.
Psychiat., 32, 14571463. Horowitz, M. J. (1983) Image Formation and Psychotherapy. New York: Aronson. Horowitz, M.D. & Reibord, S.P. (1992) Memory, emotion and response to trauma. In S-A. Christianson (ed.), The Handbook of Emotion and Memory. Hillsdale, N.J.: Lawrence Erlbaum. Houghton, G. (1994) Inhibitory control of neurodynamics: opponent process mechanisms in sequencing selective attention. In M. Oaksford & G. D. A. Brown (eds), Neurodynamics and Psychology (pp. 107155). London: Academic Press. Houghton, G. & Tipper, S. P. (1996) Inhibitory mechanisms of neural and cognitive control: Applications to selective attention and sequential action. Brain Cog., 30, 2043. Page 340
Howard, D. V. & Howard, J. H. (1992) Adult age differences in the rate of learning serial patterns: evidence from direct and indirect tests. Psychol. Ageing, 7, 232241. Howes, D. H. (1954) A statistical theory of the phenomenon of subception. Psychol. Rev., 61, 98110. Howie, D. (1952) Perceptual defence. Psychol. Rev., 59, 308315. Hudesman, J., Page, W. & Rautianen, J. (1992) Use of subliminal stimulation to enhance learning mathematics. Percept. Mot. Skills, 74, 12191224. Hume, D. (1739) A Treatise of Human Nature, Vols I and II. London: Noon. Humphreys, K. J., Asbury, A. J. & Miller, K. (1990) Investigation of awareness by homophone priming during computer controlled anaesthesia. In B. Bonke, W. Fitch & K. Millar (eds), Memory and Awareness in Anaesthesia (pp. 103112) London: Balliere. Hyman, R. (1953) Stimulus information as a determinant of reaction time. J. Exp. Psychol., 45, 188196. Idzikowski, C. & Baddeley, A. (1987) Fear and performance in novice parachutist. Ergonom., 30, 14631474. In I. G. Sarason (ed.), Test Anxiety: Theory, Research, and Application. Hillsdale, N.J.: Lawrence Erlbaum. Ingram, R. E., Partridge, S., Scott, W. & Bernet, C. Z. (1994) Schema specificity in subclinical syndrome depression: distinctions between automatically versus effortfully encoded state and trait depressive information. Cog. Ther. Res. 18, 195209. Ingram, R. & Reed, M. (1986) Information encoding and retrieval processes in depression: findings, issues and future directions. In R. Ingram (ed.) Information Processing Approaches to Clinical Psychology. Orlando, Florida: Academic Press. Ingram, R. E., Kendall, P. C., Smith T. W., Donnell. C. & Ronan, K. (1982) Cognitive specificity in emotional distress. J. Pers. Soc. Psychol., 53, 734742. Isen, A. M. (1984) Toward understanding the role of affect in cognition. In R. Wyer & T. Srull (eds), Handbook of Social Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Isen, A. M. (1985) The asymmetry of happiness and sadness in effects of memory in normal college students. J. Exp. Psychol.: Gen., 114, 388391.
Isen, A. M., Johnson, M. M. S., Merts, E. & Robinson, G. S. (1985) The influence of positive affect on the unusualness of word associations. J. Pers. Soc. Psychol., 48, 14131426. Isen, A. M., Shalker, T. E., Clark, M. & Carp, L. (1978) Affect, accessibility of material in memory, and behavior: a cognitive loop. J. Pers. Soc. Psychol., 36(1), 112. Jacoby, L. L. (1983) Remembering the data: analysing interactive processes in reading. J. Verb. Learn. Verb. Beh. 22, 485508. Jacoby, L. L. (1991) A process dissociation framework: separating automatic from intentional uses of memory. J. Mem. Lang. 30, 513541. Jacoby, L. L., Allan, L. G., Collins, J. C. & Larwill, L. K. (1988) Memory influences subjective experience: noise judgments. J. Exp. Psychol.: Learn. Mem. Cog., 14, 240247. Jacoby, L. L. & Dallas, M. (1981) On the relationship between autobiographical memory and perceptual learning. J. Exp. Psychol.: Gen., 110, 115125. Jacoby, L. L. & Kelley, C. M. (1992) A process-dissociation framework for investigating unconscious influences: Freudian slips, projective tests, subliminal perception, and signal detection theory. Curr. Dir. Psychol., 1, 174179. Jacoby, L. L., Lindsay, D. S. & Toth, J. P. (1992) Unconscious influences revealed: attention, awareness and control. Amer. Psychol., 47, 802809. Page 341
Jacoby, L. L., Ste-Marie, D. & Toth, J. P. (1993) Redefining automaticity: unconscious influences, awareness and control. In A. Baddeley and L. Weiskrantz (eds), Attention: Selection, Awareness, and Control. A Tribute to Donald Broadbent, (pp. 261282). Oxford: Clarendon Press. Jacoby, L. L. & Witherspoon, D. (1982) Remembering without awareness. Canad. J. Psychol., 36, 300324. Jacoby, L. L. & Whitehouse, K. (1989) An illusion of memory: false recognition influenced by unconscious perception. J. Exp. Psychol.: Gen., 118, 126135. Jacoby, L. L., Woloshyn, V. & Kelley, C. (1989) Becoming famous without being recognized: unconscious influences of memory produced by dividing attention. J. Exp. Psychol.: Gen., 118, 115125. James, W. (1890/1950) The Principles of Psychology. New York: Dover. Jelicic, M., Bonke, B., Wolters, G. & Phaf, R. H. (1992a) Implicit memory for words presented during general anaesthesis. Europ. J. Cog. Psychol., 4, 7180. Jelicic, M., de Roode, A., Boville, J. G. & Bonke, B. (1992b) Unconcious learning during anaesthesia. Anaesthesia, 47, 835837. Jennings, J. M. & Jacoby, L. L. (1993) Automatic vs intentional uses of memory: ageing, attention, and control. J. Exp. Psychol.: Learn. Mem. Cog., 19, 501518. Jensen, A. R. & Rohwer, W. D. (1966) The Stroop colour word test: a review. Acta Psychol., 25, 3693. Johnsen, B. H., Laberg, J.C., Cox, W. M., Vaksdal, A. & Hugdahl, K. (1994) Alcoholic
participants' attentional bias in the processing of alcohol-related words. J. Addict. Behav., 8, 111115. Johnson, E. J. & Tversky, A. (1983) Affect, generalisation, and the perception of risk. J. Pers. Soc. Psychol., 45(1), 2031. Johnson, M. H. & Magaro, P. A. (1987) Effects of mood and severity on memory processes in depression and mania. Psychol. Bull., 101, 2840. Johnson, M. K. (1985) The origin of memories. In Advances in Cognitive-Behavioural Research and Therapy, Vol. 4. New York: Academic Press. Johnson, M. K., Kim, J. K. & Risse, G. (1985) Do alcoholic Korsakoff's syndrome patients acquire affective reactions? J. Exp. Psychol.: Learn. Mem. Cog., 11, 2236. Johnson, M. K. & Hirst, W. (1993) MEM: Memory subsystems as processes. In A. F. Collins, S. E. Gathercole, M. A. Conway & P. E. Morris, (eds) Theories of Memory (pp. 241286). Hove: Erlbaum. Johnson, M. K. & Multhaup, K. S. (1992) Emotion and MEM. In S.A. Christianson (ed.) The Handbook of Emotion and Memory: Research and Theory, (pp. 3366) Hove: Erlbaum. Johnson, R. E. & Scheidt, V. J. (1977) Organisational encoding in the serial learning of prose. J. Verb. Learn, Verb. Behav., 16, 575588. Johnson-Laird, P. N. (1980) Mental models in cognitive science. Cog. Sci., 4, 71115. Johnson-Laird, P. N. (1983) Mental Models: Towards a Cognitive Science of Language, Inference and Consciousness. Cambridge: Cambridge University Press. Johnson-Laird, P. N. (1988) A computational analysis of consciousness. The control operations of consciousness. In A. J. Marcel & E. Bisiach (eds), Consciousness in Contemporary Science (pp. 357389) Oxford: Clarendon Press. Johnson-Laird, P. N., Herrmann, D. J. & Chaffin, R. (1984) Only connections: a critique of semantic networks. Psychol. Bull., 96, 292315. Johnston, W. A. & Dark, V. J. (1982) In defense of intra-perceptual theories of attention. J. Exp. Psychol.: Hum. Percept. Perf., 8, 407421. Page 342
Johnston, W. A. & Dark, V. J. (1985) Dissociable domains of selective processing. In M. I. Posner and O. S. M. Marin (eds), Attention and Performance, Vol. XI. Hillsdale, N.J.: Lawrence Erlbaum. Johnston, W. A. & Dark, V. J. (1986) Selective attention. Ann. Rev. Psychol., 37, 4375. Jonides, J. & Gleitman, H. (1972) A conceptual category effect in visual search: Percept. Psychophys., 12, 457560. Kahneman, D. (1973) Attention and Effort. Englewood Cliffs, NJ.: Prentice Hall. Kahneman, D. & Chajczyk, D. (1983) Tests of the automaticity of reading: dilution of Stroop effects by color-irrelevant stimuli. J. Exp. Psychol.: Hum. Percept. Perf. 9, 497509. Kahneman, D. & Treisman, A. (1984) Changing views of attention and automaticity. In R. Parasuraman & D. R. Davies (eds), Varieties of Attention. Orlando, FL.: Academic Press.
Kahneman, D. & Tversky, A. (1972) Subjective probability: a judgement of representativeness. Cog. Psychol., 3, 430454. Kahneman, D., Slovic, P. & Tversky, A. (eds) (1982) Judgement under Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press. Kahneman, D., Treisman, A. & Burkell, J. (1983) The cost of visual filtering. J. Exp. Psychol.: Hum. Percept. Perf., 9, 510522. Kantowitz, B. H. & Knight, J. (1976) Test tapping and time-sharing II: Use of auditory secondary task. Acta Psychol., 40, 343362. Kaspi, S. P., McNally, R. J. & Amir, N. (1995) Cognitive processing of emotional information in posttraumatic stress disorder. Cog. Ther. Res., 19, 433444. Kavanagh, D. J. & Bower, G. H. (1985) Mood and self-efficacy: impact of joy and sadness on perceived capabilities. Cog. Ther. Res., 9(5), 507525. Kaye, N. (1975) The therapeutic value of three merging stimuli for male schizophrenics. Unpublished doctoral thesis, Yeshiva University. Keele, S. W. (1973) Attention and Human Performance. Pacific Pallisades, California: Good Year. Kelly, L. P. (1995) Processing of bottom-up and top-down information by skilled and average deaf readers and implications for whole language instruction. Except. Childn, 6, 318334. Kemp-Wheeler, S. M. & Hill, A. B. (1987) Anxiety responses to subliminal experiences of mild stress. Br. J. Psychol., 78, 365374. Kemp-Wheeler, S. M. & Hill, A. B.(1988) Semantic priming without awareness: some methodological considerations and replications. Quart. J. Exp. Psychol., 40A, 671692. Kennedy, R. E. & Craighead, W. E (1988) Differential effects of depression and anxiety on recall of feedback in a learning task. Behav. Ther., 19, 437454. Kepecs, J. G. (1954) Observations on screens and barriers in the mind. Psychoanal. Quart., 23, 6277. Kieras, D. (1978) Beyond pictures and words: alternative information-processing models for imagery effects in verbal memory. Psychol. Bull., 85, 532554. Kieras, D. E. & Meyer, D. E. (1994) The EPIC architecture for modelling human informationprocessing and performance: a brief introduction. Technical Report TR-94/ONR-EPIC-1, University of Michigan, Ann Arbor. Kihlstrom, J. F. (1993) The continuum of consciousness. Consc. Cog., 2, 334354. Kinchla, R. A. (1992) Attention. Ann. Rev. Psychol. 43, 711742. Kinsbourne, M. & Hicks, R. (1978) Functional cerebral space. In J. Requin (ed.), Attention and Performance, Vol. VII. Hillsdale, N.J.: Lawrence Erlbaum. Page 343
Kinsbourne, M. & Warrington, E. K. (1962) A disorder of simultaneous form perception. Brain, 85, 461486.
Kinsbourne, M. & Warrington, E. K. (1963) The localising significance of limited simultaneous visual form perception. Brain, 86, 697702. Klapp, S. T. & Lee, P. (1974) Time of occurrence cues for 'unattended' auditory material. J. Exp. Psychol., 102, 176177. Klein, G. S. (1964) Semantic power measured through the interference of words with colour naming. Amer. J. Psychol., 77, 576588. Klein, F. B. & Kihlstrom, J. F. (1986) Elaboration organisation and the self-reference effect in memory. J. Exp. Psychol.: Gen., 115, 2638. Klein, S. B., Loftus, J. & Burton, H. A. (1989) Two self-reference effects: the importance of distinguishing between self-descriptiveness judgments and autobiographical retrieval in selfreferent encoding. J. Pers. Soc. Psychol., 56, 853865. Kleinsmith, L. J. & Kaplan, S. (1963) Paired associate learning as a function of arousal and interpolated interval. J. Exp. Psychol., 65, 190193. Klieger, D, M. & Cordner, M. D., (1990) The Stroop task as measure of construct accessibility in depression. Pers. Indiv. Diff. 11, 1927. Klinger, E. (1978) Dimensions of thought and imagery in normal waking states. J Alt. States Consc., 4, 97113. Klinger, E. (1980) Therapy and the flow of thought. In J. E. Shorr, G. E. Sobel, P. Robin & J. A. Cornella (eds), Imagery: Its Many Dimensions and Applications. New York: Plenum. Klinger, E. (1987) The interview questionnaire technique: reliability and validity of a mixed idiographic-nomothetic measure of motivation. In J. N. Butcher & C. D. Spielberger (eds), Advances in Personality Assessment, Vol 6 (pp. 3148) Hillsdale, NJ: Erlbaum. Klinger, E., Barter, S. G. & Maxeiner, M. E. (1981) Current concerns: assessing therapeutic relevant motivation. In P. C. Kendall and S. D. Hollon (eds), Assessment Strategies for Cognitive Behavioural Interventions. New York: Academic Press. Klinger, M. R. & Greenwald, A. G. (1995) Unconscious priming of association judgements. J. Exp. Psychol.: Learn. Mem. Cog., 21, 569581. Knopman, D. S. & Nissen, M. J. (1987) Implicit learning in patients with probable Alzheimer's disease. Neurol., 37, 784788 Knowlton, B. J. & Squire, L.R. (1994) The information acquired during artificial grammar learning. J. Exp. Psychol.: Learn. Mem. Cog., 20, 7991. Knowlton, B. J., Ramus, S. J. & Squire, L. R. (1992) Intact artificial grammars learning in amnesia: dissociation of abstract knowledge and memory for specific instances. Psychol. Sci., 3, 172179. Koh, S. D., Kayton, L. & Berry, R. (1973) Mnemonic organization in young non-psychotic schizophrenics. J. Abnorm. Psychol., 81, 299310. Kolers, P. A. (1983) Perception and representation. Ann. Rev. Psychol., 34, 129166. Kolers, T. A. (1962) Intensity and contour effects in visual masking. Vis. Res., 2, 277294.
Kolodner, J. (1985) Memory for experience. In G. H. Bower (ed.), Psychology of Learning and Motivation, Vol. 19. Hillsdale, N.J.: Lawrence Erlbaum. Kopelman, M. (1986) Clinical tests of memory. Br. J. Psychiat., 148, 517525. Kosslyn, S. M. (1980) Image and the Mind. Cambridge, Mass.: Harvard University Press. Page 344
Kosslyn, S. M., Ball, T. M. & Reiser, B. J. (1978) Visual images preserve metric spatial information: evidence from studies of imagery scanning. J. Exp. Psychol.: Hum. Percept. Perf., 4, 4760. Kothera, L., Fudin, R. & Nicastro, R. (1990) Effects of subliminal psychodynamic activation on dart throwing performance: Another nonreplication. Percept Mot. Skill, 71, 10151022. Kovacs, M. & Beck, A. T. (1978) Maladaptive cognitive structures in depression. Amer J. Psychiatry., 135, 525533. Krames, L. & McDonald, M. R. (1985) Distraction and depressive cognition. Cog. Ther. Res., 9, 561573. Krantz S. & Hammen, C. (1979) Assessment of cognitive bias in depression. J. Abnorm. Psychol., 88, 611619. Krueger, L. E. (1981)Is identity or regularity more salient than difference or irregularity? Paper presented at the meeting of the American Psychological Association, Los Angeles, California. Kuhn, T. S. (1962) The Structure of Scientific Revolutions. Chicago: Chicago University Press. Kuiper, N. & MacDonald, M. (1982) Self and other perception in mild depressives. Soc. Cog., 1, 223239. Kukla, A. (1972) Foundations of an attributional theory of performance. Psychol. Rev., 79, 454470. Kunst-Wilson, W. R: & Zajonc, R. B. (1980) Affective discrimination of stimuli that cannot be recognised. Science, 237, 14451452. Kuyken, W. & Brewin, C. R. (1995) Autobiographical memory fuctioning in depression and reports of early abuse. J. Abnorm. Psychol., 104, 585591. Kuyken, W. & Dalgleish, T. (1995) Autobiographical memory and depression. Br. J. Clin. Psychol., 34, 8992. LaBerge, D. (1981) Automatic information processing: a review. In J. Long and A. D. Baddeley (eds), Attention and Performance, Vol. IX. Hillsdale, N.J.: Lawrence Erlbaum. Lacy, O. W., Lewinger, N. & Adamson, J. F. (1953) Foreknowledge as a factor affecting Perceptual defence and alertness. J. Exp. Psychol., 45, 169174. Landau, R. J. (1990) The role of semantic schemata in phobic word interpretation. Cog. Ther. Res., 4, 427434. Lang, P. J. (1977) Fear imagery: an information-processing analysis. Behav. Ther., 8, 862886. Lang, P. J. (1979) A bio-informational theory of emotional imagery. Psychophysiol., 16, 495512.
Lang, P. J. (1984) Cognition in emotion: concept and action. In C. E. Izard, R. R. Zajonc & J. Kagan (eds), Emotions, Cognition and Behaviour. Cambridge: Cambridge University Press. Lang, P. J. (1985) The cognitive psychophysiology of emotion: fear and anxiety. In A. H. Tuma & J. Maser (eds), Anxiety and the Anxiety Disorders. Hillsdale, N.J.: Lawrence Erlbaum. Lang, P.J. (1987) Image as action: a reply to Watts and Blackstock (1987) Cog. Emot., 1, 407426. Lang, P. J., Kozak, M. J., Miller, G. A., Levin, D. N. & MacLean, A. (1980) Emotional imagery: conceptual structure and pattern of somato-visceral response. Psychophysiol., 17, 179192. Lang, P. J., Levin, D. N., Miller, G. A. & Kozak, M. J. (1983) Fear behaviour, fear imagery and the psychophysiology of emotion: the problem of affective response integration. J. Abnorm. Psychol., 92, 276306. Page 345
Lang, P. J., Melamed, B. G. & Hart, J. (1970) 'A psychophysiological analysis of fear modification using an automated desensitisation procedure. J. Abnorm. Psychol., 76, 229234. Lavy, E. H., van den Hout, M. & Arntz, A. (1993) Attentional bias and spider Phobia: conceptual and clinical issues. Behav. Res. Ther, 31, 1724. Lavy, E. H. & van den Hout, M. (1993) Attentional bias for appetitive cues: effects of fasting on normal subjects. Behav. Res. Ther., 31, 297310. Lavy, E. H., van Oppen, P. & van den Hout, M. (1994) Selective processing of emotional information in obsessive compulsive disorder. Behav. Res. Ther, 32, 243246. Lazarus, R. G. (1984) On the primacy of cognition. Amer. Psychol., 39, 124129. Lazarus, R. L. (1991) Emotion and Adaptation. Oxford: Oxford University Press. Lazarus, R. S. (1982) Thoughts on the relations between emotion and cognition. Amer. Psychol., 37, 10191010. Lee, I. & Tyrer, P. (1980) Responses of chronic agoraphobics to subliminal and supraliminal phobic motion pictures. J. Nerv. Ment. Dis., 168(1), 3440. Lee, I., Tyrer, P. & Horn, S. (1983) A comparison of subliminal, supraliminal and faded phobic cine-films in the treatment of agoraphobia. Br. J. Psychiat., 143, 356361. Leeper, R. W. (1935) A study of a neglected portion of the field of learning - the development of sensory organisation. J. Genetic Psychol., 46, 4175. Leigh, J. H. (1991) Information processing differences among broadcast media: review and suggestions for research. J. Advert., 20, 7175. Leight, K. A. & Ellis, H. D. (1981) Emotional mood states, strategies and state dependency in memory. J. Verb. Learn. Verb. Behav., 20, 251266. Lemaire, P. & Siegler, R. S. (1995) Four aspects of strategic change: contributions to children's learning of multiplication. J. Exp. Psychol.: Gen., 124, 8397. Leslie, A. D. (1987) Pretense and representation: the origins of 'theory of mind'. Psychol. Rev., 94, 412426.
Levene, M. (1966) Hypothesis behaviour by humans during discrimination learning. J. Exp. Psychol., 71, 331338 Leventhal, H. (1984) A perceptual-motor theory of emotion. In L. Berkowitz (ed.), Advances in Experimental Social Psychology, Vol. 17. New York: Academic Press. Leventhal, H. & Cupchik G. (1976) A process model of human judgement. J. Commun., 26, 190204. Leventhal, H. & Scherer, K. (1987) The relationship of emotion to cognition: a functional approach to a semantic controversy, Cog. Emot., 1, 328. Levine, D.S. (1994) Learning and encoding higher order rules in neural networks. Behav. Res. Meth. Instr. Compt., 27, 178182. Levine, D.S. & Aparicio, M. (1994) Neural Networks for Knowledge Representation and Inference (pp. 241268). Hillsdale, NJ: Erlbaum. Levy, L. H. (1958) Perceptual defence in tactual perception. J. Pers., 26, 467468. Levy, R. & Maxwell, A. E. (1968) The effect of verbal context on the recall of schizophrenics and other psychiatric patients. Br. J. Psychiat., 114, 311316. Lewicki, P., Czyzewska, M. & Hoffman, H. (1987) Unconscious acquisition of complex procedure knowledge. J. Exp. Psychol.: Learn. Mem. Cog., 13, 523530. Lewicki, P., Hill, T. & Bizet, E. (1988) Acquisition of knowledge about a pattern that cannot be articulated. Cog. Psychol., 20, 2437. Lewinsohn, P. M., Steinmetz, J. L., Larson, D. W. & Franklin, J. (1981) Depression-related cognitions: antecedents or consequences? J. Abnorm. Psychol., 90, 213219. Page 346
Lewis, J. L. (1970) Semantic processing of unattended messages during dichotic listening. J. Exp. Psychol., 85, 225228. Leyens, J., Cisneros, T. & Hossay, J. (1976) Decentration as a means for reducing aggression after exposure to violent stimuli. Europ. J. Soc. Psychol., 6, 459473. Lichtenstein, E. H. & Brewer, W. F. (1980) Memory for goal-directed events. Cog. Psychol., 12, 412445. Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M. & Combs, B. (1978) Judged frequency of lethal events. J. Exp. Psychol.: Human Learn. Mem., 4(6), 551578. Liepnann, H. (1908) Uber die agnostischen Storungen. Neurologisches Zentralblatt, 27, 609617. Light, L. L., Kayra-Stuart, F. & Hollander, S. (1979) Recognition memory for typical and unusual faces. J. Exp. Psychol.: Hum. Learn. Mem., 5, 212218. Linville, P. L. (1985) Self-complexity and affective extremity: don't put all your eggs in one cognitive basket. Soc. Cog., 3, 94120.
Linville, P. W. (1982) Affective consequences of complexity regarding the self and others. In M. Clark & S. Fiske (eds) Affect and Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Lipinski, J. F. & Pope, H. G. (1994) Do 'flashbacks' represent obsessional imagery? Comp. Psychiat., 35, 245247. Lloyd, G. G. & Lishman, W. A. (1975) Effect of depression on the speed of recall of pleasant and unpleasant experiences. Psychol. Med., 5, 173180. Locke, J. (1960) An Essay Concerning Human Understanding. London: T. Basset. Loewenstein D. A. & Hokanson, J. E. (1986) The processing of social information by mildly and moderately dysphoric students. Cog. Ther. Res., 10, 447460. Loftus, E. F. (1979) Eyewitness Testimony. Cambridge, Mass.: Harvard University Press. Loftus, E. F. & Kaufman, L. (1992) Why do traumatic experiences sometimes produce good memory (flashbulbs) and sometimes no memory (repression) In E. Winograd & U. Neisser (eds) Affect and Accuracy in Recall in Studies of 'Flashbulb' Memories. Cambridge: Cambridge University Press. Loftus, E. F. & Yapko, M. D. (1995) Psychotherapy and the recovery of repressed memories. In T. Ney (ed), True and False Allegations of Child Sexual Abuse: Assessment and Case Managements. New York: Brunner Mazel. Logan, G. D. (1988) Towards and instance theory of automatisation. Psychol. Rev., 95, 492527 Lonski, M. & Palumbo, P. (1978) The effects of subliminal stimulation on competitive dartthrowing performance. Unpublished, Hofstra University. Lord, C. G. (1980) Schemas and images as memory aides: two modes of processing social information. J. Pers. Soc. Psychol., 38, 257269. Lord, R. G. & Levy, P. E. (1994) Moving from cognition to action: a control theory perspective. Appl. Psychol.: Ann Int. Rev., 43, 335367. Lundh, L. G. (1979) Introspection, consciousness, and information processing. Scand. J. Psychol., 20, 223238. Lyubomirsky, S. & Nolen-Hoeksema, S. (1993) Self-perpetuating properties of dysphoric rumination. J. Pers. Soc. Psychol., 65, 339349. MacDonald, C. (1995) Classicism vs connectionism. In C. MacDonald & G. MacDonald (eds), Connectionism: Debates on Psychological Explanation. Oxford, UK: Blackwell. Mackay, D. G. (1973) Aspects of a theory of comprehension, memory and attention. Quart. J. Exp. Psychol., 25, 2240. Page 347
MacLeod, C. M. (1989) Word context during initial exposure influences degree of priming in word fragment completion. J. Exp. Psychol.: Learn. Mem. Cog., 15, 398406. MacLeod, C. (1990) Mood disorders and cognition. In M.W. Eysenck (Ed.), Cognitive Psychology: An International Review. (pp. 955). New York, Wiley. MacLeod, C. M. (1991) Half a century of research on the Stroop effect: an integrative review. Psychol. Bull. 109, 163203.
MacLeod, C. & Campbell, L. (1992) Memory accessibility and probability judgments: an experimental evaluation of the availability heuristic. J. Pers. Soc. Psychol., 63, 890902. MacLeod, C. M. & Dunbar, K. (1988) Training and Stroop-like interference: evidence for a continuum of automaticity. J. Exp. Psychol.: Learn., Mem. Cog., 10, 304315. MacLeod, C. & Hagan, R. (1992) Individual differences in the selective processing of threatening information, and emotional responses to a stressful life event. Behav. Res. Ther., 30, 151161. MacLeod, C. & Mathews, A. (1988) Anxiety and the allocation of attention to threat. Quart. J. Exp. Psychol: Hum. Exp. Psychol. 38, 659610. MacLeod, C. & Mathews, A. (1991) Cognitive-experimental approaches to the emotional disorders. In P. R. Martin (ed.), Handbook of Behaviour Therapy and Psychological Science: An Integrative Approach. Oxford: Pergamon. MacLeod, C., Mathews, A. & Tata, P. (1986) Attentional bias in emotional disorders. J. Abnorm. Psychol., 95, 1520. MacLeod, C. & Rutherford, E. M. (1992) Anxiety and the selective processing of threatening information: mediating roles of awareness, trait and state variables, and personal relevance of stimulus materials. Behav. Res. Ther., 30, 479491. MacLeod, C., Tata, P. & Mathews, A. (1987) Perception of emotionally valenced information in depression. Br. J. Clin. Psychol., 26, 6768. MacLeod, A. K., Rose, G. S. & Williams, J. M. G. (1993) Components of hopelessness about the future in parasuicide. Cog. Ther. Res., 17, 441455. MacLeod, A. K., Williams, J. M. & Bekerian, D. A. (1991) Worry is reasonable: the role of explanations in pessimism about future personal events. J Abnorm. Psychol., 100, 478486. Mahoney, M. J. (1984) Integrating cognition, affect and action: a comment. Cog. Ther. Res., 8, 585589. Makhlouf-Norris, F., Jones, H. G. & Norris, H. (1970) Articulation of the conceptual structure in obsessional neurosis. Br. J. Soc. Clin. Psychol., 9, 264274. Mandel, R. G. & Johnson, N. S. (1984) A developmental analysis of story recall and comprehension in adulthood. J. Verb. Learn. Verb. Behav., 23, 643659. Mandler, G. (1975) Consciousness: respectable, useful and probably necessary. In R. Solsol (ed.), Information Processing and Cognition: (The Loyola Symposium). Hillsdale, N.J.: Lawrence Erlbaum. Mandler, G. (1982) The structure of value: accounting for taste. In M. S. Clark & S. T. Fiske (eds), Affect and Cognition Hillsdale, N.J.: Lawrence Erlbaum. Mandler, G. (1984) Mind & Body: Psychology of Emotion and Stress. New York: Norton. Mandler, G., Nakamura, Y. & Van Zandt, B. J. S. (1987) Nonspecific effects of exposure on stimuli that cannot be recognised. J. Exp. Psychol.: Learn. Mem. Cog., 13, 646648. Mandler, J. (1975) Mind and Emotion. New York: John Wiley.
next page > Page 348
Mandler, J. (1978) A code in the node: the use of a story schema in retrieval. Dis. Process, 1, 1435. Mandler, J. (1984) Stories, Scripts and Scenes. Hillsdale, N.J.: Erlbaum. Mandler, J. M. & De Forest, M. (1979) Is there more than one way to recall a story? Child Dev., 50, 8689. Mandler, J. M. & Goodman, M. S. (1982) On the psychological validity of story structure. J. Verb. Learn. Verb. Behav., 21, 507523. Marr, D. (1982) Vision. San Francisco: W. H. Freeman. Marcel, A. J. (1978) Unconscious reading: experiments on people who do not know that they are reading. Vis. Lang., 12(4), 391404. Marcel, A. J. (1980) Conscious and preconscious recognition of polysemous words: locating the selective effects of prior verbal context. In R. S. Nickerson (ed.), Attention and performance, Vol. XII. Hillsdale, N.J.: Lawrence Erlbaum. Marcel, A. J. (1983a) Conscious and unconscious perception: experiments on visual masking and word recognition. Cog. Psychol., 15, 197237. Marcel, A. J. (1983b) Conscious and unconscious perception: an approach to the relations between phenomenal experience and perceptual processes. Cog. Psychol., 15, 238300. Marcel, A.J. (1988) Phenomenal experience and functionalism. In A.J. Marcel & E. Bisiach, (eds), Consciousness in Contemporary Science, (pp. 121158) Oxford: Clarendon press. Marcel, A.J. & Bisiach, E. (1988) Consciousness in Contemporary Science. Oxford: Clarendon Press. Marcel, A. J., Katz, L. & Smith, M. (1974) Laterality and reading proficiency. Neuropsychologia, 12, 131139 Marcel, A. J. & Patterson, K. E. (1978) Word recognition and production: reciprocity in clinical and normal studies. In J. Requin (ed.), Attention and Performance, Vol. VII, Hillsdale, N.J.: Lawrence Erlbaum. Marcus, H. (1977) Self-schemata and processing information about the self. J. Pers. Soc. Psychol., 35, 6378. Marks, D. (1972) Individual difference in the vividness of visual imagery and their effect on function. In P. Sheehan (ed.), The Nature and Function of Imagery. New York: Academic Press. Markus, H. & Nurius, P. (1986) Possibel selves. Amer. Psychol., 41, 954969. Marr, D. (1980) Visual information processing: the structure and creation of visual
representations. Phil. Trans. Roy. Soc. London, Series B, 290, 199218. Martin, D. J., Abramson, L. Y. & Alloy, L. B. (1984) Illusion of control for self and others in depressed and nondepressed college students. J. Pers. Soc. Psychol., 46, 125136. Martin, M. (1985) Neuroticism as predisposition towards depression: a cognitive mechanism. Pers. Indiv. Diff., 6, 353365. Martin, M. & Clark, D. M. (1985) Constraints on the effects of mood on accessibility. Paper given at the Symposium on Affect and Cognition, Cognitive Psychology Section, British Psychology Society, Oxford, September 1985. Martin, M. & Clark, D. M. (1986a) On the response bias explanation of selective memory effect in depression. Cog. Ther. Res., 10, 267270. Martin, M. & Clark D. M. (1986b) Selective memory, depression and response bias: an unbiased response. Cog. Ther. Res., 10, 275278. Martin, M., Horder, P. & Jones, G. V. (1992) Integral bias in naming of phobiarelated words. Cog. Emot. 6, 479486. Page 349
Martin, M., Williams, R. M. & Clark, D. M. (1991) Does anxiety lead to selective processing of threat-related information? Behav. Res. Ther., 29, 147160. Marton, F., Hounsell, D. & Entwistle, N. (1984) The Experience of Learning. Edinburgh: Scottish Academic Press. Marx, E.M. & Schulze, C.C. (1991) Interpersonal problem-solving in depressed students. J. Clin. Psychol., 47, 361367. Marzillier, J. S., Carroll, D. & Newland, J. R. (1979) Self-report and physiological changes accompanying repeated imagining of a phobic scene. Behav. Res. Ther., 17, 7177. Masling, J. M., Bornstein, R. F., Poynton, R. F., Reed, S. & Katkin, E. S. (1991) Perception without awareness and electrodermal responding: a strong test of subliminal psychodynamic activation effects. J. Mind. Behav., 12, 3347. Massaro, D. W. & Cohen, M. M. (1991) Integration versus interactive activation: the joint influence of sensory information and context in perception. Cog. Psychol., 23, 558614. Massaro, D. W. & Cowan, N. (1993) Information processing models: microscopes of the mind. Ann. Rev. Psychol.., 44, 383425. Mathews, A. M. (1971) Psychophysiological approaches to the investigation of desensitization and related procedures. Psychol. Bull., 76, 7391. Mathews, A. & Bradley, B. (1983) Mood and the self-reference bias in recall. Behav. Res. Ther., 21, 233239. Mathews, A. M. & Klug, F, (1993) Emotionality and interference with color-naming in anxiety. Behav. Res. Ther., 31, 5762.
Mathews, A, M., May, J., Mogg, K. & Eysenck, M. (1990) Attentional bias in anxiety: selective search or defective filtering? J. Abnorm. Psychol., 99, 166173. Mathews, A. & Milroy, R. (1994) Processing of emotional meaning in anxiety. Cog Emot., 8, 535554. Mathews, A. M. & MacLeod, C. (1985) Selective processing of threat cues in anxiety states. Behav. Res. Ther., 23, 563569. Mathews, A. M. & MacLeod, C. (1986) Discrimination of threat cues without awareness in anxiety states. J. Abnorm. Psychol., 95, 131138. Mathews. A. M. & MacLeod, C. (1994) Cognitive approaches to emotion and emotional disorders. Ann. Rev. Psychol. 45, 2550. Mathews, A. M., Mogg, K., Kentish, J. & Eysenck, M. (1995) Effect of psychological treatment on cognitive bias in Generalized Anxiety Disorder. Behav. Res. Ther., 33, 293303 Mathews, A. M., Mogg., K., May, J. & Eysenck, M. W. (1989) Memory bias in anxiety. J. Abnorm. Psychol., 98, 236240. Mathews, A., Richards, A. & Eysenck, M. (1989) Interpretation of homophones related to threat in anxiety states. J. Abnorm. Psychol., 98, 3134. Mathews, A. M. & Sebastian S. (1993) Suppression of emotional Stroop effects by fear arousal. Cog. and Emot., 7, 517530. Matt, G. E., Vacquez, C. & Campbell, W. K. (1992) Mood-congruent recall of affectively toned stimuli: a meta-analytic review. Clin. Psychol. Rev, 12, 227255. Matthews, G. & Southall, A. (1991) Depression and the processing of emotional stimuli: a study or semantic priming. Cog. Ther. Res., 15, 283302. Matthews, R. C., Buss, R. R., Stanley, W. B., Blanchard-Fields, F., Cho, J. R. & Druhan, B. (1989) The role of implicit and explicit processes in learning from examples: a synergistic effect. J. Exp. Psychol.: Learn. Mem. Cog., 15, 10831100. Mattia, J. I., Heimberg, R. G. & Hope, D. A. (1993) The revised Stroop color-naming task in social phobics. Behav. Res. Ther., 31, 305313. Page 350
Matussek, P. & Luks, P. (1981) Themes of endogenous and nonendogenous depressions. Psychiat. Res., 5, 235242. May, C.P., Kane, M.J. & Hasher, L. (1995) Determinants of negative priming. Psychol. Bull., 118, 3554. Mayo, P.R. (1989) A further study of the personality-congruent recall effect. Pers. Indiv. Diffs., 10, 247252. McAllister, T. W. (1981) Cognitive functioning in the affective disorders. Comp. Psychiat., 22, 572586. McCann, R. S. & Johnston, J. C. (1992) Locus of the single-channel bottleneck in dual-task performance. J. Exp. Psychol.: Hum. Percept. Perf., 18, 471484. McCauley, C. (1988) The content of awareness and top-down versus bottom-up processing. In
T.K. Srull & R.S. Wyer (eds), A Dual Process Model of Impression Formation: Advances in Social Cognition, Vol. 1, (pp. 111118). Hillsdale, NJ: Erlbaum. McClelland, J. L. (1976) Preliminary letter identification in the perception of words and nonwords. J. Exp. Psychol.: Hum. Percept. Perf., 1, 8091. McClelland, J. L. (1979) On the time relations of mental processes: an examination of systems of processes in cascade. Psychol Rev., 86, 287330. McClelland, J. L. (1995) A connectionist perspective on knowledge and development. In T.J. Simon & G.S. Halford (eds), Developing Cognitive Competence: New Approaches to Process Modelling. Hillsdale, NJ: Erlbaum. McClelland, J. L. & Jenkins, E. (1991) Nature, nurture, and connections: implications of connectionist models for cognitive development. In K. VanLehn (ed.), Architectures for Intelligence: The Twenty-second Carnegie Mellon Symposium on Cognition. Hillsdale, NJ: Erlbaum. McClelland, J. L. & Rumelhart, D. E. (1981) An interactive model of the effect of context in perception, Part I, Psychol. Rev., 88, 375407. McClelland, J. L. & Rumelhart, D. E. (1985) Distributed memory and the representation of general and specific information. J. Exp. Psychol.: Gen., 114, 159188. McClelland, J. L. & Rumelhart, D. E. (1986) Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 2: Psychological and Biological Models. Cambridge, MA: MIT Press. McDowell, J. (1984) Recall of pleasant and unpleasant words in depressed subjects. J. Abnorm. Psychol., 93, 401407. McGlynn, S. M. & Schacter, D. L. (1989) Unawareness of deficits in neuropsychological syndromes. J. Clin. Exp. Neurol., 11, 143205. McKoon, G. & Ratcliffe, R. (1992) Inference during reading. Psychol. Rev., 99, 440446. McKoon, G. & Ratcliffe, R. (1995) How should implicit memory be modelled? J. Exp. Psychol.: Learn. Mem. Cog., 21, 777784. McKoon, G., Ratcliffe, R. & Siefert, C. (1989) Textual inferences: models and measures. In D. A. Balota, G. B. F. D'Arcais & K. Rayner (eds), Comprehension Processes in Reading. Hillsdale, NJ: Erlbaum. McLemore, C. W. (1972) Imagery in desensitisation. Behav. Res. Ther., 10, 5157. McLemore, C. W. (1976) Factorial validity of imagery measures. Behav. Res. Ther., 14, 399408. McLeod, P. (1977) A dual task response modality effect: support for multi-processor models of attention. Quart. J. Exp. Psychol., 29, 651667. McLeod, P. (1978) Does probe RT measure central processing demand? Quart. J. Exp. Psychol., 30, 8389. Page 351
McNair. D. M., Lorr, M. & Droppleman, L. F. (1981) Manual for the Profile of Mood States,
2nd edition. San Diego: Edits. McNally, R. J. (1990) Psychological approaches to panic disorder: a review. Psychol. Bull. 108(3), 403419. McNally, R.J., Amir, N., Louro, C. E. Lukach, B. M., Riemann, B. C. & Calamari, J. E. (1994) Cognitive processing of idiographic emotional information in panic disorder. Behav. Res. Ther., 32, 199-122 McNally R. J., English, G. E. & Lipke, H. J. (,1993) Assessment of intrusive cognition in PTSD: use of the modified Stroop paradigm. J. Traumat. Stress., 6, 33241. McNally, R. J. & Foa, E. B. (1986) Preparedness resistance to extinction to fear-relevant stimuli: a failure to replicate. Behav. Res. Ther., 24(5), 259535. McNally, R. J. & Foa, E. B. (1987) Cognition and agoraphobia: bias in the interpretation of threat. Special Issue: Anxiety: cognitive factors and the anxiety disorders. Cog. Ther. Res., 11, 567581. McNally, R. J., Foa, E. B. & Donnell, C. D. (1989) Memory bias for anxiety information in patients with panic disorder. Cog. Emot., 3, 2744. McNally, R. J. & Heatherton, T.F. (1993) Are covariation biases attributable to a priori expectancy biases? Behav. Res. Ther., 31, 653658. McNally, R. J., Kaspi, S. P., Riemann, B. C. & Zeitlin, S. B. (1990) Selective processing of threat cues in post-traumatic stress disorder. J. Abnorm. Psychol, 99, 398402. McNally, R. J., Lasko, N. B., Macklin, M. L. & Pitman, R. K. (1995) Autobiographical memory disturbance in combat-related post-traumatic stress-disorder Behav. Res. Ther., 33, 619630. McNally, R. J., Litz, B. T. & Prassas, A. (1994) Emotional priming of autobiographical memory in post-traumatic stress disorder. Cog. Emot., 8, 351367. McNally, R. J., Reimann, B. C. & Kim, E. (1990) Selective processing of threat cues in panic disorder. Behav. Res. Ther., 28, 407412. McNally, R. J., Reimann B. C., Louro, C. E., Lukach, B. M. & Kim, E. (1992) Cognitive processing of emotional information in panic disorder. Behav. Res. Ther., 30 143149. Meichenbaum, D. (1977) Cognitive Behaviour Modification. New York: Plenum. Merikle, P.A. (1982) Unconscious perception revisited. Percept. Psychophys., 31, 298301. Merikle, P. M. & Reingold, E. M. (1990) Recognition and lexical decision without detection: unconscious perception? J. Exp. Psychol.: Hum. Percept. Perf., 16, 574583. Merikle, P. M. & Reingold, E. M. (1991) Comparing direct (explicit) and indirect (implicit) measures to study unconscious memory. J. Exp. Psychol.: Learn. Mem. Cog., 17, 224232. Merrill, E. C., Cha, K. H. & Moore, A. L. (1994) Suppression of irrelevant location information by individuals with and without mental retardation. Amer. J. Ment. Retard., 99, 207214. Merskey, H. (1995) Multiple personality disorder and false memory syndrome. Br. J. Psychol., 166, 281283
Meyer, B., Haring, M., Brandt, D. & Walker, C. (1980) Comprehension of stories and expository text. Poetics, 9, 203210. Meyer, D. E. & Kieras, D. E. (1994) EPIC computational models of psychological refractoryperiod effects in human multiple-task performers. Technical Report TR-94/ONR-EPIC-2, University of Michigan, Ann Arbor. Page 352
Meyer, D. E. & Kieras, D. E. (1995) A computational theory of human multiple-task performers. Psychol. Rev., 102. Meyer, D. E., Kieras, D. E., Lauber, E., Schumacher, E. H., Glass, J., Zurbriggen, E., Gmeindl, L. & Apfelblat, D. (1995) Adaptive executive control: flexible multiple-task performance without pervasive immutable response-selection bottlenecks. Acta Psychologica, 90, 163190. Meyer, D. E. & Kornblum, S. (eds) (1993) Attention and Performance, Vol. XIV. Synergies in Experimental Psychology, Artificial Intelligence, and Cognitive Neuroscience (pp. 245264). Cambridge, Mass: MIT Press. Meyer, D. E., Schvaneveldt, R. W. & Ruddy, M. G. (1975) Locus of contextual effects in visual word recognition. In P. M. A. Rabbitt & S. Dornic (eds), Attention and Performance, Vol. V. New York: Academic Press. Meyer, T. J., Miller, M. L., Metzger, R. L. & Borkovec, T. D. (1990) Development and validation of the Penn State Worry Questionnaire. Behav. Res. Ther., 28, 487495. Mill, J. (1829) Analysis of the Phenomena of the Human Mind. London: Baldwin. Millar, D. G. (1980) A repertory grid study of obsessionality, distinctive cognitive structure or distinctive content. Br. J. Med. Psychol., 53, 5966. Miller, E. & Lewis, P. (1977) Recognition memory in elderly patients with depression and dementia. J. Abnorm. Psychol., 86, 8486. Miller, G. A. (1962) Psychology: The Science of Mental Life. New York. Harper and Row. Miller, J. (1973) The effects of aggressive stimulation upon adults who have experienced the death of a parent during childhood and adolescence. Unpublished Doctoral Dissertation, New York University. Miller, J. (1993) A queue-series model for reaction time, with discrete-stage and continuousflow models as special cases. Psychol. Rev. 106, 702715. Miller, J. & Hackley, S. A. (1992) Electrophysiological evidence for temporal overlap among contingent mental processes. J. Exp. Psychol.: Gen., 121, 195209. Miller, J., van der Ham, F. & Sanders, A. (1995) Overlapping stage models and reaction time additivity: effects of the activation equation. Acta Psychologica, 90, 1128. Miller, J. O. (1988) Discrete and continuous models of human information processing: theoretical distinctions and empirical results. Acta Psychologica, 67, 191257. Miller, P. H. (1994) Individual differences in children's strategic behaviors: utilization deficiencies. Learn. Indiv. Diff., 6, 285307. Miller, S. M. (1987a) Monitoring and blunting: validation of a questionnaire to assess styles of
information seeking under threat. J. Pers. Soc. Psychol., 52, 345353. Miller, S. M. (1987b) To see or not to see: cognitive informational styles in the coping process. In M. Rosenbaum (ed.), Learned Resourcefulness: On Coping Skills, Self-regulation and Adaptive Behaviour. New York: Springer Press. Miller, S. M. & Mangan, C. E. (1983) Interacting effects of information and coping style in adapting to gynaecological stress: Should the doctor tell all?. J. Pers. Soc. Psychol., 45, 223236. Miller, W. R. (1975) Psychological deficit in depression. Psychol. Bull., 82, 238260. Milner, D. A. (1992) Disorders of perceptual awareness: commentary. In D. A. Milner & M. D. Rugg (eds), The Neuropsychology of Consciousness, (pp. 139158) London: Academic Press. Page 353
Minard, J. G. (1965) Response-bias interpretation of perceptual defence. Psychol. Rev., 72, 7488. Mineka, S. & Sutton, S. K. (1992) Cognitive biases and the emotional disorders. Psychol. Sci. 3, 6569. Mineo, A. & Kahneman, D. (1974) Reaction time in focussed and in divided attention. J. Exp. Psychol., 103, 394399. Minsky, M. (1975) A framework for representing knowledge. In P.H. Winston (ed.), The Psychology of Computer Vision. New York: McGraw-Hill. Miranda, J. & Persons, J. B. (1988) Dysfunctional attitudes are mood-state dependent. J. Abnorm. Psychol., 97, 7679. Miranda, J., & Persons, J. B. & Byers, C. N. (1990) Endorsement of dysfunctional beliefs depends on current mood state. J. Abnorm. Psychol., 99, 237241. Mogg, K. (1988) Processing of emotional information in clinical anxiety states. Unpublished Ph.D. thesis, University of London. Mogg, K. & Marden, B. (1990) Processing of emotional information in anxious subjects. Br. J. Clin. Psychol. 29, 227229. Mogg, K., Bradley, B. P., Williams, R. & Mathews, A. M. (1993) Subliminal processing of emotional information in anxiety and depression. J. Abnorm. Psychol., 102, 304311. Mogg, K., & Gardener, J.M., Starron, A. & Golombok, S. (1992) Recollection experience and recognition memory for threat in clinical anxiety states. Bull. Psychol. Sci., 30, 109112. Mogg, K., Kentish, J. & Bradley, B. P, (1993) Effects of anxiety and awareness on color identification latencies for emotional words. Behav. Res. Ther. 31, 559567. Mogg, K. C. & Mathews A. (1990) Is there a self-referent mood congruent recall bias in anxiety? Behav. Res. Ther., 28, 9192. Mogg, K., Mathews, A. M., Bird, C. & MacGregor-Morris, R., (1990) Effects of stress and anxiety on the processing of threat stimuli. J. Pers. Soc. Psychol., 59, 12301237. Mogg, K., Mathews, A. M., Eysenck, M. & May, J. (1991a), Biased cognitive operations in
anxiety: artefact, processing priorities or attentional search? Behav. Res. Ther. 29, 459467. Mogg, K., Mathews, A. M. & Eysenck, M. (1992.) Attentional bias to threat in clinical anxiety states. Cog. Emot. 6, 149159. Mogg, K., Mathews, A., May, J., Grove, M. Eysenck, M. & Weinman, J. (1991b) Assessment of cognitive bias in anxiety and depression using a colour perception task. Cog. Emot., 5, 221238. Mogg, K., Mathews, A. & Weinman, J. (1987) Memory bias in clinical anxiety. J. Abnorm. Psychol., 96, 9498. Mogg, K., Mathews, A. M. & Weinman, J. (1989) Selective processing of threat cues in anxiety states: a replication. Behave. Res. Ther., 27, 317323. Moore, R. G., Watts, F. N. & Williams, J. M. G. (1988) The specificity of personal memories in depression. Br. J. Clin. Psychol., 27, 275276. Morley, S. (1993) Vivid memories. Pain, 55, 5562. Morris, P., Gruneberg, M. M., Sykes, R. N. & Merrick, A. (1981) Football knowledge and the acquisition of new results. Br. J. Psychol., 72, 479483. Morris, W. N. (1992) A functional analysis of the role of mood in affect systems. In M. S. Clark (ed.) Emotional. Newbury Park: Sage. Morrow, D. G., Bower, G. H. & Greenspan, S.L. (1989) Updating situation models during narrative comprehension J. Mem. Lang., 28, 292312. Page 354
Moscovitch, M. (1982) Multiple dissociations of function in amnesia. In L. Cermak (ed.), Human Memory and Amnesia. Hillsdale, N.J.: Lawrence Erlbaum. Mueller, J. H. (1976) Anxiety and cue utilization in human learning and memory. In M. Zuckerman and C. D. Spielberger (eds), Emotions and Anxiety: New Concepts, Methods and Applications. Hillsdale, N.J.: Lawrence Erlbaum. Mueller, J. H. (1977) Test anxiety, input modality, and levels of organization in free recall. Bull. Psychonom. Soc., 9, 6769. Mueller, J. H. (1978) The effects of individual differences in test anxiety and type of orienting task on levels of organization in free recall. J. Res. Pers., 12, 100116. Mueller, J. H. (1980) Test anxiety and the encoding and retrieval of information. In I. Sarason (ed.), Test Anxiety: Theory, Research and Applications. Hillsdale N.J.: Lawrence Erlbaum. Mulder, G., Wijers, A. A., Lange, J. J. Buijink, B. M., Mulder, L. J. M., Willemsen, A. T. M. & Paans, A. M. S. (1995) The role of neuroimaging in the discovery of processing stages: A review. Acta Psychologia, 90, 6379. Muller, H. J. & Rabbitt, P. M. (1989) Reflexive and voluntary orienting of visual attention: time course of activation and resistance to interruption. J. Exp. Psychol.: Hum. Percept. Perf., 15, 315330. Munro, P. W. (1986) State-dependent factors influencing neural plasticity: a partial account of the critical period. In J. L. McClelland & D. E. Rumelhart, (eds) Parallel Distributed
Processing: Explorations in the Microstructure of Cognition, Vol 2: Psychological and Biological Processes. (Ch. 24, pp. 471502). Cambridge, MA: MIT Press. Murphy, G. E., Simons, K. D., Wetzel, R. D. & Lustman, P. J. (1984) Cognitive therapy and pharmacotherapy; singly and together in the treatment of depression. Arch. Gen. Psychiat., 41, 3341. Musen, G., & Squire, L. R. (1991) Normal acquisition of nonverbal information in amnesia. J. Exp. Psychol.: Learn. Mem. Cog., 17, 10951104. Mykel, N. & Daves, W. F. (1979) Emergence of unreported stimuli into imagery as a function of laterality of presentation: a replication and extension of research by Henley and Dixon. Br. J. Psychol., 70, 253258. Naish, P. L. M. (1985) The locus of the stroop effect: One site masquerading as two? Br. J. Psychol., 76, 303310. Nakamura, G. V. & Graesser, A. C. (1985) Memory for script-typical and script-atypical actions: a reaction time study. Bull. Psychonom. Soc., 23(4), 384396. Nasby, W. (1994) Moderators of mood-congruent encoding: self/others reference and affirmative/non affirmative judgement. Cog. Emot., 8, 259278. Natsoulas, T. (1994) A rediscovery of consciousness. Consc. Cog., 3, 223245. Navon, D. (1977) Forest before trees: the presence of global features in perception. Cog. Psychol., 9, 353383. Navon, D. (1984) Resources A theoretical soup stone? Psychol. Rev., 91, 216234. Navon, D. & Gopher, D. (1979a) Interpretation of task difficulty in terms of resources: efficiency, load, demand and cost-composition. In R. Nickerson (ed.), Attention and Performance, Vol. VIII. Hillsdale, N.J.: Lawrence Erlbaum. Navon, D. & Gopher, D. (1979b) On the economy of the human-processing system. Psychol. Rev., 86(3), 214255. Navon, D. & Margalit, B. (1983) Allocation of attention according to informativeness in visual recognition. Quart. J. Exp. Psychol., 35A, 497512. Neisser, U. (1963) The multiplicity of thought. Br. J. Psychol., 54, 114. Neisser, U. (1967) Cognitive Psychology. New York: Appleton-Century-Crofts. Page 355
Neisser, U. (1976) Cognition and Reality: Principles and Implications of Cognitive Psychology. San Francisco: W. H. Freeman. Neisser, U. & Becklen, R. (1975) Selective looking: attending to visually specific events. Cog. Psychol., 7, 480494. Nelson, D. L., Schreiber, T. A. & Holley, P. E. (1992) The retrieval of controlled and automatic aspects of meaning on direct and indirect tests. Mem. Cog., 20, 671684. Nelson, J. G. & Mazure, C. (1985) Ruminative thinking: a distinctive sign of melancholia. J. Affect. Dis., 9, 4146.
Nelson, K. (1977) Cognitive development and the acquisition of concepts. In R. C. Anderson, R. J. Spiro, & W. D. Montague (eds), Schooling and the Acquisition of Knowledge. Hillsdale, N.J.: Lawrence Erlbaum. Nelson, R. E. & Craighead, W. E. (1977) Selective recall of positive and negative feedback, self-control behaviors, and depression. J. Abnorm. Psychol., 86, 379388. Neumann, E. & DeShepper, B.G. (1992) An inhibition based fan effect: evidence for an active suppression mechanism in selective attention. Canad. J. Psychol. 46, 140. Newell, A. & Simon, H. A. (1972) Human Problem Solving. Englewood Cliffs, NJ: Prentice Hall. Newstead, S. E. & Dennis, I. (1979) Lexical and grammatical processing of unshadowed messages: a re-examination of the Mackay effect. J. Exp. Psychol., 31, 47788. Nezu, A. M., Nezu, C. M. & Perri, M. G. (1989) Problem-Solving Therapy for Depression. New York: John Wiley & Sons. Nicholson, W. M. (1958) The influence of anxiety upon learning: interference or drive increment. J. Person., 26, 303319. Niedenthal, P. M., Setterlund, M. B. & Wherry, M. B. (1992) Possible self-complexity and affective reactions to goal-relevant evaluation. J. Pers. Soc. Psychol., 63, 516. Nisbett, R. E. & Wilson, T. D. (1977) Telling more than we can know: verbal reports on mental processes. Psychol. Rev., 84, 231259. Nissen, M. J. & Bullemer, P. (1987) Attentional requirements of learning: evidence from performance measures. Cog. Psychol., 19, 132. Nissen, M. J., Willingham, D. & Hartman, M. (1989) Explicit and implicit remembering: When is learning preserved in amnesia? Neuropsychol., 27, 341352. Nolen-Hoeksema, S. (1991) Responses to depression and their effects on the duration of depressive episodes. J. Abnorm. Psychol, 100, 569582. Nolen-Hoeksema, S. & Morrow, J. (1991) A prospective study of depression and posttraumatic stress symptoms after a natural disaster: The Loma Prieta earthquake. J. Pers. Soc. Psychol., 61, 115121 Nolen-Hoeksema, S., Morrow, J. & Fredrickson, B. L. (1993) Response styles and the duration of episodes of depressed mood. J. Abnorm. Psychol. 102, 2028. Norman, D. A. & Bobrow, D. G. (1975) On data-limited and resource-limited processes. Cog. Psychol., 7, 4464. Norman, D. A. & Bobrow, D. G. (1976) On the role of active memory processes in perception and cognition. In C. N. Cofer (ed.), The Structure of Human Memory. San Francisco: Freeman. Norman, D. A. & Bobrow, D. G. (1979) Descriptions: an intermediate stage in memory retrieval. Cog. Psychol., 11, 107123. Page 356
Norman, D.A. & Shallice, T. (1986) Attention to action: willed and automatic control of behaviour. In R.J. Davidson, G.E. Schwartz & D. Shapiro (eds), Consciousness and Self-
Regulation, Vol. 4. New York: Plenum. Norton, G.R., Schaefer, E., Cox, B.J., Dorward, J. & Wozney, K. (1988) Selective memory effects in non-clinical panickers. J. Anx. Dis., 2, 169177. Nugent, K. & Mineka, S. (1994) The effect of high and low trait anxiety on implicit and explicit memory tasks. Cog. Emot, 8, 147164. Nunn, J. D., Stevenson, R. & Whalan, G. (1984) Selective memory effects in agoraphobic patients. Br. J. Clin. Psychol., 23, 195201. Oatley, K. (1988) On changing one's mind: a possible function of consciousness. In A. J. Marcel & E. Bisiach (eds), Consciousness in Contemporary Science (pp. 369390) Oxford: Clarendon Press. Oatley, K. & Johnson-Laird, P. (1987) Towards a cognitive theory of emotions Cog. Emot., 1, 2950. Oatman, L. C. (1984) Auditory evoked-potential amplitude during simultaneous visual stimulation. Presented at the Annual Meeting of the Psychonomic Society, San Antonio. O'Brien, E. J. and Albrecht, J. E. (1992) Comprehension strategies in the development of a mental model. J. Exp Psychol.: Learn. Mem Cog., 18, 777784. Ohman, A., Dimberg, U. & Esteves, F. (1989) Preattentive activation of aversive emotions. In T. Archer & L.G. Nilsson (eds), Aversion, Avoidance, and Anxiety, (pp 169193) Hillsdale, NJ: Erlbaum. Ohman, A. & Soares, J. J. F. (1993) On the automaticity of phobic fear: conditioned skin conductance responses to masked phobic stimuli. J. Abnorm. Psychol., 102, 121132. Ohman, A. & Soares, J. J. F. (1994) 'Unconscious anxiety': phobic responses to masked stimuli. J. Abnorm. Psychol., 103, 231240. Oliver, J. M. & Burkham, R. (1982) Subliminal psychodynamic activation in depression: a failure to replicate. J. Abnorm. Psychol., 92, 337342. Osman, A., Bashore, T. R., Coles, M. G. H., Donchin, E. & Meyer, D. E. (1992) On the transmission of partial information: inferences from movement-related brain potentials. J. Exp. Psychol. Hum. Percept. Perf., 18, 217232. Otto, M. W., McNally, R. J., Pollack, M. H., Chen, E. & Rosenbaum, J. F. (1994) Hemispheric laterality and memory bias for threat in anxiety disorder. J. Abnorm. Psychol., 103, 828831. Pachella, R. G. (1974) The interpretation of reaction time in information processing research. In B. Kantowitz, (ed.), Human Information Processing: Tutorials in Performance and Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Paivio, A. (1986) Mental representations: A Dual Coding Approach. Oxford: Clarendon Press. Paller, K. A. (1990) Recall and stem completion have different electrophysiological correlates and are modified differentially by directed forgetting. J. Exp. Psychol.: Learn. Mem. Cog., 16, 10211032. Paris, S. G., Lindauer, B. K. & Cox, G. L. (1977) The development of inferential comprehension. Child Devt., 48, 17281733.
Parker, G. (1989) The parental bonding instrument; psychometric properties reviewed. Psychiat. Dev., 4, 317335. Parker, G., Tupling, H. & Brown, L. B. (1979) A parental bonding instrument. Br. J. Med. Psychol., 52, 110. Page 357
Parker, K. (1977) The effects of subliminal merger stimuli on the academic performance and psychological adjustment of college students. Doctoral dissertation, New York University. Parkes, C. M. (1972) Bereavement: Studies of Grief in Adult Life. London: Tavistock. Parkin, A.J., Reid, T.K. & Russo, R. (1990) On the differential nature of implicit and explicit memory. Mem. Cog., 18, 507514. Parkinson, L. & Rachman, S. (1981) Intrusive thoughts: the effects of an uncontrived stress Adv. Behav. Res. Ther., 3, 111118. Parrott, G. W. & Sabini, J. (1990) Mood and memory under natural conditions: evidence for mood incongruent recall. J. Pers. Soc. Psychol., 59, 321336. Pashler, H. (1993) Dual-task interference and elementary mental mechanisms. In D. E. Meyer & S. Kornblum (eds), Attention and Performance, 14: Synergies in Experimental Psychology, Artificial Intelligence and Cognitive Neuroscience (pp. 245264). Cambridge, MA: Bradford. Pashler, H (1994) Dual-task interference, in simple tasks: data and theory. Psyhol. Bull. 16, 220244. Pashler, H. & Johnston, J.C. (1989) Chronometric evidence of central postponement in temporally overlapping tasks. Quart. J. Exp. Psychol. 41A, 1945. Paul, I. H. (1959) Studies in remembering: the reproduction of connected and extended verbal material. Psychol. Issues, 1(2), WHO, No. 2. Payne, D.G., Peters, L.J., Birkmire, D.P., Bonto, M.A., Anastasi, J.S. & Wenger, M.J. (1994) Effects of speech intelligibility on concurrent visual task performance. Human Factors, 36, 441475. Payne, R. W. & Hewlett, J. H. G. (1960) Thought disorders in psychotic patients. In H.J. Eysenck (ed.), Experiments in Personality, Vol. 2. London: Routledge and Kegan Paul. Perruchet, P. & Baveux, P. (1989) Correlational analyses of explicit and implicit memory performance. Mem. Cog., 17, 7786. Perruchet, P. & Pacteau, C. (1990) Synthetic grammar learning: implicit rule abstraction or explicit fragmentary knowledge? J. Exp. Psychol: Gen., 119, 264275. Perruchet, P., Gallego, J. & Pacteau, C. (1992) A reinterpretation of some earlier evidence for abstractiveness of implicitly acquired knowledge. Quart. J. Exp. Psychol., 44A, 193210. Perruchet, P., Gallego, J. & Savy, I. (1990) A critical reappraisal of the evidence for unconscious abstraction of deterministic rules in complex experimental situations. Cog. Psychol., 22, 493516. Persons, J. B. & Foa, E. B. (1984) Processing of fearful and neutral information by obsessivecompulsives. Behav. Res. Ther., 22, 26265.
Peters, L. (1992) Memory deficits in dysphoria and clinical depression. Unpublished PhD thesis, University of New South Wales. Peterson, C. & Seligman, M. E. P. (1984) Causal explanations as a risk factor for depression: theory and evidence. Psychol. Rev., 91, 347374. Philpott, A. & Wilding, J. (1979) Semantic interference from subliminal stimuli in a dichoptic viewing situation. Br. J. Psychol., 70, 559563. Pickles, A. J & van den Broeck, M. D. (1988) Failure to replicate evidence for phobic schemata in agoraphobic patients. Br. J. Clin. Psychol., 27, 271272 Plude, D. J., Enns, J. & Brodeur, D. (1994) The development of selective attention: a life-span overview. Acta Psychologica, 86, 227272. Plunkett, K. & Sinha, C. (1992) Connectionism and development theory. Br. J. Dev. Psychol., 10, 209254. Page 358
Poppel, E., Held, R. & Frost, D. (1973) Residual function of brain wounds involving the central visual pathways in man. Nature, 243, 295296. Porterfield, A. L. & Golding, S. L. (1985) A failure to find an effect of subliminal psychodynamic activation upon cognitive measures of pathology in schizophrenia. J. Abnorm. Psychol., 94, 630639. Posner, M. I. & Boies, S. W. (1971) Components of attention. Psychol. Rev., 78, 391408. Posner, M. I. & McLeod, P. (1982) Information processing models in search of elementary operations. Ann. Rev. Psychol., 33, 477514. Postman, L. (1953) On the problem of perceptual defence. Psychol. Rev., 60, 298306. Postman, L. (1954) Learned principles of organisation in memory. Psychol. Mono., 68, 374. Postman, L., Bronson, W. C. & Gropper, G. L. (1953) Is there a mechanism of perceptual defence? J. Abnorm. Soc. Psychol., 48, 215-224. Potts, R., Camp, C. & Coyne, C. (1989) The relationship between naturally occurring dysphoric moods, elaborative encoding and recall performance. Cog. Emot., 3, 197205. Poulton, E. C. (1957) On prediction in skilled movements. Psychol. Bull., 54, 467478. Powell, G. E. & Watts, F. N. (1973) Determinants of expectation in imaginal desensitisation. Percept. Mot. Skills, 37, 246. Powell, M. & Hemsley, D. R. (1984) Depression: a breakdown of perceptual defence? Br. J. Psychiat., 145, 358362. Power, M. J. & Champion, L. A. (1986) Cognitive approaches to depression: a theoretical critique. Br. J. Clin. Psychol., 25, 201212. Pratto, F. & John, O. P. (1991) Automatic vigilance: the attention-grabbing power of negative social information. J. Pers. Soc. Psychol. 61, 380391. Pressley, M. (1994) Embracing the complexity of individual differences in cognition: studying good information processing and how it might develop. Learn. Indiv. Diff., 6, 259284.
Priestley, N. M. (1995) Anxiety and depression affect threat-related memory biases differently. D.Clin. Psychol. thesis, University of Leeds. Puffet, A., Jehin-Marchot, D., Timsit-Berthier, M. & Timsit, M. (1991) Autobiographical memory and major depressive states. Europ. Psychiat., 6, 141145. Pylyshyn, Z. (1983) What the mind's eye tells the brain: a critique of mental imagery. Psychol. Bull., 80, 122. Pylyshyn, Z. (1984) Computation and Cognition: Towards a Foundation in Cognitive Science. Cambridge, Mass.: MIT Press. Quillan, M. R. (1968) Semantic memory. In M. L. Minsky (ed.), Semantic Information Processing. Cambridge, Mass.: MIT Press. Quirk, S.A. & DePrince, A.P. (1995) Backlash legislation against psychotherapists. J. Psychohist., 22, 258264. Rachman, S. (1980) Emotional processing. Behav. Res. Ther., 18, 5160. Rachman, S. (1981) The primacy of affect: some theoretical implications. Behav. Res. Ther., 19, 279290. Rachman, S. (1984) A reassessment of the 'primacy of affect'. Cog. Ther. Res., 8, 579584. Ranney, M. (1987) The role of structural context in perception: syntax in the recognition of algebraic expressions. Mem. Cog., 15, 2941. Page 359
Ranney, T.A. (1994) Models of driving behavior: a review of their evolution. Acc. An. Prev., 26, 733750. Raskin, A., Friedman, A. S., & DeMascio, A. (1982) Cognitive and performance deficits in depression. Psychopharmacol. Bull., 18, 196202. Ratcliffe, R. (1988) Continuous versus discrete information processing: modeling accumulation of partial information. Psychol. Rev., 95, 238255. Ray, C. (1979) Examination stress and performance on a colour word interference test. Percept. Mot. Skills, 49, 400402. Reber, A.S. (1965) Implicit learning of artificial grammars. Unpublished MA thesis, Brown University. Reber, A.S. (1969) Transfer of syntactic structure in synthetic languages. J. Exp. Psychol., 81, 115119. Reber, A.S. (1989) Implicit learning and tacit knowledge. J. Exp. Psychol.: Gen., 118, 219235. Reber, A.S. (1992) The cognitive unconscious: an evolutionary perspective. Consc. Cog., 1, 93133. Reber, A.S. (1993) Implicit Learning and Tacit Knowledge: An Essay on the Cognitive Unconscious. Oxford: Clarendon Press: Reber, A.S. & Allen, R. (1978) Analogy and abstraction strategies in synthetic gramar
learning: a functional interpretation. Cognition, 6, 189221. Reber, A.S., Kassin, S.M., Lewis, S. & Cantor, G.W. (1980) On the relationship between implicit and explicit modes in the learning of a complex role structure. J. Exp. Psychol.: Learn. Mem. Cog., 6, 492502. Reber, A.S., Walkenfeld, F.F. & Hernstadt, R. (1991) Implicit and explicit learning: individual differences and IQ. J. Exp. Psychol.: Learn. Mem. Cog., 17, 888896. Reed, C. L. & Farah, M. J. (1995) The psychological reality of the body schema: a test with normal participants. J. Exp. Psychol.: Hum. Percept. Perf. 21, 334343. Reed, G. F. (1969a) Under-inclusion a characteristic of obsessional personality: I. Br. J. Psychiat., 115, 781785. Reed, G. F. (1969b) Under-inclusion a characteristic of obsessional personality: II. Br. J. Psychiat., 115, 787790. Reed, G. F. (1977) Obsessional cognition: performance on two numerical tasks. Br. J. Psychiat., 130, 184185. Reed, G. F. (1985) Obsessional Experience and Compulsive Behaviour: A CognitiveStructural Approach. New York: Academic Press. Rehm, L. (1973) Relationships among measures of visual imagery. Behav. Res. Ther., 11, 265270. Reingold, E.M. & Merikle, P.M. (1988) Using direct and indirect measures to study perception without awareness. Percept. & Psychophys., 44, 563574. Reingold, E.M. & Merikle, P.M. (1990) Using direct and indirect measures to study perception without awareness. Percept. & Psychophys., 44, 563575. Reiser, B. J., Black, J. B. & Abelson, R. P. (1985) Knowledge structures in the organisation and retrieval of autobiographical memories. Cog. Psychol., 17, 89137. Reyher, J. & Smeltzer, W. (1968) Uncovering properties of visual imagery and verbal association: a comparative study. J. Abnorm. Psychol., 73, 218222. Richards, A. & French, L. (1991) Effects of encoding and anxiety on implicit and explicit memory performance. Pers. Indiv. Diff., 12, 131139. Richards, A. and French, C. C. (1992) An anxiety-related bias in semantic activation when processing threat/neutral homographs. Quart. Exp. Psychol., 45A, 503525. Page 360
Richards, A., French, C. C., Johnson, W., Naparstek, J. & Williams, J. (1992) Effects of emotion manipulation and anxiety on performance of an emotional Stroop task. Br. J. Clin. Psychol., 83, 479491. Richards, A. & Millwood, B. (1989) colour-identification of differentially valenced words in anxiety. Cog. Emot. 3, 171176. Richards, A. & Whittaker, T. (1990) Effect of anxiety and mood manipulation on autobiographical memory. Br. J. Clin. Psychol., 29, 145153. Ridderinkhof, K.R., van der Molen, M.W. & Bashore, T.R. (1995) Limits on the application of
additive factors logic: violations of stage robustness suggest a dual-process architecture to explain flanker effects on target processing. Acta Psychol., 90, 2948. Riemann, B. C. & McNally, R.J. (1995) Cognitive processing of personally relevant information. Cog. Emot., 9, 325340 Rimm, D. & Bottrell, J. (1969) Four measures of visual imagination. Behav. Res. Ther., 7, 6369. Rinck, M., Glowalla, U. & Schneider, K. (1992) Mood-congruent and mood-incongruent learning. Mem. Cog., 20, 2939. Robert, R. B. & Goldstein, I. T. (1977) The FRL Manual. Memo 409, Artificial Intelligence Laboratory, MIT. Roberts, S. & Sternberg, S. (1993) The meaning of additive reaction-time effects: test of three alternatives. In D.E. Meyer & S. Kornblum (eds), Attention and Performance. Vol. XIV: Synergies in Experimental Psychology, Artificial Intelligence, and Cognitive Neuroscience (pp. 611653) Cambridge, Mass: MIT Press. Robinson, A. & Reading, C. (1985) Imagery in phobic subject: a psychophysiological study. Behav. Res. Ther., 23, 247253. Roediger, H. L. & Blaxton, T. A. (1987) Retrieval modes produce dissociation in memory for surface information. In D. Gorfein & R. R. Hoffman (eds). Memory and Cognitive Processes: The Ebbinghaus Centennial Conference. Hillsdale, N.J.: Lawrence Erlbaum. Roediger, H. L., Knight, J. L. & Kantowitz, D. H. (1977) Inferring decay in short-term memory: the issue of capacity. Me. Cog, 5, 167176. Roediger, H. L. & McDermott, K. B. (1992) Depression and implicit memory : a commentary. J. Abnorm. Psychol., 101, 587591. Roediger, H. L. & McDermott, K. B. (1993) Implicit memory in normal human subject. In H. Spinnler and F. Boller (eds), Handbook of Neuropsychology, Vol. 8. Amsterdam: Elsevier. Roediger, H. L., Weldon, M. S., Staedler, M. S. & Riegler, G.L. (1992) Direct comparison of two implicit memory tests: word fragment and word stem completion. J. Exp. Psychol.: Learn. Mem. Cog., 18, 12511269. Rogers, T. B., Kuiper, N. A. & Kirker, W. S. (1977) Self-reference and the encoding of personal information. J. Pers. Soc. Psychol., 35, 677688. Rojahn, K. & Pettigrew, T. F. (1992) Memory for schema-relevant information: a metaanalytical resolution. Br. J. Soc. Psychol., 31, 81109. Roland, C. B. (1993) Exploring childhood memories with adult survivors of sexual abuse: concrete reconstruction and visualization techniques. J. Ment. Health Couns., 15, 363372. Rollins, H. A. & Hendricks, R. (1980) Processing of words presented simultaneously to eye and ear. J. Exp. Psychol.: Hum. Percept. Perf., 6, 99109. Roorda-Hrdlickova, V., Wolters, G., Bonke, B & Phaf, R.H. (1990) Unconscious perception during general anaesthesia, demonstrated by an implicit memory Page 361
task. In B. Bonke, W. Fitch & K. Millar (eds), Memory and Awareness in Anaesthesia (pp. 511531) London: Balliere. Rosenberg, S. T. (1977) Frame-based Test Processing. Memo 431, Artificial Intelligence Laboratory, MIT. Ross, L. (1977) The intuitive psychologist and his shortcomings: distortion in the attributional process. In L. Berkowitz (ed.), Advances in Experimental Social Psychology, Vol. 10. New York: Academic Press. Ross, L., Lepper, M. R. & Hubbard, M. (1975) Perseverance in self-perception and social perception: biased attributional processes in the debriefing paradigm. J. Pers. Soc. Psychol., 32, 880892. Ross, L., Lepper, M. R., Strack, F. & Steinmetz, J. (1977) Social explanation and social expectation: effects of real and hypothetical explanations on subjective likelihood. J. Pers. Soc. Psychol., 35(11), 817829. Ross, M. J. (1989) Self-schematic instability in mild depression. J. Soc. Behav. Pers., 4, 119126. Roter, A. (1985) Implicit processing: a developmental study. Unpublished doctoral dissertation, City University of New York. Rothbart, M. (1981) Memory processes and social beliefs. In D. L. Hamilton (ed.), Cognitive Processes in Stereotyping and Intergroup Behavior. Hillsdale, N.J.: Lawrence Erlbaum. Roy-Byrne, P. J., Weingartner, H., Bierer, L. M., Thompson, K. & Post, R. M. (1986) Effortful and automatic cognitive processes in depression. Arch. Gen. Psychiat., 43, 265267. Rozeboom, W. M. (1972) Problems in the psychophilosophy of knowledge. In J. R. Royce & W. M. Rozeboom (eds), The Psychology of Knowing. New York: London and Breech. Rubin, D. C. & Friendly, M, (1986) Predicting which words get recalled: measures of free recall, availability, goodness, emotionality and pronunciability for 925 nouns. Mem. Cog., 14, 7994. Rueckl, J.G. (1990) Similarity effects in word and pseudoword priming effects. J. Exp. Psychol.: Learn. Mem. Cog., 16, 374391. Ruiter, C. & Brosschot, J. F. (1993) The emotional Stroop effect in anxiety: Attentional bias or cognitive avoidance? Behav. Res. Ther., 32, 315319. Rumelhart, D. E. (1975) Notes on a schema for stories. In D. G. Bobrow & A. Collins (eds), Representing and Understanding: Studies in Cognitive Science. New York: Academic Press. Rumelhart, D. E. (1978) Schemata: the building blocks of cognition. In R. Spiro, B. Bruce and W. Brewer (eds), Theoretical Issues in Reading Comprehension. Hillside, N.J.: Lawrence Erlbaum. Rumelhart, D. E. & McClelland, J. L. (1986) On learning the past tenses of English verbs. In J. L. McClelland & D. E. Rumelhart (eds), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. II: Applications. Cambridge, Mass.: Bradford Books. Rush, A. J., Weissenburger, J. & Eaves, G. (1987) Do thinking patterns predict depressive symptoms?. Cog. Ther. Res., 10, 225236.
Russell, J. (1987) Comments on articles by Frijda and by Conway and Bekerian. Cog. Emot., 1, 193197. Russell, J. A. & Woudzia, L. (1986) Affective judgments, common sense and Zajonc's thesis of independence. Mot. Emot., 10, 169183. Russell, P. W. & Beekhuis, M. E. (1976) Organization in memory. J. Abnorm. Psychol., 85, 527534. Page 362
Rusted, J. L. & Dighton, K. (1991) Selective processing of threat-related material by spider phobias in a prose recall task. Cog. Emot., 5, 123132. Ruthruff, E., Miller, J.O. & Lachmann, T. (1995) Does mental rotation require central mechanisms. J. Exp. Psychol.: Hum. Percept. Perf., 21, 552570. Ryle, A. & Breen, D. (1972) Some differences in the personal constructs of neurotic and normal subjects. Br. J. Psychiat., 120, 483489. Sackeim, H. A., Packer, I. K. & Gur, R. C. (1977) Hemisphericity, cognitive set and susceptibility to subliminal perception. J. Abnorm. Psychol., 86, 624630. Sadeim, H. (1977) Self-deception: motivational determinants of the non-awareness of cognition. Doctoral dissertation, University of Pennsylvania. Salavoy, P. & Singer, J. A. (1985) Mood and recall of autobiographical memories. Paper given at APA, Los Angeles, August 1985. Sales, B. D. & Haber, R. N. (1968) A different look at perceptual defence for taboo words. Percept. Psychophys., 3, 156160. Sales, B. D., Shuman, D. W. & O'Connor, M. (1994) In a dim light: admissibility of child sexual abuse memories. Appl. Cog. Psychol., 8, 399406. Samide, J.L. (1991) A study of subliminal visual messages designed to enhance the self-esteem of alcoholic patients at two treatment centers. Diss. Abs. Int., 52, 552B553B. Sanders, A. (1979) Mental Work-load: Its Theory and Measurement. New York: Plenum Press. Sanders, A. F. (1980) Stage analysis of reaction processes. In G.E. Stelmach & J.Q. Requin (eds), Tutorials in Motor Behaviour. Amsterdam: North-Holland. Sanders, A. F. (1990) Some issues and trends in the debate in discrete vs continuous processing of information. Acta Psychol., 30, 276315. Sanford, A. J. (1983) Models, Mind and Man. Press Gang: Glasgow University. Sanford, A. J. (1985) Cognition and Cognitive Psychology. London: Weidenfeld and Nicolson. Sanford, A. J. (1990) On the nature of text driven inferences. In D. A. Balota, G. B. F. D'Arcais & K. Rayner (eds), Comprehension Processes in Reading. Hillsdale, NJ: Erlbaum. Sanford, A. J. & Garrod, S. C. (1980) A demonstration of the situational bases of text comprehension through implicit assignments of roles to entities. Unpublished manuscript, Department of Psychology, University of Glasgow.
Sanford, A. J. & Garrod, S. C. (1981) Understanding Written Language: Explorations of Comprehension beyond the Sentence. New York: John Wiley. Sanford, A. J. & Garrod, S. C. (1982) Towards a psychological model of written discourse comprehension. In J. F. Le Ny and W. Kintsch (eds), Language and Comprehension: Advances in Psychology, Vol. 9. Amsterdam: North-Holland. Sarason, I.G. (1988) Anxiety, self-preoccupation and attention. Anx. Res., 1, 37. Sarason, I. G., Sarason, B. R. & Pierce, G. R. (1994) Cognitive interference: at the intelligence-personality crossroads. In D. Saklofske & M. Zeidner (eds), International Handbook of Personality and Intelligence. New York: Plenum. Sartory, G., Rachman, S. & Grey, S. J. (1982) Return of fear: the role of rehearsal. Behav. Res. Ther., 123133. Savard, R. J., Rey, A. C. & Post, R. M. (1980) Halstead-Reitan Category Test in bipolar and unipolar affective disorders. J. Nerv. Ment. Dis., 168, 297304. Scarborough, D. L., Gerard, L. & Cortese, C. (1979) Accessing lexical memory: the transfer of word repetition effects across task and modality. Mem. Cog., 7, 312. Schacter, D. L. (1987) Implicit memory: history and current status. J. Exp. Psychol.: Learn. Mem. Cog., 13, 501518. Page 363
Schacter, D. L. (1989) On the relationship between memory and consciousness: dissociable interactions and conscious experience. In H. L. Roediger III and F. I. M. Craik (eds), Varieties of Memory and Consciousness: Essays in Honour of Endel Tulving (pp. 355390). Hillsdale, NJ: Erlbaum. Schacter, D. L. (1992) Consciousness and awareness in memory and amnesia: critical issues. In A. D. Milner & M. D. Rugg (eds), The Neuropsychology of Consciousness (pp. 179200). San Diego, CA: Academic Press. Schacter, D. L. (1993) Understanding implicit memory: a cognitive neuroscience approach. In A. F. Collins, S. E. Gathercole, M. A. Conway & P. E. Morris, (eds) Theories of Memory (pp. 387412). Hove: Erlbaum. Schacter, D. L., Cooper, L. A. & Delaney, S. M. (1990) Implicit memory for unfamiliar objects depends on access to structural descriptions. J. Exp. Psychol.: Gen., 119, 524. Schacter, D. L., McAndrews, M. P. & Moscovitch, M. (1988) Access to consciousness: dissociations between implicit and explicit knowledge in neuropsychological syndromes. In L. Weiskrantz (ed.), Thought without Language, (pp. 242278) Oxford: University Press. Schank, R. C. (1982) Reminding and memory organization: an introduction to MOPs. In W. G. Lehnert and M. H. Ringle (eds), Strategies for Natural Language Processing, Hillsdale, N.J.: Lawrence Erlbaum Associates. Schank, R. & Abelson, R. P. (1975) Scripts, Plans, and Knowledge. Advance Papers of the 4th International Joint Conference on Artificial Intelligence. Tblisi, USSR. Schank, R. & Abelson, R. (1977) Scripts, Plans, Goals & Understanding: An Enquiry into Human Knowledge Structures. Hillsdale, M.J.: Erlbaum.
Scharder, G., Davis., A., Stefanovic, S. & Christie, P. (1990) The recollection of affect. Psychol.Med., 20, 105109. Schare, M. L., Lisman, S. A. & Spear, N. E. (1984) The effects of mood variation on state dependent retention. Cog. Ther. Res., 84, 387407. Scheibe, K. E., Shaver, P R. & Carrier, N. E. (1967) Colour association values and response interference on variants of the Stroop test. Acta Psychologica, 26, 286295. Scherer, K. R (1984) On the nature and function of emotion: a component process approach. In K. R. Scherer & P. Ekman (eds), Approaches to Emotion. Hillsdale, N.J.: Lawrence Erlbaum. Schneider, W., Dumais, S. T. & Shiffrin, R. M. (1984) Automatic and control processing in attention. In R. Parasuraman & D. R. Davies (eds), Varieties of Attention. Orlando, FL.: Academic Press. Schneider, W. & Fisk, D. (1984) Automatic category search and its transfer. J. Exp. Psychol.: Learn. Mem. Cog., 10, 115. Schneider, W. & Shiffrin, R. M. (1977) Controlled and automatic human information processing: (1) Detection, search and attention. Psychol. Rev., 84, 166. Schotte, D. E., Cools, J. & Payvar, S. (1990) Problem-solving deficits in suicidal patients: Trait vulnerability or state phenomenon? J. Consult. Clin. Psychol., 58, 562564. Schotte, D. E., McNally, R. J. & Turner, M. L. (1990) A dichotic listening analysis of body weight concern in bulimia nervosa. Int. J. Eat. Dis., 9, 109113. Schvaneveldt, R. W. & Meyer, D. E. (1973) Retrieval and comparison processes in semantic memory. In S. Kornblum (ed.), Attention and Performance. Vol. IV. New York: Academic Press. Schwartz, S. (1975) Individual differences in cognition: some relationships between personality and memory. J. Res. Person., 9, 217225. Page 363
Schacter, D. L. (1989) On the relationship between memory and consciousness: dissociable interactions and conscious experience. In H. L. Roediger III and F. I. M. Craik (eds), Varieties of Memory and Consciousness: Essays in Honour of Endel Tulving (pp. 355390). Hillsdale, NJ: Erlbaum. Schacter, D. L. (1992) Consciousness and awareness in memory and amnesia: critical issues. In A. D. Milner & M. D. Rugg (eds), The Neuropsychology of Consciousness (pp. 179200). San Diego, CA: Academic Press. Schacter, D. L. (1993) Understanding implicit memory: a cognitive neuroscience approach. In A. F. Collins, S. E. Gathercole, M. A. Conway & P. E. Morris, (eds) Theories of Memory (pp. 387412). Hove: Erlbaum. Schacter, D. L., Cooper, L. A. & Delaney, S. M. (1990) Implicit memory for unfamiliar objects depends on access to structural descriptions. J. Exp. Psychol.: Gen., 119, 524. Schacter, D. L., McAndrews, M. P. & Moscovitch, M. (1988) Access to consciousness: dissociations between implicit and explicit knowledge in neuropsychological syndromes. In L. Weiskrantz (ed.), Thought without Language, (pp. 242278) Oxford: University Press.
Schank, R. C. (1982) Reminding and memory organization: an introduction to MOPs. In W. G. Lehnert and M. H. Ringle (eds), Strategies for Natural Language Processing, Hillsdale, N.J.: Lawrence Erlbaum Associates. Schank, R. & Abelson, R. P. (1975) Scripts, Plans, and Knowledge. Advance Papers of the 4th International Joint Conference on Artificial Intelligence. Tblisi, USSR. Schank, R. & Abelson, R. (1977) Scripts, Plans, Goals & Understanding: An Enquiry into Human Knowledge Structures. Hillsdale, M.J.: Erlbaum. Scharder, G., Davis., A., Stefanovic, S. & Christie, P. (1990) The recollection of affect. Psychol.Med., 20, 105109. Schare, M. L., Lisman, S. A. & Spear, N. E. (1984) The effects of mood variation on state dependent retention. Cog. Ther. Res., 84, 387407. Scheibe, K. E., Shaver, P R. & Carrier, N. E. (1967) Colour association values and response interference on variants of the Stroop test. Acta Psychologica, 26, 286295. Scherer, K. R (1984) On the nature and function of emotion: a component process approach. In K. R. Scherer & P. Ekman (eds), Approaches to Emotion. Hillsdale, N.J.: Lawrence Erlbaum. Schneider, W., Dumais, S. T. & Shiffrin, R. M. (1984) Automatic and control processing in attention. In R. Parasuraman & D. R. Davies (eds), Varieties of Attention. Orlando, FL.: Academic Press. Schneider, W. & Fisk, D. (1984) Automatic category search and its transfer. J. Exp. Psychol.: Learn. Mem. Cog., 10, 115. Schneider, W. & Shiffrin, R. M. (1977) Controlled and automatic human information processing: (1) Detection, search and attention. Psychol. Rev., 84, 166. Schotte, D. E., Cools, J. & Payvar, S. (1990) Problem-solving deficits in suicidal patients: Trait vulnerability or state phenomenon? J. Consult. Clin. Psychol., 58, 562564. Schotte, D. E., McNally, R. J. & Turner, M. L. (1990) A dichotic listening analysis of body weight concern in bulimia nervosa. Int. J. Eat. Dis., 9, 109113. Schvaneveldt, R. W. & Meyer, D. E. (1973) Retrieval and comparison processes in semantic memory. In S. Kornblum (ed.), Attention and Performance. Vol. IV. New York: Academic Press. Schwartz, S. (1975) Individual differences in cognition: some relationships between personality and memory. J. Res. Person., 9, 217225. Page 365
Shanks, D. R. & St John, M. F. (1994) Characteristics of dissociable human learning systems. Behav. Br. Sci., 17, 367447. Sheehan, M. J. (1981) Constructs and 'conflict' in depression. Br. J. Psychol., 72, 197209. Sheikh, A. A. & Jordan, C. S. (1983) Clinical uses of mental imagery. In A. A. Sheikh (ed.), Imagery: Current Theory, Research and Application. New York: John Wiley. Sheikh, A. A. & Panagiotou, N. C. (1975) Use of mental imagery in psychotherapy: a critical review. Percept. Mot. Skills, 41, 555585.
Shepard, R. N. & Cooper, L. A. (1982) Mental Images and Their Transformations. Cambridge, Mass.: MIT Press. Sher, K. J., Frost, R. & Otto, R. (1983) Cognitive deficits in compulsive checkers: an exploratory study. Behav. Res. Ther., 21, 357363. Sher, K.J., Frost, R.O., Kushner, M., Crews, T.M. & Alexander, J.E. (1989) Memory deficits in compulsive checkers: replication and extension in a clinical sample. Behav. Res. Ther., 27, 6569. Sher, K. J., Mann, B. & Frost, R. O. (1984) Cognitive dysfunction in compulsive checkers: further explorations. Behav. Res. Ther., 22, 493502. Sherman, S. J., Zehner, K. S., Johnson, J. & Hirt, E. R. (1983) Social explanation: the role of timing, set, and recall on subjective likelihood estimates. J. Pers. Soc. Psychol., 44(6), 11271143. Shevrin, H. & Dickman, S. (1980) The psychological unconscious: A necessary assumption for all psychological theory? Amer. Psychol., 35(5), 421434. Shiffrin, R. M. (1976) Capacity limitations in information processing, attention and memory. In W. K. Estes (ed.), Handbook of Learning and Cognitive Processes, Vol. 4. Hillsdale, N.J.: Lawrence Erlbaum. Shiffrin, R. M. (1985) Attention. In R. C. Atkinson, R. J. Herstein, G. Lindsey & R. C. Luce (eds), Steven's Handbook of Experimental Psychology. New York: John Wiley. Shiffrin, R. M. & Schneider, W. (1977) Controlled and automatic human information processing: II. Perceptual learning, automatic attending and a general theory. Psychol. Rev., 84(2), 127190. Shinamura, A.P. (1986) Priming effects in amnesia: evidence for a dissociable memory function. Quart. J. Exp. Psychol., 24, 490504. Shinamura, A.P. (1989) Disorders of memory. In F. Boller and J. Grafman (eds), Handbook of Neuropsychology. Amsterdam: Elsevier. Shinamura, A.P. (1993) Neuropsychological analysis of implicit memory: history, methodology and theoretical implications. In P. Graf & E.J. Masson (eds), Implicit Memory: New Directions in Cognition, Development and Neuropsychology, (pp. 265285). Hillsdale, N.J.: Erlbaum. Sidis, B. (1898) The Psychology of Suggestion. New York: Appleton. Siegel, D. R. & Romig, C. A. (1990) Memory retrieval in treating adult survivors of sexual abuse. Amer. J. Family Ther., 18, 246256. Silberman, E. K., Weingartner, H., Laraia, M., Byrnes, S. & Post, R. M. (1983) Processing of emotional properties of stimuli by depressed and normal subjects. J. Nerv. Ment. Dis., 171, 1014. Silberman, E. K., Weingartner, H. & Post, R. M. (1983) Thinking disorder in depression. Arch. Gen. Psychiat., 40, 775780. Silver, E. A. (1981) Recall of mathematical problem information: solving related problems J. Res. Maths Educ., 12, 5464.
Silverman, L. H. (1976). Psychoanalytic theory: the reports of my death have been greatly exaggerated. Amer. Psychol., 31, 621637. Page 366
Silverman, L. H. (1980) A comprehensive report of studies using the subliminal psychodynamic activation method. Psychol. Res. Bull., 20(3), 122. Silverman, L. H. (1983) The subliminal psychodynamic activation method: overview and comprehensive listing of studies. In J. Masling (ed.), Empirical Studies of Psychoanalytic Theories. Vol. 1. Hillsdale, N.J.: Lawrence Erlbaum. Silverman, L. H. (1984) Further comments on subliminal psychodynamic activations studies (revised) Unpublished manuscript. (Available from Department of Psychology, New York University, 6 Washington Place, New York, 1003.) Silverman, L. H. (1985) Comments on three recent subliminal psychodynamic activation investigations. J. Abnorm. Psychol., 94, 640643. Silverman, L. H., Bronstein, A. & Mendelsohn, E. (1976) The further use of the subliminal psychodynamic activation method for the experimental study of the clinical theory of psychoanalysis: on the specificity of relationships between manifest psychopathology and unconscious conflict. Psychother.: Theory, Res. Pract., 13, 22-16. Silverman, L. H. & Candell, P. (1970) On the relationship between aggressive activation, symbiotic merging, intactness of body boundaries and manifest pathology in schizophrenics. J. Nerv. Ment. Dis., 150, 387399. Silverman, L. H., Clinger, H., Lustbader, L., Farrel, J. & Martin, A. D. (1972) The effects of subliminal driver stimulation on the speech of stutterers. J. Nerv. Ment. Dis., 155, 1421. Silverman, L. H., Frank, S. & Dachinger, P. (1974) A psychoanalytic interpretation of the effectiveness of systematic desensitisation: experimental data bearing on the role of merging fantasies. J. Abnorm. Psychol., 83, 313318. Silverman, L. H., Kwawer, J. S., Wolitzky, C., & Coron, M. (1973) An experimental study of aspects of the psychoanalytic theory of male homosexuality. J. Abnorm. Psychol., 82, 178188. Silverman, L. H., Martin, A., Ungaro, R. & Mendelsohn, E. (1978a) Effect on subliminal stimulation of symbiotic fantasies on behaviour modification treatment of obesity. J. Consult. Clin. Psychol., 46, 432441. Silverman, L. H. & Silverman, D. K. (1964) A clinical-experimental approach to the study of subliminal stimulation. J. Abnorm. Soc. Psychol., 69, 158172. Silverman, L. H. & Spiro, R. H. (1968) The effects of subliminal, supraliminal and vocalised aggression on the ego functioning of schizophrenics. J. Nerv. Ment. Dis., 145, 5061. Silverman, L. H., Ross, D. L., Adler, J. N. & Lustig, D. A. (1978b) Simple research paradigm for demonstrating subliminal dynamic activation: effects of Oedipal stimuli on dart-throwing accuracy in college males. J. Abnorm. Psychol., 87, 341357. Simon, H. A. (1982) Comments. In S. Fiske & M. Clark (eds), Affect and Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Simons, A. D., Murphy, G. E., Levine, J. L. & Wetzel, R. D. (1986) Cognitive therapy and
pharmacotherapy for depression. Arch. Gen. Psychiat., 43, 4350. Simpson, G. B. & Kang, H. (1994) Inhibitory processes in the recognition of homograph meanings. In D. Dagenbach & T. H. Carr (eds), Inhibitory Mechanisms in Attention, Memory & Language (pp. 359382) San Diego, CA: Academic Press. Singer, M., Graesser, A. C. & Trabasso, T. (1994) Minimal or global inference during reading. J. Mem. Lang., 33, 421441. Singer, J. A. & Muffin, K. H. (1992) An experimental investigation of specificity and generality in memory narrative. Imag. Cog. Pers., 11, 233257. Singer, J. A. & Salovey, P. (1988) Mood and memory: evaluating the network theory of affect. Clin. Psych. Rev., 8, 211251. Singer, J. L. (1974) Imagery and Daydream Methods in Psychotherapy and Behaviour Modification. New York: Academic Press. Singer, J.L. (1990) Repression and Dissociation. Chicago: University of Chicago Press. Small, S. A & Robins, C. R. (1988) The influence of induced depressed mood on visual recognition thresholds: predictive ambiguity of associative network models of mood and cognition. Cog. Ther. Res., 12, 295304 Smid, H. G. O. M., Lamain, W., Hogeboom, M. M., Mulder, G. & Mulder, L. J. M. (1991) Psychophysiological evidence for continuous information transfer between visual search and response processes. J. Exp. Psychol.: Hum. Percept. Perf., 17, 696714. Smith, D. A. & Graesser, A. C. (1981) Memory for actions in scripted activities as a function of typicality, retention interval, and retrieval task. Mem. Cog., 9, 550559. Smith, E. R. & Miller, F. D. (1978) Limits on perception of cognitive processes: a reply to Nisbett and Wilson. Psychol. Rev., 85(4), 355362. Smith, G. J. W. & Henriksson, M. (1955) The effect on an established percept of a perceptual process beyond awareness. Acta Psychol., 2, 346355. Smith, G. J. W., Spence, D. P. & Klein, G. S. (1959) Subliminal effects of verbal stimuli. J. Abnorm. Soc. Psychol., 59, 167176. Smith, S. W. (1995) Perceptual processing of nonverbal-relational messages. In D.E. Hewes (ed.), The Cognitive Basis of Interpersonal Communication. Hillsdale, NJ: Erlbaum. Smolensky, P. (1995a) On the proper treatment of connectionism. In C. MacDonald & G. MacDonald (eds), Connectionism: Debates on Psychological Explanation. Oxford, UK: Blackwell. Smolensky, P. (1995b) Constituent structure and explanation in an integrated connectionist/symbolic architecture. In C. MacDonald & G. MacDonald (eds), Connectionism: Debates on Psychological Explanation. Oxford, UK: Blackwell. Smolensky, P. (1989) Connectionist modeling: neural computation/mental connections. In L. Nadel, L. A. Cooper, P. Culcover & R. M. Harnich (eds), Neural Connections, Mental Computation. Computational Models of Cognition and Perception. Cambridge, Mass: MIT Press.
Snodgrass, M., Shevrin, H. & Kopka, M. (1993) The mediation of intentional judgements by unconcious perceptions: the influence of task strategy, task preference, word meaning and motivation. Consc. Cog., 2, 169193. Snyder, M. & Swann, W. B. Jr (1978) Hypothesis-testing processes in social interaction. J. Pers. Soc. Psychol., 36(11), 12021212. Snyder, M. & Uranowitz, S. W. (1978) Reconstructing the past: some cognitive consequences of person perception. J. Pers. Soc. Psychol., 36, 941950. Soares, J. J. F. & Ohman, A. (1993a) Backward masking and skin conductance responses after conditioning to non-feared but fear-relevant stimuli in fearful subjects. Psychophys., 30, 460466. Soares, J. J. F. & Ohman, A. (1993b) Preattentive processing, preparedness and phobias: effects of instruction on conditioned electrodermal responses to masked and non-masked fearrelevant stimuli. Behav. Res. Ther., 31, 8795. Solomon, R. L. & Postman, L. (1952) Frequency of usage as a determinant of recognition thresholds for words. J. Exp. Psychol., 43, 195201. Solomon, R. L. & Corbit, J. D. (1974) An opponent-process theory of motivation: I. Temporal dynamics of affect. Psychol. Rev., 81, 119145. Page 368
Somekh, D. E. & Wilding, J. M. (1973) Perception with awareness in a dichoptic viewing situation. Br. J. Psychol., 64, 339349. Spellman, B. A. & Holyoak, K. J. (1993) An inhibitory mechanism for goal-directed analogical mapping. In Proceedings of the Fifteenth Annual Meeting of the Cognitive Science Society (pp. 947952) Hillsdale, NJ: Erlbaum. Spence, D. P. & Holland, B. (1962) The restricting effects of awareness: a paradox and an explanation. J. Abnorm. Soc. Psychol., 64, 163174. Sperling, G. (1967) Successive approximations to a model for short-term memory. Acta Psychol., 27, 285292. Spielberger, C. D, Gorsuch R. L. & Lushene, R. (1970) Manual for the State-Trait Anxiety Inventory. Palo Alto: Consulting Psychologists Press. Spielman, S. B. (1992) The effects of lateralized stimuli in a dual-task paradigm: implications for multiple resource theory. Diss. Abs. Int., 52, 5002, Spiro, R. J. (1980) Accommodative reconstruction in prose recall. J. Verb. Learn. Verb. Behav., 19, 8495. Sprock, J., Braff, D. L., Saccuzzo, D. P. & Alkinson, J. H. (1983) The relationship of depression and thought disorder in pain patients. Br. J. Med. Psychol., 56, 351360. Squire, L. R. & Frambach, M. (1990) Cognitive skill learning in amnesia. Psychobiol., 18, 109117. Squire, L. R. & McKee, R. (1992) Influences of prior events on cognitive judgements in amnesia. J. Exp. Psychol.: Learn. Mem. Cog., 18, 106115.
Srull, T. K. (1981) Person memory: some tests of associative storage and retrieval models. J. Exper. Psychol.: Hum. Learn. Mem., 47, 440463. Srull, T. K. & Wyer, R. S. (1988) A Dual Process Model of Impression Formation: Advances in Social Cognition, Vol. 1. Hillsdale, NJ: Erlbaum. Stein, N. L. & Glenn, C. G. (1979) An analysis of story comprehension in elementary school children. In R. Freedle (ed.), New Directions in Discourse Processing. Norwood, N.J.: Ablex. Stein, N. L. & Oatley, K. (eds) (1992) Basic Emotions. Hillsdale, N.J.: Lawrence Erlbaum. Steinman, S. B. (1987) Serial and parallel search in pattern vision. Perception, 16, 389398. Stent, G. S. (1981) Cerebral hermeneutics. J. Soc. Biol. Struct., 4, 107124. Sternberg, D. E. & Jarvik, M. E. (1976) Memory functions in depression. Arch. Gen. Psychiat., 33, 219224. Sternberg, R. J. (1977) Intelligence, Information Processing, and Analogical Reasoning. Hillsdale, N.J.: Lawrence Erlbaum. Sternberg, S. (1966) High-speed scanning in human memory. Science, 153, 652654. Sternberg, S. (1969) The discovery of processing stages: extensions of Donder's method. Acta Psychol., 30, 276315. Sternberg, S. (1984) Stage models of mental processing and the additive-factor method. Behav. Brain Sci., 7, 8285. Storms, M. D. & Nisbett, R. E. (1970) Insomnia and the attribution process. J. Pers. Soc. Psychol., 2, 319328. Strack, F., Schwarz, N. & Gschneidinger, E. (1985) Happiness and reminiscing: the role of time perspective, affect, and mode of thinking. J. Pers. Soc. Psychol., 49, 14601469. Strauman, T. J. (1992) Self-guides, autobiographical memory, and anxiety and dysphoria: toward a cognitive model of vulnerability to emotional distress. J. Abnorm. Psychol., 101, 8795. Page 369
Strayer, D. L. & Kramer, A. F. (1990) Attentional requirements of automatic and controlled processing. J. Exp. Psychol.: Learn., Mem. Cog. 16, 6782. Stroebe, M. S., Stroebe, W. & Hansson, R. O. (eds) (1993) Handbook of Bereavement: Theory, Research and Intervention. Campbridge: Cambridge University Press. Stroh, N., Shaw, A. M. & Washbourn, M. F. (1908) A study in guessing. Amer. J. Psychol., 1908, 243245. Stromgren, L. S. (1977) The influence of depression on memory. Acta Psychiat. Scand., 5, 109128. Strongman, K. T. & Kemp, S. (1991) Autobiographical memory for emotion. Bull. Psychon. Soc., 29, 195198. Stroop, J. R. (1935) Studies of interference in serial verbal reactions. J. Exp. Psychol., 18, 643662.
Strosahl, K. D. & Ascough, J. C. (1981) Clinical uses of mental imagery: experimental foundations, theoretical misconceptions and research issues. Psychol. Bull., 89, 422438. Suinn, R. M. (1983) Imagery and sports. In A. A. Sheikh (ed.), Imagery: Current Theory, Research and Application. New York: John Wiley. Sullivan, M. P., Faust, M. E. & Balota, D. A. (1995) Identify priming in older adults and individuals with dementia of the Alzheimer type. Neuropsychol., 9, 537555. Sun, R. (1994) Connectionist models of commonsense reasoning. In D.S. Levine & M. Aparicio (eds), Neural Networks for Knowledge Representation and Inference (pp. 241268). Hillsdale, NJ: Erlbaum. Sutherland, G., Newman, B. & Rachman, S. (1982) Experimental investigations of the relations between mood and intrusive unwanted cognitions. Br. J. Med. Psychol., 55, 127138. Sutton, S. K. & Mineka, S. (submitted) 'Naturalistic' priming of covariation bias for fearrelevant stimuli and aversive outcomes. Tait, R. & Silver, R. C. (1989) Coming to terms with major negative events. In J. S. Uleman and J. A. Bargh (eds), Unintended Thought (pp. 351382). New York: Guilford. Talbot, N. L., Duberstein P. R. & Scott, P. (1991) Subliminal psychodynamic activation, food consumption, and self-confidence. J. Clin. Psychol., 47, 813823. Tallis, F. (1993) Doubt reduction using distinctive stimuli as a treatment for compulsive checking: an exploratory investigation. Clin. Psychol. Psychother., 1, 4552. Tallis, F., Eysenck, M. W. & Mathews, A. (1991) A critical analysis of some theoretical approaches. Anx. Res., 4, 97108. Tallis, F., Eysenck, M. W. & Mathews, A. (1992) A questionnaire for the measurement of nonpathological worry. Pers. Indiv. Diff., 13, 161168. Taylor, S. E. & Crocker, J. (1981) Schematic bases of social information processing. In E. T. Higgins, C. P. Herman & M. P. Zanna (eds), Social Cognition: The Ontario Symposium, Vol. I. Hillsdale, N.J.: Lawrence Erlbaum. Teasdale, J. D. (1983a) Negative thinking in depression: Cause, effect or reciprocal relationship? Adv. Behav. Res. Ther., 5, 325. Teasdale, J. D. (1983b) Affect and accessibility. In D. E. Broadbent (ed.), Functional Aspects of Memory. London: The Royal Society. Teasdale, J. D. (1988) Cognitive vulnerability to depression. Cog. Emot., 2, 247274. Teasdale, J. D. and Barnard, P. J. (1993) Affect, Cognition and Change: Re-modelling Depressive Thought, Hillsdale, N.J.: Lawrence Erlbaum Associates. Teasdale, J. D. & Dent, J. (1987) Cognitive vulnerability to depression: an investigation of two hypotheses. Br. J. Clin. Psychol., 26, 113126. Page 370
Teasdale, J. D. & Fogarty, S. J. (1979) Differential effects of induced mood on retrieval of pleasant and unpleasant events from episodic memory. J. Abnorm. Psychol., 88, 248257.
Teasdale, J. D. & Russell, M. L. (1983) Differential effects of induced mood on the recall of positive, negative and neutral words. Br. J. Clin. Psychol., 22, 163172. Teasdale, J. D. & Spencer, P. (1984) Induced mood and estimate of past success. Br. J. Clin. Psychol., 23, 149150. Teasdale, J. D. & Taylor, R. (1981) Induced mood and accessibility of memories: an effect of mood states or of induction procedure? Br. J. Clin. Psychol., 20, 3948. Teasdale, J. D., Taylor, M. J., Cooper, Z. & Hayhurst, H. (1995) Depressive thinking: shifts in construct accessibility or in schematic mental models? J. Abnorm. Psychol., 104, 500507 Teasdale, J. D., Taylor, R. & Fogarty, S. J. (1980) Effects of induced elation-depression on the accessibility of memories of happy and unhappy experiences. Behav. Res. Ther., 18, 339346. Teder, W., Kujala, T. & Naatanen, R. (1993) Selection of speech messages in free-field listening. Neuroreport, 5, 307309. Theewues, J. (1994) Endogenous and exogenous control of visual selection. Perception, 23, 429440. Theios, J. & Amrhein, P. C. (1989) Theoretical analysis of the cognitive processing of lexical and pictorial stimuli: naming and conceptual comparisons. Psychol. Rev., 96, 524. Thomas, D. L. & Diener, E. (1990) Memory accuracy in the recall of emotions. J. Pers. Soc. Psychol., 59, 291297, Thorndike, E. L. & Lorge, I. (1944) The Teacher's Word-Book of Thirty Thousand Words. New York: Columbia University, Teacher's College, Bureau of Publications. Thorndike, E. L. (1931) Human Learning. New York: Appleton-Century Crofts. Thorndyke, P. W. (1975) Cognitive structures in human story comprehension and memory. Technical Report P-5513. Santa Monica: The Rand Corporation. Thorndyke, P. W. (1976) The role of interferences in discourse comprehension. J. Verb. Learn. Verb. Behav., 15, 437446. Thorndyke, P. W. (1977) Cognitive structures in comprehension and memory of narrative discourse. Cog. Psychol., 9, 770110. Thorpe, S. (1994) Unpublished D.Phil. University of Oxford. Tipper, S. P. (1985) The negative priming effects: inhibitory priming by ignored object, Quart. J. Exp.Psychol., 37A, 571590. Tipper, S. P., Brehaut, J. & Driver, J. (1990) Selection of moving and static objects for the control of spatially directed action. J. Exp. Psychol.: Hum. Percept. Perf., 16, 492504 Titchener, A. B. & Pyle, W. H. (1907) The effect of imperceptual shadows on the judgement of distance. Practical Amer. Philosoph. Soc., 46, 94109. Tobias, B. A., Kihlstrom, J. F. & Schacter, D. L. (1992) Emotion and implicit memory. In S. A. Christianson (ed.) The Handbook of Emotion and Memory Research and Theory. Hillsdale, N.J.: Erlbaum. Tomarken, A. J., Cook, M. & Mineka, S. (1989) Fear-relevant selective associations and
covariation bias. J. Abnorm. Psychol., 98, 381394. Tomarken, A. J., Sutton, S. K. and Mineka, S. (1995) 'Fear-relevant illusory correlations: what types of associations promote judgmental bias? J. Abnorm. Psychol., 104, 312326. Page 371
Toth, J. P., Reingold, E. M. & Jacoby, L. L. (1994) Towards a redefinition of implicit memory: process dissociations following elaborative processing and self-generation. J. Exp. Psychol.: Learn. Mem. Cog., 20, 290303. Townsend, J. T. (1974) Issues and models concerning the processing of a finite number of inputs. In B.H. Kantowitz (ed.), Human Information Processing: Tutorials in Performance and Cognition. Hillsdale, NJ: Erlbaum. Townsend, J. T. (1990) Serial vs parallel processing: sometimes they look like Tweedledum and Tweedledee but they can (and should) be distinguished. Psychol. Sci., 1, 4654. Trafimow, D. & Wyer, R. S. (1993) Cognitive representation of mundane social events. J. Pers. Soc. Psychol., 64., 365376. Trandel, D. V. & McNally, R. J. (1987) Perception of threat cues in post-traumatic stress disorder: Semantic processing without awareness? Behav. Res. Ther., 29, 451457. Treisman, A. (1986) Features and objects in visual processing. Scient. Amer., 255, 114125. Treisman, A. (1988) Features and objects: the fourteenth Bartlett Memorial Lecture. Quart. J. Exp. Psychol., 40A, 201237. Treisman, A. M. (1960) Contextual cues in selective listening. Quart. J. Exp. Psychol., 12, 242248. Treisman, A. M., Squire, R. & Green, G. (1974) Semantic processing in dichotic listening? A replication. Mem. Cog., 2, 641646. Treisman, A. & Sato, S. (1990) Conjunction search revisited. J. Exp.Psychol.: Hum. Percept. Perf., 16, 459478. Trimble, R. & Eriksen, C. W. (1966) Subliminal cues in the Muller-type illusion. Percept. Psychophys., 1, 401404. Tulving, E. & Thomson, D. N. (1973) Encoding specificity and retrieval processes in episodic memory. Psychol. Rev., 80, 352373. Tulving, E. (1983) Elements of Episodic Memory. Oxford: Oxford University Press. Tulving, E., Hayman, C. A. G. & MacDonald, C. (1991) Long-lasting perceptual priming and semantic learning in amnesia: a case experiment. J. Exp. Psychol.: Learn. Mem. Cog. 17, 595617. Tulving, E., Schacter, D. L. & Stark, H. A. (1982) Priming effects in word-fragment completion are independent of recognition memory. J. Exp. Psychol.: Learn. Mem. Cog., 8, 336342. Turk, D. C. & Salovey, P. (1986) Clinical information processing: bias inoculation. In R. Ingram (ed.), Information Processing Approaches to Clinical Psychology, Orlando, FL: Academic Press.
Turvey, M. (1973) On peripheral and central processes in vision: inferences from an information-processing analysis of masking with patterned stimuli. Psychol. Rev., 80, 152. Tversky, A. & Kahneman, D. (1973) Availability: a heuristic for judging frequency and probability. Cog. Psychol., 5, 207232. Tversky, A. & Kahneman, D. (1974) Judgement under uncertainty: heuristics and biases. Science, 185, 11241131. Tversky, A. & Kahneman, D. (1982) Judgement under uncertainty: heuristics and biases. In D. Kahneman, P. Slovic & A. Tversky (eds), Judgement under Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press. Tyler, H. R. (1968) Abnormalities of perception with defective eye movements (Balint's Syndrome) Cortex, 4,154171. Page 372
Tyrer, P., Horn, S. & Lee, I. (1978) Treatment of agoraphobia by subliminal and supraliminal exposure to phobic cine-film. Lancet, 1, 358360. Umilta, C. (1988) The control operations of consciousness. In A. J. Marcel & E. Bisiach (eds), Consciousness in Contemporary Science (pp. 334356) Oxford: Clarendon Press. Underwood, G. (1976) Semantic interference from unattended printed words. Br. J. Psychol., 67, 327338. Underwood, G. L. (1977) Facilitation from attended and unattended messages. J. Verb. Learn. Verb. Behav., 16, 99106. Underwood, G., Whitefield, A. & Winfield, J. (1982) Effects of contextual constraints and non-fixated words in a simple reading task. J. Res. Read., 5, 8999. Upper, D. & Cantela, J. R. (1979) Covert Conditioning. New York: Pergamon Press. van den Elzen, J. (1996) Cognitive disengagement in depression. Unpublished PhD thesis, the University of Western Australia. van den Hout, M., Tenney, N., Huygens, K. & Merckelbach, H. (1995) Responding to subliminal threat cues is related to trait anxiety and emotional vulnerability: a successful replication of MacLeod & Hagan (1992) Behav. Res. Ther., 33, 451454. van Selst, M. & Merikle, P. A. (1993) Perception below the objective threshold? Consc. Cog, 2, 194203. Velten, E. (1968) A laboratory task for induction of mood states. Behav. Res. Ther., 6, 473482. Vestre, N. D. & Caulfield, B. P. (1986) Perception of neutral personality descriptions by depressed and nondepressed subjects. Cog. Ther. Res., 10, 3136. Volans, P. J. (1976) Styles of decision-making and probability appraisal in selected obsessional and phobic patients. Br. J. Soc. Clin. Psychol., 15, 305317. von Benedek, L. (1992) The mental activity of the psychoanalyst. Psychother. Res., 2, 6372. von Wright, J. N., Anderson, K. & Stenman, U. (1975) Generalisation of conditioned GSRs in dichotic listening. In P. M. A. Rabbitt & S. Dorric (eds), Attention and Performance. New York: Academic Press.
Walker, C. H. & Meyer, B. J. F. (1980) Integrating different types of information in text. J. Verb. Learn. Verb. Behav., 19, 263275. Walker, C. H. & Yekovich, F. R. (1984) Script-based inferences: effects of text and knowledge variables on recognition memory. J. Verb. Learn. Verb. Behav., 23, 357370. Walker, P. & Myer, R. R. (1978) The subliminal perception of movement and the course of autokinesis. Br. J. Psychol., 69, 225231. Walley, R. E., McLeod, B. E. & Weiden, T. D. (1994) Increased attention to the irrelevant dimension increases interference in a spatial Stroop task. Canad. J. Exp. Psychol., 48, 467492. Ware, R. C. (1995) Scylla and Charybdis: Sexual abuse or 'false memory syndrome'? J. Anal. Psychol., 40, 522. Warren, C. & Morton, J. (1982) The effect of priming on picture recognition. Br. J. Psychol. 73, 117129. Warren, R. E. (1972) Stimulus encoding and memory. J. Exp. Psychol., 94, 90100. Warren, R. E. (1974) Association, directionality and stimulus encoding. J. Exp. Psychol., 102, 151158. Warrington, E. K. & Weiskrantz, L. (1974) The effect of prior learning on subsequent retention in amnesic patients. Neuropsychologia, 12, 419428. Wason, P. C. & Shapiro, D. (1971) Quart. J. Exp. Psychol., 23, 63. Page 373
Wason, P. C. (1971) Problem-solving and reasoning. Br. Med. Bull., 27, 206210. Watkins, P. C., Mathews, A., Williams, D. A. & Fuller, R. D. (1992) Mood congruent memory in depression: emotional priming or elaboration? J. Abnorm. Psychol., 101, 581586. Watson, D. & Clark, L. A. (1984) Negative affectivity: the disposition to experience aversive emotional states. Psychol. Bull., 96(3), 465490. Watson, J. B. (1925) Behaviourism. New York: W. W. Norton. Watts, F. N. (1974) The control of spontaneous recovery of anxiety in imaginal desensitisation. Behav. Res. Ther., 12, 5759. Watts, F. N. (1979) Habituation model of systematic desensitisation. Psychol. Bull., 86, 627637. Watts, F. N. (1983) Affective cognitive: sequel to Zajonc and Rachman. Behav. Res. Ther., 21, 8990. Watts, F. N. (1984) Cognitive paradigms for the study of depression. Paper presented at British Psychological Society Conference, Warwick, April 1984. Watts, F. N. (1985) Individual-centred cognitive counselling for study problems. Br. J. Couns. Guid., 13, 238247. Watts, F. N. (1986) Cognitive processing in phobias. Behav. Psychother., 14, 295301. Watts, F. N. (1989) Experimental abnormal psychology. In G. Parry & F. N. Watts (eds), Skills
and Methods in Mental Health Research. Brighton: Lawrence Erlbaum. Watts, F. N. (1993) Problems of memory and concentration. In C. G. Costello (ed.), Symptoms of Depression. New York: John Wiley. Watts, F. N. (1995a) An information processing approach to obsessional checking. Clin. Psychol. Psychother., 2, 6977. Watts, F. N. (1995b) Depression and anxiety. In A. Baddeley, F. N. Watts & B. Wilson (eds), Handbook of Memory Disorders. Chichester: John Wiley. Watts, F. N. & Blackstock, A. (1987) Lang's theory and emotional imagery. Cog. Emot., 1, 391405. Watts, F. N. & Cooper, Z. (1989) The effects of depression on structural aspects of the recall of prose. J. Abnorm. Psychol., 98, 150153. Watts., F. N. & Dalgleish, T. (1991) Memory for phobia-related words in spider phobias. Cog. Emot., 5, 313329. Watts, F. N., Dalgleish, T., Bourke, P. & Healy, D. (1990) Memory deficit in clinical depression: processing resources and the structure of materials. Psychol. Med., 20, 345349. Watts, F. N., Herbert, J., Moore, G. F. & Levey, A. (1986a) Approaches to studying, personality and examination. Pers. Indiv. Diff., 7, 243245. Watts, F. N., Morris, L. & MacLeod, A. K. (1987) Recognition memory in depression. J. Abnorm. Psychol., 96, 273275. Watts. F. N., McKenna, F. P., Sharrock, R. & Trezise, L. (1986b) Colour naming of phobia related words. Br. J. Psychol., 77, 97108. Watts, F. N., MacLeod, A. K. & Morris, L. (1988a) Associations between phenomenal and objective aspects of concentration problems in depressed patients. Br. J. Psychol., 79, 241250. Watts, F. N., MacLeod, A . K. & Morris, L. (1988b) A remedial strategy for memory and concentration problems in depressed patients. Cog. Ther. Res., 12, 185193. Watts, F. N. & Sharrock, R. (1985a) Description and measurement of concentration problems in depressed patients. Psychol. Med., 15, 317326. Watts, F. N. & Sharrock, R. (1985b) Relationships between spider constructs in phobics. Br. J. Med. Psychol., 58, 149153. Page 374
Watts, F. N. & Sharrock, R. (1987) Cued recall in depression. Br. J. Clin. Psychol., 26, 149150. Watts, F. N., Sharrock, R. & Trezise, L. (1986) Detail and elaboration in phobic imagery. Behav. Psychother., 14, 115123. Watts, F. N., Trezise, L. & Sharrock, R. (1986) Processing of phobic stimuli. Br. J. Clin. Psychol., 25, 253261. Wegman, C. (1985) Psychoanalysis and Cognitive Psychology. New York: Academic Press. Weingartner, H. & Silberman, E. (1982) Models of cognitive impairment: cognitive changes in
depression. Psychopharm. Bull., 18, 2742. Weingartner, H., Cohen, R. M., Murphy, D. L., Martello, J. & Gerdt, C. (1981) Cognitive processes in depression. Arch. Gen. Psychiat., 38, 4247. Weiskrantz, L. (1986) Blindsight: A Case Study and Implications. Oxford: University Press. Weiskrantz, L. (1993) Search for the unseen. In A. Baddeley & L. Weiskrantz (eds), Attention: Selection, Awareness and Control. A Tribute to Donald Broadbent (pp. 235245) Oxford: Clarendon Press. Weiskrantz, L., Warrington, K., Sanders, M. D. & Marshall, J. (1974) Visual capacity of the hemianopic field following a restricted occipital ablation. Psychonom. Sci., 2, 7576. Weldon, M. S. & Roediger, H. L. (1987) Altering retrieval demands reverses the picture superiority effect. Mem. Cog., 15, 269280. White, J. D. & Carlston, D. E. (1983) Consequences of schemata for attention, impressions, and recall in complex social interactions. J. Pers. Soc. Psychol., 45, (3), 538549. White, P. (1980) Limitations on verbal reports of internal events: a refutation of Nisbett and Wilson and of PEM. Psychol. Rev., 87(1), 105112. Whitehead, A. (1973) Verbal learning and memory in elderly depressives. Br. J. Psychiat., 123, 203208. Whitehead, A. (1974) Factors in the learning deficit of elderly depressives. Br. J. Soc. Clin. Psychol., 13, 201208. Whitfield, C. (1995) Memory and Abuse: Remembering and Healing the Effects of Trauma. Deerfield Beach, FL: Health Communications. Whyte, L. L. (1978) The Unconscious Before Freud. London: Longwood. Wickelgren, W. (1976) Network strength theory of storage and retrieval dynamics. Psychol. Rev., 83, 466478. Wickens, C. D. (1979) The structure of attentional resources. In R. Nickerson (ed.), Attention and Performance, Vol. VIII. Hillsdale, N.J.: Lawrence Erlbaum. Wickens, C. D. (1984) Processing resources in attention. In R. Parasuraman and D. R. Davies (eds), Varieties of Attention. Orlando, FL.: Academic. Wickens, C. D. & Boles, D. B. (1987) Display formatting in integration and nonintegration tasks. Human Factors, 29, 395406. Williams, A. (1938) Perception of subliminal visual stimuli. J. Psychol., 6, 187199. Williams, J. M. G. (1992) The Psychological Treatment of Depression: A Guide to the Theory and Practice of Cognitive Behaviour Therapy. London: Routledge. Williams, J. M. G. (1994) Interacting cognitive subsystems and unvoiced murmurs. Cog. Emot., 8, 571579. Williams, J. M. G. (1996) Depression and the specificity of autobiographical memory. In D. C. Rubin (ed.), Remembering our Past: Studies in Autobiographical Memory. Cambridge: Cambridge University Press.
Page 375
Williams, J. M. G. & Broadbent, K. (1986a) Autobiographical memory in attempted suicide patients. J. Abnorm. Psychol., 95, 144149. Williams, J. M. G. & Broadbent, K. (1986b) Distraction by emotional stimuli: use of a Stroop task with suicide attempters. Br. J. Clin. Psychol., 25, 101110. Williams, J. M. G. & Dritschel, B. H. (1988) Emotional disturbance and the specificity of autobiographical memory. Cog. Emot., 2, 221234. Williams, J. M. G. & Dritschel, B. H. (1992) Categoric and extended autobiographical memories, in M. A. Conway, D. C. Rubin, H. Spinnler & W. A. Wagenaar (eds), Theoretical Perspectives on Autobiographical Memory, (pp. 391412). Dordecht, Boston & London: Kluwer Academic Publishers. Williams, J. M. G., Ellis, N. C., Tyers, C., Macleod, A. K. & Rose, G. (1996) Specificity of autobiographical memory and imageability of the future. Mem. Cog., 24, 116-125. Williams, J. M. G. & Marker, H. (1991) Money hidden and rediscovered in; subsequent manic phases: a case of state-dependent re-enactment. Br. J. Psychiat., 159, 579581. Williams, J. M. G., Mathews, A., & MacLeod, C. (1996) The emotional Stroop task and psychopathology. Psychol. Bull., 120, 324. Williams, J. M. G. & Nulty, D. D. (1986) Construct accessibility, depression and the emotional Stroop task: Transient emotion or stable structure? Pers. Indiv. Diff., 7, 485491. Williams, J. M. G. & Oaksford, M. (1992) Cognitive science and psychopathology: from experiments to connectionism. In D. Stein & G. Young (eds), Cognitive Science and Clinical Problems. New York: Academic Press. Williams, J. M. G. & Scott, J. (1988) Autobiographical memory in depression. Psychol, Med., 18, 689695. Williams, J. M. G. & Teasdale, J. D. (1982) Facilitation and helplessness: the interaction of perceived difficulty and importance of a task. Behav. Res. Ther., 20, 161171. Williams, J. M. G., Watts, F. N., MacLeod, C. & Mathews, A. (1988) Cognitive Psychology and Emotional Disorders. Chichester: John Wiley. Williams, K. W. & Durso, F. T. (1986) Judging category frequency: automaticity or availability? J. Exp. Psychol.: Learn. Mem. Cog., 12, 387396. Williams, M. D. & Hollan, J. D. (1981) Process of retrieval from very long-term memory. Cog. Sci., 5, 87119. Williams, W. H, Williams, J. M. G. & Ghadiali, E. (submitted) Specificity of autobiographical memory following closed head injury. Wine, J. (1980) Cognitive-attentional theory of test anxiety. In I. G. Sarason (ed.), Test Anxiety, Theory, Research and Application. Hillsdale, NJ.: Lawrence Erlbaum. Winter, D. A. (1983) Logical inconsistency in construction relationships: Conflict or complexity? Br. J. Med. Psychol., 56, 7987. Winthorpe, C. & Rabbitt, P. A. (1988) Working memory capacity, IQ, age, and the ability to
recount autobiographical events. In M. M. Gruneberg, P. E. Morris & R. N. Sykes (eds) Practical Aspects of Memory; Current Research and Issues II (pp. 175179). Chichester: John Wiley. Wolpe, J. (1958) Psychotherapy by Reciprocal Inhibition. Stanford: Stanford University Press. Woods, D. L., Alho, K. & Algazi, A. (1993) Intermodal selective attention: evidence for processing in tonotopic auditory fields. Psychophys., 30, 287295. Worthington, A. G. (1964) Effect of Worthing subliminal structural cues on reproductions of a simple line drawing. Percept. Mot. Skills, 19, 823882. Page 376
Wright, J. & Mischel, W. (1982) Influence of affect on cognitive special learning person variables. J. Pers. Soc. Psychol., 43, 901914. Yantis, S. & Jones, E. (1991 Mechanisms of selective attention: temporally modulated priority tags. Percept. Psychophys., 50, 166178. Yekovich, F. R. & Thorndyke, T. W. (1981) An evaluation of alternative functional models of narrative schemata. J. Verb. Learn. Verb. Behav., 20, 454469. Yonelinas, A. P. & Jacoby, L. L. (1995) Dissociating automatic and controlled processes in a memory search task: beyond implicit memory. Psychol. Res. 57, 156165. York, D., Borkovec, T. D., Vasey, M. & Sten, R. (1987) Effects of worry and somatic anxiety induction on thoughts, emotion and physiological activity. Behav. Res. Ther., 25, 523526. Zajonc, R. B. (1980) Feeling and thinking: preferences need no inferences. Amer. Psychol., 35, 151175. Zajonc, R. B. (1984) On the primacy of affect. Amer. Psychol., 39, 117123. Zajonc, R. B., Pietromonaco, P. & Bargh, J. (1982) Independence and interaction of affect and cognition. In M. S. Clark and S. T. Fiske (eds), Affect and Cognition. Hillsdale, N.J.: Lawrence Erlbaum. Zander, J. R. & McNally, R. J. (1988) Bio-informational processing in agoraphobia. Behav. Res. Ther., 26, 421429. Zaragoza, M. S. & Mitchell, K. J. (1995) Empirical psychology and the repressed memory debate: current status and future directions. Consc. Cog., 4, 116119. Zarantonello, M., Slaymaker, F., Johnson, J. & Petzel, T. (1984) Effects of anxiety and depression on anagram performance, rating of cognitive interference, and the negative subjective evaluation of performance. J. Clin. Psychol., 40, 2025. Zimbardo, P. G., Cohen, A., Weisenberg, M., Dworkin, L. & Firestone, I. (1969) The control of experiment pain. In P. G. Zimbardo (ed.) The Cognitive Control of Motivation. Glenview, Illinois: Scott Foreman. Zimmerman, M. & Coryell W. (1986) Reliability of follow-up assessments of depressed inpatients. Arch. Gen. Psychiat., 43, 468470. Zuroff, D. C., Colussy, S. A. & Wielgus, M. S. (1983) Selective memory and depression: a cautionary note concerning response bias. Cog. Ther. Res., 7, 223232. Page 377
Author Index A Abelson, R. P., 164, 216 Abrams, M., 247 Abramson, L. Y., 198 Adamowicz, A., 34 Alba, J. W., 12, 211 Albrecht, J. E., 218 Alexander, J. E., 54 Algazi, A., 21 Alho, K., 21 Ali, M. A., 256 Alkinson, J. H., 50, 52 Allan, L. G., 238 Allen, R., 244 Allison, J., 252 Alloy, L. B., 194, 195, 198, 218 Allport, D. A., 69 Allport, T. A., 253, 267 Alpert, J. L., 270 Altman, F., 68 Altman, G. T. N., 248 Amin, J., 199 Amir, N., 89, 90, 100, 113, 159, 184, 285, 293 Amrhein, P. C., 25 Anastasi, J. S., 20 Anastasides, P., 228 Anderson, J. R., 11, 24
Anderson, K., 251 Anderson, M. C., 24, 31 Anderson, M. P., 179 Anderson, R. C., 151, 216, 217 Andrade, J., 256 Andreasen, N. C., 52 Andrews, B., 139 Apfelbalt, D., 181 Apfelblat, D., 20 Arian, S., 274 Arntz, A., 92, 99, 103, 111, 199, 200, 203 Asarnow, R. F., 40, 52 Asbury, A. J., 256 Ascough, J. C., 179, 187 Ashbrook, P. W., 20, 26, 45, 46, 55, 63, 65, 66 B Baars, B. J., 261, 265 Babkoff, H., 35 Baddeley, A., 237, 278, 282 Baddeley, A. D., 20, 59 Baddeley, A. D, 134 Baddeley, A. D., 165 Bailey, D. E., 34 Bainbridge, L., 37 Baker, J. E., 63 Baker, L. L., 175 Balay, J., 274 Ball, T. M., 170 Balota, D. A., 24 Bamford, K. A., 52
Banaji, M. R., 222, 223 Banquet, J. P., 40 Bara, B. G., 36 Bargh, J., 3 Bargh, J. A., 79, 224, 252, 253 Barker, J. E., 63 Barlow, D. H., 181, 183 Barnard, P., 289, 302 Barnard, P. J., 12, 15, 18, 140, 144, 145, 148, 166 Baron, M. G., 186 Baron, R., 46 Barter, S. G., 77, 128, 131, 132 Barthol, R. P., 253 Bartlett, F. C., 36, 37, 210 Basden, B., 242 Basden, D., 242 Bashore, T. A., 28 Bashore, T. R., 26 Baveux, P., 239 Page 378
Beck, A. T., 10, 18, 36, 52, 106, 143, 186, 190, 218, 268, 286, 287 Becker, A. S., 64 Becker, R., 159, 288 Beech, A. R., 24 Beekhuis, M. E., 60, 62 Bekerian, D. A., 203, 271 Belli, R. F., 271 Benjamin, C., 275 Berent, S., 47 Bernet, C. Z., 152, 218, 222 Berry, D., 244
Berry, D. C., 243, 244, 245 Berry, R., 60 Bettelheim, B., 303, 304 Bickley, W., 177 Bierer, L. M., 61 Bird, C., 84, 87, 96, 97, 103, 105, 107, 129 Birkmore, D. P., 20 Bisiach, E., 261 Bizet, E., 244 Bjork, E. L., 24 Bjork, R. A., 24 Black, J., 211, 214, 215, 217 Black, J. B., 164 Blackburn, I. M., 48 Blackstock, A., 57, 173, 174 Blaney, P. H., 143, 196, 218 Blankstein, K. R., 53 Blaxton, T. A., 161, 162 Bless, H., 203 Block, N., 170, 171, 172, 232, 256, 261 Block, R. I., 256 Bobrow, D. G., 28, 164, 212 Boies, S. W., 79 Boles, D. B., 20 Bolles, R. C., 20, 34 Bolten, S., 34 Bonanno, G. A., 256 Bond, R. A., 253 Bonke, B., 256 Bonn, J., 88
Bonto, M. A., 20 Booker, J., 251 Bootzin, R. R., 81 Borgeld, T. W., 275 Borkovec, T. D., 179, 181, 182, 300 Bornstein, R. F., 252, 257, 275 Bottrell, J., 179 Bourke, P., 63 Boville, J. G., 256 Bower, G., 211, 214, 215, 217, 222, 223, 224, 227, 284 Bower, G. H., 11, 24, 36, 106, 134, 135, 137, 138, 142, 143, 147, 176, 202 Bowers, J. S., 160, 242 Bradley, B., 88, 96, 105, 108, 113, 120, 129 Bradley, B. P., 153, 154, 224, 225, 285, 286, 287, 289 Bradshaw, J. L., 251 Braff, D. L., 50, 52 Brand, N., 58 Bransford, J. D., 216 Breen, D., 180 Brehaut, J., 24, 51, 109 Bressler, J., 252 Brewer, M., 34 Brewer, W. F., 142 Brewin, C. R., 11, 139, 164, 166, 286 Brittlebank, A. D., 164, 167, 286 Broadbent, D., 244, 245, 288, 289 Broadbent, D. E., 19, 21, 22, 23, 26, 31, 32, 37, 50, 53, 75, 81, 109, 249 Broadbent, K., 86, 99, 164 Broadbent, M. H. P., 51, 53, 109 Brodeur, D., 21
Brody, N., 245 Bronson, W. C., 81 Bronstein, A., 274 Brooks, L., 244 Brosschot, J. F., 108 Brown, L. B., 138 Bruch, M. A., 68 Bruner, J. S., 23, 34 Buijink, B. M., 27 Bullemer, P., 249 Bundensen, C., 22 Burgess, I. S., 77 Burgess, P., 264 Burke, A., 145, 163 Burke, M., 158, 288, 290 Burkell, J., 22, 251 Burkham, R., 274 Burnett, J. W., 53 Burton, H. A., 223 Butler, G., 185, 202, 204, 228 Button, E., 180, 181 Byers, C. N., 205 Byrne, A., 157, 162, 290, 295 Byrne, D., 23, 46, 47, 49, 50, 81 Byrnes, S., 57 Caine, E. D., 52 Calamari, J. E., 90, 100, 113, 184 Calev, A., 46, 56, 60, 62 Calvo, M. G., 26, 65, 66, 229 Camp, C., 61
Campbell, D., 48 Campbell, L., 87, 93, 97, 112, 130 Campbell, W. K., 141, 203 Cancienne, J., 159, 285, 288 Candell, P., 274 Cane, D. B., 92, 96, 196, 284 Cantela, J. R., 186 Cantor, G. W., 244 Carlson, R., 246 Carpenter, K. M., 201 Carpenter, P., 39 Carr, T. H., 257, 259 Carrier, M. L., 19 Carrier, N. E., 85 Carroll, D., 179 Carroll, J. M., 34 Carroll, J. S., 201 Caspar, F., 33 Casscells, W., 193 Cassens, G., 45, 52 Cassiday, K. L., 89, 98, 99, 108 Caulfield, B. P., 196 Cave, K. R., 22 Cermak, L. S., 69 Cha, K. H., 24 Chaffin, R., 12 Chajczyk, D., 251 Challis, B. H., 295 Chamberlin, C. J., 211 Chambers, D., 171
Champion, L. A., 11, 36, 169 Chan, C., 248 Channon, S., 63 Chapman, J. P., 45, 195, 199, 208 Chapman, L. J., 45, 195, 199, 208 Charniak, E., 35 Chase, C., 34 Chase, W. G., 172 Chazran, S., 62 Cheal, M., 29 Cheesman, P., 255 Chen, E., 159, 287, 288, 290 Cherry, E. C., 250 Chin, C. Y., 295 Chmiel, N., 37 Choi, S., 20 Christian, J., 177 Christianson, S. A., 144, 145, 146, 147 Christie, P., 147 Church, B. A., 242 Cialdini, R. B., 201 Claeys, W., 152, 288 Clark, A., 32 Clark D. A., 36, 52 Clark, D. A., 182 Clark, D. M., 87, 88, 89, 90, 94, 95, 98, 105, 119, 130, 137, 138, 139, 142, 153, 154, 182, 228, 284 Clark, H. H., 34, 172 Clark, L. A., 82 Clayton, K. T., 177
Clements, C., 159, 285, 293 Clinger, H., 274 Cloitre, M., 159, 285, 288, 295 Clore, G. L., 197, 202 Cohen, 229 Cohen, A., 234 Cohen, J. D., 114, 115, 118, 120, 122, 129 Cohen, M. M., 18 Cohen, N. J., 107, 249, 252 Cohen, P. R., 11, 58, 147, 176 Cohen R. M., 58, 60, 62, 63 Colby, C. A., 58 Colby, C. L., 35 Coles, M. G. H., 28 Collins, J. C., 238 Colussy, S. A., 57 Combs, B., 192 Constans, 202 Conway, A. R. A., 39 Cook, E. W. III, 174 Cook, M., 199, 200 Cooper, L. A., 170, 242 Cooper, R., 33, 39, 64 Coover, J. E., 252 Corbit, J. D., 126 Cordner, M. D., 92 Cornell, D. G., 47 Coron, M., 274 Corteen, R. S., 251 Cortese, C., 238
Coryell, W., 146 Courrieu, P., 32 Covington, M. V., 7 Cowan, N., 18, 28 Cowan, P., 179 Cowen, L. W., 270 Page 380
Cowey, A., 256 Cox, W. M., 100 Coyne, C., 61 Coyne, I. C., 218 Craig, F. I. M., 69 Craig, M. J., 59, 69 Craighead, W. E., 196, 198 Craik, K. J. W., 37, 59 Craske, M. G., 183 Crews, T. M., 54 Cromwell, R. L., 82 Cronholm, B., 55 Crossman, E. R. S. W., 18 Crow, T. J., 77 Cuthbert, B. D., 174 Czerwinski, M., 253 Czyzewska, M., 244 D Dachinger, P., 274 Dagenbach, D., 257, 259 D'Agostino, P. A., 253, 257 Dalgleish, T., 7, 63, 93, 102, 116, 151, 152, 164, 286, 291 Dallas, M., 238, 241
Damle, A., 49 Danaher, B., 179 Danion, J. M., 62 Dannenbaum, S. E., 57 Dark, S., 21, 22 Darke, S., 26 Daves, W. F., 253 Davey, G. L., 300 Davidson, D., 215 Davies, S., 89, 111, 286 Davis, A., 147 Davis, C., 251 Davis, H., 223 Dawes, R. M., 64 Dawkins, K., 87, 95, 107 De Houwer, J., 154, 284 De Jong, R., 19 de Roode, A., 256 de Silva, P., 52, 182 de-Jong, P., 199, 200 Debner, J. A., 258, 259 Deffenbacher, J. L., 182 Defoe, C. G., 20 Delaney, S. M., 242 DeMarco, S., 35 DeMascio, A., 52 Dember, W. N., 50 DeMonbreun, B. G., 196 Dennett, D. C., 232, 261 Dennis, I., 252
Denny, E. B., 148, 149, 150, 286, 295 Dent, J., 197, 205, 225 DePree, J. A., 182 DePrince, A. P., 271 Derry, P. A., 223 DeShepper, B. G., 24 Deutsch, 21 Dewey, G., 246 Dewick, H., 185 Di Caprio, N. S., 177 Dickman, J., 233, 236, 245 Diener, E., 146 Dienes, Z., 244, 248 Dighton, K., 151 Dimberg, U., 271 Dixon, N. F., 81, 252, 253 Dixon, W. A., 53 Doctor, R. M., 68 Dombeck, M. J., 90 Donchin, E., 28 Donnell, C. D., 159, 226, 288 Donnelly, E. F., 52 Doren, B., 143 Dorfman, D. D., 81 Dorfman, J., 242 Dornic, S., 66 Dreyfus, H. F., 171 Dritschel, B. H., 165, 289 Driver, J., 24, 51, 109 Duberstein, P. R., 275
Dulany, D. E., 246 Dulany, E., 81 Dumais, S. T., 41 Dunbar, D. C., 57 Dunbar, K., 107, 114, 115, 120, 122, 129 Duncan, J., 22 Duncker, K., 34 Dunlap, K., 252 Dunn, D., 251 Durso, F. T., 192 Dutta, A., 20 Dutton, F., 285 Dworkin, L., 234 Dwyer, M., 253 Dyck, D. G., 225 Dyckman, J., 179 Dziadosz, T., 228 Page 381
E East, M. P., 182 Ebbinghaus, H., 237 Edmunson, E. D., 70 Eelen, P., 154, 284 Ehlers, A., 89, 111, 286 Elliot, C. L., 62, 149, 285 Ellis, H. C., 20, 26, 45, 46, 55, 58, 60, 63, 64, 65, 66, 68, 70 Ellis, N. C., 166, 167 Elman, J. L., 32 Emerson, E., 77 Emery, G., 10, 18, 106, 186, 218, 268
Emmelkamp, P. M. G., 138 Engle, R. W., 39 English, G. E., 89, 99 Enns, J., 21, 30 Enright, S. J., 24 Entwistle, N., 60 Erdelyi, M., 5, 21, 23 Erdelyi, M. H., 81, 177 Erdley, C. A., 253 Eriksen, C. W., 21, 28, 81, 252 Erlelyi, M. H., 263 Erwin, P, G., 56, 60 Eskridge, T. C., 33 Estes, W. K., 31 Estevaz, A., 229 Esteves, F., 271 Evans, J., 168 Eysenck, M., 89, 96, 102, 110, 284, 285, 288 Eysenck, M. W., 18, 21, 26, 45, 46, 49, 50, 59, 65, 66, 84, 96, 109, 129, 155, 156, 157, 160, 162, 181, 182, 205, 206, 226, 228, 229, 244, 287, 290, 291, 295 F Farah, M. J., 211, 255 Farne, M., 252 Farrel, J., 274 Faust, M. E., 24, 35 Fehrer, E., 253 Fennell, M. J. V., 49 Ferrier, I. N., 164, 167, 286 Feske, U., 89, 99 Fielder, K., 34 Finger, R., 69
Firestone, I., 234 Fischer, S. C., 39 Fischhoff, B., 192 Fischman, M. W., 49, 58 Fisher, C., 253 Fisk, D., 40 Fitzgerald, J. M., 146 Flett, G. L., 53 Foa, E. B., 20, 78, 89, 90, 99, 102, 152, 159, 180, 202, 226, 228, 288 Foa, E. G., 5, 54 Fodor, J. A., 19, 26, 32 Fogarty, S. J., 136, 137, 138, 141, 158, 308 Folkard, S., 37 Forgas, J. P., 203 Forster, K., 253 Forster, P. M., 251 Foulds, G. A., 48, 49 Fowler, C. A., 253 Fox, E., 23, 51, 88, 109 Frambach, M., 247 Franks, B., 33 Franks, J. J., 187, 274 Fransson, A., 60 Fredrickson, B. L., 183 Freeman, J. T., 81 French, C. C., 87 French, L., 153, 155, 207, 284, 285 Frensch, P. A., 39 Freud, S., 270 Friedman, A. B., 20, 52
Friedman, A. S., 52 Frierson, H. T., 68 Frisby, J. P., 34 Fromholt, P., 139, 141, 148, 286 Frost, R. O., 53, 54, 55 Fudin, R., 275 Fuller, R. D., 148, 149, 150, 285, 295 Furnham, A., 87, 95, 107 G Galassi, J. P., 68, 69 Galbraith, M., 153, 154, 286, 287, 289 Gallego, J., 246 Ganzer, V. J., 68 Gardiner, J. M., 84, 96, 287 Gargano, G., 242 Garrod, S. C., 34, 36, 212, 213, 217 Gaskill, S. J., 20 Geary, D. C., 39 Gelder, M., 228 Gellatly, A. R. H., 252 Geller, V., 85 Gemar, M., 93, 101, 224, 225, 284 Gerard, L., 238 Gerdt, C., 60, 62, 63 Gerlsma, C., 138 Page 382
Gernsbacher, M. A., 24, 229 Gerrig, R. J., 284 Ghadiali, E., 165 Ghoneim, M. M., 256
Gilligan, S. G., 11, 135, 137, 138, 142, 222, 223 Glass, J., 20, 181 Glass, R. M., 49, 58 Gleitman, H., 30 Glowalla, U., 144 Glucksberg, S., 252 Gmeindl, L., 20, 181 Godden, D. R., 134 Gogel, W. C., 34 Goldberg, B., 177 Goldiamond, I., 81 Golding, S. L., 275 Goldstein, M. J., 253 Golinkoff, M., 62 Golombok S., 84, 88, 96, 287 Gomez, R. L., 246 Goodge, A., 248 Goodrich, S. J., 271 Goodwin, A. M., 138 Goodwin, F. K., 52 Gopher, D., 19 Gordon, G., 252 Gordon, P. K., 53 Gorham, D. R., 52 Gotlib, I., 218, 284 Gotlib, I. H., 4, 11, 52, 58, 86, 92, 96, 139, 196, 284 Govier, E., 251 Graesser, A. C., 211, 212, 213, 214, 228 Graf, P., 41, 122, 123, 124, 237, 238, 241, 242, 278, 279, 282, 283 Grange, D., 62
Granholm, E., 40 Gray, J. A., 118 Grayboys, T., 193 Green, G., 250 Greenberg, L., 4, 10, 15, 18 Greenberg, M. S., 143, 186, 218, 286, 287 Greenberg, R. L., 106 Greene, R. L., 62, 149, 285 Greenspoon, S. L., 36 Greenwald, A. G., 222, 223, 253, 257, 258, 259 Gregory, M., 23, 81 Gregory, W. L., 201 Griffin, J. A., 50 Gropper, G. L., 81 Grosberg, J. M., 81 Grove, M., 110 Guirguis, M., 93, 101, 224, 225, 284 Gunther, W., 40 Gur, R. C., 253, 267 Gustafson, R., 275 H Haber, R. N., 34, 81, 263 Hackley, D., 28 Hackmann, A., 228 Haden, P. E., 238 Hadley, R. F., 33 Hagan, R., 93, 112, 130, 271 Hamilton, V., 66 Hammen, C., 225, 228 Hammen, C. L., 4, 11
Hansson, R. O., 185 Harber, K. D., 185 Hardaway, R. A., 275 Hardin, T. S, 62, 71 Hardin, T. S., 285 Harper, R. S., 34 Harrell, R. M., 35 Harris, R. S., 274 Hart, J., 173 Hartman, M., 247 Harvey, J. M., 228 Hasher, L., 11, 18, 23, 40, 59, 143, 211, 278 Haspel, K, C., 274 Hastel, F. W., 49, 58 Hayes, N., 249 Hayman, C. A. G., 295 Healy, D., 4, 63 Heatherton, T. F., 200 Heimberg, R. G., 90, 91, 103, 113 Hemsley, D. R., 50, 82, 141 Henderson, L., 28 Hendricks, R., 20 Henley, S. H. A., 252, 253, 267 Henriksson, M., 252 Henry, G, M., 39, 57 Heppner, P. P., 53 Herbert, J., 60 Hermans, D., 154, 284 Hernstadt, R., 248 Herrmann, D. J., 12
Hershenson, M., 34, 263 Hertel, P., 285 Hertel, P. T., 61, 62, 71 Page 383
Heuer, F., 145 Hewlett, J. H. G., 47, 48 Hickey, D. T., 39 Hicks, R., 19 Hicks, W. E., 18, 19 Higgins, E. T., 224 Hill, A. B., 92, 244, 253, 255, 271, 285 Hill, T., 200 Hillyard, S. A., 21 Hines, D., 253 Hinton, G. E., 30 Hirsch, C., 229 Hirst, W., 15, 20, 289, 297, 298, 299, 314 Hiscock, M., 179 Hitch, G., 20 Hochberg, J. E., 22 Hockey, G. R. J., 67 Hoffman, H., 145, 244 Hogeboom, M. M., 26 Holender, D., 251, 254 Hollan, J. D., 164 Holland, B., 252 Holley, P. E., 40 Holmes, D. A., 275 Holmes, D. S., 108 Holyoak, K. J., 24
Homme, L. E., 186 Hood, J. E., 224 Hope, D. A., 90, 91, 103, 113 Hopkins, M., 181, 182 Horder, P., 91 Horn, S., 273 Horowitz, L. M., 93, 101, 224, 225, 284 Horowitz, M. J., 175, 183, 184 Houghton, G., 24, 37, 125, 126 Hounsell, D., 60 Howard, D. V., 248 Howard, J. H., 248 Howell, S. K., 168 Howes, D. H., 81 Howie, D., 81 Hubbard, M., 193 Hudesman, J., 275 Humphrey, S. M. S., 59 Humphreys, G. W., 22 Humphreys, K. J., 256 Hunt, R. R., 148, 149, 150, 286, 295 Huygens, K., 42 Hyman, R., 18 I Idzikowski, C., 59 Ilai, D., 91 Ingram, R. E., 218, 222 Inman, V. W., 57 Inz, J., 181 Isen, A. M., 142, 143, 196, 197
Ivry, R., 107, 114, 115, 120, 122, 129, 249 J Jackel, L., 183 Jacoby, L. L., 40, 122, 124, 204, 238, 239, 240, 241, 242, 243, 257, 258, 259, 278, 282, 283, 291, 292 Jarvik, M. E., 39, 55, 58 Jaster, H. R., 68 Jeavons, A., 228 Jehin-Marchot, D., 164, 286 Jelicic, M., 256 Jenkins, E., 32 Jennings, J. M., 243 Jensen, A. R., 85 Jessup, B. A., 175 John, O. P., 101, 108, 111 Johnson, E. J., 192, 201 Johnson, J., 197 Johnson, M. H., 45 Johnson, M. K., 15, 55, 56, 64, 146, 187, 188, 216, 289, 297, 298, 299, 305, 314 Johnson, M. M. S., 142, 143, 196, 197 Johnson, W., 87 Johnson-Laird, P., 6, 12, 26, 36, 112, 264, 269, 314 Johnson-Laird, P. N., 169, 171, 188 Johnston, J. C., 19, 22, 53 Johnston, W. A., 21 Jolles, J., 58 Jones, E., 21 Jones, G. V., 90 Jones, H. G., 180 Jones, J. L., 51, 53, 109
Jones, L. N., 77 Jones, S., 49 Jonides, J., 30 Jordan, C. S., 175, 178 Just, M. A., 39 K Kaflowitz, N. G., 68 Kahneman, D., 18, 22, 65, 167, 191, 251, 312 Kallmen, H., 275 Kalmar, D., 20 Kane, M. J., 23 Kang, H., 24 Page 384
Kantowitz, D. H., 18, 19 Kaplan, S., 2 Kaspi, S. P., 89, 98, 99, 108 Kassin, S. M., 244 Katkin, E. S., 275 Katz, A. N., 284 Katz, L., 253 Kaufman, L., 184 Kavanagh, D. J., 202 Kaye, N., 274 Kayton, L., 60 Keele, S. W., 19, 107, 114, 115, 120, 122, 129, 249 Kelley, C. M., 204, 239, 240, 242 Kelly, L. P., 34 Kemp, S., 146 Kemp-Wheeler, S. M., 253, 255, 271 Kentish, J., 88, 89, 96, 102, 105, 107, 108, 113, 120, 129, 285, 288 Kepecs, J. G., 176
Kerr, S. A., 295 Ketzsger, R. L., 181 Keuther, M., 68 Kieras, D., 170, 173 Kieras, D. E., 20, 181 Kihlstrom, J. F., 123, 222, 232 Kim, E., 90, 99 Kim, J. K., 146 Kinchla, R. A., 21 Kinsbourne, M., 19 Kirker, W. S., 221, 223 Klapp, S. T., 252 Klein, F. B., 123, 222 Klein, G. S., 85, 252 Klein, S. B., 223 Kleinsmith, L. J., 2 Klieger, D. M., 92 Klinger, E., 77, 100, 128n, 131, 132, 175, 179 Klinger, M. R., 253, 257, 259 Klug, F., 88, 99, 101, 107, 224, 225, 286, 287, 291 Klumpp, G., 203 Knight, J. L., 18, 19 Knopman, D. S., 247 Knowles, T. H., 92 Knowlton, B. J., 247 Koehler, C. K., 228 Koh, S. D., 60 Kolers, P. A., 34 Kolers, T. A., 253 Kolodner, J., 164
Kopelman, M., 58 Kopka, M., 255 Kosslyn, S. M., 170, 171 Kothera, L., 275 Kovacs, M., 219 Kowalski, D. J., 214 Kozak, M. J., 5, 78, 89, 99, 102, 173 Kramer, A. F., 40 Kramer, J. J. A. M., 138 Krames, L., 65 Krantz, S., 228 Kravetz, S., 35 Kroecker, L., 81 Kuhn, K., 17 Kuiper, N. A., 221, 223, 224 Kujala, T., 21 Kukla, A., 67 Kunst-Wilson, W. R., 238, 252 Kushner, M., 54 Kutas, M., 21 Kuyken, W., 164, 166, 286 Kwan, L., 248 Kwawer, J. S., 274 L LaBerge, D., 79 Lachman, T., 19 Lamain, W., 26 Lamber, E., 20 Landau, R. J., 180 Lane, J. W., 63, 64
Lang, P. J., 5, 173, 174 Lange, J. J., 27 Laraia, M., 57 Larsen, P., 139, 141, 148, 286 Larsen, S., 139, 141, 148, 286 Larwill, L. K., 238 Lasko, N. B., 164, 166, 289 Lauber, E., 181 Lavy, E. H., 90, 92, 99, 101, 103, 111 Law, D. J., 39 Law, J., 295 Lawler, P., 77 Lawrence, R. H., 146 Lay, R., 63 Layman, B., 192 Lazarus, R. L., 2, 3 Lee, I., 273 Lee, P., 252 Leeper, R. W., 34 Leigh, J. H., 20 Page 385
Leight, K. A., 58, 64 Lemaire, P., 39 Lepper, M. R., 193 Leslie, A. D., 262 Levene, M., 52 Levenkron, J. C., 82 Leventhal, H., 3, 14, 278 Levey, A., 60 Levin, D. N., 173
Levine, D. S., 32 Levy, L. H., 87 Levy, P. E., 37 Lewicki, P., 200, 244 Lewis, J. L., 79, 250 Lewis, P., 57 Lewis, S., 244 Liberman, K., 20 Lichtenstein, S., 192 Liebowitz, M. R., 285, 288, 295 Lindsay, D. S., 293 Linville, P. L., 221 Lipinski, J. F., 270 Lipke, H. J., 89, 99 Lishman, W. A., 57, 136, 138, 158, 308 Lisman, S. A., 135 Litz, B. T., 89, 100, 164, 166, 289 Liu, T. J, 253 Lloyd, G. G., 136, 138, 158, 308 Lockhart, R. S., 59 Loftus, E. F., 145, 184, 223, 271 Loftus, G. R., 145 Logan, G. D., 29 Lombardi, W. J., 253 Lonski, M., 274 Lord, C. G., 222 Lord, R. G., 37 Lorge, I., 81 Louro, C. E., 90, 99, 100, 113, 184 Lovibond, P. F., 199
Lukach, B. M., 90, 99, 100, 113, 184 Luks, P., 48 Lundh, L. G., 22 Lustbader, L., 274 Lyon, D. R., 29 Lyubomirsky, S., 183 M McAllister, T. W., 55 McAndrews, M. P., 241 McCann, C. D., 86, 92 McCann, R. S., 19 McCarthy, P. R., 89, 90, 99 McCauley, C., 33 McClelland, C. L., 28, 30, 32, 34 McClelland, J. L., 107, 114, 115, 120, 122, 129 McDermott, K. B., 237, 292, 294, 295, 296 MacDonald, C., 31, 295 MacDonald, M., 224 McDonald, M. R., 65 McDowell, J., 135 McFarland, A. D., 63, 64 McGlynn, S. M., 265 MacGregor-Morris, R., 84, 87, 96, 97, 103, 105, 129 Mackay, D. G., 251, 252 McKee, R., 241 McKenna, F. P., 86, 102, 177 Macklin, M. L., 164, 166, 289 McKoon, G., 36, 230, 242 McLachlan, A. L,, 284 MacLean, A., 173
McLemore, C. W., 179 Macleod, A. K., 166, 167, 229 McLeod, B., 21 MacLeod, C., 23, 27, 33, 42, 68, 71, 79, 82, 83, 85, 86, 87, 88, 93, 94, 97, 105, 111, 112, 113, 114, 120, 121, 124, 129, 130, 137, 148, 150, 152, 153, 154, 155, 157, 163, 168, 197, 203, 218, 271, 280, 281, 284, 285, 287, 292, 310, 313, 206160 McLeod, P., 19, 21, 28, 57 McNally, R. J., 53, 78, 79, 89, 90, 93, 98, 99, 100, 101, 102, 107, 108, 113, 152, 159, 164, 166, 174, 184, 200, 202, 226, 228, 285, 287, 288, 289, 290, 293 McNeil, D. W., 174 Magaro, P. A., 45, 55, 56, 64 Magill, R. A., 211 Mahoney, M. J., 4 Makhlouf-Norris, F., 180 Malhotra, A., 34 Mandler, G., 3, 122, 123, 124, 212, 238, 261, 262, 263, 278, 279, 282 Mandler, J., 212, 233 Mangan, C. E., 23 Mann, R., 53, 54 Marcel, A. J., 22, 252, 253, 258, 259, 261, 266, 267, 269 Marcus, H., 224 Marden, B., 87, 96, 98, 102 Page 386
Marder, S. R., 40 Margalit, B., 82 Margraf, J., 89, 111, 159, 286, 288 Marker, H., 87, 88, 89, 94, 95, 98, 105, 119, 130, 134 Marks, D., 178 Markus, H., 221, 227 Marr, D., 34 Marshall, C. R., 34, 39 Martello, J., 60, 62, 63
Martin, A. D., 274 Martin, D. J., 198 Martin, M., 87, 88, 89, 91, 94, 95, 98, 105, 119, 130, 141, 153, 154, 284 Marton, F., 60 Marx, E. M., 53 Marzillier, J. S., 179 Masling, J. M., 275 Massaro, D. W., 18, 28 Masson, M. E. J., 41 Masure, C., 48 Mathews, A., 23, 27, 33, 42, 51, 79, 82, 83, 84, 86, 87, 88, 94, 95, 96, 97, 99, 101, 103, 106, 107, 109, 110, 111, 112, 113, 114, 119, 120, 121, 124, 129, 130, 137, 148, 149, 150, 152, 154, 155, 156, 157, 158, 163, 168, 181, 182, 202, 204, 205, 206, 226, 228, 229, 280, 281, 284, 285, 288, 290, 295, 310 Mathews, A. M., 87, 88, 89, 90, 96, 97, 102, 103, 105, 113, 155, 156, 157, 160, 179, 218, 224, 225, 226, 244, 284, 285, 286, 287, 288, 291, 307 Matt, G. E., 141 Matthews, A., 284 Matthews, G., 110, 111, 154 Mattia, J. I., 90, 91, 103, 113 Matussek, P., 48 Matuzas, W., 49, 58 Maxeiner, M. E., 77, 128, 131, 132 Maxwell, A. E., 63 May, C. P., 23 May, J., 51, 109, 110, 155, 156, 157, 160, 228, 229, 244, 284, 287 Mayer, J. D., 135 Mayo, P. R., 158, 288 Meichenbaum, D., 178, 186 Melamed, B. G., 173, 174 Mendelsohn, E., 274 Merckelbach, H., 199, 200
Merckelbach, N., 42 Merikle, P. A., 251, 254, 255, 256 Merikle, P. M., 238, 253 Merrill, E. C., 24 Merskey, H., 270 Merts, E., 142, 143, 196, 197 Meyer,, 181 Meyer, D. E., 20, 23, 28 Meyer, R. R., 253 Middleton, H., 228 Miklowitz, D. J., 225 Millar, D. G., 180 Millar, K., 256 Miller, E., 57 Miller, F. D., 235, 236 Miller, G. A., 173, 233 Miller, J., 28, 274 Miller, J. O., 19, 25, 28, 39, 45, 47, 48 Miller, K., 256 Miller, S. M., 23, 82 Millwood, B., 87, 95, 96, 111, 286 Milner, D. A., 256 Milroy, R., 112, 120, 307 Minard, J. G., 81 Mineka, S., 153, 154, 160, 161, 199, 200, 285, 287, 288, 293, 294 Minsky, M., 36, 216, 266 Miranda, J., 205 Mischel, W., 196 Mitchell, K. J., 271 Miyake, A., 39
Mogg, K., 51, 84, 87, 88, 89, 95, 96, 97, 98, 99, 102, 103, 105, 106, 107, 108, 109, 110, 111, 113, 120, 129, 150, 153, 154, 155, 157, 160, 226, 228, 229, 244, 284, 285, 286, 287, 288, 289, 291 Monteiro, K. P., 135, 142 Moore, A. L., 24 Moore, G. F., 60 Moore, R. G., 164 Morley, S., 146 Morris, W. N., 7, 57, 68, 71 Morrow, D. G., 36, 217, 227 Morrow, J., 183 Morton, J., 23 Moscovitch, M., 241 Mueller, J. H., 59, 60, 226 Mulder, G., 26, 27 Mulder, L. J. M., 26, 27 Muller, H. J., 21 Page 387
Multhaup, K. S., 146, 297, 305 Munro, P. W., 125 Murdock, I. B., 152 Murdock, T. B., 89, 90, 99 Murphy, D. L., 39, 52, 57, 58, 60, 62, 63 Musen, G., 242 Myer, R. R., 252 Mykel, N., 253 N Naatanen, R., 21 Nakamura, Y., 211, 212, 213, 214, 238 Naparstek, J., 87 Nasby, W., 143, 286
Natsoulas, T., 81, 261 Navon, D., 19, 66, 82 Neisser, U., 22, 34, 187, 233, 253, 280 Nelson, D. L., 40, 70 Nelson, J. G., 48 Nelson, R. F., 196, 198 Neubauer, R. M., 200 Neumann, E., 24 Newell, A., 31 Newland, J. R., 179 Newman, B., 182 Newstead, S. E., 252 Newton, R. E., 34 Nezu, A. M., 53 Nezu, C. M., 53 Nicastro, R., 275 Nicholson, W. M., 66 Niedenthal, P. M., 221, 226 Nigal, D., 62 Nijman, J., 199 Nisbett, R. E., 233, 234, 267 Nissen, M. J., 247, 249 Nolen-Hoeksema, S., 183 Norman, D. A., 28, 164, 212, 264 Norris, H., 180 Nugent, K., 153, 154, 160, 161, 285, 287, 288, 293, 294 Nulty, D. D., 86, 92 Nunn, J. D., 150, 277 Nurius, P., 221, 227 O
Oaksford, M., 116, 125 Oatley, K., 6, 112, 262, 314 Oatman, L. C., 21 O'Brien, E. J., 218 O'Connor, M., 271 Ohman, A., 271 Oliver, J. M., 274 O'Loughlin, S., 168 Omelich, C. L., 7 Osman, A., 28 Otto, M. W., 159, 287, 288, 290 Otto, R., 54, 55 Ottosson, J. -O., 55 Ouardirhi, S. E., 40 P Paans, A. M. S., 27 Pachella, R. G., 25 Packer, I. K., 253, 267 Pacteau, C., 246 Page, W., 275 Paivio, A., 170 Paller, K. A., 242 Palumbo, P., 274 Parker, G., 138 Parker, K., 275 Parkes, C. M., 180 Parkin, A. J., 242 Parkinson, L., 76, 80, 182 Parkinson, S. R., 57 Parrot, G. W., 140
Parrott, G. W., 306 Partridge, S., 152, 218, 222 Pashler, H., 19 Patterson, K. E., 252 Payne, D. G., 20, 47, 48 Pellegrino, J. W., 39 Pennebaker, J. W., 185 Perrett, A., 153, 154, 286, 287, 289 Perri, M. G., 53 Perruchet, P., 239, 246 Persons, J. B., 54, 180, 205 Peters, L. J., 20, 62, 63 Pettigrew, T. F., 214 Petzel, T., 197 Phaf, R. H., 256 Philpott, A., 251 Pichert, J. W., 151, 216 Pickar, D., 58 Pickles, A. J., 150, 277 Pierce, G. R., 21 Pietromonaco, P., 3, 79 Pitman, R. K., 164, 166, 289 Pittman, T. S., 252 Pitts, M., 251 Plude, D. J., 21 Plunkett, K., 32 Pollack, M. H., 159, 287, 288, 290 Polson, M. C., 20 Page 388
Pope, H. G., 270 Porterfield, A. L., 275
Posner, M. I., 28, 79 Post, R. M., 52, 57, 61 Postman, L., 23, 34, 81 Potts, R., 61 Poulton, E. D., 37 Powell, G. E., 187 Powell, M., 82 Power, M. J., 11, 36, 169 Poynton, R. F., 275 Prassas, A., 89, 100, 164, 166, 289 Pratto, F., 101, 108, 111 Pressley, M., 39 Proctor, R. W., 20 Pruzinsky, T., 181, 182 Puffet, A., 164, 286 Pyle, W. H., 252 Pylyshyn, Z., 170, 171 Pylyshyn, Z. W., 32 Q Quirk, S. A., 271 R Rabbitt, P. A., 165 Rabbitt, P. M., 21 Rachman, S., 4, 76, 80, 182, 184 Radcliffe, W. N., 77 Raiser, B. J., 170 Ramus, S. J., 247 Ranney, M., 34, 37 Rapee, R. M., 90, 183 Raskin, A., 52
Ratcliffe, R., 28, 36, 230, 242 Rauner, M., 203 Rautianen, J., 275 Ray, C., 86, 93 Reading, C., 173 Reber, A. S., 41, 243, 244, 247, 248 Reed, C. L., 211 Reed, G. F., 45, 54 Reed, S., 275 Rehm, L., 179 Reid, T. K., 242 Reimann, B. C., 89, 90, 93, 98, 99, 100, 101, 107, 108, 113 Reingold, E. M., 238, 240, 253, 256 Reisberg, D., 145, 171 Reiser, B. J., 164 Resnick, R. A., 30 Revelle, W., 59 Rey, A. C., 52 Reyher, J., 175 Richards, A., 87, 95, 96, 111, 155, 158, 163, 205, 206, 207, 228, 229, 285, 286, 287, 288, 290 Richards, J. C., 228 Ridderinkof, K. R., 26 Riegler, G. L., 242 Riemann, B. C., 159, 184, 293 Rimm, D., 179 Rinck M., 144, 159, 288 Risse, G., 146 Rittenauer-Schatka, H., 203 Roberts, S., 25 Robertson, M. M., 63
Robertson, R. R., 229 Robertson, S. A., 77 Robins, C. R., 80, 284 Robinson, A., 173 Robinson, E., 182 Robinson, G. S., 142, 143, 196, 197 Rocklin, T., 59 Rodrigez, A., 60, 64, 70 Rodrigues, J., 34 Roediger, H. L., 18, 161, 162, 237, 242, 292, 294, 295, 296 Rogers, T. B., 221, 223 Rohwer, W. D., 85 Rojahn, K., 214 Roland, C. B., 270 Rollins, H. A., 20 Romig, C. A., 270 Roorda-Hrdlickova, V., 256 Rose, G., 166, 167 Rose, G. S., 197, 229 Rose, K. C., 143 Rosebeaum, J. F., 287, 288, 290 Rosenbaum, J. F., 159 Ross, L., 193, 208 Ross, M. J., 225 Roter, A., 248 Roth, W. T., 89, 111, 286 Rothbart, M., 214 Rothenfluh, T., 33 Roy-Byrne, P. J., 61 Rozeboom, W. M., 177
Ruddy, M. G., 23 Rude, S. S., 61, 71 Rueckl, J. G., 242 Ruiter, C., 108 Rumelhart, D. E., 30, 32, 34, 36, 266 Rush, A. J., 10, 186, 218, 268 Page 389
Russell, J., 176 Russe, M. L., 142, 283 Russell, P. W., 60, 62 Russo, R., 242 Rusted, J. L., 151 Rutherford, 42 Rutherford (92), 271 Ruthruff, E., 19 Ryle, A., 180 S Sabini, J., 140, 306 Saccuzzo, D. P., 50, 52 Sackeim, H. A., 253, 267, 275 Safer, M. A., 144, 146, 147 Safran, J. D., 4, 5, 15 St John, M. F., 245 Salavoy, P., 137, 138, 139, 208 Sales, B. D., 81, 271 Salkovskis, P. M., 228 Samide, J. L., 275 Sanders,, 19, 25, 28 Sanders, A., 19, 28 Sanders, A. F., 25
Sanford, A. J., 33, 34, 36, 212, 213, 217 Sanft, H., 143 Sarason, B. R., 21 Sarason, I. G., 21, 65 Sato, S., 29 Savard, R. J., 52 Savy, I., 246 Sawyers, P. K., 253 Scarborough, D. L., 238 Schacter, D. L., 41, 146, 160, 237, 239, 241, 242, 251, 264, 265, 270, 296 Schank, R., 36, 216 Schapiro, D., 37 Schare, M. L., 135 Scheibe, K. E., 85 Scherer, K., 278 Scherer, K. R., 2, 3, 5 Schneider, K., 144 Schneider, L., 295 Schneider, W., 40, 41, 277 Schoenberger, A., 193 Scholing, A., 138 Schotte, D. E., 53, 78 Schreiber, T. A., 40 Schuh, E. S., 258 Schultz, D. W., 28 Schulze, C. C., 53 Schumacher, E. H., 20, 181 Schvaneveldt, R. W., 23, 246, 253 Schwartz, S., 59, 60 Schwarz, N., 197, 202, 203
Schwarzer, R., 7 Schweickert, R., 20 Sclienger, J. L., 62 Scott, J., 164, 167, 286 Scott, J. D., 70 Scott, P., 275 Scott, W., 152, 218, 222 Searle, J., 171, 261 Sebastian, S., 90, 103, 111, 119, 120, 121, 130, 286 Segal, Z., 33, 68 Segal, Z. V., 92, 93, 101, 224, 225, 284 Seger, C. A., 41, 243 Seibert, P. S., 68 Seifert, C., 36 Selfridge, W., 28 Semin, G. R., 34 Sergeant, H., 177 Servan-Schreiber, D., 118 Setterlund, M. B., 221, 226 Shadbolt, N., 34 Shadick, R., 181, 182 Shallice, T., 26, 37, 68, 262, 263, 264 Shanks, D. R., 245 Sharrock, R., 57, 68, 69, 86, 91, 99, 102, 111, 177, 178, 179, 180, 313 Sharver, R., 68 Shaver, P. R., 85 Shaw, A. M., 252 Shaw, B. F., 10, 186, 218, 224, 268 Shear, M. K., 159, 285, 288 Sheehan, M. J., 180
Sheikh, A. A., 175, 178 Shepard, R. N., 170 Sher, K. J., 53, 54, 55 Shevrin, H., 233, 236, 245, 255, 274 Shiffrin, R. M., 18, 22, 40, 41, 277 Shinamura, A. P., 241, 242, 294 Shoyer, B., 89, 90 Shuman, D. W., 271 Sidis, B., 252 Siegel, D. R., 270 Siegle, R. S., 39 Silberman, E., 39, 52, 57 Silver, E. A., 34 Silver, R. C., 185 Silverman, D. K., 82 Silverman, L. H., 82, 274 Page 390
Simon, H. A., 31 Simons, A., 203 Simpson, G. B., 24 Singer, J. A., 137, 138 Singer, J. L., 82, 186 Singer, L., 62 Singer, M., 228 Sinha, C., 32 Slade, R., 253 Slade, S., 146 Slaymaker, F., 197 Slovic, P., 191 Small, S. A., 80, 284
Smallberg, S. A., 58 Smeltzer, W., 175 Smid, H. G. O., 26 Smith, D. A., 214 Smith, E., 253 Smith, E. R., 235, 236 Smith, G. J. W., 252 Smith, M., 253 Smith, M. J., 40 Smith, S. W., 35 Smolensky, P., 30, 32, 33 Snodgrass, M., 255 Snyder, M., 21, 194 Soares, J. J. F., 271 Solomon, R. L., 81, 126 Somekh, D. E., 252 Southall, A., 110, 154, 284 Spear, N. E., 135 Spellman, B. A., 24 Spence, D. P., 252 Spencer, P., 197 Spielman, S. B., 20 Spinakis, A., 40 Spiro, R. J., 274 Sprock, J., 50, 52 Squire, L. R., 241, 242, 247 Squire, R., 250 Srull, T. K., 36, 214 Staedler, M. S., 242 Stark, H. A., 239
Starron, A., 84, 88, 96, 287 Ste-Marie, D., 240 Stefanovic, S., 147 Stein, N. L., 6 Steinman, S. B., 29 Stenman, U., 251 Stent, G. S., 34 Stephens, M. W., 81 Stern, R., 182 Sternberg, S., 25, 27, 29, 39, 55, 58 Stevenson, R., 150, 277 Stillings, N. A., 256 Stoerig, P., 256 Storms, M. D., 234 Strack, F., 203 Strauman, T. J., 221, 226 Strayer, D. L., 40 Stroebe, M. J., 185 Stroebe, W., 185 Stroh, N., 252 Stromgren, 58 Strongman, M. S., 146 Stroop, J. R., 85 Strosahl, K. D., 179, 187 Suarwez, R., 47 Suinn, R. M., 178 Sullivan, M. P., 24 Sum Ping, S. T., 256 Sun, R., 32 Sutherland, G., 182
Sutton, S. K., 199 Swann, W. B. Jr, 194 Sweeney, J. A., 62 Swindell, A., 228 T Tabachnick, N., 194, 195 Tait, R., 185 Talbot, N. L., 275 Tallal, P., 34 Tallis,, 181, 182 Tallis, F., 55 Tarka, M., 177 Tassinary, L., 253 Tata, P., 82, 84, 121, 154, 280, 281, 284, 310 Taylor, R., 137, 138 Teasdale, J. D., 7, 12, 15, 18, 49, 67, 136, 137, 138, 139, 141, 142, 143, 144, 145, 148, 153, 158, 166, 177, 197, 205, 219, 225, 226, 283, 289, 302, 308 Teder, W., 21 Tenney, N., 42 Theewues, J., 21 Theios, J., 25 Thomas, D. L., 146 Thomas, J. C., 34, 60, 64, 70 Thomas, R. L., 63, 64 Thompson, K., 61 Thomson, D. N., 12 Page 391
Thoresen, C., 179 Thorndike, E. L., 81 Thorpe, S., 146, 310 Timsit, M., 164, 286
Timsit-Berthier, M., 164, 286 Tipper, S. P., 24, 37, 51, 109 Titchener, A. B., 252 Tobias, B. A., 146 Tomarken, A. J., 199, 200 Tota, M. E., 224, 253 Toth, J. P., 240, 293 Totterdell, P., 37 Townsend, J. T., 30 Trabasso, T., 228 Trafimow, D., 211 Trandel, D. V., 79 Treisman, A., 18, 21, 22, 29, 46, 250, 251 Treisman, A. M., 114 Trezise, L., 86, 91, 102, 111, 177, 178, 179, 180, 313 Trimble, R., 252 Truchon, C., 93, 101, 224, 225, 284 Tulving, E., 12, 177, 239, 294, 295 Tupling, H., 138 Turk, D. C., 208 Turner, M. L., 53, 78 Turner, T. J., 211, 214, 215 Turvey, M., 28, 253 Tversky, A., 167, 191, 192, 201, 312 Tyers, C., 166, 167 Tyrer, P., 273 U Uhlenhuth, E. H., 49, 58 Umilta, C., 261 Underwood, G., 251
Underwood, G. L., 79 Upper, D., 186 Uranowitz, S. W., 21 V Vacquez, C., 141 van Berkum, J., 181 van den Broeck, M. D., 150, 277 van den Elzen, J., 39 van den Hout, M., 42, 90, 91, 92, 99, 101, 103, 111, 203 van der Ham, F., 28 van der Molen, M. W., 26 van Oppen, P., 90, 101 van Selst, M., 255 Van Voorhis, B. A., 21 van Zandt, B. J. S., 238 Varner, C. J., 64 Vasey, M., 182 Vella, D. D., 92, 101, 224 Velten, E., 135, 136, 142 Vestre, N. D., 196 Vochatzer, K. G., 21 Vokey, J. R., 244 Volans, P. J., 54 von Benedek, L., 147 von Wright, J. N., 251 W Waldman, I. N., 52 Walkenfeld, F. F., 248 Walker, P., 252 Walley, R. E., 21
Ware, R. C., 271 Warm, J. S., 50 Warren, C., 23 Warren, R. E., 85 Warrington, E. K., 123 Washbourne, M. F., 252 Wason, P. C., 37, 193 Watkins, P. C., 148, 149, 150, 285, 295 Watson, D., 81 Watson, J. B., 232 Watts, F. N., 3, 4, 6, 7, 45, 46, 53, 55, 56, 57, 60, 63, 64, 68, 69, 71, 83, 86, 90, 91, 99, 102, 111, 113, 124, 137, 148, 150, 151, 152, 157, 163, 164, 168, 173, 174, 177, 178, 179, 180, 182, 187, 291, 313 Wegman, C., 5 Weiden, T. D., 21 Weingartner, H., 39, 52, 57, 58, 60, 61, 62, 63, 70 Weinman, J., 88, 95, 99, 106, 110, 111, 150, 155, 226, 286 Weisenberg, M., 234 Weiskrantz, L., 123, 256 Welb, A., 185 Weldon, M. S., 242 Wenger, M. J., 20 Whalan, G., 150, 277 Wherry, M. B., 221, 226 White, P., 235 Whitefield, A., 251 Whitehead, A., 57 Whitehouse, K., 257, 259 Whitfield, C., 271 Whittaker, T., 87, 158, 163, 288, 290 Whyte, L. L., 231
Page 392
Wickelgren, W., 28 Wickens, C. D., 19, 20 Wieglas, M. S., 57 Wijers, A. A., 27 Wilding, J., 251, 252 Wiley, J. G., 39 Wilhelmsen, A. L., 257, 259 Willard-Schroeder, D., 62 Willemsen, A. T. M., 27 Williams, A., 252 Williams, D. A., 148, 149, 150, 285, 295 Williams, J., 87 Williams, J. M. G., 4, 11, 33, 67, 83, 86, 88, 89, 92, 94, 95, 98, 99, 105, 113, 114, 116, 119, 120, 124, 125, 130, 134, 137, 138, 140, 148, 150, 152, 153, 154, 157, 163, 164, 165, 166, 167, 168, 197, 203, 229, 281, 284, 285, 286, 287, 288, 289, 302, 309, 310, 313 Williams, M. D., 164 Williams, R., 88, 113 Williams, R. M., 87, 130 Williams, S. K. W., 192 Williams, W. H., 165 Willingham, D., 247 Wilson, B,, 165 Wilson, T. D., 233, 234, 267 Wine, J., 46, 65, 68 Winfield, J., 251 Winter, D. A., 180, 181 Winthorpe, C., 165 Witherspoon, D., 122, 124, 241, 278, 282 Wolfe, J. M., 22, 52 Wolfe, L., 45
Wolford, G., 253 Wolitzky, C., 274 Woll, S. B., 214 Woloshyn, V., 240, 242 Wolpe, J., 186 Wolters, G., 256 Wood, B., 251 Woods, D. L., 21 Worthington, A. G., 252 Wright, J., 196 Wyatt, R, J., 52 Wyer, R. S., 36, 211 Y Yantis, S., 21 Yapko, M. D., 271 Yerevanian, B. I., 52 Yonelinas, A., 295 Yonelinas, A. P., 40 York,, 182 Z Zacks, R. T., 18, 40, 59, 143, 278 Zajonc, R. B., 3, 42, 238, 252 Zander, J. R., 174 Zaragoza, M. S., 271 Zarantonello, M., 197 Zawada, S. L., 50 Zeitlin, S. B., 89, 98, 99, 108, 159, 285, 288 Zimbardo, P. G., 234 Zimmerman, M., 146 Zimmermann, M. A., 62
Zola, M., 45, 52 Zurbriggen, E., 20, 181 Zuroff, D. C., 57 Page 393
Subject Index A abstracting ability, 51-2 accessibility hypotheses, memory structure, 224-5 activation see also network theories anxiety, 219 consciousness, 262-3 danger schemata, 86 emotional Stroop task, 116-18 input differences, 130 memory, 11 schemata, 219 selective attention, 23-4 ADM see affective decision mechanism affect see emotion affective decision mechanism (ADM), 284 agoraphobia ambiguity, 228 concern-related bias, 77-8, 83, 98 emotional imagery, 174 memory bias, 277 mood-congruent recall, 150 nonconscious processing, 273 subjective risk, 202
Alzheimer's disease, nonconscious learning, 247 ambiguity, schemata, 227-30 amnesia biased retrieval, 282 conceptual priming, 294 nonconscious learning, 247 nonconscious memory, 241 anaesthesia, nonconscious perception, 256 anorexia, personal constructs, 180 anxiety, 6-7, 315-16 ambiguity, 207, 228-30 attention, automatic versus strategic, 42 attentional bias, 50-1 Stroop task, 86-94 autobiographical memory bias, 158-9 Beck's treatment, 4 behavioural treatment, 5 cognition, 6-7, 45 cognitive, impairments, 44-5 cognitive effort, 60-2 cognitive impairments clustering, 59 inhibitory processing, 24 processing resources, 65-6 remedial strategies, 70-1 concern-related bias, 77-8, 83, 98 digit span, 59 encoding, 59-60 focus, 109 imagery
emotional, 174 phobic, 179 integrative model, 291-4 learning rates, 125 memory, 55, 152-5, 158-9 bias, 111, 277, 285, 288 network theory, 11 mood, activation, 219 multiple-entry memory, 299 nonconscious perception, 271 norepinephrine, 118 normal subjects, 152-5 parallel distributed processing, 117-18 phenomenology, 310-11 preoccupation, 7, 46 processing priority, 307-10 resource allocation, 46 integrative model, 280-1 schemata, 218 self-schema theory, 226 self-talk 186 Page 394
anxiety (cont.) Spielberger Trait Anxiety Inventory, 82 state versus trait, 106-7 Stroop task, 94-8, 105 subjective risk, 202, 203-4 subliminal Stroop task, 113 threat stimuli, 83-5 worry, 178 arousal
emotional, cognitive impairment, 2, 46-7 emotional disorders, 5, 7 associative networks see also network theories emotional disorders, 11 attention, 315-16 see also awareness; preoccupation anxiety, 6-7 automatic versus strategic processes, 42 cognitive impairments, 49-51 maintenance, 109-10 modulator, 114 negative priming effect, 24 nonconscious perception, 250-60 obsessive-compulsive disorder, 53-5 selection patterns, 22-3 selective see selective processing spider-avoidance, 146 worry, 299-300 attention elaboration hypothesis, schemata, 214 attentional bias, 73-6 concerns, 83, 98, 129 depression, 8-9, 49-50, 284, 308 dissociation, 122 emotion, parallel distributed processing, 129-30 emotional Stroop interference, 94-8 experimental paradigms, 8-9 facilitated performance, 76-82 integrative model, 283-5 later recollection, 125-8
limits, 110-14 mechanisms, 106-32 parallel distributed processing, 114-22 psychopathology, 128-30 Stroop task performance, 85-94 subliminal Stroop task, emotional disturbance, 130 task-irrelevant processing, 107-8 attentional draining, secondary probe stimulus, 79 attribution theory, 234 Autobiographical Recollections Method, 137 avoidance cognitive effort, 108 phobic anxiety, 5 awareness, 3 see also attention attentional bias, 74 memory, 122-3 phenomenal, 74 B backward masking, 50 nonconscious perception, 253-5, 257, 267 BDI see Beck Depression Inventory Beck Depression Inventory (BDI), 61 behaviourism, 18 bereavement, working through, 183 BESA see brain electrical source analysis bias see also attentional bias; memory clinical judgements, 208-9 cognitive, 299-301
concern-related, 76-8, 83, 98, 129 emotional events, 199-200 integrative model, 277-83 judgements, 193-5 mood-congruent stimuli, 277 response, 75, 80 retrieval, 282-3 trait versus state feature, 128 blanking, 69 blindsight, 256 body positions, schemata, 211 brain electrical source analysis (BESA), 27 brain imaging technology, 27 C capacity limitations, 18-21, 46, 107 multiple resource theory, 19-20 CAS see Conscious Awareness System cascade models, information processing, 28 CBF see cerebral blood flow cerebral blood flow (CBF), 27 circadian rhythms, depression, 4 Page 395
Claparede phenomenon, 122-3, 146 cognition, 7-10 see also information-processing paradigm bias, memorial versus non-memorial elaboration, 299-301 capacity nonconscious learning, 249 nonconscious memory, 242-3 nonconscious perception, 259
capacity limitations, 18-21 connectionist versus symbolic architectures, 31-3 differential deficit hypotheses, 45 dissociation, 13-14 emotion, 2-5 impairments, 2, 44-71 abstracting ability, 51-2 attention, 49-51 control hierarchies, 38 depression, circadian rhythms, 4 experiential correlates, 67-9 memory, 55-64 performance speed, 47-8 problem solving, 52-3 processing resources, 46, 65-7 remedial strategies, 69-71 multiples levels, 13-15 parallel-distributed processing models, 31-3 processing automatic versus strategic, 39-42 bottom-up versus top-down, 33-7 discrete versus continuous, 27-9 serial versus continuous, 29-30 stages, 24-7 verbal reports, 223-7 resources, differentiation, 20 social context, 4 cognitive avoidance, phobic anxiety, 5 cognitive effort, 60-2, 119
see also repression processing resources, 65 repression/avoidance, 108 cognitive load manipulations, information processing, 40 cognitive loop, mood, 196 comprehension bottom-up versus top-down processing, 34-5 schemata, 36 concept-driven processing, 33-7 Concern Dimensional Questionnaire, 131-2 concerns, attentional bias, 76-8, 83, 98, 129 confirmation, 193-4 connectionism, cognitive architecture, 31-3 Conscious Awareness System (CAS), 264-5 consciousness, 261-8 nonconscious distinction, 231-3 verbal reports, 223-7 construct accessibility, Stroop task, 86 continuous flow, cognitive processing, 28 control hierarchies, 37-9 covariation emotional events, 199-200 judgements, 194-5 coverant control method, 186 cross-talk, 20 D dart-throwing performance, subliminal messages, 274 data-driven processing, 33-7 depression, 6-7, 316 abstracting ability, 52
ambiguity, 228-30 attentional bias, 8-9, 284, 308 Stroop task, 86, 92-4 attentional dysfunction, 49-50 autobiographical memory, 136-41 circadian rhythms, 4 cognitive impairments, 45 effort, 60-2, 67 materials structuring, 62-4 performance speed, 47-8 problem solving, 52-3 processing resources, 65-6 remedial strategies, 70-1 concern-related bias, 101 distraction, 48-9, 185 encoding, 60 evaluative judgement, 196-7 hierarchical model, 15 interactive cognitive subsystems, 302-3 judgements of contingency, 198-9 memory control hierarchies, 39 Page 396
direct versus indirect tests, 148-50 function of, 55-6, 62, 64 network theory, 11 nonconscious, 270 omission errors, 57 over-general, 164 responsiveness, 56-8 memory bias, 136-9, 146, 225, 277, 286
diurnal variation, 140 experimental materials, 141-4 explicit, 289-90 hedonic, 44 mood, activation, 219 motivational depression, 67 perceptual thresholds, 284 personal constructs, 180 phenomenology, 310-11 preoccupation, 7-8, 46 priming, integrative model, 280 psychotic, physiological arousal, 46-7 resource allocation, 46, 281 retrieval bias, integrative model, 282, 3 ruminative cycles, 166 schemata, 218 self-schema theory, 223-6 short-term memory, 58-9 Stroop task, 96-7 subjective risk, 202, 205 subliminal stimuli, 274 subliminal Stroop task, 113 treatment, linguistic representation, 186 worry, 178, 182 depressive interlock, 166 depressive retardation, 48-9 descriptionalism, 170-2 desensitisation emotional imagery, 174, 178-80 relaxation, 179
dichotic listening emotional stimuli, 76-9 nonconscious perception, 250-1 parallel distributed processing, 121-2, 130 Differential Activation Hypothesis, 205 differential deficit, cognitive performance, 45 differential processing, selective attention, 22 digit span, 58 disruption emotional Stroop task, 86-94 visual dot-probe paradigm, 82-5 dissociation, attentional bias, 122 dissonance theory, 234 distraction attention focus, 109-10 depression, 185 Stroop task, emotional stimuli, 86 dopamine, schizophrenia, 118 dot-probe task disruption/facilitation, 82-5 integration, 124 integrative model, 279-80 parallel distributed processing, 121-2, 130 dual-task studies, cognitive impairments, 66 E elaboration, 279 depression, 287 memorial versus non-memorial, 299-301 mental representations, 124 emotion
see also mood assessment, 175-8 awareness, 3, 42 cognition, 2-5, 13-15 excitation-inhibition mechanism, 126 hierarchical model, 14 interactive cognitive subsystems, 303 models of, 5-7 network theory, 11-13, 176 encoding, 59-60 bias, 279-82 data-driven versus concept-driven, 161-3 mood-congruence effect, 142-3 nonconscious memory, 242 perceptual versus conceptual, 294-6 specificity framework, 12 state-dependent retrieval, 135 environment, attentional bias, 74 ERPs see event-related brain potentials event-related brain potentials (ERPs), 27 examination stress Page 397
anxiety, 84 mood, 129 treatment, 69 executive control procedures capacity limitations, 20 negative priming effect, 24 expectations, reality monitoring, 187 expertise, Stroop task, 101-3, 116
F facilitation, visual dot-probe paradigm, 82-5 false-recognition paradigm, 59 familiarity, fearfulness, 78 fear familiarity hypothesis, 78 multiple-entry memory, 299 neuromodulatory influence, 118 filtering, 50 attentional dysfunction, obsessional patients, 54 schemata, 213-14 flashbulb memory phenomena, 145 focus, distractibility, 109-10 frames, schemata, 215-18 G GAD see general anxiety disorder general anxiety disorder (GAD) attentional dysfunction, 51 attentional focus, 109, 129 emotional Stroop task, 94-8 memory bias, 111, 155-60, 287-8 worry, 182-3 global workspace, consciousness, 265-6 goals, control hierarchies, 39 H habituation theory, 180 hedonic memory bias, 44 hemispheric laterality, mnemonics, 290 heuristics, 191-3 availability, 167
'How do I feel about it?', 197-8, 203 Human Associative Memory Network theory, 11 hypervigilance, 112 hypnosis, mood manipulation, 135, 137 I ICS see interacting cognitive subsystems imagery emotion relationship, 179-80 emotion versus emotional information, 176 emotional, 172-4 characteristics, 178-81 rehearsal for action, 178 retrieval of emotional material, 175-8 theories, 170-4 treatment techniques, 186-9 visual perception, 187 worry, 181-2 imagination, outcome probabilities, 200-1 information-processing paradigm, 16-43 activation versus inhibition, 23-4 anxiety versus depression, 287 attention, 115 automatic versus strategic, 39-42 bottom-up versus top-down, 33-7 capacity limitations, 18-21, 46 cascade models, 28 cognitive avoidance, 5 cognitive deficits, 39 cognitive load manipulations, 40 cognitive versus connectionist architectures 31
consciousness, 262-7 control hierarchies, 37-9 discrete versus continuous, 27-8 emotion, 5-6 experimental paradigm, 13 implicit/explicit distinction, 41 individual strategies, 39 integrative model, 277-83 memory bias, 112 nonconscious, 231-3 clinical implications, 268-75 learning, 243-50 memory, 237-43 perception, 250-60 perceptual hypotheses, 267 verbal reports, 233-7 processing stages, 24-7 retrieval, schemata, 36 serial versus continuous, 29-30 social processes, 4 trait versus state effects, 307-10 inhibition, information-processing paradigm, selective attention, 23-4 integration, mental representations, 123-4 Page 398
integrative model attentional bias, 283-5 data inconsistencies, 288-94 encoding, perceptual versus conceptual, 294-6 memory, 285-8 intensity, selective processing, 21
interacting cognitive subsystems (ICS), 302-6 interference memory bias, 110-12 subliminal stimuli, 112-14 introspection perceptual hypotheses, 267 verbal reports, 235 J judgements clinical, de-biasing, 208-9 contingency, 198-200 covariation, 194-5 heuristics, 191-3 mood, 195-8 subjective risk, 200-6 K knowledge connectionist models, 32 implicit/explicit distinction, 41 perceptual hypotheses, 266-7, 269-70 schemata, 35-7, 211-12 L language analysis, 35 individual strategies, 39 negative priming effect, 24 parallel-distributed processing models, 32 processing bottom-up versus top-down, 34
stages, 28 retrieval of emotional material, 175-8 learning accuracy, 247-8 confidence, 247-8 connectionist models, 32 conscious, artificial grammar, 246 implicit, 243 individual strategies, 39 nonconscious, 243-50 individual differences, 248-9 psychopathology, 272 rates, anxiety, 125 selective, meta-analysis, 141-2 state-dependent, 135 M mania, cognitive performance, 45 matching tasks, 46 MEM see multiple-entry memory memory see also retrieval accessibility hypotheses, 224-5 activation, 11 anxiety, 55, 152-5, 158-9 bias, 315-16 agoraphobia, 277 anxiety, 288 depression, 225 hedonic, 44 interference, 110-12
mood, 136-7, 225 prioritisation model, 112 categoric versus extended, 165 childhood, mood effects, 138-9 cognitive impairments, 45, 55-64 cognitive load manipulation, 40 context, 133-4 depression, 7, 136-41 control hierarchies, 39 diurnal variation, 140 experimental materials, 141-4 depressogenic information, 205 direct tests, integrative model, 286-8 direct versus indirect, 148-50, 161 emotional events, normal subjects, 144-6 emotions, 146-7 explicit, integrative model, 289-91 general anxiety disorder, 155-8 general versus specific, 164-8 implicit, 237 implicit versus explicit, 160-3 implicit/explicit distinction, 41, 278-9 indirect tests, 238-9 integrative model, 285-6 mechanisms, emotional disturbances, 124 mood effects, 137-8 mood-congruency, 139-41, 150, 226 integrative model, 288-9 multiple-entry model, 297-306
negative priming effect, 24 nonconscious, 237-43 Page 399
psychopathology, 270-1 overgeneral, 301 panic disorder, 159 perceptual, 122-3 post-traumatic stress disorder, 184-5 processing, 28, 59 repressed, 271 responsiveness, 56-8 schemata, 36-7 self-reference, 221-3 storage, individual strategies, 39 system/process distinction, 296 working through, 183, 185 mental models, 171 self-schemata, 227 mental representations, 31 metacognition, 261 metaknowledge, nonconscious learning, 248 MIP see mood induction procedure mnemonic interlock, 166 mood cognition, 3 cognitive impairment studies, 45 cognitive loop, 196 congruency, 95 encoding, 142-3 interactive cognitive subsystems, 306 memory, 139-41, 150, 222
network theory, 147-8 retrieval, 142-3 evaluative judgements, 195-8 memory, 133-5, 137-8 childhood, 138-9 normal subjects, 144-6 memory bias, 136-7, 225 network theory, 12 representations, semantic versus episodic, 177 selective recall, 9 state versus trait, 307-10 resource allocation, 281 subjective risk, 201-4 trait emotionality, 204-6 variation, memory, 140 mood induction procedure (MIP), 136 motivation, excitation-inhibition mechanism, 126 motivational depression, resource capacity, 67 Motivational Structure Questionnaire (MSQ), 100 motor skills learning, 37 processing stages, 28 schemata, 211 MSQ see Motivational Structure Questionnaire multiple personality disorder, nonconscious memory, 270 multiple resource theory, capacity limitations, 19-20 multiple-entry memory system (MEM), 297-9 cognitive bias, 299-301 conscious reflective processes, 296 interacting cognitive subsystems, 302-6
overgeneral recall, 301-2 N negative priming, attentional focus, 109 network theories see also activation attentional bias, 106 emotional concepts, 176 emotional disorders, 11-13 memory rehearsal, 139 retrieval, 167 mood congruency, 147-8 subjective risk, 201-2 selective processing, 24 neural networks see parallel distributed processing neurosis see also obsessive-compulsive neurosis normal subjects, 152-5 personal constructs, 180 neurotransmitters, emotional disorders, 118, 308 neutral tasks, interpretation of performance, 44 non-verbal communicative signals, processing, 35 norepinephrine anxiety, 118 learning rates, 125 O obsessive-compulsive disorder attentional bias, 78 attentional dysfunction, 53-5
Page 400
cognitive performance, 45 concern-related bias, 101-2 nonconscious memory, 270 personal constructs, 180 Stroop interference, 90 occulomotor control, 35 opponent process theory, 125-6 override, Stroop interference, 120 P panic disorder ambiguity, 228 concern-related bias, 99 explicit conceptual bias, 295 fluency processing, 293 indirect memory tests, 285 memory bias, 159, 285, 288 Stroop interference, 89-90 parachuting anxiety, 59 parallel contingent account, cognitive processing, 28 parallel distributed processing (PDP) model attentional bias, 114-22, 129-30 cognitive architecture, 31-3 integration, 124-5 opponent processes, 126 priming, 125 trait versus state effect, 308-9 parasuicide, overgeneral memory, 164 partial copy model, schemata, 214 pattern identification, processing stages, 28
PDP see parallel distributed processing perception see also vision attentional bias, 75 memory bias, 278-9 nonconscious, 250-60 backward masking, 253-4, 257, 267 psychopathology, 271-2 subliminal stimuli, 252-3 processing, bottom-up versus top-down, 34 visual, imagery, 187 perceptual defence, 80-2 perceptual hypotheses, consciousness, 266-7, 269-70 perceptual processing, 33-7 performance, facilitated, attentional bias, 76-82 phobia cognitive avoidance, 5 dissociation, 13-14 illusory correlation effect, 199 judgement, 190 memory bias, 111 nonconscious memory, 270 nonconscious perception, 271-2 personal constructs, 180 processing priority, 307 social, Stroop interference, 90-1, 113 specific, Stroop interference, 91-2 pictorialism, 170 pigeon holing, attentional dysfunction, obsessive-compulsive disorder, 54 post-traumatic stress disorder (PTSD)
attentional bias, 79 concern-related bais, 98 memory bias, 289 memory phenomena, 184-5 over-general memory, 164 Stroop interference, 89 preoccupation see also attention cognitive impairments, 46 emotional disorders, 7-8, 46 imagery, 179 priming, 23-4, 279 affective, 154 anxiety, 287 conceptual/perceptual dissociation, 161 hot versus cold, 284 negative, attentional focus, 109 parallel distributed processing 125 perceptual, 294 selective processing, 21 prioritisation, memory, 112 problem solving bottom-up versus top-down, 34 cognitive impairments, 52-3 cognitive processing, 37 judgement, 193-4 process dissociation procedure, nonconscious memory, 239-40, 243, 258 psychological unconscious, 233, 236, 250, 260 psychopathology interactive cognitive subsystems, 303
nonconscious learning, 272 Page 401
nonconscious memory, 270-1 nonconscious perception, 271-2 subliminal stimuli, 273-5 PTSD see post-traumatic stress disorder punishment, neuromodulatory influence, 118 R reality monitoring, expectations, 187 reasoning negative priming effect, 24 parallel-distributed processing models, 32 repetition compulsion, trauma, 183 representations emotion, 176 emotional imagery, 180-1 linguistic, 177, 186 process, 171 psychological treatments, 186-9 real/imaginary world distinction, 188 schemata, 36-7 semantic versus episodic, 177 repressed memory debate, 271 repression, 82 cognitive effort, 108 semantic versus episodic representations, 177 response training, 173-4 responsiveness, memory, 56-8 retrieval, 164-8 see also memory biased, 282-3
concept-driven, 161-3 data-driven, 161-3 effect of mood, 134-5 individual strategies, 39 mood-congruence, 135, 142-3, 150, 164, 177 response training, 173-4 strategic control, 165-6 risk, subjective, 200-6 S saliency, 21 SAS see Supervisory Attentional System scene analysis, bottom-up processing models, 34 schema pointer/tag, 214-15 schemata, 10-11, 210-12 ambiguity, 227-30 attentional bias, 106 bottom-up versus top-down interactions, 212-15 bottom-up versus top-down processing, 35-7 clinical theory, 218-20 danger, 86, 98-9 threat tag, 118 frames and scripts, 215-18 self-knowledge, 220-3 schizophrenia cognitive deficit model, 118 subliminal stimuli, 274 scientific paradigms, 17-18 scripts, schema theory, 215-18 SCRs see skin conductance responses secondary tasks, cognitive impairments, 48
selective processing, 21-3 see also attention self-guides see self-schema self-knowledge, 226-7 self-schema, 220-3 self-reference memory, 221-3 mood-congruence effects, 143 self-relevant threat, Stroop interference, 93 self-schema depression, 223-6 self-knowledge, 220-3 semantic network theories see network theories semantic processing, attentional bias, 21, 79-80 sensitivity see attentional bias Sirhan Sirhan, memory, 133 skill limitations, resource capacity, 20 skin conductance responses (SCRs), nonconscious perception, 271 snake-avoidance see also phobia attentional bias, Stroop task, 91, 102-4 memory bias, 111 override, 120 social situations, schemata, 36 SPA see subliminal psychodynamic activation spatial tasks, individual strategies, 39 Sperling task, cognitive impairments, depression, 58 spider-avoidance see also phobia attention focus, 146
Page 402
attentional bias, Stroop task, 86, 91-2 concern-related bais, 99 memory bias, 111, 287 mood-congruent recall, 150-2 Spielberger Trait Anxiety Inventory, 82 stimuli subliminal psychopathology, 273-5 Stroop interference, 112-14 stress management, visual imagery, 178 Stroop interference cognitive effort, 108 nonconscious perception, 251 override, 120 stress levels, 97-8 studies of recovered patients, 96-7 trait versus state emotion, 94-5 Stroop paradigm, 51 concern-related bias, 98-105 emotional, 9 anxiety inventories, 95-6 attentional bias, 85-94, 110-14 conflict model, 114-16 PDP model, 116-21, 128-30 response bias, 155 subliminal attentional bias, 130 interference, 112-14 study approaches, 60
stuttering, subliminal stimuli, 274 subliminal psychodynamic activation (SPA) procedure, 274-5 subliminal stimuli, Stroop interference, 112-14 Supervisory Attentional System (SAS), consciousness, 264 symbol processing account, cognitive architecture, 31-3 T task difficulty, cognitive impairments, 67 task structure, capacity limitation, 19 task-irrelevant processing, 72, 119 attentional bias, 107-8 task-irrelevant thoughts, cognitive impairments, 67-9 text comprehension, bottom-up versus top-down processing, 34 therapy see also treatment cognitive, automatic thingking, 268-70 connectionist framework, 33 thought automatic, reports, 268 knowledge structures, 169 threat, memory bias, 111-12 thresholds auditory, attentional bias, 76-80 perceptual, 161, 255 depression, 284 visual, attentional bias, 80-2 transfer, therapeutic, 186-8 trauma mnemonic overlock 167 mood repair, 140 repetition compulsion, 183
treatment see also therapy behavioural versus cognitive, 4 cognitive deficits, obessional-compulsive disorder, 55 emotional versus cognitive, 4 examination anxiety, 69 hierarchical model, 15 imagery-based, 178, 180 perceptual hypotheses, 313-14 phobic anxiety, 5 representations, 186-9 tunnel memory, 146 V Velten procedure, 135 verbalisation, linguistic representations, 186 vividness, 179 W word recognition, processing stages, 28 working through, emotional processing, 183-6 worry attention, 299-300 imagery, 181-2 physiological arousal, 182 subjective risk, 203 Index compiled by Indexing Specialists, Hove, Sussex Page 403
The Wiley Series in CLINICAL PSYCHOLOGY Carlo Perris, Willem A. Arrindell and Martin Eisemann (Editors) Parenting and Psychopathology Chris Barker, Nancy Pistrang and Robert Elliott Research Methods in Clinical and Counselling Psychology
Graham C.L. Davey and Frank Tallis (Editors) Worrying: Perspectives on Theory, Assessment and Treatment Paul Dickens Quality and Excellence in Human Services Edgar Miller and Robin Morris The Psychology of Dementia Ronald Blackburn The Psychology of Criminal Conduct: Theory, Research and Practice Ian H. Gotlib and Constance L. Hammen Psychological Aspects of Depression: Toward a Cognitive-Interpersonal Integration Max Birchwood and Nicholas Tarrier (Editors) Innovations in the Psychological Management of Schizophrenia: Assessment, Treatment and Services Robert J. Edelmann Anxiety: Theory, Research and Intervention in Clinical and Health Psychology Alastair Agar (Editor) Microcomputers and Clinical Psychology: Issues, Applications and Future Developments Page 404
Bob Remington (Editor) The Challenge of Severe Mental Handicap: A Behaviour Analytic Approach Colin A. Espie The Psychological Treatment of Insomnia Martin Herbert Clinical Child Psychology: Social Learning, Development and Behaviour David Peck and C.M. Shapiro (Editors) Measuring Human Problems: A Practical Guide Roger Baker (Editor) Panic Disorder: Theory, Research and Therapy Friedrich Fösterling Attribution Theory in Clinical Psychology Anthony Lavender and Frank Holloway (Editors) Community Care in Practice: Services for the Continuing Care Client John Clements Severe Learning Disability and Psychological Handicap