The Metaphysics of Memory
PHILOSOPHICAL STUDIES SERIES VOLUME 111
Founded by Wilfrid S. Sellars and Keith Lehrer Edi...
32 downloads
872 Views
2MB 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
The Metaphysics of Memory
PHILOSOPHICAL STUDIES SERIES VOLUME 111
Founded by Wilfrid S. Sellars and Keith Lehrer Editor Stephen Hetherington, The University of New South Wales, Sydney, Australia Senior Advisory Editor Keith Lehrer, University of Arizona, Tucson, AZ, U.S.A. Associate Editor Stewart Cohen, Arizona State University, Tempe, AZ, U.S.A. Board of Consulting Editors Lynne Rudder Baker, University of Massachusetts, Amherst, MA U.S.A. Radu Bogdan, Tulane University, New Orleans, LA, U.S.A. Marian David, University of Notre Dame, Notre Dame, IN, U.S.A. John M. Fischer, University of California, Riverside, CA, U.S.A. Allan Gibbard, University of Michigan, Ann Arbor, MI, U.S.A. Denise Meyerson, Macquarie University, NSW, Australia François Recanati, Institut Jean-Nicod, EHESS, Paris, France Mark Sainsbury, University of Texas, Austin, TX, U.S.A. Stuart Silvers, Clemson University, Clemson, SC, U.S.A. Barry Smith, State University of New York, Buffalo, NY, U.S.A. Nicholas D. Smith, Lewis & Clark College, Portland, OR, U.S.A. Linda Zagzebski, University of Oklahoma, Norman, OK, U.S.A.
The titles published in this series are listed at the end of this volume.
Sven Bernecker
The Metaphysics of Memory
Sven Bernecker University of California at Irvine Department of Philosophy Irvine, CA 92697-4555 USA
ISBN 978-1-4020-8219-1
e-ISBN 978-1-4020-8220-7
Library of Congress Control Number: 2008921236 © 2008 Springer Science + Business Media, B.V. No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Printed on acid-free paper 9 8 7 6 5 4 3 2 1 springer.com
Acknowledgments
This book is a revised version of the Habilitation thesis which I submitted to the University of Munich in January 2002. I thank Eckart Förster, Verena Mayer, Ulises Moulines and Wilhelm Vossenkuhl for helpful comments on the thesis. Some of the ideas in this book have been tried out in graduate seminars at the University of Munich and the University of Minnesota at Minneapolis as well as at a couple of talks. The talks took place at the 2nd biannual Principia Symposium at the Federal University of Santa Catarina in Florianópolis, Brazil in 2001 and at the Bertrand Russell Society meeting at the Eastern Division meeting of the American Philosophical Association in Philadelphia in 2002. I am grateful to the audiences on those occasions. Moreover, I would like to thank Thomas Baldwin, Peter Baumann, Dorothea Debus, Fred Dretske, Tom Noerper, Wade Savage and Marya Schechtman for useful conversations about some of the material in this book. Chapter 3 is based on my paper ‘Russell on Mnemic Causation’, Principia: An International Journal of Epistemology 5 (2001). I am grateful to the editors of Principia for their permission to make use of this material. Grateful acknowledgment is made to Kegan Paul for permission to reprint two figures from C.D. Broad, The Mind and its Place in Nature (London: Kegan Paul, Trench, Trubner and Co., 1925), Figs. 3.1 and 3.2 in section 3.4. Finally I owe thanks to my wife, Narghes. She managed to remain cheerful in the face of occasional moods of frustration and depression and was a rich source of encouragement. The book is dedicated to her.
v
Contents
1 Setting the Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 1.2 1.3
The Classification of Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Analysis of Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 2 7 9
Part I Memory Causation 2 The Motivation of the Causal Theory of Memory . . . . . . . . . . . . . . . . . 2.1 2.2 2.3 2.4 2.5 2.6
17
Two Arguments Against the Causal Theory of Memory . . . . . . . . . . An Argument from Conditionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . An Argument from Personal Identity . . . . . . . . . . . . . . . . . . . . . . . . An Argument from Temporal Forgetting . . . . . . . . . . . . . . . . . . . . . . Critique of Pure Retention Conditions . . . . . . . . . . . . . . . . . . . . . . . Critique of Evidential Retention Conditions . . . . . . . . . . . . . . . . . . .
18 20 22 24 26 28
3 An Argument for Memory Traces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.1 3.2 3.3 3.4
Contiguous Causation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What Are Memory Traces? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Verifiability of Memory Traces . . . . . . . . . . . . . . . . . . . . . . . . . Against Mnemic Causation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32 34 38 42
4 From Traces to Recall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
4.1 4.2 4.3 4.4
Approaches to Memory Causation . . . . . . . . . . . . . . . . . . . . . . . . . . Retrieval Cues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Traces, Cues and Suggestibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Operative Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47 50 52 55
vii
viii
Contents
Part II Objects of Memory 5
6
7
The Primary Objects of Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
5.1 5.2 5.3 5.4 5.5
Representative Versus Direct Realism . . . . . . . . . . . . . . . . . . . . . . . The Co-temporality Objection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Argument from Hallucination. . . . . . . . . . . . . . . . . . . . . . . . . . The Causal Theory and Direct Realism . . . . . . . . . . . . . . . . . . . . . . A Humean Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62 68 72 74 76
Against Representative Realism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
6.1 6.2 6.3 6.4
81 87 93 97
The Need for Memory Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Varieties of Memory Markers . . . . . . . . . . . . . . . . . . . . . . . . . . Remembering and Trying to Remember . . . . . . . . . . . . . . . . . . . . . The Validation of Ostensible Memory . . . . . . . . . . . . . . . . . . . . . . .
Skepticism, Externalism, and Closure . . . . . . . . . . . . . . . . . . . . . . . . . 105 7.1 7.2 7.3 7.4 7.5 7.6
Skepticism and Metajustification . . . . . . . . . . . . . . . . . . . . . . . . . . . Internalism and the Acquisition of Justification . . . . . . . . . . . . . . . The Problem of Forgotten Evidence . . . . . . . . . . . . . . . . . . . . . . . . Internalism About Continuous Justification. . . . . . . . . . . . . . . . . . . Externalism About Continuous Justification . . . . . . . . . . . . . . . . . . Russell’s Hypothesis and Closure . . . . . . . . . . . . . . . . . . . . . . . . . .
106 109 113 117 122 126
Part III Truth in Memory 8
The Factivity Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 8.1 8.2 8.3 8.4 8.5 8.6
9
Motivation of the Truth Condition . . . . . . . . . . . . . . . . . . . . . . . . . . Truth in Reproductive Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . Truth in Meta-representational Memory . . . . . . . . . . . . . . . . . . . . . The Xerox Model of Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adaptive Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Function of Adaptive Variation. . . . . . . . . . . . . . . . . . . . . . . . .
137 140 142 144 146 152
Diachronic Content Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 9.1 9.2 9.3 9.4
The Relativity of Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Principles of Semantic Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . Rectifying Memory Claims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Attitudinal Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
156 158 164 165
Contents
10
ix
The Pragmatic Dimension of Memory . . . . . . . . . . . . . . . . . . . . . . . . . 169 10.1 10.2
Epistemic Contextualism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 Memorial Contextualism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Chapter 1
Setting the Stage
This book investigates central issues in the philosophy of memory. Does remembering require a causal process connecting the past representation to its subsequent recall and, if so, what is the nature of the causal process? Of what kind are the primary intentional objects of memory states? How do we know that our memory experiences portray things the way they happened in the past? Given that our memory is not only a passive device for reproducing thoughts but also an active device for processing stored thoughts, when are thoughts sufficiently similar to be memory-related? The book defends a version of the causal theory of memory, argues for direct realism about memory, proposes an externalist response to skepticism about memory knowledge, and develops a contextualist account of the factivity constraint on memory. The causal theory of memory has it that a past representation and its subsequent recall are connected by a causal process involving memory traces. Direct realism maintains that the primary intentional objects of memory are past events rather than present internal representations (experiences) thereof. We have to internally represent a past event to remember it; but what we remember is the past event, not the internal representation. Memory is indirect in the sense that it involves internal representations; but memory is not indirect in the sense of involving a prior awareness of something other than the past event. Externalism about memory knowledge is the position that a memory belief is justified if it has the property of being trutheffective. This property of being truth-effective may, for example, consist in the belief being formed or sustained by a reliable method or process. No more than this is necessary for justification. So, whether the subject takes his belief to be trutheffective doesn’t add anything to the memory belief’s epistemic status.1 Finally given contextualism about the factivity of memory, remembering is pragmatically sensitive in that it depends on the conversational context whether the content of a memory state must be the same as the content of the past representation from which the memory state causally derives, or whether the two diachronic content tokens need only be sufficiently similar.
1
Here and throughout the book gender-unspecific reference is made with “he” and its cognates, which may be read as “she or he,” “her and him,” etc.
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
1
2
1 Setting the Stage
The book consists of three parts. The first part (chapters 2–4) elaborates and defends the causal theory of memory. The second part (chapters 5–7) argues for direct realism and sets forth an externalist account of memory knowledge. The third part (chapters 8–10) develops a contextualist account of the truth conditions for remembering.
1.1
The Classification of Memory
Before I explore the metaphysics of memory, I have to restrict the range of the inquiry to a set of phenomena which can be handled within the limitations of this study. That is accomplished by the development of a new typology for memory. The new typology is based on a grammatical classification of kinds of memories in terms of objects of the verb “to remember.” After explaining this classification and characterizing the kinds of memory discussed in this study, I point out how my classification relates to the standard philosophical distinction of kinds of memory and the distinctions favored by psychologists. There are four kinds of remembering. One can remember (i) objects (e.g., remembering Fido), (ii) properties (e.g., remembering Fido’s floppy ears), (iii) events (e.g., remembering Fido’s biting the mailman), and (iv) facts (e.g., remembering that Fido bit the mailman). Consequently, I will speak of object-, property-, event-, and fact memory. Fact memory is also called “propositional memory.”2 In addition to these four constructions of memory attributions, “to remember” is frequently used with wh-clauses, i.e., clauses beginning with “who,” “whom,” “what,” “where,” “when,” and “why.” We can say things like “S remembers what Fido did,” “S remembers whom Fido bit,” “S remembers why Fido bit the mailman,” etc. Yet the combinations of “remember” with wh-clauses do not refer to a distinct kind of memory, but are nothing but incomplete attributions of propositional memory. Saying that S remembers what Fido did is just another way of saying that S remembers that Fido did such-and-such. And saying that S remembers whom Fido bit is tantamount to saying that S remembers that Fido bit so-and-so.3 Remembering facts implies the deployment of concepts. If you remember that Fido bit the mailman, you have to have the concepts Fido and mailman and apply them to what you remember. You cannot remember that Fido bit the mailman unless you understand who Fido is and what it means to bite someone. Object-, property-, and event memory, however, do not require the possession of concepts. A child can remember Fido’s floppy ears long before he acquires the concepts Fido and ear. By saying “S remembers x”, where “x” stands for a person, an object, an event or a property one doesn’t say anything about how S describes or classifies x.
2 Dretske (1969, ch. 2; 1979; 1981a, ch. 6) and Warnock (1965) propose an analogous distinction of types of seeing. 3 The suggestion that wh-constructions of “remember” are reducible to fact memory is taken from Aho and Niiniluoto (1990, pp. 419–421). When “remember” is combined with how-clauses, as in “S remembers how to swim,” it usually expresses practical (or procedural) memory.
1.1 The Classification of Memory
3
In these statements, x is referentially transparent. If you remember, say, Marilyn Monroe you thereby also remember Norma Jean Mortenson, regardless of whether you know that “Marilyn Monroe” and “Norma Jean Mortenson” are coextensive. Attributions of fact memory, on the other hand, are, at least sometimes, referentially opaque. You can remember that the Marilyn Monroe was blond without remembering that Norma Jean Mortenson was blond. Substituting co-extensive terms may change the truth value of attributions of propositional memory. One can remember that p without the original awareness that gives rise (via memory traces) to the state of propositional memory being propositional awareness as opposed to object-, property-, or event awareness. To cut the topic to a manageable size, I will concentrate on fact memory of previous fact awareness. Broadening the study to include object-, property-, and event memory would demand a discussion of the difference between conceptual and non-conceptual content, a muchdiscussed topic of contemporary philosophy of mind. Yet since the conceptual/ non-conceptual content distinction is not specific for memory, it would lead us away from our guiding question “What is memory?.” Apart from the distinction between object-, property, event, and fact memory, there is the distinction between the veridical (or factive) and the ostensible usage of “remembering.”4 Given the ostensible usage, you can remember Fido when Fido was hallucinated. It is the veridical sense which is of interest here. As I use the phrase, “remember that p” implies that the proposition in question is true. Thus, you cannot remember that Fido bit the mailman if he didn’t do it. Yet it may seem to you as if you remember that Fido bit the mailman, and you can remember it seeming to you that Fido bit the mailman. Analogously, object-, property-, and event memory are governed by an existence condition: you cannot remember Fido’s floppy ears and his biting the mailman without Fido having floppy ears and without his having bitten the mailman.5 Furthermore, there is the distinction between the occurrent and the dispositional sense of “remembering.” When we say of S that he remembers that p we either say of him that he is currently engaged in propositional remembering, or that he is disposed to remember that p if suitable circumstances occur. According to the dispositional usage, we can say of S that he remembers that p even while he is sound asleep. I will concentrate on the occurrent use of “remembering,” for the dispositional sense seems to be parasitic on the occurrent sense. It is meaningless to say of someone that he remembers that p unless he did, at some point, somehow manifest his memory of p.
4 It is important not to confuse “fact memory” or “factual memory” in the sense of propositional memory with “factive memory” in the sense of veridical memory. There are a number of stock expressions which indicate ostensible memory: “as S recalls it, p,” “according to my best recollection, p,” “I seem to remember that p,” and so on. 5 The cases of object memory I have in mind are those where the object is expressed by a singular term. Problems arise when the object is expressed by a quantifier. Does remembering no one imply that no one was there?
4
1 Setting the Stage
In the course of the study, it will be necessary to introduce further distinctions of kinds of memories, such as the distinction between inferential and non-inferential memory (section 8.2) and the distinction between memory de re and memory de dicto (section 9.2). These distinctions will be explained at their appropriate places in the body of the text. The primary object of this study is the factive and occurrent usage of “remembering that,” where this refers to non-inferential fact memory de re of previous fact awareness. There are two kinds of fact memories. When I remember that p, “p” can stand for a simple fact (e.g., that Fido bit the mailman) or for a previous representational state of mine (e.g., that I believed that Fido bit the mailman). When I remember that Fido bit the mailman, the content of my memory state is a reproduction of one of my previous thought contents. When I remember that I believed that Fido bit the mailman, however, the memory content refers not only to the content of my past belief but also to the attitude I had towards that content (in this case the attitude of believing). The memory content is a meta-representation of a past content. The content of the metarepresentation consists of the content of the past thought and an extra bit of content, the attitudinal component (e.g., I believed). By referring to the former attitude, the meta-representational memory content represents the past content as a representation. To drive home the difference between meta-representation and simple contentreproduction, consider photography. Content-reproduction may be compared to prints made from a negative. Prints made from the same negative contain the very same amount of information. Meta-representation can be compared to a given photo A showing a table with a given photo B lying on top of it. The content of photo A not only contains the content of photo B, but it represents this content as a photographic representation; it is a photographic meta-representation. For lack of a better term, I will call fact memories of one’s own representational states meta-representational memories. Fact memories of other things than one’s own mental states I label reproductive memories. As I use the term, “meta-representational memories” are memories of propositional attitudes which I had. I cannot have a meta-representational memory of something that someone else believed.6
6 What I call “meta-representational memory” is sometimes labeled “episodic memory.” Following Tulving, psychologists use “episodic memory” to characterize that faculty which allows us to receive and store “information about temporally dated episodes or events [which we have experienced in the past], and temporal-spatial relations among these events” (Tulving 1972, p. 385; cf. 1983, part I). Roughly speaking, episodic memory is that kind of memory that allows us to report episodes that once happened to us. It enables us to answer questions such as “What were you doing yesterday morning?” or “When did you last go to the theater?” “Semantic memory”, on the other hand, is the store of general knowledge about the world, concepts, rules, and language. The characteristic feature of semantic memory is that it can be used without reference to the events that account for its formation in the first place. Semantic memory can also store information about ourselves. When asked what we do, what our opinions are, and so on, we do not have to remember specific past experiences in order to answer. Instead we can give a general account of ourselves. Thus, whereas semantic memory involves retrieval of the fact acquired during a given learning episode, the corresponding episodic memory would involve remembering something about the specific learning episode itself, namely the context in which the fact was acquired.
1.1 The Classification of Memory
5
The scope of the distinction between meta-representational and reproductive memory is more restricted than it initially seems. Conative states can be recalled only by means of meta-representational states, for I cannot remember the content of my conative states without remembering their attitudes. Suppose I wanted Fido to bite the mailman. What I remember later on is not just Fido, bite! but that I wanted him to bite the mailman. The same is true for certain cognitive states. For example, I cannot remember what it is that I regretted without stating that I regretted it. Likewise, whenever I recall what I doubted I have to state that I doubted it. But there are other types of cognitive states where the distinction between metarepresentational and reproductive remembering does apply. Paradigm examples of such types of cognitive states are thoughts, beliefs, inferences, and memories. I can entertain or express the belief that Fido bit the mailman by thinking or saying “Fido bit the mailman;” without mentioning the attitude. When recalling thoughts, beliefs and memories, I have the choice between merely reproducing the content or representing my previous mental state. Another problem with the distinction between meta-representational and reproductive memory is that, though it is grounded on a difference in the way we speak of remembering, verbal memory reports are not always reliable indicators of whether someone occupies a state of meta-representational memory or reproductive memory. Sometimes reproductive memories are phrased in a way typical of metarepresentational memories. Suppose, for example, someone asks you “What did you think about such-and-such?”, you may respond “I thought that p” even though you are engaged in reproductive memory. Rather than referring to the attitudinal component of a past mental state of yours, the “I thought” simply takes up the formulation of the question you are answering. And just as there are instances where reproductive memories are expressed as meta-representational memories, there are instances of the reverse case.7 How does the typology of kinds of memory proposed here relate to the standard philosophical typology? Usually philosophers differentiate between propositional (or factual) memory, personal (or experiential) memory, and practical (or habit) memory.8 Practical memory is remembering how to do something where this refers to an acquired skill. Examples of practical memory are remembering how to swim and remembering how to ride a bicycle. Propositional memory is remembering that p, were “p” stands for some proposition. And the characteristics of personal memory are, first, that one can remember only what one previously experienced and, second, that it involves imagery. Regarding the former feature, Don Locke writes,
7 These latter cases have to be set apart from instances of ostensible memory: suppose you claim to remember that p. Shortly afterwards you find out that p is false. Instead of admitting that you didn’t remember that p, you may retreat to the claim that you thought that p. See section 9.3. 8 See, for example, Furlong (1951, pp. 72–91), Hamilton (1998, p. 285), Johnson (1983, pp. 376– 377), Locke (1971, ch. 5–7), Malcolm (1963, pp. 203–222), Martin and Deutscher (1966, pp. 161–164), O’Connor and Carr (1982, pp. 116–121), Pollock (1974, pp. 181–188), and Smith (1966, ch. 3).
6
1 Setting the Stage personal memory is restricted to those cases in which the claim to remember something incorporates a claim to have experienced it…. [P]ersonal memory consists in bringing some previously experienced thing to mind, thinking about it again, and going over what it was like (1971, pp. 71, 76).
Since, when an agent personally remembers something, his memory of it was caused by the experience he had of the thing, it is impossible to personally remember things that happened before one’s birth or that will happen in the future. Concerning the second feature of personal memory, Charles Martin and Max Deutscher declare: If someone is asked whether he remembers what he did last Friday at lunchtime, he may be able to say that he went down the street. Yet he may feel scarcely in a position to say that he remembers actually going down the street. What he needs in order to be able to say that he does [personally] remember going down the street is at least more detailed remembering that … certain things happened when he went down the street…. [T]his addition of detail must be due to the original perception.9
Contrary to first impression, the distinction between personal and propositional memory is far from clear. If it is understood as a grammatical distinction, it doesn’t make much sense, for, obviously, memories of personal experiences can also be remembered in propositional form. And if it is read as the distinction between memories with images and memories without images, it is equally problematic because propositional memories may, of course, be associated with images.10 Some philosophers construe the relation between personal and factual memory along the lines of Bertrand Russell’s distinction between knowledge by acquaintance and knowledge by description. Something is known by acquaintance when there is direct experience of it; it is known by description if it can only be described as a thing with such-and-such properties. Personal memory is memory by acquaintance and its objects are not facts or propositions but people, places, things, events, situations. John Pollock, for example, writes:
9
1966, pp. 162–163. Martin and Deutscher call factual memory “remembering that1” and personal memory “remembering that2.” In addition to factual and personal memories Malcolm posits perceptual memories (1963, pp. 204–221). Malcolm defines personal memory as follows: “A person, B, personally remembers something, x, if and only if B previously perceived or experienced X and B’s memory of x is based wholly or partly on his previous perception or experience of x” (ibid., p. 215). He defines perceptual memory as follows: “B perceptually remembers x if and only if B personally remembers x and B can form a mental image of x” (ibid., p. 219). Thus, Malcolm’s notion of perceptual memory is synonymous with how most philosophers define the notion of personal memory. 10 Psychological studies have revealed that the frequency of memory images varies greatly from one person to another (e.g., Nigro and Neisser 1983). Some people report that their mental lives are replete with imagery as vivid and detailed as the actual scenes they recall. There are even some, eidetikers, who can apparently summon up such images and then “read off” information from them in a very impressive way. For others, imagery is not so clear and vivid, although images are at least generally recognizable. For still others imagery is uniformly vague, dim and fleeting. And, for a final small group, there seems to be no imagery at all.
1.2 The Analysis of Memory
7
Personal memory is appropriately described as memory-of rather than memory-that…. We can express personal memories using gerund clauses…. These gerund clauses can be said to describe “states of affairs.” We can then say that personal memory is always memory of a state of affairs. Personal memory always has the form “I remember its being the case that p.”11
As far as I can see, this is the only coherent reading of the factual/personal distinction. But now it turns out that the traditional notion of factual memory is synonymous with my concept of fact memory and the traditional notion of personal memory is a cluster concept for object-, property-, and event memory. Before terminating this section, let me briefly point out how my typology of remembering fits in the psychological taxonomy of memory. The most fundamental psychological distinction of forms of memory is the one between short-term memory (also known as “working memory”) and long-term memory. The short-term memory stores information for not more than about 15 seconds. Once in the shortterm memory, information can meet one of two fates: it can be transferred to the long-term memory, or it can be forgotten. Long-term memory, as the name suggests, is the memory system that is responsible for storing practically limitless amounts of information for such long periods of time that its limits are unknown. Within long-term memory psychologists distinguish between associative and non-associative learning. Non-associative learning encompasses elementary forms of behavioral plasticity such as habituation and sensitization. Within associative learning there is the distinction between explicit (conscious, direct, or declarative) memory and implicit memory. Implicit memory is expressible only in behavior. The main forms of implicit memory are priming, conditioning, and procedural learning. My notions of object-, property-, event-, and fact memory belong to the class of explicit longterm memory.
1.2
The Analysis of Memory
The preliminary analysis of reproductive memory looks something like this: S’s representing at t2 that p qualifies as reproductive remembering only if (1) S represented at t1 that p* (2) p and p* are identical or sufficiently similar
11
1974, p. 184. Broad (1925, p. 222) and Malcolm (1963, p. 207) also hold that propositional memories are of facts whereas personal (or perceptual) memories are of objects, properties or events. Although Malcolm maintains that the objects of personal (or perceptual) memories are not facts, he argues that these memories entail factual memory (ibid., pp. 212–214). His argument can be phrased in five steps: (1) A person S personally remembers object x only if S can form an image of x. (2) If S has an image of x, he can correctly describe that image of x. (3) If S can correctly describe the image of x, he can correctly describe facts learned in the past. (4) If S can correctly describe facts learned in the past, he factually remembers that p, where “p” is a true sentence which describes these facts. (5) If S personally remembers x, then S factually remembers that p. Saunders’ (1965b) critique of this argument caused Malcolm (1977, p. 16n) to discard it.
8
1 Setting the Stage
(3) S’s representing at t1 that p* is the cause of S’s representing at t2 that p (4) p is true at t2 (5) p* is true at t1 And the preliminary analysis of meta-representational memory reads: S’s representing at t2 that he represented (at t1) that p, where “represented” is a placeholder for a non-factive attitude qualifies as an instance of meta-representational memory only if (1) S represented at t1 that p* (2) p and p* are identical or sufficiently similar (6) S represents at t1 that p* is the cause of S’s representing at t2 that he represented (at t1) that p12 Condition (1) may be labeled the previous representation condition, condition (2) the content condition, conditions (3) and (6) the causation condition, and conditions (4) and (5) the truth conditions. Chapter 9 explicates the notion of similarity employed in condition (2) and chapter 4 makes an attempt to say what kinds of causal relation conditions (3) and (6) call for. A terminological note: The value of the index in the subscript to “t” determines whether the time referred is in the past or the present: the relatively biggest number indicates the present. So here “t2” is the present and “t1” is the past. When there is more than one past time involved, “t1” indicates the distant past, “t2” the close past and “t3” the present. With respect to reproductive memory, the previous representation condition ensures that one can remember only that of which one was previously aware.13 In the case of meta-representational memory, this condition guarantees that the embedded content of one’s memory state is something of which one was previously aware. The function of the causal condition is to exclude re-learning and guarantee that the representation had in remembering is a retained representation. The idea is that a claim to remember something implies, not merely that the rememberer represented it in the past, but that his current representation is in some way due to his past representation. Finally, the truth condition makes sure that one can remember only what is the case. The factivity constraint on meta-representational memory need not be stated as a separate condition for it is implied by the past representation condition (1) and the causal condition (4).
12 The reason “at t1” is in brackets is that we frequently remember our previous propositional attitudes without recalling the exact time when we first entertained these attitudes. 13 Negative memories (e.g., remembering that you didn’t lock the door) seem to conflict with the previous representation condition of reproductive memory. Yet, as will be explained in section 8.2, negative memories are a kind of inferential or impure memory. What I call the “previous representation condition” Shoemaker (1970, p. 269) calls the “previous awareness condition” and others call it the “witness condition.”
1.3 The Plan
9
Calling remembering a form of representing is meant as a dissociation from the epistemic theory of memory, which states that remembering is the reciting of previously acquired and stored knowledge. In sections 2.6 and 7.3, I argue that propositional memory requires neither that the previous nor that the present state of awareness qualifies as knowledge. The way I use “represent” is intended to include various kinds of attitudes towards factual awareness; knowing is among these attitudes but it is not the only one. Given the above analyses, memory presupposes personal identity. In the case of reproductive memory, the presupposition of personal identity is entailed by the fact that the same variable “S” is used to refer to the subject who represented at t1 that p and the subject who represents at t2 that p. Regarding meta-representational memory, not only S must be numerically identical between t1 and t2 but also he must identify his present person-stage as being identical with the past person-stage referred to by the memory content. The relation between personal identity and memory is a fascinating issue, but it is too complex to be exhaustively discussed in this study. The topic of personal identity will be dealt with only in passing. Section 2.3 is a critique of Shoemaker’s argument that unless we interpret remembering as involving a causal process, we cannot account for the role played by the notion of memory in judgments of personal identity.
1.3
The Plan
As was mentioned at the beginning, this study consists of three parts. The first part (chapters 2–4) elaborates and defends the causal condition of memory. The second part (chapters 5–7) argues for direct realism and sets forth an externalist account of memory knowledge. The third part (chapters 8–10) develops a contextualist account of the truth condition of memory.
Part I: Memory Causation According to the causal theory of memory, a past representation and its subsequent recall are connected by a causal process involving memory traces. What is novel about the treatment of the causal theory of memory here is that it is shown to be not only plausible but indispensable. The philosophical literature on memory contains a number of direct arguments for and against the causal theory of memory. In chapter 2, I argue that none of these direct arguments is successful. The only way of vindicating the causal theory of memory is by means of an indirect argument, an argument to the best explanation. What are alternatives to the causal account of memory? Some philosophers define memory simply as retained belief. But this account appears coherent only as long as one doesn’t ask what is involved in the retaining of a piece of information.
10
1 Setting the Stage
When this question is raised, proponents of this approach have to concede that their account cannot provide an answer. But if the process underlying retention of information cannot be explained, memory becomes a magical faculty. Another alternative to the causal theory of memory is the evidential approach. The idea behind this approach is that for a mental state to qualify as memory it not only has to be a piece of knowledge but its justificatory factors must be the same as those supporting the original piece of knowledge that has been retained. The evidential approach fails, I argue, because of its commitment to the epistemic theory of memory, i.e., the thesis that we can remember only what we previously knew. No doubt, memory implies truth, but it does not imply epistemic justification. After having argued that there is no viable alternative to the causal theory of memory, I proceed by examining the nature of memory causation. According to the standard view, the causal process connecting a past representation and its subsequent recall involves intermediary memory traces (or engrams). However, in The Analysis of Mind, Russell argued that since the physiological evidence for memory traces is not yet quite conclusive, it is prudent to come up with an account of memory causation – he calls it “mnemic causation” – that manages without the stipulation of memory traces. According to the theory of mnemic causation, a past experience is directly causally active over a temporal distance; past experiences are proximal causes of states of recalling. The task of chapter 3 is to refute the theory of mnemic causation, to explain the notion of a memory trace, and to establish that the stipulation of memory traces is indispensable for the analysis of memory. The idea is that the stipulation of memory traces allows us to explain certain features of our intuitive notion of memory that cannot be accounted for by the theory of mnemic causation. What are memory traces? On the intentional level of explanation, traces are dispositional beliefs or subdoxastic states – depending on whether their content is conceptual or non-conceptual. On the computational level, traces are formulae in the language of thought or patterns of activation in a neural network – depending on whether one assumes the classical or the connectionist approach. It is shown that, with respect to memory, classical computationalism and connectionism shouldn’t be viewed as rivals but as complementary theories. Chapter 4 sets forth a positive characterization of the kind of causation involved in information retrieval. Given that we frequently remember only after being presented a suitable retrieval cue (or prompt), what is it to remember something upon being prompted, rather than merely repeating back the prompt itself? The thesis is that the content of a memory trace at least has to be a necessary component of a condition that is independently sufficient for bringing about the state of recounting.
Part II: Objects of Memory An obstinate preconception has it that the causal theory of memory is committed to representative (or indirect) realism which, in turn, inevitably leads to skepticism about memory knowledge. My goal is to show, first, that the causal theory is not
1.3 The Plan
11
committed to representative realism, second, that direct realism is to be preferred over representative realism and, third, that even if one were to accept representative realism one would not have to give in to skepticism. Chapter 5 explains the distinction between direct and representative realism about memory and argues that, contrary to what many philosophers think, the causal theory of memory is perfectly compatible with direct realism. Direct realism is the denial of representative realism, i.e., the view that remembering that p involves primary awareness of a mental representation of p. According to direct realism, remembering involves the possession mental representations but what we are directly aware of in memory are not these representations but what it represented by them. In addition to showing that direct realism is compatible with the causal theory, chapter 5 criticizes two further objections to direct realism – the co-temporality objection and the objection from hallucination. Chapter 6 continues to defend direct realism about memory by arguing that the alternative position – representative realism – inevitably leads to skepticism about memory knowledge. The representative theory of memory claims that what we are directly aware of when we remember, are not the things we experienced in the past, but present internal representations of those past things. The difficulty for the representative theory is that restricting what we are directly aware of to internal representations, equally restricts what we can know about the external world. For if all we are directly aware of are our representations or ideas, how can we know that there are external things at all, much less that these external things are the causes of our present ideas? In other words, how can we be sure that our seeming memories portray things the way they happened in the past? In response to the skeptical challenge many proponents of representative realism have maintained that one can tell, by reflection alone, whether the representation one is currently entertaining originates from the memory or from some other mental faculty, such the imagination. And since there is an a priori knowable feature of authentic memory representations that stamps them as such, we can tell whether or not a particular memory claim amounts to knowledge. Such memory markers are said to provide us with something from which it can be ‘read off’ that an experience is a memory experience, rather than an experience of an act of imagining. Chapter 6 examines a variety of accounts of memory markers and shows that none of them are convincing. It also discusses alternative strategies for neutralizing skepticism about memory knowledge by way of validation of ostensible memories and finds all of them wanting. Skepticism about memory knowledge assumes internalism about justification, that is, the view that all of the factors required for a belief to be justified must be cognitively accessible to the subject and thus internal to his mind. The most promising strategy for fending off skepticism about memory knowledge available to the representative realist is to reject internalism and embrace externalism. The externalist holds that some of the justifying factors may be external to the subject’s cognitive perspective. Given externalism, we need not be able to respond to the skeptical query. The fact that we don’t know (or justifiably believe) that our ostensible memories amount to knowledge does not mean that they do not amount to
12
1 Setting the Stage
knowledge. A person who knows something does not have to know that what he has in his evidential base amounts to knowledge. As long as he in fact satisfies the conditions of knowing something, there is nothing more he has to do in order to know. No skeptical worry gets started. Chapter 7 is an elaboration and defense of externalism about memory knowledge. I argue that externalism provides not only a plausible account of coming to justifiably believe something that is subsequently committed to memory but also a convincing account of continuing to justifiably believe something one remembers.
Part III: Truth in Memory A claim to the effect that one reproductively remembers that p is veridical if, first, p is the case and, second, at some point in the past, one represented (thought) that p. The truth of a meta-representational memory claim that one thought that p, however, requires only that one thought that p. Unless one’s past attitude towards p was factive, p doesn’t have to be the case. Chapter 8 motivates and explains the factivity constraint on memory. Various sources of error in metarepresentational and reproductive memory are distinguished. Meta-representational and reproductive memories share a truthmaker. Regardless of whether one remembers that p or whether one remembers having represented that p, it has to be the case that, at some point in the past, one represented that p. In both cases the (embedded) memory content must have been previously entertained. And the truth of memory claims depends on the match between the (embedded) memory content and the past thought content from which it causally derives. Since a match can be more or less perfect, the question arises to what extent past and present thought contents have to be alike for the latter to qualify as remembering. According to the standard view, the content of a reproductive memory claim or the embedded content of a meta-representational memory content, respectively, has to be a token of the very same type of content as the original thought with which it is connected by traces. Since, in this view, memory is conceived of as a photocopier producing duplicates of past contents, this view may be called the “xerox model of memory.” Chapter 8 is devoted to showing that the xerox model is incomplete, for it captures only one aspect of what our memory is supposed to do. Our memory is not only a passive device for reproducing information, but also an active device for reconstructing information. Of course, when the reconstructive nature of memory gains the upper hand, the distinction between remembering and imagining becomes blurred. But remembering does allow for the moderate transformation of the stored contents. Memory contents may count as veridical even though they are not exact duplicates of the past contents they derive from but contain some reconstructive elements. The aim of chapter 9 is to investigate the extent to which two diachronic content tokens may differ from one another while one of them is still able to count as a memory of the other. Given that the context is such that remembering does not require the regurgitation of contents, what is the permissible range of aberration between the original thought content and the memory content? When are two diachronic
1.3 The Plan
13
content-tokens sufficiently similar for the later to be able to qualify as a memory of the earlier? May, for example, my belief that a dog bit someone count as a memory of my previous belief that Fido bit the mailman? I propose a number of principles of semantic similarity. The general idea behind these principles is that, in the process of remembering, the informational content stored in traces may stay the same or may, to some degree, decrease, but it may not increase. Chapter 10 argues for contextualism about truth in memory, that is, the view that the truth condition of memory is pragmatically sensitive in the sense that it is contextually dependent on whether memory requires literal reproduction or whether it allows for the moderate reconstruction of previously learned information. The function of memory is to bring past thought contents to bear on present circumstances. Depending on the context, realizing this function may require either that the memory content be type-identical with the content of the thought it derives from or that the two content-tokens be merely similar. Hence, the truth condition of memory is to some degree sensitive to the function of memory, and the function of memory, in turn, is to some degree sensitive to the context of remembering. Saying that the function of remembering depends on the context can mean either of two things. The issue is whether the varying standards a subject must live up to in order to be remembering are relative to the context of the subject himself or to the context of the attributor of memory. I argue that both are the case. The conclusion of chapter 10 is that remembering is a relationally absolute concept. To remember that p means to accurately represent a previously had thought or experience. In this respect memory is an absolute concept. Yet what counts as an accurate representation may be, to some degree, relative to the interests and purposes of the rememberer. In this respect memory is a context-dependent concept.
Part I
Memory Causation
Chapter 2
The Motivation of the Causal Theory of Memory
The idea behind the causal theory of memory is that a claim to remember something implies, not merely that the rememberer represented it in the past, but that his current representation is in some way due to, that it came about because of, his past representation. The memory state must not only correspond to, but must also stand in an appropriate causal relation to, his past representation. The goal of this and the subsequent two chapters is to motivate and explicate the causal condition of memory. Since it is the same causal process that underlies meta-representational and reproductive memory, the meta-representation/reproduction distinction will be of no importance in what follows. The idea that remembering involves a causal process is far from novel. The majority of philosophers hold that memory calls for a causal condition. Among the proponents of a causal interpretation of memory are G.E.M. Anscombe, David Armstrong, B.S. Benjamin, Alvin Goldman, Wolfgang von Leyden, and Sydney Shoemaker: I think that it would be true to say that the everyday view of remembering is simply that it is the final stage of a causal process and that the memory is some sort of causal device or mechanism (Benjamin 1956, p. 323). It is part of the meaning of memory that, when it is correct, it is causally dependent upon a previous perception…. [There is] a memory process or causal chain stretching continuously from the occurrence of x and the original experience of x up to the present recollection of x (Leyden 1961, pp. 31, 42). Remembering, like perceiving, must be regarded as a causal process. S remembers p at time t only if S’s believing p at any earlier time is a cause of his believing p at t (Goldman 1967, p. 360). [A] claim to remember a past event implies, not merely that the rememberer experienced such an event, but that his present memory is in some way due to, that it came about because of, a cognitive and sensory state the rememberer had at the time he experienced the event…. [M]emory must not only correspond to, but must also stand in an appropriate causal relationship to, a past cognitive and sensory state of the rememberer (Shoemaker 1970, p. 272). The original witnessing of a remembered event is a cause of any present memory (even a false memory) of that event (Anscombe 1981b, p. 130).
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
17
18
2 The Motivation of the Causal Theory of Memory When we remember an event, that event must play a part in bringing about the later recall (Armstrong 1987, p. 464).1
Though there is widespread agreement that memory requires some kind of causal process it is far from clear how memory causation should be construed. There are numerous causal accounts of memory which differ from one another in two respects. First, some theories claim that memory causation is contiguous while others deny it. Second, different theories of memory propose different interpretations of causal dependence. Among the interpretations of the causal dependence of memory are analyses in terms of necessary and sufficient conditions, as well as counterfactual and probabilistic analyses. Before we can discuss the exact nature of memory causation we need to concern ourselves with a more fundamental issue: is it indeed necessary to analyze memory as a causal process? Is a causal condition – however it might be spelled out – indispensable for explaining memory? This is the issue the present chapter is concerned with. The philosophical literature on memory contains five direct arguments that are supposed to either confirm or refute the causal theory. I will show that none of them is successful. The only successful argument in favor of a causal interpretation of memory is an indirect argument to the effect that alternative accounts of memory are flawed or at least considerably less plausible than the causal theory. Sections 2.1–2.4 discuss direct arguments for and against the causal theory. In sections 2.5 and 2.6 non-causal accounts of memory are explained and evaluated.
2.1
Two Arguments Against the Causal Theory of Memory
In his influential paper “A Definition of Factual Memory” Norman Malcolm develops two arguments against the causal interpretation of memory – an argument from logical dependence and an argument from the nomological character of causality. The aim of this section is to critically examine both arguments. Malcolm’s first objection against the causal interpretation of memory is pointed at in the following remark: “One might object to the idea that the supposed effect is causally dependent on the supposed cause, for the reason that the ‘effect’ is logically dependent on the ‘cause’” (1963, pp. 231–232). What Malcolm seems to have in mind is an objection inspired by David Hume which can be summed up as follows2: The relation between cause and effect is a metaphysical relation rather than a logical one. For, if all one knew was the effect, one could not infer the cause, or vice versa. In the case of remembering, however, the present representation is logically dependent on the past
1 The causal interpretation of memory is also endorsed by Ayer (1956, pp. 135–136), BraddonMitchell and Jackson (1996, p. 7), Dancy (1985, p. 195), Pojman (1995, p. 223), Saunders (1965a, p. 283), Smith (1966, pp. 80–81), and Wiggins (1967, pp. 43–58; 1980, pp. 219–220). 2 I follow Siebel’s (2000, pp. 157–158) reading of Malcolm.
2.1 Two Arguments Against the Causal Theory of Memory
19
representation from which it causally derives. From “S remembers that p” it follows with logical necessity that “Sometime in the past, S represented that p.” Thus, Malcolm concludes, the kind of relation characteristic of memory cannot be causation. The problem with Malcolm’s reasoning is that it confuses events with their descriptions. The truth of a causal statement depends on the description used to pick out the causally relevant properties of the event involved. For example, the statement “S is intoxicated because he drank a liter of clear liquid” is (at best) inconclusive since it doesn’t specify the causally effective elements of the event which forms the cause. What caused S’s intoxication is not his ingestion of a liter of clear liquid, but the ingestion of a liter of, say, gin. Any two causally related events can be described so that one is logically dependent on the other. To achieve this all one has to do is to include the description of the cause in the description of the effect, or vice versa. For example, just because “Oscar is angry due to Bert’s rude remark” logically implies “Bert made a rude remark” doesn’t mean that Bert’s remark is not the cause of Oscar’s anger. Notwithstanding that the two events (Bert’s remark and Oscar’s anger) can be described so that one implies the other, one is the cause of the other. Hence, Malcolm’s contention that the relation between cause and effect may not be a logical relation is incorrect.3 (I will return to the issues discussed here in section 5.5.) Malcolm’s second objection to the causal theory of memory concerns the nomological character of causality. Already Hume held that wherever one event causes another, there is a causal law which relates types of events of which the two particular events are instances. In Hume’s words, “we may define a cause to be an object, followed by another, and where all objects similar to the first are followed by objects similar to the second” (1748, p. 76). Given the nomological character of causality and given the causal interpretation of memory, it follows that memory is a law-governed process. But, Malcolm holds, by making causal claims about memory we don’t thereby commit ourselves to there being causal laws. He declares: There is an important sense of “cause” in which a singular causal statement of the form “x caused y” implies a general proposition of the form “in like circumstances, whenever x then y.” But this meaning of “cause” cannot be involved in factual memory, since in saying that someone remembers that p, we are certainly not committing ourselves to the truth of the general proposition that “In like circumstances, whenever a person has previously known that p then he knows that p” (1963, p. 232).
And Carl Ginet goes one step further than Malcolm by challenging the truth of the generalization “In like circumstances, whenever a person has previously known that p then he knows that p.” According to Ginet, the concept of memory is a counterexample to the assumption that any sort of causal connection requires that there be general laws under which its instances are subsumable:
3 Malcolm could try to restate his point in this manner: Whenever event A causes event B, there will be available descriptions of both events which do not imply one another. As far as I can see this principle is valid.
20
2 The Motivation of the Causal Theory of Memory It seems clear that a particular experience and later matching memory-impression could satisfy the condition I have given for memory-connection without there being any true nomic generalization of the form “Whenever a person has such-and-such an experience, and certain further circumstances obtain (then or thereafter), then the person at suchand-such later time has such-and-such a memory-impression” (1975, pp. 166–167; 1988, pp. 166–167).
Unfortunately, Ginet doesn’t explain why he is skeptical of the possibility of causal laws regarding memory. Presumably he holds that it is hopeless to try to predict whether a person will later on remember a representation he is having now. If at all, such predictions can be made only for a particular individual in a particular situation. Now, I grant Ginet that it would be very difficult to come up with a causal law that warrants predictions concerning what and when someone remembers. However, this strikes me to be merely a problem of complexity. I don’t see why it should be in principle impossible to develop such laws of memory. Regardless of Ginet’s scruples concerning the possibility of laws of memory, psychologists have known for more than a century that remembering is governed by lawlike regularities. Hermann Ebbinghaus (1885), the pioneer of scientific studies on memory discovered that the effects of practice on memory retrieval are regular. To study the relationship between practice and learning, Ebbinghaus taught himself series of nonsense syllables, consonant-vowel-consonant triplets such as jev, zud, and vam. He employed lists of 16 such syllables which he practiced for 0, 8, 16, 24, 32, 42, 53, or 64 repetitions, and then 24 hours later measured what he remembered by seeing how many additional trials were necessary for him to completely master the list to a point at which he could recite it through without error. What is interesting about these results is that the relation between the amount of time spend learning on the first day and the additional learning time on the consecutive day is extremely regular: The amount of learning is directly proportional to the time spent in learning. A single repetition on the first day lasting less than 7 seconds gives a saving of over 12 seconds on re-learning on the second day.
2.2
An Argument from Conditionals
After having examined two arguments against the causal theory of memory let’s look at an argument in favor of memory causation. In their seminal paper “Remembering” Charles Martin and Max Deutscher develop such an argument. They rest their case on two thought experiments: suppose a painter paints what he takes to be an imaginary scene of a farmyard. The painter’s parents see the finished painting and recognize it as an accurate representation of an actual farmyard, often visited by the painter as a boy. Martin and Deutscher claim that while it is not true to say of the artist that he believes that the picture is a picture of a house he has seen in the past – for he takes himself to be inventing the scene on the canvas – it is true
2.2 An Argument from Conditionals
21
to say of him that he remembers the farmyard – since his remembering it is the best explanation for his correctly representing it. They declare: The onlookers were compelled to the conclusion that the painter was remembering something he saw in childhood. It would have been unreasonable for them to think that he would have done what he did if it had not been for some particular past observation…. [T]he only reasonable explanation of the fact that the painter put details, colors, people and so on into his picture, just as he saw them only once in his childhood, is that he was remembering that scene from his childhood (1966, pp. 175–176).
And Martin and Deutscher give a second example: suppose a person has apparent memories of something from early childhood, and wonders whether he really remembers it. There is no doubt that what he describes did happen and that he witnessed it but he wonders whether his now giving the story could be completely explained by what he later heard from his parents about his childhood. Whether he has been told about it in the meanwhile, how young he was at the time, and whether he has seen things very similar at many other times are all relevant to deciding whether he actually remembers the event. These facts are the same as those which are used to decide whether or not he would have given the story if he had not witnessed the event in his childhood. To decide that he would not have done so it to decide that his past witnessing is causally necessary for his present account (1966, p. 176).
Martin and Deutscher’s interpretation of these thought experiments can be schematized as follows: (1) If S had not represented that p in the past, he would not represent that p now. (2) S’s past representing that p in the past is a causally necessary for his present presenting that p. (3) Facts relevant to supporting (1) are the same as those relevant to supporting (2). The crucial step in this reasoning is the third premise which states that reasons for the truth of the counterfactual statement (1) are ipso facto reasons for the truth of the causal statement (2). Shoemaker makes this step explicit when he writes that he “can see no reason for doubting that counterfactuals [such as ‘If S had different experiences at t1 he would have corresponding differences in what he remembers at t2’] assert causal connections” (1970, p. 282). It seems that what Deutscher, Martin, and Shoemaker claim is that by explaining the occurrence of memory states in terms of counterfactuals we are committed to the idea that there is a causal connection between the original thought and its subsequent recall. The above argument for the causal theory of memory fails because counterfactual statements are not equivalent to causal statements. Even if one’s present retelling of a past event is counterfactually dependent on one’s previous having witnessed that event, this doesn’t have to mean that the witnessing is the cause of the retelling. The sort of dependency expressed by counterfactuals is considerably broader than strictly causal dependency and causal dependency is only one among the heterogeneous group of dependency relationships that can be expressed by counterfactuals. To drive this point home consider examples of counterfactual dependencies that do not warrant the stipulation of a causal process: “If Oscar had not been born in 1950,
22
2 The Motivation of the Causal Theory of Memory
he would not have reached the age of 21 in 1971” and “If my sister had not given birth, I would not have become an uncle.”4
2.3
An Argument from Personal Identity
Shoemaker’s paper “Persons and their Past” contains another argument to the effect that a causal component in the analysis of remembering is essential. Shoemaker’s strategy is to argue that unless we understand the notion of memory as implying a causal (or quasi-causal) process, we cannot account for the role played by the notion of memory in judgments of personal identity. The argument is based on the following hypothetical case: Let us suppose that the brain from the body of one man, Brown, is transplanted into the body of another man, Robinson, and that the resulting creature – I call him “Brownson” – survives and upon regaining consciousness begins making memory claims corresponding to the past history of Brown rather than that of Robinson. We can also suppose that Brownson manifests personality traits strikingly like those previously manifested by Brown and quite unlike those manifested by Robinson…. I think that most people would want to say that Brownson is (is the same person as) Brown (1970, p. 282).
Shoemaker maintains that it is natural to take Brownson’s autobiographical “memories” of Brown as evidence of his numerical identity with Brown. Our intuition that Brown is the same person as Brownson, Shoemaker argues, presupposes a causal analysis of memory and counts against any non-causal notion of retention. Why is this so? Let’s suppose, for the sake of the argument, the reason Brownson has autobiographical “memories” of Brown is that Brownson non-causally retained experiences acquired by Brown in the past. For Brownson to be able to non-causally retain experiences of Brown, Brownson and Brown have to be numerically identical. But if Brownson’s retaining experiences of Brown presupposes personal identity, then memory – analyzed in terms of non-causal retention – cannot, without circularity, be offered as evidence for the personal identity between Brownson and Brown. Thus,
4 Kim 1973, p. 571. Squires (1969, p. 181) points out that sometimes conditional statements are used not to assert but to rule out a causal explanation. Consider the statement “If the curtains had not been indigo yesterday, they would not have been indigo today.” This counterfactual, Squires maintains, is significantly different from the causal condition “the curtains are indigo today, because they were indigo yesterday.” The counterfactual formulation “indicates directly that nothing has happened since yesterday…. It does not assert a causal connection. Rather, it denies that certain kinds of causal explanation are necessary,” by correcting the impression that the curtains became indigo since yesterday or that the color has faded since yesterday. Furthermore, Malcolm claims that even when we say things like “S remembers that p because he knew that p” the “because” doesn’t refer to a causal connection. It has, he argues, an essentially negative function. Its force is to rule out other reasons why the person in question may now know that p – reasons such as that he has re-learned or been told that p. “To say that A knows that p because he previously knew that p implies that A has not just now learned over again that p. This brings out, in part, the negative sense of the ‘because’” (1963, pp. 233–234; cf. Warnock 1987, pp. 45–46).
2.3 An Argument from Personal Identity
23
the non-causal retention-view of memory cannot provide a non-circular explanation of the idea that Brown and Brownson are numerically identical. The intuition that Brownson’s autobiographical “memories” of Brown indicate that Brownson is the same person as Brown calls for a causal connection between Brown’s past experiences and Brownson’s seeming memories of these experiences; so at least Shoemaker argues. Shoemaker’s argument hinges on the claim that a non-causal theory of memory does while a (quasi-)causal theory of memory does not presuppose personal identity. Let’s start with the latter part of Shoemaker’s claim, viz. the idea that quasicausal remembering does not presuppose the diachronic numerical identity of the rememberer. It is a common objection to the psychological continuity theory of personal identity, first raised by Joseph Butler, that it is circular: it is circular to appeal to autobiographical memory in explicating our numerical identity through time because autobiographical memory already presupposes personal identity.5 There are two aspects of the claim that autobiographical memory presupposes personal identity. First, there is the claim that the causation condition constitutive of remembering cannot hold in the absence of personal identity. The causal condition presupposes an identity of the subject remembering an experience and the subject who had the experience remembered. There is, second, the claim that remembering having had experience E is necessarily remembering oneself having had E. A reference to oneself, now remembering, enters into the content of the memory experience. I can remember having E in the past only if I in fact had E in the past. In response to the circularity objection, Shoemaker develops a strategy to explain the assumed identity between Brown and Brownson without employing a notion – like the notion of memory – which already presupposes personal identity. He coins a term of art, quasi-memory, which, unlike memory, does not imply personal identity and defines personal identity in terms of quasi-remembering. Leaving aside the details of Shoemaker’s account, if the notion of memory is replaced with that of quasi-memory, a non-circular explanation can be given of why we are inclined to take Brownson’s “memories” as evidence of his identity with Brown. Brown is identical with Brownson because Brownson’s memory claims are causally related to Brown’s past experiences in essentially the same ways as a normal person’s memory representations are causally related to his past experiences. As was explained above, Shoemaker argues that analyzing the personal identity between Brown and Brownson by means of a non-causal notion of retention of thoughts and experiences is circular since this notion presupposes personal identity. This is false. For just as one can define a notion of memory causation which doesn’t entail personal identity, namely, quasi-memory, one can define a notion of noncausal retention which doesn’t entail personal identity, one may call it quasi-retention (cf. Ginet 1975, pp. 169–171; Schumacher 1976). A person S quasi-retains an
5 For recent endorsements of the circularity objection, see Evans (1982, ch. 7), Noonan (1989, ch. 8), and Schechtman (1990).
24
2 The Motivation of the Causal Theory of Memory
experience if (i) there is a correspondence between his present state in which he has the representation of the previous experience and a past state in which the experience was acquired and (ii) the past state need not have been a state of S. Apart from condition (ii) the correspondence implied by quasi-retention is just like that which exists in the normal retention of thoughts and experiences. And since explaining the numerical identity of Brown and Brownson by means of the notion of quasi-retention is not circular, the notion of non-causal quasi-retention can be employed in an account of personal identity. Shoemaker’s argument in favor of the causal theory of remembering crumbles.
2.4
An Argument from Temporal Forgetting
In his paper “Memory Unchained,” Roger Squires puts forward two claims, first, that memory is the retention of knowledge (rather then merely true belief) and, second, that this retention need not be analyzed in terms of a causal process since remembering that p is simply one’s persisting ability or disposition to produce knowings-that-p. Here I am interested only in the second thesis – that the retention of knowledge characteristic of memory is non-causal. Remembering can be understood as a disposition (a power, ability, or propensity) or as a distinctive occurrence. When we say of S that he remembers that p we can either say of him that he is currently engaged in propositional remembering, or that he is disposed to remember that p if suitable circumstances occur. According to the dispositional usage of “remembering-that,” we can say of S that he remembers that p even while he is sound asleep. Squires, unlike myself, concentrates on the dispositional use of “remember.” According to Squires, the chief motivation of the causal theory of memory is the account of temporal forgetting. Frequently, we observe an event, thereby acquiring the ability to represent it, then forget about it, and later remember what we had forgotten. It looks as if we have thereby regained knowledge which we had lost in the meantime. To explain the appearance of discontinuity of knowledge we seem to be forced to stipulate a causal chain (consisting of traces) which spans the temporal gap between the lost and regained ability. Squires’ point is that the problem of temporal forgetting, which (on his reading) causal theorists believe compels us to adopt their analysis, is spurious and contrived. He maintains that the appearance of discontinuity in temporal forgetting depends on an ambiguity in the notion of retaining an ability. Once this ambiguity is brought to light, temporal forgetting can be explained without having to introduce an intermediary causal process. Squires invites us to consider the following thought experiment: suppose that a teddy bear has the capacity to squeak when pressed. Due to the age of the toy, this capacity is dependent on the atmospheric humidity. The bear can squeak in the dry summer but cannot squeak in the damp winter months. When we observe the bear throughout the course of a year, it might seem as if the bear’s capacity to squeak is discontinuous for it is lost in winter and regained in summer. However, Squires
2.4 An Argument from Temporal Forgetting
25
claims, the phrases “in winter” and “in summer” have an ambiguous role. On the reading of these phrases which gives rise to the appearance of discontinuity, they are temporal qualifications, such that the claim “the bear can squeak” concerns the toy’s ability to squeak whenever it is pressed. But Squires holds that this is not the only reading of the phrases “in winter” and “in summer.” Squires takes these phrases to qualify the ability, with the result that the ability to squeak in winter is not the same as the ability to squeak in summer. But then there really is a continuous possession of the retained ability, for even in winter the toy can squeak in summer. Squires concludes that “bridging the retention gap” by stipulating an intermediary causal process “is a superfluous piece of philosophical engineering” (1969, p. 184). Then Squires applies this point to the retention of knowledge. He declares that “when we forget something, and then remember it, the capacity we lost is not the capacity we keep” (1969, p. 186). What he presumably means is that the difference between permanent and temporal forgetting is the difference between losing the capacity to know that p given any circumstances and losing the capacity to know that p given particular circumstances. Temporal forgetting is not the loss of a relevant ability tout court, but only the loss of a relevant ability in certain circumstances. The appearance of discontinuity in temporal forgetting is due to the absence of the circumstances on which one’s ability to know that p is dependent.6 Squires’ non-causal account of temporal forgetting is not convincing. The problem is that it renders the difference between temporal and permanent forgetting meaningless. Any ability we have and retain we have only in particular circumstances. Since none of our abilities are such that we have them in any circumstances, we never forget permanently. All cases of human forgetting are instances of temporal
6 Squires illustrates his point with another example. It is an example which Martin and Deutscher (1966, p. 174) have invented to motivate their causal analysis of memory. Oscar experiences a car accident. His memory of the accident is wiped out by a second car accident that Oscar is involved in shortly after the first accident. At this point, Oscar doesn’t remember anything about the first accident. Again, shortly afterwards, a hypnotist produces in Oscar the belief that he had been in a car accident at a certain time and place. By sheer coincidence the hypnotist’s description of the accident matches Oscar’s first accident. After having been hypnotized Oscar correctly believes that he has been in a car accident of a certain kind at a certain time and place. But does he remember having been in a car accident? According to the causal theory of memory, Oscar’s belief concerning the accident fails to qualify as memory because it wasn’t caused by his original experience of the accident. Since Squires rejects a causal analysis of remembering, he needs to provide a different explanation for why Oscar’s belief of the accident doesn’t classify as memory. According to Squires, Oscar does not remember the accident because, after the hypnosis, “he does not possess the same ability as he had before, for several reasons. His post-hypnotic ability does not date from the time of the accident. It is unlikely that he will be able to recall the accident in the same varied ways as before. Indeed, it is doubtful if he knows what occurred, despite his correct narrative. He has done the right thing for the wrong reason, like someone who mistakes a picture for a window and reports (as it happens) correctly that there is an oak tree outside” (1969, pp. 190–191).
26
2 The Motivation of the Causal Theory of Memory
forgetting. Therefore, it is hardly informative to be told by Squires that we can lose a capacity we never have – the capacity to know that p, given any circumstance.7
2.5
Critique of Pure Retention Conditions
The job of the causal condition of memory is to exclude re-learning and guarantee that the representation had in remembering is a retained representation. We saw that none of the arguments found in the literature succeed in either confirming or refuting the causal theory of memory. Thus apparently the only way of vindicating the causal theory is by showing that alternative accounts of memory retention are flawed or at least considerably less plausible than the causal theory. The alternative accounts of memory retention are the pure retention condition and the evidential retention condition. The aim of this and the following section is to demonstrate that these non-causal accounts of memory are not viable alternatives. Advocates of the pure retention condition and the evidential retention condition sign up to the epistemic theory of memory, that is, the view that memory is retained (or long-standing) knowledge, that is knowledge which was previously acquired and preserved. In his classical monograph Memory, Don Locke notes that the epistemic theory of memory “has come to be the standard contemporary account of the nature of memory” (1971, p. 39). Robert Audi, for example, claims that “if you remember that we met, you know that we did. Similarly, if you remember me, you know me” (1998, p. 67). (I will return to the epistemic theory of memory in section 7.2.) Given the epistemic theory, fact memory consists of three conditions: S remembers at t2 that p only if, first, S knows at t2 that p, second, S knew at t1 that p and, third, S’s knowledge at t2 that p is retained rather than re-learned or re-acquired. When the epistemic theory of memory is combined with the causal theory of memory the retention condition reads something like this: (4) S knows at t2 that p because S knew at t1 that p. The pure and evidential retention conditions take the place of the causal condition (4). The pure retention condition comes in three different flavors:
7
Lewis 1983, p. 27. Annis (1980, p. 330), Don Locke (1971, pp. 62–63), and Deutscher (1989, p. 61) argue that Squires’ non-causal theory of memory is unable to account for prompted memory after a period of temporal forgetting. Don Locke, for example, writes: “But if someone reminds me of some fact I had completely forgotten, and would not have remembered if he had not reminded me, it does not seem correct to say that I possessed that knowledge continuously. Something has been retained, something that enables me to say when reminded, “Ah yes, I remember now.” But that something does not seem to be a piece of knowledge, because, before I was reminded, I no longer knew the fact in question” (1971, pp. 62–63). This objection is unfair for it assumes that the possession of an ability implies that one can employ it at any time and in any context. Yet, Squires claims that there is a context-sensitive notion of abilities and capacities. Squires could respond to Locke’s objection by saying that what I retained was the ability to know in the circumstance of being prompted in a certain way.
2.5 Critique of Pure Retention Conditions
27
(5) At t2 S has not forgotten that p. (6) There is no time t1,5, between t2 and t1, when S did not know that p. (7) At t2 S believes that at t1 he knew that p. The prototype of the evidential retention condition reads as follows: (8) S’s grounds for knowing at t2 that p are identical to S’s grounds for knowing that p at t1. Let’s start discussing the pure evidential conditions (5)–(7). Condition (5) goes back to Gilbert Ryle. Ryle doesn’t offer an explicit definition of memory. Yet, in his discussion of memory in The Concept of Memory he notes that “remembering something means having learned something and not forgotten it” (1949, p. 272). If this passage may be read as an attempt to define fact memory, we get condition (5). Having “not forgotten” that p serves to rule out as fact remembering those cases in which a person comes to know that p, forgets it, and then re-learns it afresh. There are two reasons why condition (5) fails. First, there are instances where a person fact remembers something at t2 but yet has forgotten it in the interval between prior learning at t1 and t2. Second, condition (5) may contain a vicious circularity, when seen in the context of a definition of fact memory. The notion of remembering is defined by way of appeal to the notion of forgetting. But it seems plausible that the notion of forgetting can be adequately defined only by way of appeal to the notion of remembering. Condition (6) which is Stanley Munsat’s (1967, pp. 33–36) contribution doesn’t fair any better than (5), for it is too stringent. Condition (6) is too stringent because it cannot account for instances where a person remembers that p but has forgotten that p in the interval between the prior and the present knowing. Yet, we frequently forget something temporarily but are able to remember it later. Condition (7) stems from E.M. Zemach (1968, p. 529). The problem with this condition is that it is both too stringent and two liberal. It is too stringent for it rules out cases where someone remembers that p but fails to believe that he previously knew that p. Maybe there are good reasons for excluding cases in which a person knows that p from memory, but is unaware of having known that p previously, but Zemach doesn’t tell us what they are. Condition (7) is also too liberal in that it counts certain instances of re-learning as instances of fact memory. Suppose at t1 I learn that p. Then I completely forget this fact. At t2 I read in a book and again come to know that p. Moreover, I come to believe that I once knew that p. I am not reminded that p for I have completely forgotten it and now have learned it afresh. But I knew that p at t1, know it at t2, and believe at t2 that I knew it at t1. Zemach’s conditions for remembering are fulfilled, but I don’t remember. But even if there was a pure retention condition that overcomes all kinds of counterexamples I would still not be convinced. The reason is that pure retention conditions appear coherent only as long as one doesn’t ask what is involved in the retaining of a piece of knowledge. When this question is raised, proponents of this approach have to concede that their “analysis” cannot provide an answer. But if the process underlying retention of knowledge cannot be explained, memory
28
2 The Motivation of the Causal Theory of Memory
becomes a magical faculty. For, as Deutscher declares, in claiming “the continuity of capacities, [we] are always committed to the continuity of some processes adequate to the continuity” (1989, p. 62). And as far as I can see, the causal theory of memory is the best candidate to account for the process underlying retention of knowledge.
2.6
Critique of Evidential Retention Conditions
For knowledge to have been retained from t1 to t2, what one knows at t2 must, in some clear way, be the same thing one knew at t1. The evidential retention condition attempts to explicate the idea of sameness of knowledge in terms of sameness of reasons or justification. “Sameness of knowledge, if it is to count as memory, requires sameness of grounds” (Holland 1974, p. 362). The idea is that for a mental state at t2 to classify as memory it has to be a piece of knowledge and its justificatory factors must be the same as those supporting the original piece of knowledge that has been retained from t1. In other words, for S to remember that p he must know at t2 that p, he must have known at t1 that p, and his reasons for believing that p at t1 must be the same as the reasons for believing that p at t2. The evidential retention condition is a relatively recent development, finding first expression in a paper by Andrew Naylor (1971). Apart from Naylor, David Annis (1980), Alan Holland (1974, pp. 362–371), John Pollock (1974, pp. 198– 203, 1986, pp. 46–58), and Shoemaker (1967, pp. 271, 272) propose versions of the evidential retention condition.8 The biggest obstacle to the evidential approach to memory is to specify what it means for two pieces of knowledge to be supported by the same grounds. Under what conditions has a person the same grounds he originally had? In a naive sense of the expression “same grounds,” a person remembers only if the grounds supporting his true belief at t2 are of the same number and kind as the grounds which supported his true belief at t1. Yet, this interpretation is open to counterexamples. It may happen that one’s reasons for believing that p at t1 are only a subset of one’s reasons at t2 because one acquired a few additional reasons in between t1 and t2. But then one’s grounds for believing that p at t1 and at t2 are not the same. The grounds at t2 include items not included in the set of original grounds. The evidential retention condition can be patched up to account for cases where sameness of grounds is violated because of an enlarged set of grounds. One can specify that the set of grounds for S’s knowing at t1 that p is the same as, or a subset of, the set of grounds for S’s knowing that p at t2. This is how Naylor defines the evidential retention condition: at t2 S remembers that p from t1 only if
8 Margolis (1977, pp. 191–192) erroneously claims that Malcolm is committed to an evidential retention condition.
2.6 Critique of Evidential Retention Conditions
29
(1) there is a set of grounds a subset of which consists of (i) only those grounds S has at both t1 and at t2 for being sure that p, and (ii) enough such grounds to make it reasonable at both t1 and t2 for S to be sure that p …, and (2) there is no time prior to t1 such that S has a set of adequate original grounds dating from that time (1971, p. 33).
Given this evidential retention condition, cases in which a person’s evidence for p at t2 outstrip his evidence at t1 can be reckoned as cases of remembering, provided that the original grounds are contained within the present grounds. But there remains another problem for the evidential approach. The above evidential condition doesn’t rule out certain re-learning-counterexamples. The evidential condition doesn’t rule out instances were a person knew at t1 that p on grounds G and again knows at t2 that p on grounds G, but fails to remember that p, since he had completely forgotten that p in the interval between t1 and t2. Yet, there are ways to get around this problem. Mark Sherouse (1979, p. 149), for example, proposes the following revision of the evidential retention condition which blocks re-learning counterexamples: S’s grounds for knowing that p at t1 are identical to, or a subset of, S’s grounds for knowing that p at t2, and, there is no time t1,5 between t1 and t2 when S didn’t know that p, and, there is no time t1,5 between t1 and t2 when S knows that p on grounds that are not identical to, or a subset of, his grounds for knowing that p at t2.
Notwithstanding Naylor’s and Sherouse’s amendments, the evidential approach is still open to counterexamples. For just as we can gain new reasons between t1 and t2, we can lose some, or even all, of our original grounds by receiving defeating evidence. It is possible that S knows at t1 that p, remembers at t2 everything he knew at t1, and yet fail to know at t2 that p – even if he continues to believe that p, and p continues to be true. The reason is that there might be new evidence encountered along the way that points to a relevant alternative S cannot exclude. To drive this point home, suppose that at t1 S met his father and noticed that he was wearing a blue pullover (Saunders 1965a, pp. 282–283). Because he then knew this fact about his father, S holds at t2 the (true) memory belief that his father wore a blue pullover. Now suppose that S is told by all who were present at t1 that his father wore a green pullover. Moreover, S is presented with convincing but misleading evidence to the effect that his father wore a green pullover. In this case, the piece of fresh evidence undermines S’s knowledge that his father wore a blue pullover. For even on a reliabilist conception of knowledge, to know that his father wore a blue pullover, S must be able to rule out the possibility that he wore a green one. And since S cannot exclude this possibility, he fails to know that his father wore a blue pullover, despite the fact that he remembers it. Thus, memory may transform knowledge into merely true belief. (I will return to this example of memory without present justification in section 7.2.) It is beyond doubt that both knowledge and memory imply truth. Just as you can know that p only if p is true you can remember only what is the case. “S remembers that p” entails that p is true. If not-p, then S may think he remembers that p, but cannot actually remember that p. Truth is a component of both knowledge and memory. But memory, unlike knowledge, does not imply justification. The above example illustrates that it is possible to remember p without being justified in believing p. The evidential retention condition fails because it rests on the epistemic theory of memory which is unconvincing.
Chapter 3
An Argument for Memory Traces
After it has been established that a causal condition – however it might be spelled out – it indispensable for explaining remembering, we can proceed by examining the exact nature of memory causation. According to the standard view, the causal process connecting a past representation and its subsequent recall involves intermediary memory traces (or engrams). The stipulation of memory traces is motivated by the contention that between any two diachronic mental events there have to be a series of intermediary events, each of which causes the next, and each of which is temporally contiguous to the next. However, in The Analysis of Mind, Bertrand Russell argued that since the physiological evidence for memory traces is not yet quite conclusive, it is prudent to come up with an account of memory causation that manages without the stipulation of memory traces. Russell’s name for this notion of memory causation is “mnemic causation”. According to the theory of mnemic causation, a past experience is directly causally active over a temporal distance; past experiences are proximal causes of states of recalling. The aim of this chapter is to refute the theory of mnemic causation and to establish that the stipulation of memory traces is indispensable for the analysis of memory. Sections 3.1 and 3.2 motivate the standard view that causality implies contiguity and explicate the notion of a memory trace. Section 3.3 is a discussion of the verifiability of memory traces. Section 3.4 provides a detailed account of Russell’s theory of mnemic causation. The concept of mnemic causation is defended against the widespread objection that direct causal action at a distance in time is impossible because a cause cannot operate when it has ceased to exist. Given that Russell’s notion of mnemic causation does not face any crucial difficulties and given that there is no conclusive empirical evidence for the existence of memory traces, how should we determine whether or not to stipulate memory traces? It is argued that the apparent tie between the theory of mnemic causation and the theory of contiguous causation can be resolved when memory traces are treated as theoretical constructs.
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
31
32
3 An Argument for Memory Traces
3.1
Contiguous Causation
Saying that a representation of p qualifies as a memory only if it is causally connected to a previous representation of p, raises the question of how one should conceive of the causal connection. David Hume is the dominant philosopher of cause and effect. He notoriously maintained that two events are related as cause and effect if they meet each of three individually necessary and jointly sufficient conditions: first, priority of cause to effect, second, contiguity in space and time, and, third, constant conjunction or necessary connection (1739, pp. 73–78). Each one of these conditions has spawned an enormous and still ongoing debate. Fortunately, we need to concern ourselves only with two of the three conditions. We can abstract from the first condition that a cause must precede its effect in time. For in the case of memory causation, it is indisputable that the cause can neither be simultaneous with its effect, nor temporally posterior to its effect. Regarding the condition of temporal and spacial contiguity, Hume declared in the Treatise of Human Nature, “nothing can operate in a time or place, which is ever so little remov’d from those of its existence” (1739, p. 75; cf. pp. 170, 173).1 The reason Hume was led to stipulate that contiguity is a necessary condition of causation is that if cause and effect were not contiguous, some factor could intervene and prevent the effect, even though the cause had occurred. As Hume himself noticed, we are often not aware of the continuous causal paths connecting cause and effect. A switch on the wall is some distance from the electric light overhead that it controls; pushing a button on an alarm clock makes it ring 7 hours later. Where contiguity appears to be lacking, Hume held that we find, upon closer examination, that they are connected by a chain of causes such that the effect is finally caused by an event that is contiguous with it. Tho’ distant objects may sometimes seem productive of each other, they are commonly found upon examination to be link’d by a chain of causes, which are contiguous among themselves, and to the distant objects; and when in any particular instance we cannot discover this connexion, we still presume it to exist (1739, p. 75).
1 A number of Hume scholars have argued that Hume did not take contiguity to be a necessary condition for causation (cf. Flage 1985a, pp. 179–186; Stroud 1977, pp. 43–44). Hume admitted that we do not get an impression of contiguity every time we observe a pair of objects that we take to be related as cause and effect. In such case, we only “presume” that there is contiguity nevertheless. In claiming that we may merely “suppose” contiguity to be essential to causation, “according to the general opinion,” Hume can be read to be giving little more than an enumeration of the common assumptions regarding causation. Nothing he said demonstrates that “A causes B” implies “A and B are contiguous.” Furthermore, in the Enquiry Concerning Human Understanding no appeal is made to contiguity in the definition of causation. Beauchamp and Rosenberg (1981, pp. 194–195), however, maintain that Hume used “succession” and “contiguity” as near synonyms. They argue that the fact that the Enquiry only incorporates succession into the definition of “cause” and makes no mention of contiguity does not mean that Hume believed that causation may not be contiguous.
3.1 Contiguous Causation
33
In other words, by making a distinction between remote and proximate causes, we may say that the remote cause is connected with the effect through a chain of causes, the last one being the proximate cause. And that the proximate cause is that event that is contiguous with and produces the effect. Hume concluded that contiguity is “essential” to causality. The causal path has no spatial or temporal gaps or breaks. Today, the prevailing view is still that causality implies contiguity. Ernest Nagel, for example, writes: The [causal] relation has a temporal character, in the sense that the event said to be the cause precedes the effect and is also “contiguous” with the latter. In consequence, when events separated by a temporal interval are said to be causally related, they are also assumed to be connected by a series of temporally adjacent and causally related events (1961, p. 74).
And A.J. Ayer declares, “it is fairly generally assumed that in the cases where the cause can be represented as an event which precedes the effect, the two events must be temporally contiguous” (1972, p. 135). Like any philosophical position, the thesis of contiguous causation is open to criticism. An important objection lodged against this thesis stems from quantum mechanics. Quantum mechanics seems to permit non-instantaneous action at a distance where no energy exists in the space across which the action occurs. But setting aside physics, of which I know nothing, there is another interesting objection to the thesis of contiguous causation. In his early paper, “On the Notion of Cause,” Russell developed the simultaneity paradox which is supposed to show that a cause cannot be temporally contiguous with its effect. The simultaneity paradox has the form of a reduction. Russell assumes that cause and effect are temporally contiguous and shows that when this thesis is conjoined with the idea of necessary connection, it entails that cause and effect are contemporaneous. The argument conceives of cause and effect as events that take time and are divisible into atomic units. Consider, for example, the breaking of a cup on the kitchen floor. The cause begins with the knocking over by one’s elbow, and encompasses the downward hurling until the impact. The effect begins with the first impact on the floor, and ends with distribution of the pieces of china on the floor. Russell maintains that a real cause takes place only in the instant directly adjoined to the instant in which the effect begins. And the real effect takes place only in the instant right after the final unit of the cause-event has ceased to exist. For, if there is the slightest interval between the cause and effect, “something may happen during the interval which prevents the expected result,” even though “the cause” had occurred.2 In the example at hand, if the cause were some event prior to the moment of impact on the kitchen floor (e.g., the knocking over), then any number of occurrences may intervene (e.g., the sudden disappearance of gravity),
2 “On the Notion of Cause” was Russell’s presidential address to the Aristotelian Society in 1912. In his Lowell Lectures delivered in Boston in 1914 he held essentially the same views.
34
3 An Argument for Memory Traces
altering the normal course of events so that the cup does not break. But if the effect may not have happened, then the purported cause is not a necessary condition for the occurrence of the effect and hence not a genuine cause. Hence, if causation is analyzed in terms of necessary conditions, cause and effect must be perfectly simultaneous; there may be no temporal interval between cause and effect. Prima facie it is easy to rebut Russell’s simultaneity paradox and to hold on to the idea of contiguous causation. All one has to do is to reject the Humean analysis of causation in terms of necessary conditions. When the thesis of contiguous causation is combined with, say, a probabilistic notion of causation, the simultaneity paradox evaporates. Given probabilistic causation, one event’s causing another does not require that the former determines the latter, but only that it makes it more probable than it would otherwise have been. However, this strategy of circumventing the simultaneity paradox misses the whole point. For the point of the simultaneity paradox, I reckon, is not to refute the doctrine of contiguous causation but to point out that contiguous causation presupposes a contentious claim about the nature of time, namely that time is discrete. To say that time is discrete is tantamount to saying that time has a granular structure, with there being a smallest quantum of time. To see that contiguous causation assumes the discreteness of time, suppose that time were infinitely divisible, like real numbers are. Assuming the continuity of time, there would be a time interval separating any two causally related events which are not simultaneous. And, during this interval, something could happen which prevents the effect from occurring, although the cause had occurred. Hence, arguing for the thesis of contiguous causation requires, among other things, an argument for the discreteness of time. The formulation of such an argument would, of course, go beyond of the scope of this study.
3.2
What Are Memory Traces?
Notwithstanding the difficulties surrounding the idea of contiguous causation, this idea forms the basis of most philosophical accounts of memory. It underlies all those theories of memory which explicate the causal process connecting a past experience and a subsequent recall by means of a continuous memory trace (or a sequence of traces, respectively). The trace hypothesis states that between any two diachronic mental events there is a series of intermediary events, each of which causes the next, and each of which is temporally contiguous to the next. A rigorous definition of the trace hypothesis would tell us exactly how short the time span between mental event A and mental event B has to be, for A to act as the direct cause of B. (In the case of short-term visual memory, the temporal distance between the original experience and the remembering may be as short as 0.25 seconds.) Common sense speaks in favor of the stipulation of memory traces. For how else can past representations (or experiences) act at a temporal distance, if not through a continuous trace (or a series of traces)? How can there be direct causes remote in space and time? If contiguity were not a necessary component of memory, remembering,
3.2 What Are Memory Traces?
35
it seems, would have to rely on a magical process bearing some resemblance to telepathy and clairvoyance. Furthermore, without causal continuity, a past thought or event would somehow have to track one’s spatiotemporal path, to ensure that it could, at any time, become causally active as one moves around. Such longdistance tracking of past thoughts or events seems unlikely to be “direct” in any intuitive sense. Memory traces are designed to account both for the propagation of information and for the production of states of recall. The stipulation of memory traces allows us to understand how past experiences can exert causal influence long after they have ceased to exist. Furthermore, by means of postulating traces, we can explain the transmission of information through time. Corresponding to the two aspects of causal processes – production and propagation – there are two distinct aspects of the notion of a memory trace: a mental and a physical aspect. Insofar as memory traces produce states of recall, they may be purely physical states. To account for the production-aspect of memory causation, it suffices to conceive of traces in purely physical terms. From a physicalist point of view, memory traces are structural modifications at synapses (i.e., the area where the axon of one neuron connects with the dendrite of another neuron) that affect the ease with which neurons in a neural network can activate each other.3 Insofar as memory traces communicate information, they have to be capable of bearing content, and hence, they have to be mental states. What kind of mental states are memory traces? The representational nature of traces varies depending on the kinds of memory they give rise to. Fact memory of previous fact awareness presupposes the possession of the relevant concept (cf. section 1.1). Therefore, it is reasonable to assume that the traces constitutive of fact memory of previous fact awareness are dispositional beliefs. Object-, property-, and event memory, however, do not require that the rememberer possesses the concepts necessary for expressing the memories in question. I could, for example, remember having seen Fido on the sofa, even if I didn’t possess the concepts dog and sofa. Due to my limited conceptual abilities, I wouldn’t know what it is that I remember, but I would remember the
3 Cf. Dudai (1989) and Squire (1987). Shoemaker (1970, p. 282), Wiggins (1980, pp. 219–220), and Williams (1973, pp. 75–76) argue that the causal theory of memory conflicts with the concept of a mind as an immaterial substance. According to the causal theory, for someone to remember something there must be some appropriately characterized causal chain which links his present representation to his past observation. The idea of this causal chain is unpacked into the idea of a physical trace. The crucial step of the argument is the claim that the idea of a structurally complex memory trace belonging to an immaterial substance is incoherent. Memory traces have to have a physical basis. The rest of the argument is a simple reduction. Let’s suppose dualism were true and we existed independent of our bodies. We could not have any memories of experiences dating from our embodied state, for there is nothing to play the role of the causal chain required by the causal theory of memory. A disembodied person could not be said to remember at all; and given that the ability to remember is an important aspect of what it means to be a person, it follows that it is impossible for a person to be disembodied. Hence, given the causal theory of memory, the suggestion that a person existed independent of his body is incoherent.
36
3 An Argument for Memory Traces
event nevertheless. My non-conceptual memory would allow me to discriminate this event from others, even if not by thinking or speaking of it as involving a dog and a sofa. Memory traces constitutive of object-, property, and event memories may contain non-conceptual information. Subconscious states capable of transmitting non-conceptual content are commonly called subdoxastic states.4 In sum, traces of fact-memory of previous fact awareness are dispositional beliefs, while traces of object-, property-, and event memory are subdoxastic states.5 It a basic assumption of cognitive science that there are three levels of explanation for any intentional process. First, there is the intentional level of everyday psychology, in which we talk about people’s beliefs, memories and other such intentional states. Second, there is the computational level which explains how intentional states are realized by means of computational operations. Finally there is the level of physical implementation or realization. The three-floor model of the mind is of course inspired by the distinction between the hardware, the software, and the system interface of a computer. After having discussed the intentional and the physical aspect of memory traces we still need to concern ourselves with the computational aspect? The computational account of traces depends on whether one endorses the classical or the connectionist approach. The classical approach holds that mental representations are symbolic structures that have semantically evaluable constituents and mental processes are rule-governed manipulations of them. This position lends itself particularly well to account for meta-representations of past contents (e.g., I remember that I believed that p). On the assumption that thoughts are syntactically structured, we can conceive of mental meta-representation in analogy to linguistic metarepresentation. Memories of one past mental states are to be conceived of in analogy to direct quotation. Connectionism has it that mental representations are realized by patterns of activation in a network of model neurons and mental processes consist of the spreading activation of such patterns. Given this model, information is not stored by a formulae in an internal code with a specific location. Rather information is encoded through a change in the strengths of connections between nodes. What is stored in memory is a set of changes in the instructions neurons send each other,
4 Other labels for subdoxastic states are “implicit knowledge,” “tacit knowledge,” “proto-knowledge,” “unconscious knowledge,” and “subpersonal states.” The problem with characterizing subdoxastic states as “unconscious” is that unconscious (or subconscious) states can, in principle, be made conscious while subdoxastic states cannot. The problem with the label “knowledge” for subdoxastic states is that the information transmitted by them may be both false and unwarranted. Knowledge, however, implies truth and justification. 5 Since remembering doesn’t imply knowing, Siebel concludes that subdoxastic states may transmit false information (2000, pp. 236–257). I disagree. Of course, subdoxastic states may propagate “misinformation,” yet the states of recall based on these subdoxastic states cannot classify as memories, for memory implies truth. It is not the truth-condition but the justification-condition which distinguishes remembering from knowing (cf. section 2.6).
3.2 What Are Memory Traces?
37
affecting what patterns of activity can be constructed from given inputs. Since many items of information can be represented over the same set of neurons and connections, connectionist networks store efficiently. Remembering occurs when an input “travels” through an already established activation pattern. Each memory is a product of the activity of the total neural network. As David Rumelhart and Donald Norman say, “information is not stored anywhere in particular. Rather it is stored everywhere” (1981, p. 3). An immediate consequence of connectionism is that memories are deeply sensitive to context. Context produces slight differences in activity patterns and the corresponding memory may be subtly different on different occasions (though these difference may not be noticed by the rememberer). An important argument in favor of connectionist models of memory stems from neurology. It is a common fact that brain damage may not result in a sudden loss of certain kinds of information but that the performance of the memory system becomes slowly worse. Psychologists refer to this phenomenon as graceful degradation. Connectionist models also exhibit smooth degradation in the face of a “lesion,” that is, the removal of processing nodes and alteration of connection weights. This suggests that memory traces are distributed across many different brain cells, rather than located in one specific cluster of cells (Rose 1992). One of the fundamental characteristics of memory is its reconstructive nature. Our recall of events and thoughts is frequently not literal; rather, we reconstruct memories of past events that contain inferences and these inferences, may be indistinguishable from real happenings. Assuming the connectionist model, it is easy to explain the reconstructive nature of our memory. Since memories are encoded in the connection weights, memories are shared over the same hardware. Retrieval is more a matter of reconstructing information than going to a discrete location to find it. It is therefore not surprising that connectionist networks are good in modeling pattern completion. Simple examples of pattern completion are when we identify letters and words even though they are presented only for a split second or they are presented incompletely. Connectionist networks have the astonishing capacity to take such an ambiguous stimulus and to quickly give an unambiguous response. As connectionist computers we don’t need to perceive every letter in a string of words to be able to read it. Consider, for instance, the following sentence: I cxn rxplxce xvexy txirx lextex of x sextexce xitx an x, anx yox stxll xan xanxge xo rxad xt – ix wixh sxme xificxltx (Anderson 1995, p. 62). A feature of our memory related to pattern completion is content addressability. For example, if I tell you that I am trying to remember the name of J.F. Kennedy’s blond girlfriend who later married Arthur Miller, you can, no doubt, provide a great deal of associated information in addition to the name “Marilyn Monroe.” You may, for example, remember the missile crisis, Kennedy’s assassination, and the movie Some Like It Hot. This way of getting into memory, by matching the current contents of experience to similar contents in memory and then retrieving associated information, is our primary way of remembering. Given connectionism, the content addressability of memory is due to the fact that the incoming pattern of activation has matching parts to a previous pattern and that this is sufficient to reactivate other parts of the pattern. Content addressable memory is characteristic of humans, but
38
3 An Argument for Memory Traces
hard to achieve in classical architectures, where items are typically accessed on the basis of knowing in what register they were stored. Finally, a feature of human memory which connectionism is particularly apt in accounting for is cross-talk. Cross-talk occurs when, in an attempt to activate a particular memory trace, similarity to another trace leads to the alternative trace being reinstated rather than the “intended” trace. An example of cross-talk is when someone goes upstairs to his bedroom with the intention of changing out of his T-shirt into a pullover but instead undresses completely (cf. Reason and Mycielska 1982). Such errors are expected to emerge from a connectionist model of memory where a net for a planned action overlaps with the net for a very familiar yet different sequence. The activation of the less familiar trace might trigger the more familiar and so more strongly weighted trace.
3.3
The Verifiability of Memory Traces
After having explained the notion of a memory trace we can turn to Russell’s critique of the trace hypothesis. Two of the fifteen lectures that constitute Russell’s Analysis of Mind deal with the issue of memory traces. Both of these lectures are, in large part, a response to the work of the zoologist Richard Semon, whom Russell regarded as “the best writer on mnemic phenomena” (1921, p. 83). Semon was a passionate advocate of the trace hypothesis. Although he admitted that science had not progressed enough to say what memory traces are and how they work, Semon was deeply convinced that traces consist in something organic, in “a material alteration” (1909, pp. 138–139). Commenting on this aspect of Semon’s position, Russell writes: Concerning the nature of an engram, Semon confesses that at present it is impossible to say more than that it must consist in some material alteration in the body of the organism…. It is, in fact, hypothetical, invoked for theoretical uses, and not an outcome of direct observation. No doubt physiology, especially the disturbances of memory through lesions in the brain, affords grounds for this hypothesis; nevertheless it does remain a hypothesis (1921, p. 85).
This passage illustrates Russell’s acute awareness of the fact that, in his time, the trace hypothesis was empirically underdetermined. Russell holds that the empirical evidence for traces is “not quite conclusive” (1921, pp. 86, 92), but he regards it to be “quite possible” that some day physiological memory traces will be discovered. Given “the present state of physiology,” Russell writes, “the introduction of the engram does not serve to simplify the account of mnemic phenomena,” and it is therefore prudent to settle for a less speculative account of memory causation, an account formulated “in terms, wholly, of observable facts.”6
6
1921, p. 85. “I do not wish to urge that [mnemic] causation is ultimate, but that, in the present state of our knowledge, it affords a simplification, and enables us to state laws of behavior in less hypothetical terms than we should otherwise have to employ” (ibid.).
3.3 The Verifiability of Memory Traces
39
Russell’s point is that if traces exist, they must be shown to exist and not simply postulated to support the account of memory causation. As long as traces cannot be proven empirically, one should refrain from assuming their existence. And this is exactly what Russell did in The Analysis of Mind. Because of the inconclusiveness of the physiological evidence in favor of memory traces, Russell sees the necessity to develop a notion of memory causation that manages without the stipulation of memory traces. Russell’s name for this notion of memory causation is mnemic causation. Before examining the concept of mnemic causation in the next section, I want to take a closer look at the contention that the physiological evidence for memory traces fails to be conclusive.7 Russell is right in maintaining that brain-injured patients tend to have deficits that are rarely sufficiently pure, or specific enough, to draw any interesting conclusions regarding the locations of the memory system in the brain. Granted that the disturbances of memory through brain damage fail to provide conclusive evidence for the trace hypothesis, what would constitute conclusive evidence? Three types of evidence suggest themselves: neurobiological, introspective and conceptual evidence. Russell presumably holds that for neurobiology to prove the trace hypothesis, it has to establish the identity between memories, on one hand, and structural modifications of the synapses, on the other. A neurosurgeon could then, by artificially structuring synapses, bring about certain memories; and by removing those synapses he could erase those memories. In the 1940s the neurosurgeon Wilder Penfield carried out brain operations on epileptic patients in order to relieve their intractable seizures. Patients were first fully anaesthetized; the appropriate area of the skull was then removed, and the brain exposed. Consciousness was then restored, with only a local anaesthetic being maintained. During the operation, it was necessary to stimulate the surface of the brain with an electrode. Penfield noted that electrical stimulation in the temporal lobes resulted in patients having memory flashbacks. One patient said, “I just heard one of my children speaking … it was Frank, and I could hear the neighborhood noises,” and another, “something brings back a memory. I can see the Seven-Up bottling company.”8 In general, these recollections were vivid, detailed, and concerned with seemingly insignificant past events. Though suggestive, Penfield’s experiments fail to provide conclusive evidence for the hypothesis of physical memory traces. First of all, of the 520 patients who
7 It is not entirely clear whether Russell develops the theory of mnemic causation as an abstract possibility or whether he takes it to be true. Passages in lecture 4 of The Analysis of Mind suggest that he adopts mnemic causation while passages in lecture 15 suggest that he endorses the trace theory. Pincock (2006) tries to give an explanation for Russell’s wavering stand towards the theory of mnemic causation. Apparently Russell read Semon’s work while writing the lectures that were later published as The Analysis of Mind. The tension within The Analysis of Mind could be due to Russell’s progressive understanding of the trace theory. 8 The full transcript of the recordings made during electrical stimulation of the temporal lobes can be found, for example, in Rosenfield (1988, pp. 201–212).
40
3 An Argument for Memory Traces
received electrical stimulation in the temporal lobes, only 40 reported having memory flashbacks. Second, Penfield failed to test whether repeated stimulation of the same area of the lobes gave rise to the same type of memory. Third, subsequent studies have shown that such memory flashbacks occur only when the limbic structures (generally believed to be essential for emotional experiences) are activated (cf. Gloor et al. 1982). Since the 1970s, a number of fascinating non-invasive diagnostic methods for studying brain operations have been developed. These techniques make it possible to “observe” in vivo how the human brain subserves cognition. For example, positron emission topography (PET) and functional magnetic resonance imaging (fMRI) measure changes in blood flow in various regions of the cortex. The underlying assumption is that areas of particular high blood flow are especially involved in a particular mental activity. Despite the greatly improved methods for studying brain processes, it hasn’t been possible to establish a one-to-one correspondence between neural activation patterns, on the one hand, and psychological attitudes and intentional contents, on the other.9 Among other things, the problem is that the data delivered by neural imaging are not as fine-grained as the attitudes and contents of our mental states. None of the available techniques can account for the difference, say, between the memory that triangles have three sides and the belief that triangles have three angles. Given today’s science, it is not possible to rule out the possibility that two memories (had by the same subject) which differ in content supervene on the same brain process and thus are realized by the same trace. As long as such cases cannot be ruled out, Russell’s claim that empirical evidence for the trace hypothesis is inconclusive, stays valid. Granted that third-person data of neurobiology don’t provide reliable evidence for the existence of memory traces, maybe first-person data from introspection can do the job.10 Yet we do not have introspective access to our brain states. If anything, it is the intentional contents of memory traces that can be detected via introspection. But is it really the case that we have direct introspective access to the contents of our memory traces? Are memory traces transparent to the mind? The answer is negative. Memory traces can give rise to conscious and introspectable memory states, but they are removed from consciousness. They are the opaque entities that explain the coming about of (potentially) transparent memory states. What we are able to become aware of, are not the traces, but the states of recall they give rise to.
9 Cacioppo et al. (2003) pp. 654 and 655 calls it a “category error” to assume “that the organization of cognitive phenomena maps in a 1:1 fashion into the organization of the underlying neural substrates.” It is “not necessarily true” that “when a region of differential brain activation is active during a specific information-processing operation, this brain region is the neural substrate for the information-processing operation.” 10 My notion of introspection is not committed to any particular cognitive process by which a person becomes aware of his internal states.
3.3 The Verifiability of Memory Traces
41
Memory traces are opaque intentional states that represent past events or experiences, and, when activated, can give rise to conscious thought and conscious behavior. An immediate consequence of the fact that traces are removed from consciousness is that I cannot tell, by reflection alone, whether the state I am in is a state of remembering rather than of imagining. For to know this I would have to be able to rule out the possibility that the state I occupy wasn’t caused by a memory trace. But this I cannot do by reflection if traces are in principle unconscious. (We will return to the issue discussed here in chapter 6.) Since neither the third-person data provided by neurobiology, nor the firstperson data of introspective reports, provide conclusive evidence in favor of the trace hypothesis some philosophers have tried to establish the existence of traces by an a priori argument. They argue that the concept of a trace is implied by the notion of remembering. Among those who hold that the existence of memory traces is knowable a priori is Martha Kneale, who writes: It is involved in the ordinary notion of memory or recollection that the memory event should have as a part-cause the occurrence of the event recollected. This is what makes it so easy for us to accept the story of brain traces as a physiological condition of remembering. They fill in the gaps in the causal chain which is felt to be necessary to explain recollection (1972, p. 2).
It takes only a little thought though to see that the concept of a trace is not implied by the concept of remembering. Our language of memory makes no reference to physical traces. Traces are not what we mean when we talk about remembering, any more than the secretion of digestive juices is (part of) what we mean by “eating,” even though this is something that takes place when we eat.11 How else is it possible that children learn what “remembering” and “eating” mean long before they learn anything about biology? Traces are a feature of the dominant hypothesis about memory, rather than a feature of the concept of memory. What is implicit in the language of memory is only that there be some form of causal link between the past and the present representation. Just what sort of link it is has to be established by other means.
11
This analogy is borrowed from Malcolm (1963, p. 237). The critique of the idea that memory traces are knowable a priori echo some early objections to the psycho-physical identity theory. When the identity theory was first advanced, it was objected that if mental states are identical with brain states how had this fact eluded attention for so long (cf. Smart 1959, p. 152). Identity theorists replied by pointing out that psycho-physical identities are something we discover from observation and experience, not something that could be ascertained a priori or by merely investigating the meaning or concepts involved. The concept of, say, pain and the concept of C-fiber excitations are distinct and independent concepts, and this explains how it is possible for someone to know a lot about pains but nothing about C-fiber excitations. Psycho-physical identities are empirical truths (assuming they are truths at all) which depend on scientific research. For this reason, the status of psycho-neural identities is like that of theoretical identities in the sciences, e.g., “temperature in gases is mean molecular kinetic energy,” “lightning is electrical discharge” and “water is H2O.”
42
3 An Argument for Memory Traces
3.4
Against Mnemic Causation
Having realized that the empirical evidence for the existence of memory traces is “not quite conclusive,” Russell went ahead and proposed an account of memory causation which manages without the stipulation of memory traces – mnemic causation. While the theory of contiguous memory causation maintains that traces are the proximate cause of states of recall, the theory of mnemic causation holds that a past representation is directly causally relevant over a temporal distance. Both theories of causation are divided over the question of whether cause and effect may be separated by a time gap. The contiguity theory regards causation at a temporal distance to be impossible and therefore postulates traces that are produced by the past representation and persist into the present. The theory of mnemic causation, on the other hand, is prepared to accept that cause and effect don’t have to be contiguous. Given mnemic causation, a past experience is not only “part of a chain of causes leading to the present event” but also (together with some retrieval cue) “the proximate cause” of the state of recalling.12 “Mnemic causation” amounts to direct causal action at a distance in time. In C.D. Broad’s The Mind and Its Place in Nature, a chapter is devoted to the notion of a memory trace. Broad characterizes the difference between the theory of contiguous memory causation and Russell’s theory of mnemic causation as follows: On the trace theory, if you were to take a cross-section of the history of the experient’s body and mind anywhere between the past experience and the stimulus you would find something, viz., the trace, which corresponds to and may be regarded as the representative of the past experience. On Mr. Russell’s theory … these intermediate slices, though relevant and necessary, would contain nothing which corresponds to and represents the past experience…. Although there is continuity between the total cause and the effect …, yet there is no continuity between the effect and each independently necessary factor in the cause. The original experience is not joined on to the memory either directly; or by … some special persistent which represents it (1925, pp. 458–459).
In addition, Broad illustrates the difference between both accounts of causation by means of two diagrams (1925, pp. 444–445). Dots stand for momentary events, circles for memory images, crosses for persistent memory traces, full arrows for
12
Russell 1921, pp. 78, 85. Russell gives the following illustration of mnemic causation: “you smell peat-smoke, and you recall some occasion when you smelt it before. The cause of your recollection, so far as hitherto observable phenomena are concerned, consists both of the peat-smoke (present stimulus) and of the former occasion (past experience). The same stimulus will not produce the same recollection in another man who did not share your former experience, although the former experience left no observable traces in the structure of the brain. According to the maxim “same cause, same effect,” we cannot therefore regard the peat-smoke alone as the cause of your recollection, since it does not have the same effect in other cases. The cause of your recollection must be both the peat-smoke and the past occurrence” (1921, pp. 78–79). Thus it is the past olfactory experience and not some present representative of the past experience (like a trace) that causes the recollection. The past experience is the proximal cause of the recall.
3.4 Against Mnemic Causation
43
causal relations, and dotted arrows for cognitive relations. Moreover, “e” stands for a past event, “t” for a trace, “s” for a stimulus (or prompt), “i” for a memory image, and “m” for the memory of event e. The trace hypothesis is represented by Fig. 3.1. Here the past event e brings about a persisting trace t which, at some point, is activated by a prompt s, and produces the memory image i, which represents e. Russell’s theory of mnemic causation is represented by Fig. 3.2. Here the past event e and the present prompt s together directly produce the memory image i, which represents the past event e. Incidentally, Ludwig Wittgenstein seems to have had some sympathy for Russell’s theory of mnemic causation. He writes: I saw this man years ago: now I have seen him again, I recognize him, I remember his name. And why does there have to be a cause of this remembering in my nervous system? Why must something or other, whatever it may be, be stored up there in any form? Why must a trace have been left behind? Why should there not be a psychological regularity to which no physiological regularity corresponds? If this upsets our concept of causality then it is high time it was upset.13
First and foremost this passage is directed against the idea that intentional phenomena can be reduced to brain processes. But this is not the only target. Wittgenstein also seems to dismiss the need for mediative causality between past representations and present memories thereof. That Wittgenstein was toying with the idea of direct causation at a temporal distance becomes apparent in when he declares: “there is
Fig. 3.1 The trace hypothesis
Fig. 3.2 Russell’s theory of mnemic causation
13
This passage is published as §610 of Zettel (1967a) and as §905 of the Remarks on Philosophy of Psychology I (1980).
44
3 An Argument for Memory Traces
something like action at a distance here – which shocks people. The idea would revolutionize science” (1993, §411).14 An undeniable advantage of the theory of mnemic causation over the trace theory is that it has the economy of postulating the existence of fewer relations of causal relevance. Nevertheless, most philosophers, including myself, find the theory of mnemic causation implausible. The standard objection to the theory of mnemic causation is that a cause cannot operate when it has ceased to exist. Without causal contiguity, a past event would somehow have to leap to the present to cause one’s present memory activity. Broad (1925, p. 452) formulates this objection as follows: According to the theory of mnemic causation my perception of a town which I visited last year literally produces a memory of this event whenever a suitable stimulus acts on me. But the perception is long past and is in no sense continued into the present. It has ceased to exist itself, and nothing now exists which can be regarded as a continuation of it. How then can it do anything now?
This objection to the theory of mnemic causation is not convincing, for it rests on what might be called the activity theory of causation, that is, the view that causes engage in an activity in order to bring about effects. On this view, a past experience cannot be the direct cause of a present state of recall, for the past experience cannot do anything, once it has ceased to exist. However, there are good reasons to reject the activity theory of causation. Causes do not act or operate. For A to cause B, all that is required is that whenever A takes place, B takes place, and whenever B takes place, A has taken place; whether or not A still exists by the time B happens is irrelevant. Russell emphasizes that “any case of sufficiently frequent sequence will be causal” (1912a, p. 185; 1921, pp. 88–89, 93). Once causation is defined in terms of regular sequences, nothing stops us from countenancing causal action at a distance in time. Moreover, the idea that a cause can bring about an effect after it has ceased to exist is not unique to the theory of mnemic causation. On the theory of contiguous causation, cause and effect are also separated by a finite time interval. The interval is much smaller than in the case of mnemic causation, but it must exist, for otherwise cause and effect become indistinguishable (see section 3.1). Thus, if Broad’s objection were a good one, it would refute both the theory of mnemic causation and the theory of memory traces. Given that the standard objection to mnemic causation isn’t successful what should we make of this account? The notion of mnemic causation flies in the face of the everyday contention that the time span between cause and effect has to be much shorter than the interval between learning and retention. Yet mnemic causation has the economy of postulating fewer relations of causal relevance between a past experience and its subsequent recall than the trace theory. To see what is wrong
14
In addition to Wittgenstein himself some of his disciples adopted the notion of mnemic causation: Annis (1980, p. 331), Ginet (1975, pp. 166–169; 1988, pp. 166–167), Heil (1978, pp. 68–69), Malcolm (1977, p. 187), and Shope (1973, pp. 317–322).
3.4 Against Mnemic Causation
45
with Russell’s theory of mnemic causation it is instructive to have a closer look at his motive for developing the theory. As was explained above, Russell argues that because the empirical evidence for the existence of memory traces is inconclusive we should settle for an account of memory causation that manages without the stipulation of traces. Two comments. First, science never produces conclusive evidence for anything. There is always room for skepticism. Bearing this in mind, the empirical evidence in favor of memory traces is as good as scientific evidence gets. Second, Russell’s argument in favor of the theory of mnemic causation only goes through if we assume that unless the existence of memory traces can be empirically proven beyond doubt, philosophers are not entitled to stipulate the existence of traces for the explanation of memory causation. But this assumption misconstrues the role that traces play within philosophical accounts of memory. Memory traces are employed as theoretical entities, that is, as devices introduced in the context of a theory to explain some more accessible phenomenon. The status of the concept memory trace is like that of equator or center of gravity. And when memory traces are taken to be theoretical constructs, to find fault with the theory of memory traces is to cast doubt on our need to postulate traces in order to account for remembering. Conversely, arguments in favor of the trace hypothesis have to demonstrate that the concept of a trace allows us to explain certain features of the notion of memory that cannot be explained otherwise.15 The problem with the theory of mnemic causation is that it cannot explain where an how information is stored within the memory process. Clearly, without the assumption of memory traces, information storage cannot be conceived as a causal process. But how can we make sense of the idea of trace-free retention of information? Presumably the only way of spelling out the idea of trace-free retention is in terms of dispositions. According to the dispositional account of retention, all that is required to remember an event is that, in virtue of having witnessed the event, one acquired a disposition to represent it, a disposition which one retained and now exercises by thinking of the event. Remembering that p would be simply one’s persisting disposition to produce tokens of the thought that p in certain circumstances; there would be no need for a causal connection between one’s past witnessing and one’s present representation. The most detailed defense of a dispositional account of retention is due to Squires (1969) and was discussed in section 2.4.16 The dispositional analysis of information storage appears coherent only as long as one doesn’t ask what is involved in the retention of dispositions such as the
15
In my (2001, pp. 167–175) I argue that the stipulation of memory traces is indispensable for analyzing memory causation. My argument claims that the causal relation underlying information encoding supports counterfactuals, while the causal relation underlying information retrieval does not. I no longer believe in the latter of these claims. 16 Russell himself suggested a dispositional account of information storage: “memories, as mental facts, arise from time to time, but do not, so far as we can see, exist in any shape while they are “latent.” In fact, when we say that they are “latent,” we mean merely that they will exist under certain circumstances” (1921, p. 86).
46
3 An Argument for Memory Traces
disposition of recounting a past event. When this question is raised, Russell would have to concede that the trace-free notion of memory retention cannot provide an answer. But if the cognitive process underlying trace-free retention of information cannot be explained, memory becomes a magical process bearing some resemblance to telepathy and clairvoyance. Thus, the account of information storage based on the trace theory is not only more convincing than the account of information storage based on mnemic causation but also, when spelled out, the latter collapses into the former. The very idea of memory storage calls for the stipulation of traces.
Chapter 4
From Traces to Recall
The aim of the present chapter is to provide a positive account of the causal relation responsible for information retrieval from memory. Anticipating the conclusion, for a memory trace to give rise to genuine remembering its content must be an independently sufficient condition (that is not preempted by another independently sufficient condition) of a state of recounting or at least a necessary condition of such an independently sufficient condition. Section 4.1 discusses various approaches to memory causation. These approaches are the analysis in terms of necessary and sufficient conditions, the counterfactual analysis, and the probabilistic analysis. Section 4.2 provides a general characterization of retrieval cues. Given that we may remember only after being prompted what is it for us to remember an event upon being prompted, rather than merely parroting the prompting? This is the topic of section 4.3. This section also discusses the question of whether the recountings of a highly suggestible agent may qualify as remembering. Finally, in section 4.4 I distinguish my account of the causal collaboration between traces and prompts from an account proposed by Charles Martin and Max Deutscher.
4.1
Approaches to Memory Causation
Easily the most favored account of causation is in terms of sufficient conditions. But this account doesn’t work for memory causation. The reason a past representation isn’t causally sufficient for its subsequent recall is that there are frequently other factors which are necessary for the recall. Often we remember an experience or event only after being prompted. The prompt may have the form of a question, but it can also involve some of very information that is contained in the memory evoked by the prompt. Consider the following example by Bede Rundle (1986, pp. 305–306): suppose you are shown a photograph dating back to your schooldays. You ask someone who one of the pupils is, and you are told his name. “Of course,” you say, “now it comes back to me.” Although it took the prompt to bring back the memory, there is no doubt in your mind that you do remember the name. It is quite unlike a case where the name does not “ring a bell,” and where you might S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
47
48
4 From Traces to Recall
accept that the pupil in the photograph had the name given but have no indication to claim it as a memory. Since most of our rememberings depend on some kind of cues, a past representation is rarely causally sufficient for its subsequent recall. Memory causation may be analyzed in terms of necessary conditions rather than sufficient conditions. But also this analysis is too narrow. The reason past thoughts are frequently not causally necessary for their subsequent representation has to do with the problem of causal overdetermination. Consider the following example1: suppose you are taking part in a family reunion and are trying to remember a distant relative’s name. At the very moment when you are about to remember that the relative is called “Bert” your partner who is next to you blurts out “Bert, good to see you!” Your memory that the person’s name is “Bert” is causally overdetermined by your previous knowledge of this fact and by your partner’s blurting. Since each factor is a sufficient condition of your present knowledge that the relative is called “Bert,” neither factor is necessary for your knowing this. Intuitively, however, you do remember that the person is called “Bert,” even though your past knowledge of this fact isn’t necessary for your present knowledge. Another approach to causation in terms of necessary and/or sufficient conditions has been proposed by John Mackie (1965). Mackie’s idea is that while a cause is normally neither a sufficient nor a necessary condition, it is always at least an insufficient but necessary part of an unnecessary but exclusively sufficient condition of some event. Mackie calls this kind of condition inus condition. To see that the inusanalysis of causation is still too strict to cover all kinds of memory causation, reconsider the previous example: suppose that the exclusively sufficient condition of your present knowledge that the relative’s name is “Bert” is a conjunction consisting of four elements: (i) your previous knowledge that he is called “Bert,” (ii) various neurophysiological conditions, (iii) your partner’s blurting out the name “Bert,” and (iv) your ears working correctly. Neither (i) nor (iii) are necessary conditions of the exclusively sufficient condition. Thus, on Mackie’s account, neither the previous knowledge nor your partner’s blurting qualifies as the cause of your present knowledge that the relative’s name is “Bert.” Still another way to spell out causal dependence is in terms of counterfactuals. According to the counterfactual analysis, what makes an event a cause of another is that if the cause-event had not occurred, the effect-event would not have. Usually, the truth of counterfactuals is spelled out in terms of relations of similarity between close possible worlds. In a nutshell, the statement “If A had no happened, then B would not have occurred” is true just in case there is a possible world in which both A and B are true and which is more similar to the actual world than any possible world in which A is true but B is false. The possible-worlds-interpretation of counterfactuals yields a weaker notion of causal dependence than the analysis of causation in terms of necessary and/or sufficient conditions, for conditionals need not hold in all possible worlds but only in those closest to the actual world.
1 This example is taken from Loeb (1974, p. 52). Martin and Deutscher (1966, pp. 178–179) and Pears (1975a, p. 35) develop their own examples of causal overdetermination.
4.1 Approaches to Memory Causation
49
Probability theory yields an even weaker notion of memory causation. Given probabilistic causation, one event’s causing another doesn’t require that the former determines the latter, but only that it makes it more probable than it would otherwise have been. That is, an event A is a probabilistic cause of an event B if the probability of the occurrence of B, given that A has occurred, is greater than the antecedent probability of B: P(B/A) > P(B). This implies that event B must be more probable given the presence rather than the absence of the cause: P(B/A) > P(B/¬A). To illustrate the probabilistic analysis of memory causation consider an Ebbinghaus-style learning situation2: a subject is given a number between one and eight – say, 7 – and on the next day is told to repeat the number. The probability of the subject’s responding with the correct number (7), given that he has heard the number on the previous day is one or almost one. This probability is definitely higher than the probability of the subject responding with the correct number (7), given that he didn’t hear the number on the previous day. Since he might respond with any of the first eight numbers, the probability of him responding “seven” is only 1/8. Hence, P(B/A)[≈ 1] > P(B) [≈ 1/8]. The probabilistic analysis of memory causation has the advantage that it is weak enough to capture all instances of remembering. The disadvantage, however, is that being told that the connection between the thought at t1 that p and the recounting at t2 that p is such that the occurrence of the recounting, given the past thought, is more likely than if the past thought had not occurred is not informative, unless we are given precise values for the probabilities in question. By how much does the conditional probability of the occurrence of the recounting have to exceed the probability of the occurrence of the recounting in general? Unless this question is answered, the probabilistic analysis of memory causation is too vague to be of any real use in distinguishing genuine memory causation from a variety of spurious causal routes. The challenge of the causal theory of memory is to define the causal process responsible for information retrieval so that is neither too broad nor too narrow. The notion of a necessary and sufficient condition, I reckon, is the most promising tool for analyzing the dependence of states of recounting on memory traces. Yet since remembering is often dependent on retrieval cues, memory traces are rarely sufficient conditions for bringing about states of recall. And when a memory trace and a retrieval cue are simultaneous sufficient conditions of the recounting of some past event, the trace doesn’t even qualify as a necessary condition of the recounting. So assuming that it is possible at all to capture the causal relation between traces and states of recall by means of the notions of a necessary and of a sufficient condition, these notions have to be combined in some complex manner. In section 4.3, I will examine ways in which prompts and traces may cooperate to produce memory states. But before going into this issue, I will give a general characterization of the concept of a retrieval cue.
2
This example is borrowed from Rosen (1975, pp. 5–6).
50
4.2
4 From Traces to Recall
Retrieval Cues
Psychologists distinguish between three different kinds of retrieval – free recall, cued recall, and recognition. “Free recall” means that subjects try to remember information without any external information to help them. In the case of cued recall, subjects are given some explicit information (which is not identical with the stored information) to help them remember. In recognition, a subject is presented with the very piece of information he has previously learned and then asked whether it is the piece he is trying to remember. Recall is a more severe test of a person’s ability to remember than recognition because in the former the subject must supply all those characteristics which are necessary to distinguish the required item from all other similar items. In recognition, he only need be able to supply those characteristics which suffice to distinguish the item from the alternatives presented to him.3 And just as recognition is easier than cued recall, cued recall is easier than free recall. Retrieval cues are snippets of information that allow us to access a memory trace. To be prompted means to be re-exposed to some relevant information. Prompting can occur intentionally or unintentionally and it can occur with or without our awareness. For example, a person is prompted if he just happens to read a fictional story which by chance matches something in his own past, or if he sees some event very much like another which he previously saw, or sees an object in much the same state as he saw it previously. Smells and tastes are particularly powerful cues. Probably the most famous literary illustration of this fact comes from Marcel Proust’s Remembrance of Things Past where he describes how the taste and smell of a Madeleine cake soaked in tea brings back with enormous vividness memories of his childhood. After smells and tastes, the two most powerful kinds of cues are verbal and visual reminders. Verbal reminders can consist of just one word or of an entire narrative. Analogously, visual reminders can be single pictures or series of pictures. Verbal reminders are often richer than visual reminders which usually touch the original representation at only one point. In addition to the difference between verbal reminders and sensory cues, C.D. Martin and Max Deutscher (1966, pp. 182–183) introduce the useful distinction between complete (or full) and strict prompting. Complete prompting has to do with the amount of information given. A subject is completely prompted when he cannot correctly reproduce more about the relevant past event (or proposition) than was supplied by the prompt. The prompted information is identical with the stored information. Strict prompting has to do with the mode in which the information is presented – whether or not it is presented as a piece of autobiographical memory.
3 The demands of the recognition task can be increased by increasing the number of alternatives that are presented to the subject. This is what the police does when they arrange an identification parade rather than asking a witness whether he recognizes a single subject. The recognitional task can be made even more difficult by letting the subject know that the correct answer might not be one of the alternatives presented. This sometimes done in multiple choice exams.
4.2 Retrieval Cues
51
A subject in need of strict prompting will remember a certain event (or will recognize a cue representing the event) but will not remember that it was him who previously witnessed the event. Most of the time strict prompting encompasses complete prompting, but it doesn’t have to. It might be possible for a person to remember an event in the impersonal mode so that he only needs to be told that it was him who represented the event. Martin and Deutscher point out that linguistic but not visual prompts can be strict. In the case of linguistic prompts the event-awareness caused by the prompting may contain no recognitional factor – we may be unable to learn from the prompt that we previously experienced that event. Sensory cues, however, don’t allow for strict prompting because whenever we see something which we have seen before, then, even if we cannot give any detail of what we saw before which we did not gain from being prompted, the representation brought about by the prompt contains the idea that what we have seen something like this before. “No matter how much detail is supplied by observation of a replica of what has been previously observed, there is always, in such cases, the additional ‘detail’ of recognition” (Martin and Deutscher 1966, pp. 182–183). David Pears endorses this point when he declares that the difference between a verbal reminder and a sensory cue is that the subject discovers for himself that something is a sensory cue, and this discovery is an achievement of memory, whereas he does not have to discover for himself that what another person says is a reminder when it is explicitly offered as such…. [S]ensory cues … do not undermine the self-sufficiency of the causation of experience-memory as much as verbal reminders undermine it (1975a, pp. 39–40).
It seems reasonable to say that even in cases where we need strict and complete prompting before we are able to recount a certain event, it is possible that we should be genuinely remembering. Of course, it seems lame when someone is asked “Well, what do you remember?” and all he can say is “Ah, … what you just told me.” But despite the fact that we might not be willing to attribute memory to this person, he might still remember – the prompting might have enabled him to “look” into a section of his past. To drive this point home, consider the following example4: suppose Oscar cannot remember what he did the previous afternoon. Bert tells him a long story about driving through certain villages, eating certain dishes, visiting gothic churches, etc. Then Oscar says, “Now it all comes back” and retells the whole story once again. Does Oscar genuinely remember? If he were able to provide additional details over and above the ones provided by Bert’s story, there wouldn’t be the slightest doubt that the answer is “yes.” But suppose that what Oscar was given in the prompt is practically the whole content and that there is no room for adding corroborative detail. I see no reason for denying that Oscar is remembering the outing. Already the mere fact that he can repeat such a long story, might be taken as evidence that he remembers the activities of the previous afternoon. Thus, the fact that a person requires complete and strict prompting before he can recount certain events cannot be ruled out as inadmissible for genuine memory.
4
The example is adopted from Martin and Deutscher (1966, p. 184).
52
4.3
4 From Traces to Recall
Traces, Cues and Suggestibility
Given that a person has to be fully and strictly prompted in order to remember, what is it for him to remember the event upon being thus prompted, rather than merely repeating the prompting itself? Intuitively speaking, he remembers the event if his memory traces are causally connected in a suitable fashion with his acceptance of the prompting. But what is meant by “suitable fashion”? In what way and to what degree may retrieval cues contribute to the production of memories? As was explained in section 4.1, I attempt to analyze the dependence of states of recall onto traces by means of the notion of a necessary and sufficient condition. With this aim in view, let’s begin by enumerating the possible forms of causal cooperation between traces and cues. A terminological note. To distinguish genuine memory causation from a variety of spurious causal routes, I need a term to refer to a person’s activity of reporting past events which leaves open whether the report qualifies as genuine memory or whether it is nothing but the repeating back of a previously presented retrieval cue. In this and the following two section I will use the verb “to recount” to play this role. There are four ways in which retrieval cues and memory traces can cooperate in the formation of states of recounting (Pears 1975a, p. 34). First, when the memory trace and the retrieval cues are independently sufficient and simultaneous causes of the recounting, it is a case of causal overdetermination. Second, when the trace and the cues are independently sufficient but successive causes of the recounting, it is a case of preemptive or superseding causation. When the causation of the trace and that of the cues are sufficient to produce the recounting only jointly, two cases need to be distinguished: when the two causal chains interact before they jointly bring about their effect, it is a case of causal reinforcement; when the two causal chains do not interact but jointly bring about their effect, then it is a case of causal supplementation. Let’s examine these four kinds of causal contribution of retrieval cues in turn.
Causal Overdetermination Consider a variation of the story about the accident-prone Oscar discussed in section 2.4: after having experienced a car accident, Oscar tells Bert about the accident he was involved in. Then Oscar has a second accident as a result of which he temporarily forgets everything about the first accident. When Bert finds that Oscar can no longer remember the first accident, he offers to tell him those details which Oscar had told him in the period between the first and the second accident. At the very moment when Bert is about to retell the story of the accident, Oscar’s memory traces revive. Though the revival of the traces and the cue happen simultaneous they are not causally related. Bert’s beginning to retell the story of the first accident is not the cause of Oscar’s suddenly remembering the accident. Oscar’s recounting of the accident is overdetermined since there are two independently sufficient causes – his memory traces and the prompt provided by Bert.
4.3 Traces, Cues and Suggestibility
53
What should we say about this case? Since each of the two causes is sufficient and since they happen at the very same moment it follows that Oscar’s recounting of the story of his accident is shared between them. Following the majority of philosophers of memory, I believe that the verdict must be that Oscar’s recounting is an instance of genuine memory.
Preemptive Causation Reconsider the case of the accident-prone Oscar. Everything is like in the above example except that the two independently sufficient causes – the prompt provided by Bert and the activation of Oscar’s traces – do not occur simultaneously but in succession. To evaluate this case we need to know which of the two causes comes first. If the retrieval cues occur shortly before the reactivation of the memory trace, Oscar’s recounting of the first accident does not qualify as remembering. If, however, Oscar’s traces are revived shortly before Bert provide reminders, Oscar does remember.
Causal Supplementation In the case of causal supplementation traces and prompts are sufficient to produce a state of recounting only jointly. There is independent causation both through traces and through prompts but only together they are sufficient to produce the recounting. Let’s modify the story of the accident-prone Oscar to illustrate causal supplementation: suppose his memory traces representing the first of the two accidents are too weak to make Oscar recount the accident. Bert’s report of the accident, by itself, also doesn’t suffice to make Oscar represent the prompted story as something that happened to him. Now suppose that Bert’s starting to tell the story of the accident and the revival of Oscar’s memory trace take place simultaneously but independently. The two causal chains do not interact but they jointly bring about their effect, namely Oscar’s recounting of the story of the accident. The arrows with crosses stand for independently insufficient but jointly sufficient conditions. Does causal supplementation yield genuine memory? This is not easy to say for only half of the causal history of Oscar’s recounting is of the right sort – the causal efficacy transmitted via traces. Yet I see no good reason not to classify states of recounting brought about by causal supplementation as legitimate and genuine cases of memory. It would be too stringent to demand that traces be independently sufficient conditions of recounting.
Causal Reinforcement Causal reinforcement and causal supplementation have in common that traces and prompts produce states of recounting only jointly. What distinguishes causal supplementation from causal reinforcement is that in the former case the retrieval cues don’t act as reminders but make an independent contribution to the causation of the recounting. In causal reinforcement the two causal lines interact before they
54
4 From Traces to Recall
produce their final effect. They merge and produce the recounting together. Thus, the last stage of the causal process leading up to the recounting is normal. The abnormality occurs at the penultimate stage: a weak trace is strengthened by a weak prompt. Like in the case of causal supplementation, I am inclined to say that states of recounting deriving from causal reinforcement do classify as genuine memories.5 Let us take stock. We have seen that three out of four ways of how traces and cues can cooperate in the production of states of recounting qualify as genuine remembering. When traces and cues are independently sufficient conditions of the recounting, the causation by the traces has to occur either shortly before, or simultaneous with, the causation of the cues. When traces and cues are jointly sufficient conditions, the resulting states of recounting always classify as remembering. Thus, for a memory trace to give rise to genuine remembering it must be an independently sufficient condition (which is not preempted by another independently sufficient condition) of a state of recounting or at least a necessary condition of such an independently sufficient condition. Before terminating this section we need to consider another factor, apart from prompts and traces, that can cooperate in the formation of a state of recounting – suggestible states. One way to enhance the suggestibility of an agent is to subject him to hypnosis. Hypnosis involves a form of suggestibility, in which the hypnotized individual abdicates responsibility for his actions on the assumption that he is now under the control of the hypnotist. Hypnotized individuals are, for example, more prone to accept directions, no matter how irrational. They are also less likely to distinguish an instruction as coming from another rather than themselves, and so will tend to act on another person’s ideas as though they were their own. In extreme cases, individuals may become what has been called “honest liars,” believing strongly in implanted or imagined “recollection.” One such example is reported by David Spiegel (1995, p. 139): A highly hypnotizable businessman was instructed in hypnosis that there was a communist plot to take over the television media, and that he would see three names on a sheet of paper. He was then interviewed on camera by a well-known television reporter. He elaborated a tale of intrigue regarding such a plot, replete with names and dates of meetings. He hallucinated three names when confronted with a blank sheet of paper, saying: “I know him and him, but not that one in the middle.” When pushed harder by the interviewer, he stated: “He is a terrible man – maybe he has gotten to you already,” thereby elaborating a paranoid defense of his position. As soon as the hypnosis was ended, he recanted the story, viewing it as an amusing experiment.
5 According to Pears, states of recounting due to causal supplementation and to causal reinforcement are only half-memories. Pears raises the interesting question how one can tell in one’s own case whether a retrieval cue acts as an independent causal contribution or as a reminder. A psychological criterion is out of question, for one could have the impression of being reminded by a prompt when, in fact, one is not. Because the distinction in question is not transparent to introspection Pears concludes that “neurology is the only possible source of evidence which would discriminate between a case in which the two causal lines do not interact before they produce their final effect and a case in which they interact at some earlier point” (1975a, p. 38).
4.4 Operative Conditions
55
In another study conducted by Jean-Roch Laurence and Campbell Perry (1983), 13 of 27 hypnotized individuals who were told that they had heard an accident reported this suggestion as real memories 7 days later. And some persisted in this false belief even after they were informed that their apparent memories had been the product of hypnotic suggestion. Just as hypnotized subjects can be made to accept an imagination as a piece of autobiographical memory they can be made to not recognize an event as having been witnessed before. Suggestibility can be enhanced not only by means of hypnosis but also by means of hypnotic drugs (also called “truth drugs”) such as sodium amytal. Given the presence of a suggestible state, a state of recounting can be due not only to two but also to three factors – traces, prompts, and the suggestible state. Now the question is whether it is possible for a suggestible agent to remember not only when he is not prompted but also when he is prompted. Since someone occupying a suggestible state will accept any prompt as an item of his memory and since memory implies truth, I will focus on cases where a highly suggestible agent is prompted with a story he in fact witnessed in the past. Thus, the question to consider is whether the state of recounting of a highly suggestible subject may qualify as remembering, even when he was veridically prompted. Suppose the suggestible state together with the prompt, on the one hand, and the memory trace, on the other, are independently sufficient conditions of the formation of a state of recounting. I hold that if the causation by the traces occurs either shortly before, or simultaneous with, the causation by the suggestible-state-cumprompt, the recounting does qualify as remembering. Next suppose that the suggestible state together with the prompt, on the one hand, and the memory trace, on the other, are independently insufficient but jointly sufficient conditions of the recounting. In this case it seems to me that as long as the memory trace is a necessary condition of the jointly sufficient condition, the resulting state of recounting does classify as remembering. In sum, as long as prompts are veridical, cases of recounting due to traces, prompts and suggestible states can be interpreted along the lines of cases of recounting due to only traces and prompts.
4.4
Operative Conditions
Given the complexity of an adequate account of memory causation in terms of necessary and sufficient conditions, Martin and Deutscher propose a change of vocabulary. Instead of talking about traces “causing” memory state they talk about traces being “operative” in bringing about memories. An operative condition may be necessary or sufficient but need to be neither. “A condition may be operative in producing another, even though the result would have been obtained at the same time by another method, had the operative condition not been present” (Martin and Deutscher 1966, p. 179). In an attempt to tell apart those cases of prompting that are and those that are not compatible with remembering Martin and Deutscher introduce the operators
56
4 From Traces to Recall
“operative for” and “operative for and in.” The notion of something being operative for something else is defined in the following manner: if A produces B and B produces C, then A is operative for the circumstance B which is operative for C. Although A is not directly involved in bringing about C, by transitivity, it produces C. But B’s production of C is independent of what brought about B – whether it was A or some other factor D. Thus, given B, A is irrelevant for the production of C. Deutscher provides the following example for something being “operative for” something else: A brick is thrown according to the pre-arranged signal of a low-pitched whistle. It is partly because of the whistle that the brick is thrown. The brick then breaks a window. Thus the low-pitched whistle is, by transitivity, an operative condition for the window’s being broken. But it is not operative in bringing it about that the window is broken in the circumstance that a brick was thrown at it (1969, p. 110).
The operator operative for and in is defined in this way: if A produces B and B produces C and if, in addition, A produces some factor B* other than B which acts together with B to bring about C, then A is operative for C in (as well as for) the circumstance B, which is operative for C. What distinguishes “operative for” from “operative for and in” is that in the latter case, given B*, A (via B) is still relevant for the production of C. The following example helps to grasp the meaning of “operative for and in”: A tennis ball is thrown according to the pre-arranged signal of a high-pitched whistle. It is partly because of the whistle that the ball is thrown. The ball may not be thrown hard enough to break the window by itself. The additional effect of the sound waves from the whistle may also be needed. Thus the high-pitched whistle is operative both for and in the circumstance that the ball is thrown (1969, p. 110).
In the case of the brick-example, given that the brick is thrown towards the window, the operative condition for the brick being thrown – the sound of the whistle – is irrelevant for the breaking of the window. In the tennis-ball case, however, given that the tennis ball is thrown, the sound of the whistle is still relevant because the sound waves are partly responsible for the breaking of the glass. After having distinguished between the operators “operative for” and “operative for an in” Martin and Deutscher forward the following rule: In those cases where prompting is operative for the representation, [a person’s] past experience of the thing represented is operative in producing the state (or the successive set of states) in him which is finally operative in producing the representation, in the circumstances in which he is prompted (1966, p. 185).
This is supposed to be a general criterion for telling apart cases where promptings give rise to genuine memories from cases where a person simply parrots the prompted information. The idea is that a state of recounting brought about by a strict and complete prompt classifies as a memory only if, unless the person had the relevant past experience, he wouldn’t accept the prompt. Though I applaud Martin and Deutscher’s contention that remembering is compatible with the relevant past experience being operative in bringing about a present recounting only upon prompting, they fail to distinguish between two different cases both of which are subsumable under the label “a past experience is
4.4 Operative Conditions
57
operative for a memory-impression, in the circumstances in which one is prompted.” I have labeled these cases “causal supplementation” and “causal reinforcement” (cf. section 4.3). What is more, Martin and Deutscher erroneously assume that whenever prompts form independently sufficient conditions, the state of recounting brought about by the prompts does not qualify as a genuine memory. Yet, as was shown in section 4.3, if the independent sufficient causation of the prompt occurs simultaneously with the independently sufficient causation of a memory trace, there is no reason not to classify the resulting state of recounting as an instance of remembering.
Part II
Objects of Memory
Chapter 5
The Primary Objects of Memory
After having developed and defended a version of the causal theory of memory it is time to explore its metaphysical and epistemological consequences. In particular I will examine the obstinate preconception according to which the causal theory of memory is committed to representative (or indirect) realism and is incompatible with direct (or naive) realism. John Sutton, for example, declares: The idea that a “trace” acquired in past experience somehow “represents” that experience, or carries information about it, is at the heart of “representative” or “indirect” realism in the philosophy of memory…. In contrast to the representative realist, direct realists and others hostile to the memory trace claim that in the act of remembering I am in direct contact with past events (2004; see also 1998, p. 280).
Another philosopher who give expression to the worry that direct realism about memory is incompatible with the causal theory of memory is William Earle. Given direct realism, Earle claims, “[t]he relation between the conscious subject and its object is that of presence and not physical influence, touching, or any mode of causation whatsoever…. [T]hat past event … stands in my presence, and presence is not a physical relation” (1956, p. 26). Since the causal theory of memory is highly intuitive, if it turns out to be incompatible with direct realism, this would speak against direct realism. The goal of this chapter is to show that, contrary to what Sutton, Earle and others think, the causal theory of memory is perfectly compatible with direct realism. In addition, I dispel other objections to direct realism about memory. In chapter 6 I will continue to defend direct realism by criticizing the conception of a memory marker which is a central component of the representative theory of memory. Section 5.1 starts out explaining the distinction between direct and representative realism concerning both perception and memory. Sections 5.2 and 5.3 analyze and criticize two objections to direct realism about memory – the co-temporality objection and the objection from hallucination. Finally sections 5.4 and 5.5 undermine altogether three arguments to the effect that direct realism is incompatible with the causal theory of memory.
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
61
62
5.1
5 The Primary Objects of Memory
Representative Versus Direct Realism
The distinction between representative realism and direct realism about perception is more developed than the corresponding distinction about memory. It is therefore prudent to begin by discussing perception. In its broadest formulation, representative realism about perception claims that, though there is a world of mind-independent objects that causes us to have experiences, we are not directly or immediately aware of these external objects. What we are directly aware of are the effects these objects have on us, namely representations of the external reality. Representations – or ideas (Descartes and Locke), impressions (Berkeley, Hume), sense-data, sensa or sensibilia (Russell, Moore, and Austin), percepts (Firth) and appearances – are the messengers that stand between us and physical objects such as tables and chairs. Just as images appearing on a television screen represent their remote causes, the sense data that occur in the mind – the representations of which we are directly aware – represent their external causes. Ordinary perception of external objects, then, turns out to be a type of indirect perception. Direct realism, on the other hand, maintains that in perception we are directly aware of what J.L. Austin (1962, p. 8) calls “moderate-sized specimens of dry goods” – trees, rocks, cars and the like, i.e., things that, unlike representations, continue to exist when we are no longer aware of them. We perceive physical objects directly, without a mental intermediary. Direct realism is the denial of representative realism.1 The distinction between direct and indirect realism about perception has an interesting history. There was a time when perception was understood to be of things themselves, not of our ideas of things. This is what we find in Aristotle and Aquinas, who maintain that the mind or understanding grasps the form of the material object without the matter. What we perceive directly, on this view, are material objects. This changed in the seventeenth century with Descartes and Locke, who can be read as saying that the primary objects of perception are not things external to the mind but sense-data. Sense-data are the messengers that stand between us and physical objects such as tables and chairs. While indirect realism was the standard view of early modern philosophers, nowadays direct realism is, once again, in fashion. Though there are still some indirect realists around (cf. Jackson 1977; Lowe 1992; O’Shaughnessy 1980; Robinson 1994) most contemporary theories of perception on the market see themselves as falling in the direct realism camp. What are sense-data? They are defined as private, non-physical entities that resemble the physical objects by which they are caused. Representative realists differ among themselves about the question of how much sense-data resemble external objects. Some take sense-data to possess both primary qualities (solidity, extension,
1
Some people distinguish between naive direct realism and scientific direct realism. The naive direct realist claims that the objects of perception possess all the properties that we usually perceive them to have. The scientific direct realist claims that some of the properties an object is perceived as having are dependent on the perceiver, and that unperceived objects should not be thought of as retaining them. I will be concerned only with naive direct realism.
5.1 Representative Versus Direct Realism
63
shape, and mobility) and secondary qualities (color, sound, smell, taste), others (such as John Locke) think that they resemble their causes only in respect of primary qualities, and others assert that sense-data exhibit only a structural isomorphism to primary qualities. The most problematic aspect of the doctrine of sense-data is the fact that they are supposed to possess the various sensory qualities that a person experiences. The primary motivation for claiming that sense-data resemble physical objects is the argument from hallucination. Suppose someone hallucinates an orange. His experience may be indistinguishable from the experience he has when he actually sees an orange. When hallucinating, he is not aware of an orange. But he is aware of something, namely a mental image of an orange, or a sense-datum of an orange. Since the experience of an orange-sense-datum is, subjectively speaking, exactly like seeing an orange, it is plausible to suppose that seeing an orange is really seeing an orange-sense-datum. The only difference is that when one actually sees an orange, the experience is caused by a real orange. So even in the veridical perception of an orange, what one is directly aware of is an orange-sense-datum; or so the representative realist argues. Representative realism about perception not only claims that sense-data are the objects of perception, but also that they are the direct or primary objects of sensory awareness. It claims that our epistemic access to physical objects is somehow less good than our epistemic access to our sense-data. The picture is one in which our representations are epistemically prior to external objects, in the sense that I can perceive the former without perceiving the latter, but not vice versa. The thesis of epistemic priority is clearly presupposed by David Hume in the following passage: It is a question of fact, whether the perceptions of the senses be produced by external objects, resembling them: how shall this question be determined? By experience surely; as all other questions of a like nature. But here experience is, and must be entirely silent. The mind has never anything present to it but the perceptions, and cannot possibly reach any experience of their connection with objects. The supposition of such a connexion is, therefore, without any foundation in reasoning (1748, p. 153).
Different representative realists offer different analyses of our epistemic privilege with respect to sense-data. John Stuart Mill (1843, ch. 1, §2), for example, explains the privileged status of knowledge of sense-data in terms of non-inferentiality: knowledge of sense-data is non-inferential whereas knowledge of physical objects is inferential. The obvious problem with this way of spelling out the direct/representative realism distinction is that it doesn’t square with how things feel from the inside. Frequently our sensory awareness of physical objects does not seem to be the product of inference. To defend the interpretation of the direct/representative realism distinction under discussion, one would have to maintain that even though we are not conscious of making inferences when we are aware of physical objects, there are unconscious inferences at work. Another way of explaining the privileged access to sense-data is by appeal to the idea of infallibility: for something to be immediately experienced the awareness of this thing must be certain, indubitable or infallible. Our sensory awareness of external
64
5 The Primary Objects of Memory
objects can be subject to error while this is not so with respect to sense-data. There is no possibility of “misperceiving” sense-data. As Hume says “since all actions and sensations of the mind are known to us by consciousness they must necessarily appear in every particular what they are, and be what they appear” (1739, p. 190). Similarly, H.H. Price maintains that, if it seems to him as if he is seeing a tomato, while it is open to doubt that it is really a tomato he is seeing, he cannot be wrong that “there exists a red patch of a round and somewhat bulgy shape … and that this whole field of colour is directly present to [his] consciousness” (1932, p. 3). The notorious problem with this line of reasoning, first noticed by Wilfried Sellars (1963), is that for a conscious experience to be reportable it must be conceptually classified and the conceptual classification of experiences can be subject to error. It is unlikely that there are objects whose esse is their percipi. The two criteria for immediacy of awareness – non-inferentiality and infallibility – are frequently conjoined. A case in point is Frank Jackson. On his reading of representative realism, the reason we don’t directly perceive external objects is that we perceive objects in virtue of perceiving surfaces. We commonly see things in virtue of seeing other things: I see the aircraft flying overhead in virtue of seeing its underside (and the aircraft is not identical with its underside); I see the table I am writing on in virtue of seeing its top; I first see England on the cross-channel ferry in virtue of seeing the white cliffs of Dover; and so on and so forth…. It follows, therefore, that I see an opaque physical object in virtue of seeing a part of it…. Now for our definition: x is a mediate object of (visual) perception (for S at t) if S sees x at t, and there is a y such that (x ≠ y and) S sees x in virtue of seeing y.2
It is an undeniable fact that we see external objects in virtue of seeing their surfaces. But why should acknowledging this undeniable fact commits us to saying that we do not perceive external objects directly? Jackson’s argument seems to be the following: provided we live in an epistemically friendly environment, having a perceptual experience of a part of the aircraft’s underside is sufficient for perceiving an aircraft. In epistemically unfriendly environments, however, one could have the very same perceptual experience of the aircraft’s underside without thereby perceiving an aircraft. So by maintaining that we perceive external objects only in virtue of perceiving portions of their surfaces Jackson is saying that one could perceive a portion of an object’s surface without thereby perceiving the object.
2
1977, p. 19. Armstrong doesn’t endorse representative realism, but he gives a characterization of this view which is very similar to one put forward by Jackson. Armstrong writes: “Consider seeing an apple. Whether the perceptions produced by the apple be veridical or illusory, it is clear that it is not the apple as a whole which acts upon the eyes to produce these perceptions. If, for example, the back half of the apple had been cut away, this would have had no effect upon the resultant perception…. Furthermore, we could obviously pare away a good deal more of the apple and yet the object left might still have exactly the same perceptual effect…. If we consider these facts, there seems to be a clear sense in which when, as we say, somebody sees an apple, then ‘in truth and strictness’ as Berkeley would put it, they see something far less. They see a much smaller object and see only a selection of the properties of that smaller object…. [W]e see that an apple is seen in virtue of seeing much less than the apple” (1979, p. 87).
5.1 Representative Versus Direct Realism
65
Following Jackson, we can spell out the direct/representative realism distinction as follows. The representative realist claims that a person can be occupying the very same sensory awareness states in both veridical and non-veridical scenarios. The reason awareness of sense-data is said to be epistemically privileged vis-à-vis awareness of physical objects is that one can know how things appear to oneself regardless of whether there is anything corresponding to one’s sensory awareness.3 According to direct realism, however, knowledge of one’s sensory awareness is not any more secure than knowledge of external objects. The reason there is no epistemic asymmetry is that a veridical sensory awareness state is defined such that it differs in kind from its non-veridical counterpart – even if the two are subjectively indistinguishable. The direct realist therefore rejects the idea that first-person indistinguishability entails identity of experiential kind. The discussion of memory has closely followed the discussion of perception. Just as philosophers have debated the question whether perception is a direct awareness of objects or an inferential procedure, so it has been debated whether memory provides mediate or immediate awareness of the past. Representative theorists of memory have been in the majority until fairly recently. Among the advocates of the representative theory are E.J. Furlong (1948), David Hume (1739, pp. 8–10), William James (1890, I, ch. 16), John Locke (1694, pp. 149–155), and Bertrand Russell (1921, ch. 9). According to representative realism about memory, one remembers something not by way of being directly aware of that thing, but rather a mediating image which represents that thing. To remember is to undergo a certain sort of mental experience. It is to experience a mental representation – usually called a “memory-image” – which reproduces some past sense-experience. The memory-image provides us with the information we are then said to remember. So it is a distinctive feature of representative realism that remembering involves primary awareness of memoryimages. The classic view of memory-images includes three distinct claims: (i) memory-images are mental entities which serve as objects for a direct memory; (ii) a memory-image of something X shares numerous properties with some prior perception of X; (iii) a memory-image of something X is causally linked with some prior perception of X. The standard objection to the representative theory states that memory-images are not a necessary component of remembering; it is possible to remember an event without experiencing the appropriate representative image of it. In response to this objection, advocates of the representative theory can point out three things. First, memory-images are dispensable only in the case of fact memory, yet representative realism is mainly concerned with personal memory. Though I can remember that Fido bit the mailman without having an appropriate image of that fact before my
3 Given this way of drawing the direct/representative realism distinction disjunctive theories of perception turn out to be indirect. The disjunctive theory holds that an appearance statement such as “It looks to S as if he perceives an object X” should be analyzed as a disjunctive statement “Either S perceives X, or it only seems to S as if he perceives X.”
66
5 The Primary Objects of Memory
“mind’s eye” I cannot remember Fido’s biting the mailman without forming images of the scene. “Memory demands … an image,” Russell (1921, p. 186) declares. When I remember what I ate for breakfast this morning, “[t]he process of remembering will consist of calling up images of my breakfast” (p. 175). Second, the word “image” need not be taken literally. “Memory-images” are not necessarily representations of a visual kind. Instead “memory-image” stands for whatever way something is represented in memory. It is therefore more appropriate to speak about “memory-data” than “memory-images.” Third, just because we may not be explicitly aware of memorydata and take ourselves to be directly acquainted with the past events themselves, this should not be counted against representative realism. Like so many other cognitive processes memory is not transparent from the first-person perspective. Just as sense-data are said to resemble the physical objects by which they are caused, memory-data constitutive of personal memory (and meta-representational memory) are said to resemble the original experiences, not only in their informational content, but also in their phenomenal character, their subjective “feel.” Some even go as far as claiming that both the memory-datum and the memory trace which brings it about are structural analogues of the original experience. C.B. Martin and Max Deutscher, for example, claim a memory trace is “an analogue which contains at least as many features as there are details which a given person can relate about something he has experienced” (1966, p. 190). The structural analogy between the original experience, on the one hand, and the persisting trace and the memorydatum, on the other hand, is conceived along the lines of the analogy between musical sounds and the number and the deflection of the grooves in a vinyl record. For each variation in the pitch, tempo, or volume of the music there is a variation in the grooves of the record. To my mind, however, the idea of traces as structural analogues is as implausible as claiming that a virus carrying a contagious disease from one patient to another must be structurally analogous to the symptoms of the illness it transmits. Representative realism about memory can be seen as a part of a larger foundationalist epistemological enterprise. The representative theory is meant to provide an answer to the question of how memory experiences provide us with knowledge of the past. The reason memory-data play a prominent role in representative realism is that they are supposed to be the self-evident and experiential basis for memory knowledge. If justified beliefs must be based on facts known with complete certainty, and if no fact about the past can be known with complete certainty, it appears that memory beliefs can be reasonable only if they are grounded on certain knowledge of facts about the present.4 Memory-data are supposed to be knowable with absolute certainty (Shoemaker 1967, p. 268). Remembering cannot be defined simply in terms of awareness of images or experiences, since we have images of many things we do not remember. The need to discern memory-data from other kinds of experiences is particularly pressing if one
4
Here and elsewhere I use the term “memory belief” to apply to retained beliefs.
5.1 Representative Versus Direct Realism
67
wants to base memory knowledge on awareness of memory experiences. Representative realists maintain that one can tell, by reflection alone, whether a particular image one is having is one of memory, or one of the other faculties of the mind, such as perception or imagination. They maintain that there is a feature of memory-data that stamps them as such. This feature of memory-data is commonly referred to as the “memory marker” or “memory indicator.” Memory markers are a priori knowable properties of memory-data by which they can be distinguished from other mental phenomena. Memory markers have been described by representative realists in a number of ways, as the feeling of warmth and intimacy (James 1890, vol. I, p. 650), the feeling of familiarity and pastness (Brandt 1955, p. 80; Broad 1925, p. 271; Plantinga 1993, p. 59; Russell 1921, p. 163), the force and vivacity of memory-data (Harrod 1942, p. 5; Hume 1739, pp. 9, 85; Jacoby and Whitehouse 1989), their embeddedness (Johnson et al. 1988) and their spontaneity and involuntariness (Furlong 1951, p. 98). For a detailed discussion of memory markers see section 6.2. Direct realism about memory differs from the representative theory primarily over the question of what it is that is directly experienced in remembering. According to the direct realist, we do not remember the past by virtue of being aware of an image presenting the past to us, rather our awareness of the past is direct. Although remembering something requires the having of memory-data and though these memory-data determine the way the thing appears to us, there is no reason to suppose we are aware of these memory-data themselves. I am aware of the past event by internally representing the event, not by being aware of the internal representation of the event. Memory-data, according to the direct realist, do not function as the primary object of awareness, but are merely the vehicle of the remembered information. As John Laird puts it, “memory does not mean the existence of present representatives of past things. It is the mind’s awareness of past things themselves” (1920, p. 56).5 The realist theory derives most of its plausibility from the phenomenology of remembering. When we remember something, what we are aware of is just that thing, and nothing further. This motivating sentiment of the realist view is well captured in the following remark from Thomas Reid: Suppose that once, and only once, I smell a tuberose in a certain room, where it grew in a pot, and gave a very grateful perfume. Next day I relate what I saw and smelled. When I attend as carefully as I can to what passes in my mind in this case, it appears evident that the very thing I saw yesterday, and the fragrance I smelled, are now the immediate objects of my mind, when I remember it …. Philosophers indeed tell me, that the immediate object
5 1920, p. 56. I agree with Searle’s assessment of the dispute between direct and representative realism: “[T]he traditional sense data theorists were correct in recognizing that we have experience, visual and otherwise, but they mislocated the intentionality of perception in supposing that experiences were the objects of perception, and the naive realists were correct in recognizing that material objects and events are characteristically the objects of perception, but many of them failed to realize that the material objects can only be the objects of visual perception because the perception has an intentional content, and the vehicle of the intentional content is a visual experience” (1983, p. 61).
68
5 The Primary Objects of Memory of my memory and imagination in this case, is not the past sensation, but an idea of it, an image, phantasm, or species, of the odour I smelled: that this idea now exists in my mind, or in my sensorium; and the mind, contemplating this present idea, finds it a representation of what is past or of what may exist; and accordingly calls it memory, or imagination. This is the doctrine of the ideal philosophy; … Upon the strictest examination, memory appears to me to have things that are past, and not present ideas, for its object (1764, p. 28).
Apart from John Laird and Thomas Reid, direct realism about memory is defended, among others, by Samuel Alexander (1920, I, pp. 113–116), Henri Bergson (1896), William Earle (1956, pp. 10–16), Bertrand Russell (1912b, pp. 114–115) and A.D. Woozley (1949, pp. 60–67). Since, on the realist view, what one is directly aware of in memory is the past event in propria persona, and not some representation of it, one of the difficulties of the representative theory of memory disappears – the difficulty of explaining how we can be justified in inferring the occurrence of a past event from the occurrence of a present memory-datum. If what we are directly aware of is the past event itself, and not a present image which serves as a representation of it, no such inference is required. Though direct realism makes some problems disappear, it gives rise to others. One of the notorious problems of direct realism is to explain our direct acquaintance with, or experience of past events. Representative realists as we have seen, claim that awareness of memory-data is direct in the sense that it is not to the same extend error-prone as awareness of physical events. Obviously the notion of directness employed by direct realists differs from the homonym notion employed by representative realists. Our cognitive access via memory to past events is neither infallible nor incorrigible. Presumably then the notion of direct awareness employed by the direct realist should be read as meaning “non-inferential awareness”: to remember a past event is not to infer its past existence on the basis of present evidence or background knowledge. However, the claim that the statement “We have direct awareness of past events” is tantamount to “We have awareness of past events that is not grounded on present evidence” does not explain our direct awareness of the past; it simply rejects the explanation offered by the representative realist (Malcolm 1976, pp. 5–6). If it is non-inferentiality that direct realism asserts, it is first and foremost a negative theory which states that knowledge of sensory awareness is no more secure than knowledge of external objects. Yet given this interpretation of direct realism, it follows that representative and direct realists talk past one another by employing different notions of direct awareness. According to representative realism, direct awareness is awareness that is immune to brute error; according to direct realism, direct awareness is non-inferential awareness.
5.2
The Co-temporality Objection
As was explained in the beginning of this chapter, my aim is not only to explain direct realism about memory but also to defend it. To this end I will start by discussing one of the most obvious objections to the view that the immediate object of
5.2 The Co-temporality Objection
69
remembering is the event remembered. How, it is asked, can we be in direct touch with events which occurred and ended a long time ago? How can the direct object of my present state of remembering be something that has ceased to exist? This objection to direct realism dates back to William Hamilton, Thomas Reid’s editor, who, in commenting on Reid’s claim that “it is by memory that we have an immediate knowledge of things,” commented: “An immediate knowledge of a past thing is a contradiction. For we can only know a thing immediately, if we know it in itself, or as existing; but what is past cannot be known in itself, for it is non-existent” (Reid 1895, I, p. 339n). Norman Malcolm makes essentially the same point when he notes that it is an implication of “direct awareness,” in its philosophical use, “that if B is ‘directly’ aware of X, then B and X coexist…. The implication, in philosophical language, of this would be that if Robinson now remembers last week’s earth tremor, then the tremor now exists” (1976, p. 5). If one can have only direct awareness of what is contemporary with the awareness itself, as Hamilton and Malcolm suggest, then direct realists must provide some explanation of how past, even non-existent things can co-exist with the episode of remembering. It might seem as if direct realism rests upon the supposition that an event continues to exist after it has happened and that it is accordingly an “attempt to immortalize events” (Furlong 1951, pp. 27–28). The crux of this objection to direct realism is the supposition that direct awareness implies co-temporality. If this supposition were correct, it would count against not only direct realism about memory but also direct realism about perception. For also in the case of perception we have to allow that what is directly perceived is not contemporary with the act of perceiving it. The reason is that perceiving a physical object is a causal process that takes time. Consequently, it is possible that by the time we become visually aware of an object, the object no longer exists; it ceased to exist during the time light was being transmitted to the eye. But even when the object ceases to exist before we become aware of anything, we are aware of something. Since we cannot perceive into the past, this something of which we are aware cannot be the physical object itself but must be a sense-datum. The conclusion of this time-lag argument is that direct realism about perception is to be rejected. Why should it be necessary that the content of perception is contemporaneous with the act of perceiving? Why can we not perceive something that no longer exists? Obviously, if something no longer exists, we cannot presently perceive it as it presently is. But from this truism it does not follow that it is impossible to presently perceive a no-longer existent object as it used to be. Among those who endorse this way of rejecting the time-lag argument is George Pitcher who writes: [A direct realist] can simply insist that the finite speed of light does not entail that we do not directly see things and states of affairs in the “external world,” but only that we must see them as they were some time ago. We see real physical things, properties and events, all right, but we see them late, that is all. According to a direct realist, it is a mere prejudice of common sense – and one on which the time-lag argument trades – that the events, and the states of objects, that we see must be simultaneous with our (act of) seeing them (1971, p. 48).
70
5 The Primary Objects of Memory
According to this line of reasoning, the time-lag argument does not show that we never perceive physical things; it shows only that we never perceive things that coexist with us. What we perceive are past objects. Howard Robinson finds the idea that we never at time t perceive a physical object which obtains at t intolerable. He thinks that “[a]nything which wholly or partly constitutes something else must exist at the same time as (or, at least, overlap in time with) the thing it wholly or partly constitutes” (1994, p. 81), and therefore the direct realist’s content of perception – the external object – must be contemporaneous with the act of perceiving. Denying the contemporaneity thesis, Robinson maintains, involves confusing intentional objects qua logical feature of a state and intentional object qua content. The [direct realist’s] argument would be that if I think of Charles I, Charles I constitutes the content of that act, but that monarch no longer exists. This is mistaken because the content of the episode of thought is not constituted by the logical object of the thought, but by some vehicle – probably words – by which I think it. I cannot make any sense of the idea that a past object could constitute the content of a current mental state without the aid of a current form of representation (1994, pp. 81–82).
Robinson’s worry is groundless, for direct realism is not committed to denying that cognition involves “a current form of representation.” But according to direct realism these representations are the vehicles of perception rather than its objects. And to say that perception directly acquaints us with a past object is not to say that it provides some inexplicable immediate awareness or experience of the past object. We are directly aware of the past object in the sense of not having to rely on present evidence. Let’s return to the co-temporality objection to direct realism about memory. The problem with this objection is not only the supposition that direct awareness implies co-temporality but also the presumption that when an event is past, it ceases to exist. This presumption goes by the name of presentism. According to presentism, the future and past are unreal and only those things that exist at the present moment really exist. The only strictly true statements of the form “A is F at t” are ones of the form “A is F now.” That being so, the word “now” in such statements is really redundant and can be dropped without loss of informative content (cf. Prior 1970). The major problem facing presentism is this: If only the present is real, where does this leave statements about the past? How does presentism enable us to affirm, for example, that there were dinosaurs, while denying that there were unicorns? The presentist could argue that the statement “It is not now true that there were dinosaurs” means that “The states of affairs that exist at the present moment do not contain any truth-makers for the statement that there were dinosaurs.” The latter statement is certainly true. So following this proposal, the statement “There were dinosaurs” the presentist would translate as “ ‘There are now dinosaurs’ was true.” However, this translation is inconsistent with presentism. For how can a presentist make sense of the idea that a sentence was true? If only present-tense states of affairs exist, there is no way of making sense of a past time. The idea of a statement’s
5.2 The Co-temporality Objection
71
having been true in the past only shifts the problem from that of referring to enduring objects to that of referring to moments of time. Both enduring objects and moments of time are unreal, according to the presentist. In the end, presentists seem to have to analyze statements about the past in terms of statements about the present. The statement “There were dinosaurs” is true now, because its truthmaker doesn’t consist in some past states of affairs, but rather in some present states of affairs, for example, fossils. Yet the problem with this position is that it flies in the face of the natural intuition that the past is fixed and that it does not change in light of present events. There can be evidence, at one time, for the occurrence of some past event, but the evidence may disappear at some later time. According to the presentist, the disappearance of the evidence can turn a once-true statement about the past event into a false one. This view is certainly strange.6 Events do not cease to exist when they cease to be present or when there ceases to be evidence for them. Once an event has happened, it exists eternally; the only thing that still happens to it is that it retreats into the more and more distant past. On this eternalist view of time and change, past events are always “there” waiting to be remembered. Given eternalism, there is no good reason why past events should not be able to enter into a relation with certain present events that they become objects of direct acquaintance. There is no reason, in other words, why a cognitive relation such as remembering should not bridge a temporal gap and connect a present mental event with a past event. I agree with H.H. Price when he declares: “Nothing prevents a past event from being presented to consciousness, though it may be a contradiction to say that a past event could be present in the sense of ‘now existing.’”7 Existence includes past existence, and there is therefore no metaphysical difficulty about direct acquaintance with the past (1920, p. 50). I conclude that the co-temporality objection does not succeed in establishing that what is no longer a present existent is incapable of becoming an object of direct remembering.
6 Lukasiewicz develops an existentialist vindication of the presentist treatment of statements about the past: “We should not treat the past differently from the future. If the only part of the future that is now real is that which is causally determined by the present instant, and if causal chains commencing in the future belong to the real of possibility, then only those parts of the past are at present real which still continue to act by their effects today. Facts whose effects have disappeared altogether, and which even an omniscient mind could not infer from those now occurring, belong to the realm of possibility. One cannot say about them that they took place, but only that they were possible. It is well that it should be so. There are hard moments of suffering and still harder ones of guilt in everyone’s life. We should be glad to be able to erase then not only from our memory but also from existence. We may believe that when all the effects of those fateful moments are exhausted, even should that happen only after our death, then their causes too will be effaced from the world of actuality and pass into the realm of possibility. Time calms our cares and brings us forgiveness” (1967, pp. 38–39). 7 1936, pp. 25–26. See also Broad 1925, pp. 251–252; Harvey 1941, pp. 161–162; Laird 1920, p. 50.
72
5 The Primary Objects of Memory
5.3
The Argument from Hallucination
If in remembering I am directly in touch with the past events, as direct realists claim, why are my “memories” so often indeterminate, or accompanied by incorrect details or outright mistaken? Given that how things appear to me in (ostensible) memory is frequently different from how they are, what reason is there to say that the direct objects of memory are the past events rather than my representations thereof? It seems as if direct realism is committed to its being impossible for a memory experience to differ from the event remembered. E.J. Furlong formulates the so-called “argument from hallucination” as follows: The illusory phenomena of memory show that the object as visualized cannot always be numerically identical with the objects as originally seen …. But, it may then be argued, there is no mark whereby we could distinguish between those rememberings in which the object as visualized is numerically identical with the object as seen and those in which it is not. Therefore in no case does this identity hold: the object visualized and the object originally seen are always numerically different (1954, p. 137).
Obviously this objection to direct realism about memory is analogous to the problem the direct realist about perception has accounting for illusions and hallucinations. Since the argument from hallucination, as applied to memory, is very similar to the argument from hallucination, as applied to sense perception, and since much more has been written about the latter than the former it is reasonable if I begin with the former argument. The generic formulation of the argument from hallucination, as applied to sense perception, looks like this: (1) Whenever perceptual experience occurs, there is something that is being perceived. (2) When hallucinating we are not perceiving physical objects. (3) The objects of hallucinations are sense-data. (4) The experiences one has in veridical perception and those one has in visual hallucination are intrinsically indistinguishable. (5) An experience of something can be intrinsically indistinguishable from an experience of sense-datum only if that thing is itself a sense-datum. (6) Therefore, sense-data are the direct objects not only of visual hallucination but also of veridical perception. The upshot of this argument is that when we ordinarily see macroscopic objects, we see them through seeing something else directly: something not seen by seeing anything else. What we see directly are the sense-data. With the exception of (2), every single premise of the argument from hallucination has been questioned. With respect to premise (4), it might be questioned whether there is no experiential difference between veridical and non-veridical perception. Unlike perceptions, hallucinations are chaotic and fantastic. J.L. Austin, for example, declares: [I]t is simply not true that seeing a bright green after-image against a white wall is exactly like seeing a bright green patch actually on the wall, or that seeing a white wall through
5.3 The Argument from Hallucination
73
blue spectacles is exactly like seeing a blue wall; or… that seeing a stick refracted in water is like seeing a bent stick (1962, p. 49).
Yet this objection misses the point. First, there can surely be hallucinations that are intrinsically indistinguishable from veridical perceptions. Second, premise (4) requires only that there be general qualitative similarities between perceptual and hallucinatory contents (Robinson 1994, p. 88). Next consider premise (5). Why should the possibility that hallucination can mimic perception tell us anything about the objects of perception? Why should two things that are experientially indistinguishable have to be of the same kind? It seems this move would render, for example, all forged banknotes genuine, and vice versa. The direct realist holds that, even granted that there is no experiential difference between veridical and non-veridical perception, it doesn’t follow that what we are aware of in regular perception is the same as what we are aware of in hallucination and dreams. In ordinary perception we are aware of physical objects. In hallucination, if we are aware of anything, we are aware of mental entities. Regarding premises (1) and (3), one can argue that there is no reason to suppose that in cases of hallucination there must be some object present that actually has the experienced qualities. Why couldn’t the victim of hallucination simply be in a state of seeming to experience such an object without the object’s actually being there? Premises (1) and (3) rest on what is commonly called the “phenomenal principle”: if there sensibly appears to a subject to be something which possesses a particular sensible quality, then there is something of which the subject is aware which does possess that sensible quality. Some philosophers have taken the phenomenal principle to be obvious. H.H. Price, for example, declares that “when I say ‘this table appears brown to me’ it is quite plain that I am acquainted with an actual instance of brownness” (1932, p. 63; see also Broad 1925, p. 238). It is the phenomenal principle in combination with the argument from hallucination which motivates the view that sense-data possess standard sensible qualities. Given the plausible thesis that perception is a form of mental representation, the phenomenal principle does not hold. It is not generally the case that if a mental state represents that G is F, then there must actually be something which is F. So the direct realist can resist the inference from the fact that it seems as if a sensory quality is instantiated, to the conclusion that it is instantiated. The argument from hallucination, as applied to memory, fails for the same reasons as its perceptual counterpart. It is a mistake to suppose that because I sometimes seem to remember things that did not occur, the primary object of memory cannot be the remembered event itself but must be some representation of the event, a memory-datum. When making a memory claim, I describe an event as I remember it, as it appears to myself in my remembering, and this may very well be different from what actually occurred. Yet the fact that something appears different from what it really is does not rule out that it is not the thing itself that is the primary object of my memory. In other words, the fact that a state of imagining may appear to be a state of remembering, or vice versa, does not mean that I cannot be in direct contact with past events when I remember. How can the direct realist about memory account for hallucination and illusion? He must say that although the experiences may be subjectively the same, what we
74
5 The Primary Objects of Memory
are aware of in genuine memory is different from what we are aware of in hallucination. An ostensible memory and its genuine counterpart are not about the same thing. What we are aware of in genuine memory is different from what we are aware of in ostensible memory. In genuine memory we are aware of past events. In ostensible memory, if we are aware of anything – and some direct realists would deny that we are (it only seems as though we are) – we are aware of mental entities, viz. memory-data. Instances of memory hallucinations do not show that the primary objects of memory are internal representations rather than external events.
5.4
The Causal Theory and Direct Realism
As was explained in the introduction to the chapter, there is a notorious worry that direct realism about memory is incompatible with the causal theory. Given that the causal theory of memory is highly intuitive, if it were incompatible with direct realism, this would speak against direct realism. Fortunately, however, direct realism is not incompatible with the causal theory. The goal of this and the following section is to undermine the incompatibility thesis, that is, the view that direct realism about memory is incompatible with the causal theory of memory. To this end I will examine a variety of arguments for the incompatibility thesis and show that none of them holds up to scrutiny. In his recent book The Threefold Cord Hilary Putnam sketches an argument to the effect that the causal theory of perception is “wholly incompatible” with direct realism about perception (1999, p. 12). Direct realism, Putnam maintains, holds that in perception we are in genuine cognitive contact with reality, that our perceptual experiences are “ab initio encounters with a public world” (1999, p. 40). Yet by postulating causal intermediaries between us and the world the causal theory “makes it impossible to see how persons can be in genuine cognitive contact with a world at all” (1999, p. 11). If perception involves a long and complex causal series of events, the external world is not the direct object of perception. Putnam proposes to resolve the (alleged) tension between direct realism and the causal theory of perception by rejecting the latter.8 And if he is right that the causal theory of perception is “wholly incompatible” with direct realism about perception, the same goes for the causal theory of memory and direct realism about memory. Presumably then
8 Two commentarial points. First, Putnam is not so much concerned with direct realism but with what he calls “natural realism,” a version of direct realism. Here I shall not be concerned with Putnam’s natural realism (which echos the views of James, Wittgenstein, Austin and McDowell), which has already been challenged by Fodor (2000) who criticizes Putnam for urging us to accept a position which he never adequately explains or defends. Second, Szubka (2002) interprets Putnam as saying that direct realism is incompatible with reductive versions of the causal theory of perception, which hold that we can give an exhaustive account of perception in causal terms and can safely dispense with unreduced intentional notions. Non-reductive causal theories, Szubka maintains, escape Putnam’s arguments and are fully compatible with direct realism.
5.4 The Causal Theory and Direct Realism
75
Putnam is also willing to sacrifice the causal theory of memory for direct realism about memory. It doesn’t take much thought to see what is wrong with Putnam’s argument for the incompatibility thesis. Although sensory perception of external objects requires the having (and thereby the existence) of causal intermediaries (sense-data or internal representations) and though such intermediaries in fact determine the way external objects appear to us, there is no reason to suppose we perceive these intermediaries themselves. We perceive, say, a table by internally representing a table, not by perceiving an internal representation of a table. We have to have causal intermediaries to perceive the external world; but what we perceive is the external world, not these causal intermediaries. Putnam seems to be mistaken in supposing that the causal intermediaries are the direct intentional objects of perception rather than merely the vehicles of the perceptual information.9 What is true for perception also applies to memory. Though remembering something may require the having of memory-data, there is no reason to suppose we are aware of these memory-data themselves. I am aware of a past event by internally representing the event, not by being aware of the internal representation of the event. Memory-data do not function as the primary objects of awareness, but are merely the vehicles of the remembered information. Memory is indirect in the sense that it involves a series of causal intermediaries between the past event and the memory experience (memory-datum). But from this it does not follow that memory is indirect in the sense of involving a prior awareness of something other than the past event. Another argument for the incompatibility thesis is put forward by Howard Robinson (1994, p. 86). Applied to memory, the argument looks like this: (7) The memory experience occurs at the end of a causal chain running from the external object to the subject’s brain. (8) The content of the memory experience is part of, or internal to, the experience. (9) If something occurs at a certain location anything which is part of or internal to it occurs at the same location. (10) Therefore, the content of memory experience occurs at the end of a causal chain running from the external object to the brain. (11) Events located at opposite ends of a causal chain cannot be identified. (12) Therefore, the content of the memory experience is not identical with the past event remembered. The point of Robinson’s argument is that the causal theory locates perceptual contents inside the subject’s brain whereas direct realism wants them to be located outside the subject’s skull.
9 This objection to the incompatibility thesis is not new. For example, Strawson declares: “We take ourselves to be immediately aware of real, enduring physical things in space, things endowed with visual and tactile properties …. The immediacy which common sense attributes to perceptual awareness is in no way inconsistent with the causal dependence of [visual experience] on [the things we perceive]” (1979, p. 53).
76
5 The Primary Objects of Memory
What, if anything, is wrong with this argument for the incompatibility thesis? First, conclusion (12) misrepresents the direct realist’s position. Direct realism does not identify memory contents with past events in the sense of reducing the latter to the former; that’s the position of the phenomenalist. According to direct realism, the intentional object of memory “is” the past event in the sense of “being constituted by,” but not in the sense of “being reducable to.” Second, the direct realist rejects premise (8) on the ground that the content of the memory experience – the object of memory – is not in the brain but outside the subject’s head. Properly understood, direct realism about memory is committed to externalism about memory content, that is, the thesis that memory contents depend, at least in part, on relations that the subject bears to certain conditions in the external world. If externalism is true, two people can have memory experiences which are alike in intrinsic nature and are hence introspectively indistinguishable but which nonetheless differ in content. As was explained in section 5.1, direct realism also maintains that a veridical sensory awareness state differs in content from its non-veridical counterpart, even if the two states are subjectively indistinguishable. Given that direct realism holds that a person’s memory contents are – to use Putnam’s phrase – “not in the head,” it is incorrect to say, as premise (8) does, that the contents of memory experiences occur inside the brain. But if memory contents are just as much outside the subject’s skin as the events which are remembered, the incompatibility argument (7)–(12) crumbles.
5.5
A Humean Argument
As was mentioned in section 2.1, Hume famously held that the relation between cause and effect is not a necessary one and that causal relations cannot be known a priori. For Hume these two claims were two sides of the same coin. To say that two events are necessarily connected is just another way of saying that, if one knows that one event occurs, one can predict on the basis of a priori reasoning that the other event occurs as well. A logically necessary connection between two events is one that supports an a priori inference from one event to the other. Today we know that Hume conflated analyticity with a priority. To say that it cannot be known a priori whether A caused B is quite different from saying that even if A caused B, it might not have done so. Just as there are statements that are necessarily true without being knowable a priori, there are statements that are knowable a priori even though they are not necessarily true. According to Hume, whether there is a causal relation between A and B can be learned only a posteriori. Any event can be conceived to follow from the occurrence of a given cause; and the only way in which a particular effect can be inferred from a cause is on the basis of experience, in particular by observation of a regularity between events of the same type. This is the Humean requirement that a cause and its effect are independent existences or distinct events:
5.5 A Humean Argument
77
When I see, for instance, a Billiard-ball moving in a straight line towards another; even suppose motion in the second should by accident be suggested to me, as the result of their impulse; may I not conceive, that a hundred different events might as well follow from that cause? May not both these balls remain at absolute rest? May not the first ball return in a straight line, or leap off from the second in any line or direction? All these suppositions are consistent and conceivable. Why then should we give the preference to one, which is no more consistent or conceivable than the rest? All our reasonings a priori will never be able to show us any foundation for this preference. In a word, then, every effect is a distinct event from its cause. It could not, therefore, be discovered in the cause, and the first invention or conception of it, a priori, must be entirely arbitrary. And even after it is suggested, the conjunction of it with the cause must appear equally arbitrary; since there are always many other effects, which, to reason, must seem fully as consistent and natural. In vain, therefore, should we pretend to determine any single event, or infer any cause or effect, without the assistance of observation and experience.10
When applied to memory, the Humean claim that causal relations can subsist only between distinct events demands that the past event causing (via chain of traces) the memory experience be independent from the experience in two respects: first, it is impossible to know a priori what kind of memory experience a given event will give rise to (or what kind of past event caused a given memory experience); second, there is no logical relation between the past event and the memory experience. Now, it might be argued that direct realism about memory is incompatible with the causal theory of memory because it violates the Humean requirement of distinct existences. The argument I have in mind can be stated as follows: (13) Cause and effect are Humean “distinct events.” (14) According to the causal theory of memory, the content of the memory experience is a Humean distinct event from the object of memory. (15) According to direct realism, the content of the memory experience is not a Humean distinct event from the remembered object. (16) Therefore, direct realism is incompatible with the causal theory. To determine whether there is anything to this argument it is vital to examine premise (15). Given direct realism, is it logically possible for a particular memory experience to exist without the corresponding past event which caused it to have existed? Is it possible to have a memory experience even though nothing corresponds to it in the past? The answer to this question depends on whether Hume is understood to require type-type or token-token independence between cause and effect. Following
10
1748, pp. 29–30. And in the Treatise Hume writes: “[A]s all distinct ideas are separable from each other, and as the ideas of cause and effect are evidently distinct, ‘twill be easy for us to conceive any object to be non-existent this moment, and existent the next, without conjoining to it the distinct idea of a cause or productive principle. The separation, therefore, of the idea of a cause from that of a beginning of existence is plainly possible for the imagination; and consequently the actual separation of these objects is so far possible that it implies no contradiction or absurdity” (1739, pp. 79–80).
78
5 The Primary Objects of Memory
Gerald Vision (1997, p. 45), I suppose the independence Hume has in mind is token-token independence. Thus the question is whether it is logically possible that S has a memory experience of token X without X’s having existed? The answer is obviously “no.” Since memory experiences are individuated according to their contents and since memory contents are determined by certain external conditions, it follows that the experience of a genuine memory cannot be the same as the experience of an ostensible memory. The direct realist will, of course, not deny that genuine and ostensible memory may be indistinguishable from the subject’s point of view. Yet he will insist that introspectively indistinguishable memory experiences may nevertheless differ in content and hence be objectively different. So there is indeed a tension between direct realism, on the one hand, and the Humean idea that causal relations are not logically necessary.11 Next consider the epistemic reading of the Humean requirement of distinct existences: Does direct realism violate the Humean requirement in the sense that either cause or effect can be inferred a priori from the other? In our case, the cause is a past event X and the effect is a memory experience of X. The issue is whether the occurrence of X is inferrable a priori from the occurrence of the memory experience of X. Before this question can be addressed it must be disambiguated. Is the question about events or their descriptions? The reason this disambiguation is important is because the a priori knowability of a causal statement partly depends on the description used to pick out the causally relevant properties of the events involved. And since we can redescribe causes in terms of effects and vice versa, there is no prohibition on a priori relations between causes and effects. Instead of saying “A caused B” we can say “The cause of B caused B” thereby generating a statement which is knowable a priori (Davidson 1980, p. 14). Given that cause and effect can be redescribed in terms of each other, the Humean distinct-existences requirement is presumably about events rather than their descriptions. Following William Child (1994, p. 157), we can reformulate Humean requirement of distinct existences as follows: If A causes B, then the fundamental description of A and B must be such that, under those descriptions, neither one can be inferred a priori from the other. Given these propaedeutic remarks, let’s consider whether direct realism about memory is compatible with the refined Humean requirement of distinct existences. If the effect is characterized as “S’s having a memory experience of X,” is it possible to tell a priori that X occurred? In other words, is there a fundamental description of X and the memory experience of X such that neither can be inferred a priori from the other? Clearly no such fundamental description is available to the direct realist. The effect is indeed inferrable a priori from its cause: if it is known that S has a memory experience of X, then the past presence of X can be inferred, by
11
Strawson declares that “the correctness of the description of a perceptual experience as the perception of a certain physical thing logically requires the existence of that thing; and the logical is thought to exclude the causal connection, since only logically distinct existences can be causally related” (1979, pp. 51–52).
5.5 A Humean Argument
79
reflection alone. Even though direct realism doesn’t allow for a priori inference from putative causes to putative effects, it does allow for a priori inference from putative effects to putative causes. Note that, unlike direct realism, representative realism has no problem in accounting for the Humean distinct-existences requirement. According to representative realism, there is an intermediate step between X’s causal influence and S’s remembering X – namely a memory-datum which can be described as “It seems to S as if he remembers X.” The introduction of this intermediate step ensures that cause and effect are distinct events in Hume’s sense. It is not possible to a priori infer from the fact that it seems to S as if he remembers X that X did obtain (Vision 1993, p. 353). Provided that direct realism violates both readings of the Humean distinct-existences requirement (the analyticity reading and the a priority reading), what should be conclude from this? One option is to reject direct realism, another option is to deny the Humean requirement of distinct existences. I opt for the latter strategy.12 In my view, the presence of logical relations does not bar the presence of causal relations. Consider sexual reproduction. The formation of a new person out of a sperm and an egg is a causal process if anything is. But is the child-parent relation only a causal one or is it also a necessary one? The answer to this question depends on whether one’s parents could not have been other than those individuals they actually are. The question is not whether one could turn out to have parents other than those commonly believed to be one’s parents. Rather the question is whether one could have parents other than those, whoever they may be, one actually has. Saul Kripke (1980, pp. 110–113) has famously argued that having certain parents is an essential property of a person and that every possible world that contains me will contain my parents. That would mean that the child-parents relation is both a causal and a logical one. Not everyone agrees with Kripke’s essentialist idea that an object cannot have had an origin other than its actual origin. But the reason for rejecting essentialism about causal origin is not, I suppose, driven by Hume’s distinctness requirement on cause and effect. The Humean worry that if two events are related in all possible words there is no room left for causal efficacy among them turns out to be a pseudo-problem. We can always re-describe the effect in a way as to make it an entailment of the cause.
12 According to Strawson, the Humean distinct-existences requirement applies to causal relations between physical objects but not to causal relations between a physical object and its perception. “[T]he requirement [of distinct existences] holds for causal relations between distinct objects of perception; but not for the relation between perception and its object” (1979, S. 52). Searle thinks that “intentional causation” is not governed by the Humean distinct-existences requirement. “[I]n every case [of intentional causation] there seems to be a logical or internal connection between cause and effect. And I do not mean just that there is a logical relation between the description of the cause and the description of the effect; but rather that the cause itself quite independently of any description is logically related to the effect quite independently of any description” (1983, p. 121).
80
5 The Primary Objects of Memory
But from this it does not follow that causation is a myth. For even if we chose to describe the effect-event in a different manner it would still follow the cause with the same regularity as before. Causation is a relation between events. Logical relations, however, hold between propositions and linguistic entities. And just because there is a logical relation between the descriptions of two events does not preclude that the events themselves stand in a causal relation. In sum, there are no good reasons to suppose that direct realism is incompatible with the causal theory of perception.
Chapter 6
Against Representative Realism
In the previous chapter we have discussed some of the common objections to direct realism about memory – the co-temporality objection, the argument from hallucination, and the claim that direct realism is incompatible with the causal theory of memory. It turned out that none of these objections holds up under scrutiny. In this chapter I will continue to defend direct realism by showing that the alternative position, representative realism, has most unfortunate epistemological consequences. Representative realism inevitably leads to skepticism about memory knowledge. Section 6.1 explains the skeptical problem which is brought about by representative realism about memory. A popular response to the skeptical challenge involves the posit of so-called “memory markers,” that is, intrinsic features of (seeming) memory experiences which indicate that they provide information about the past. Sections 6.2 and 6.3 examine a variety of accounts of memory markers and show that none of them are convincing. Section 6.4 discusses alternative strategies for neutralizing skepticism about memory knowledge by way of validation of ostensible memories and finds all of them wanting. In the end, the only promising strategy available to the representative realist for neutralizing skepticism about memory knowledge is to adopt epistemic externalism. This proposal is elaborated in chapter 7.
6.1
The Need for Memory Markers
The most widely canvassed objection to representative realism about perception is that it makes the external world unknowable. If what we are immediately aware of in sensory experience is not an external object but a sense-datum, then we are cut off from the external world by a “veil of perception” which neither sense nor reason can pierce. For given the existence of illusions, hallucinations and dreams, it is obvious that the properties of physical objects cannot simply be “read off” from the properties of sense-data. Given representative realism about perception, how can we gain knowledge of external objects on the basis of immediate awareness of sense-data? Following John Locke, representative realists argue that the beliefs about the external world that we S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
81
82
6 Against Representative Realism
arrive at on the basis of our immediately experienced sense-data are justified by means of a causal or explanatory inference. In a nutshell, the idea is that, first, the complicated order of our sense-data demands some explanation and, second, that the best explanation of the complicated order is that the sense-data are caused by the external world which they represent. Given this inference to the best explanation, we are justified in supposing that the world is the way our sense-data represent it to be. Perceptual knowledge rests on the explanation of our immediately experienced sense-data as being caused by, and systematically reflecting, the nature of external objects. According to Locke (1694, pp. 630–639), there are two features of sense-data that call for an explanation: sense-data are said to be involuntarily experienced as well as systematically ordered and coherent. Locke’s favorite explanation is in terms of sense-data being caused by the external world: the reason sense-data simply occur without any choice or control on the part of the person having them is that they originate in the world rather than our imagination. Likewise, the reason sensedata are, by and large, coherent is that the world of independent material objects that gives rise to sense-data is systematically ordered. Now even if it is granted that the involuntary nature and systematicity of our sense-data demand some sort of explanation it is an open question whether the explanation provided by the representative realist is indeed the best one. Prima facie it may seem that Berkeley’s explanation of sense-data in terms of God is just as good as Locke’s explanation in terms of the external world. Laurence BonJour (2007) has tried to improve on Locke’s justification of perceptual beliefs about the external world by pointing out that a sequence of sensedata is at times incomplete and fragmentary. The most straightforward explanation for why a sequence of sense-data is incomplete and fragmentary, according to BonJour, is that there is an independent realm of objects outside our experience, which has its own patterns of order, and that our perceptual contact with that realm is partial and intermittent. Notwithstanding the intuitive plausibility of BonJour’s proposal, it is unlikely that representative realists will be able to come up with conclusive reasons for preferring an explanation of sense-data in terms of the external world over alternative explanations, such as Berkeley’s view that sense-data are caused by God. The representative realist about perception is faced with the difficult task of reasoning his way to the external world. Thomas Reid realized that the same problem exists for representative realism about memory. According to representative realism about perception, what we are directly aware of when we perceive things are not the objects in the external world but internal representations which are produced in our minds by those external things. Similarly, the representative theory of memory claims that what we are directly aware of when we remember, are not the things we experienced in the past, but present internal representations of those past things. The difficulty for both representative theories is that in restricting what we are directly aware of to internal representations, they equally restrict what we can know about the external world. For if all we are directly aware of are our representations or ideas, how can we know that there are external things at all, much less that these
6.1 The Need for Memory Markers
83
external things are the causes of our present ideas? Reid saw that the representative theory cuts us off from the only way we have of telling whether a particular experience is caused externally and what caused it. He writes: It may be observed, that the common system, that ideas are the only immediate objects of thought, leads to scepticism with regard to memory, as well as with regard to the objects of sense…. Ideas are said to be things internal and present, which have no existence but during the moment they are in the mind. The objects of sense are things external, which have a continued existence. When it is maintained that all that we immediately perceive is only ideas or phantasms, how can we, from the existence of those phantasms, conclude the existence of an external world corresponding to them? … The same difficulty with regard to memory naturally arises from the system of ideas; … For since ideas are things present, how can we, from our having a certain idea presently in our mind, conclude that an event really happened ten or twenty years ago, corresponding to it? There is the same need of arguments to prove, that the ideas of memory are pictures of things that really did happen, as that the ideas of sense are pictures of external objects which now exist. In both cases, it will be impossible to find any argument that has real weight. So that this hypothesis leads us to absolute scepticism, with regard to those things which we most distinctly remember, no less than with regard to the external objects of sense (1785, pp. 289–290).
The question which occupied John Locke, arising out of representative realism about perception “How shall the Mind, when it perceives nothing but its own Ideas, know that they agree with Things themselves?” (1694, p. 563), has its counterpart within the representative theory of memory: How do we know that our memory experiences portray things the way they happened in the past? It seems that to know that a memory-datum of mine represents what actually happened I would have to first establish what happened and then check the datum against the facts. But how can I do this if, as the representative theory insists, the direct objects of memory are internal representations? To discover whether something is an authentic memory-datum, I would have to inspect it from an external point of view, but, according to the representative theory, the only way of finding out what happened is via memory-data. Thus the representative theorist finds himself imprisoned within his memory-data, with no way of confirming that the data reveal the past, as they have to if he is to have memory knowledge. He doesn’t seem to be able to justify a claim to remember without falling into a petitio principii. Many philosophers have followed Reid in charging that the representative theory renders memory knowledge impossible. Bertrand Russell (1921, p. 159), for instance, took Reid’s criticism of representative realism as a basis for his claim that it is possible to rule out, on the basis of memory experiences, the skeptical hypothesis that the world sprang into being 5 minutes ago, exactly as it then was, with a population that “remembered” a wholly unreal past. (This skeptical hypothesis will be discussed in section 7.5.) In response to Reid’s and Russell’s skepticism about memory knowledge, many advocates of representative realism have maintained that a subject can tell, by reflection alone, whether a particular internal representation of his qualifies as a genuine memory – whether a memory-datum is authentic. The subject is said to know a priori whether a representation he is currently experiencing originates from the memory or from some other mental faculty, such the imagination. The prime
84
6 Against Representative Realism
motivation for this claim is, of course, the attempt to meet the skeptic’s challenge. If there is an a priori knowable feature of authentic memory-data that stamps them as such, we can tell whether or not a particular memory claim amounts to knowledge. This feature of memory-data that stamps them as authentic has been called “memory marker” or “memory indicator.” A memory-marker is a property of a memory-datum which indicates that it represents what has happened in the past and thus provides information about the past. Memory markers provide the subject with something from which it can be “read off” that an experience is a memory experience, rather than an experience of an act of imagining. R.F. Holland explains the reasoning behind the stipulation of memory markers as follows: [I]t seems that the connexion with a past reality which distinguishes memory from imagination does not merely have to exist; it also has to announce its existence to the person who remembers at the time when he remembers. Had it existed clandestinely … we should never have learned to distinguish them: the difference between them would have been useless, would have meant nothing to us. So it seems necessary to add that, whenever anything comes into our minds, we must have some means of knowing in which cases the connexion with reality exists and in which cases it does not. That is to say, it seems that there must be some mark or sign whereby a remembering state of mind can be distinguished from an imagining state of mind (1954, p. 465).
Memory markers are meant to allow us to know whether and when we remember something as opposed to imagining it; they are the signs by means of which the connection with the past announces itself. Memory markers are supposed to be the cure for skepticism about memory knowledge. It is important to note that skepticism about memory knowledge assumes internalism about justification. In its broadest formulation, internalism about justification is the view that all of the factors required for a belief to be justified must be cognitively accessible to the subject and thus internal to his mind. Something is internal to one’s mind so long as one is aware of it or could be aware of it merely by reflecting. Beliefs are the sort of thing that are internal to the mind. Internalists therefore hold that it is the subject’s other beliefs which are relevant to the question of whether he is justified in believing something. Externalism about justification, on the other hand, is simply the denial of internalism, holding that some of the justifying factors may be external to the subject’s cognitive perspective. A belief is justified if it has the property of being truth-effective. This property of being truth-effective may, for example, consist in the belief’s being produced by a reliable method or process. No more than this is necessary for justification. Whether the subject takes his belief to be truth-effective doesn’t add anything to the belief’s epistemic status. The chief motive for internalism is epistemic deontology, that is, the idea that when one is justified in holding a belief, one has fulfilled one’s epistemic duties and obligations in forming that belief. The epistemic duties arise from the goal of believing what is true and not believing what is false. The deontological notion of justification is normative; justification is not something that happens to a person; it is something one achieves by one’s own efforts. If, therefore, justification is a function of meeting obligations, the factors that determine whether a belief is justified must be internal to the subject’s mind.
6.1 The Need for Memory Markers
85
The skeptic about memory knowledge raises the question how you can know (or justifiably believe) that the memory experiences on which you base your memoryclaims are in fact an accurate, a reliable, guide to the past. How can you be sure that the things you seem to remember are not just your inventions but happened in the past? If there is a positive answer to the skeptical question, then, given internalism about justification, it will involve reference to a priori knowable memory makers that stamp memories as such. The externalist, on the other hand, can acknowledge that he does not know that his memory-data are indeed accurate rather than fabricated, but this does not matter; this does not undermine his memory knowledge. For to know something on the basis of a memory-datum a person does not have to know (or justifiably believe) that the memory-datum is a reliable indicator of the past; it just has to be a reliable indicator of the past. If the memory-datum is reliable, it gives us memory knowledge; if it is not, then it does not. (The externalist response to skepticism will be discussed in sections 6.4 and 7.1.) There are two ways to understand the skeptical question “How do you know that you remember something, as opposed to imaging it?” According to one interpretation, I am asked to prove that the events as recounted by me did in fact occur. On the other interpretation, I am asked to establish that, as a matter of psychology or phenomenology, I was not making things up; I was not fancying. I was describing what I sincerely believed took place. Given these two readings of the skeptical question and given that memory markers are meant to provide an answer to the skeptical question, there are two corresponding readings of memory markers. On the one hand, memory markers are supposed to set apart veridical memories from ostensible ones (as well as products of the imagination). On the other hand, memory markers are said to enable us to differentiate between two kinds of cognitive acts – (veridical or ostensible) remembering and imagining. Frequently representative realists don’t say which of these interpretations of memory markers they have in mind. Discussing Hume’s account of memory markers, David Pears explains: One distinction is between remembering and imagining regarded as two things that we do, two different activities. For example, my wife visits me in hospital and asks me to remember what happened when I drove through the red stop-light…. The other distinction is a subdivision of cases where I am trying to remember. I may remember the accident correctly…. Alternatively, I may get it all wrong, in which case she would say, “No, you are just imagining that it happened like that.” … It is one thing to distinguish two activities, and quite another thing to take one of them and distinguish between its correct and incorrect performances (1990, pp. 40–41; cf. Shoemaker 1967, p. 267).
The distinction between remembering and imagining as two kinds of activities is, as Pears puts it, a distinction concerning meaning; the distinction between correct and incorrect remembering is a distinction concerning the truth or success of an act of meaning. Pears’ conclusion is that Hume and other proponents of representative realism fail “to start with questions about meaning and settle them first, before going on to questions about truth” (1990, p. 40). It is one thing to know what kind of cognitive activity one is engaging in; it is quite another to know whether one succeeds in this activity.
86
6 Against Representative Realism
The difference between veridical and ostensible memories is an external one: veridical memories are related to reality in ways that ostensible memories are not. Unless one adopts an anti-realist account of truth such as the coherence theory or the pragmatic theory, external or relational features of memory-data cannot be detected “from within” and hence cannot be known a priori. Thus memory markers cannot both be a priori knowable and, at the same time, indicate the difference between veridical and ostensible memory. In fact, none of our factive attitudes are such that we can know, by reflection alone, whether we are having them. For example, knowledge is not transparent because the subject cannot discriminate knowledge from mere belief just on the basis of reflection. And just as I cannot know a priori whether I know some worldly proposition, I cannot know a priori whether I genuinely remember something about the past. Memory markers at best indicate the kind of cognitive activity one is engaged in, but not whether one succeeds in this activity. How can one tell whether a particular representation one is having is an instance of (veridical or ostensible) remembering or imagining? For the representative theorist the answer is that one can tell on the basis of the intrinsic (qualitative) properties of the memory-data themselves or, at least, on the basis of intrinsic properties invariably associated with the memory-data. The alternative position would be that we rely on extrinsic properties of memory-data to differentiate them from imagination-data. The problem with the alternative view is two-fold. First, it is phenomenologically inaccurate, for we do not proceed in this manner. Rather than inquiring, collecting evidence, and drawing conclusions, we seem to know straight-off whether a representation is of the memory or of the imagination. Second, the use of extrinsic properties to determine the cognitive activity underlying a given representation involves reliance on memory. The reliance on external properties must ultimately be based on the assumption of memory’s general trustworthiness. But the trustworthiness of memory is precisely the object of the skeptical doubt. In resting the distinction between memory-data and imagination-data on extrinsic properties, the advocate of representative realism opens himself to the charge of begging the question. There are at least two ways in which the use of extrinsic properties to determine the cognitive activity underlying a given representation one is having relies on the trustworthiness of memory. First, the subject can bring the testimony of other people to bear in determining the nature of his representation only if he has good reason to count that testimony as accurate. And, whatever reason he has for so counting it must depend, in the end, on what he has learned about them in the past and so on his memory. Second, once the subject has gathered external evidence, he must draw his conclusions on the nature of his representation and the truth of his claims. If his inference is complex, he will no doubt have to rely on his memory for knowing the earlier parts of the argument. Thus, if the reasoning is lengthy, he can be sure of the conclusion only if his memory testifies to the soundness of the earlier steps in the inference. In sum, if a positive (non-skeptical) solution to skepticism about memory knowledge is to be had, then, given representative realism and epistemic internalism, we will have to seek out intrinsic properties of memory-data that stamp them as such. What are those memory markers?
6.2 The Varieties of Memory Markers
6.2
87
The Varieties of Memory Markers
Arguably the most famous discussion of memory markers is that of David Hume. Hume proposed two criteria by which to distinguish ideas of memory from ideas of pure imagination. The formal criterion has to do with the amount of free-play we might have with our ideas.1 In recalling something we are not free; rather we are bound by the necessity of how things actually were at the time when we experienced them. Memory not only has the function to “preserve the simple ideas” but also to preserve “their order and position” (1739, pp. 9–10). In imagination, however, we are free to alter and transpose, join and divide, as we like. This freedom concerns both the temporal order of the original impressions and the non-temporal arrangement and complexities of the impressions. “[T]he imagination is not restrain’d to the same order and form with the original impressions; while the memory is in a manner tied down in that respect, without any power of variation.” And Hume adds that “winged horses, fiery dragons, and monstrous giants” which we meet in fables and poems “put this entirely out of question” (1739, pp. 9–10). The ideas of the memory and ideas of the imagination are both complex in that they are composed of simple ideas and “all our simple ideas in their first appearance are deriv’d from simple impressions.”2 Both kinds of ideas trace their ancestries to simple impressions. While the simple ideas composing a complex idea of the imagination might originally have been caused by any number of simple and temporally separated impressions, the causal history of a complex idea of the memory must be traceable to a single complex impression which resembles that idea. Thus the formal criterion is basically a causal thesis: for a complex idea to be an idea of the memory, it must be causally derived from a complex impression that it resembles and represents. The formal criterion fails for two reasons, one of which was already realized by Hume himself. First, this criterion is not useable, for “it being impossible to recall the past impressions, in order to compare them with our present ideas, and see whether their arrangement be exactly similar” (1739, p. 85). Clearly, we cannot tell on the basis of introspectively available evidence whether an idea preserves the order of the original impressions simply because the original impressions are no longer available for comparison. To make this comparison we should have to use external evidence such as photographs, diaries, etc. And, of course, if we use external sources for comparing the past to our present ideas about it, the ideas would
1 “Formal criterion” is Flage’s term (1985a, pp. 168, 186n); Noxon (1976, p. 271) calls it an “epistemic criterion.” 2 Hume 1739, p. 4. Given the proposed interpretation of the formal criterion, there cannot be simple memory ideas. Ideas of the memory and imagination must be complex. This is Flage’s position (1985a and b) and he cites Hume (1739, p. 85) in support of it. For a differing opinion see Traiger (1985, p. 168–169) and Friedman (1993).
88
6 Against Representative Realism
be useless as sources of information, since we need only to inspect the actual past for our knowledge of it; an inference from present ideas to the past would be unnecessary. Second, it doesn’t seem right to say that when we remember things, as opposed to imagining them, we always remember them in the order in which they were originally experienced. I may very well remember last night’s dinner by first remembering the desert, then the main course, and finally the appetizer. Hume’s second criterion for the distinction of memory ideas from ideas of imagination – the phenomenal criterion – employs the same concepts by which he also distinguishes impressions from ideas and beliefs from conceptions: liveliness and vivacity. Hume does not define “vivacity” but wherever he explains it he does it by suggesting that greater vivacity consists in brighter, more vivid, and more precise detail.3 According to Hume, there are differences in the degree of vivacity between a perceptual experience of X, a memory of X, and a fantasy of X occurring; and, indeed, this is the only difference between these kinds of ideas. The perception of the event is particularly vivacious, the memory less so, and the fantasy of the event is the least vivacious of the three. The ideas of memory, Hume insists, are “much more lively and strong than those of the imagination” and have a “superior force and vivacity,” that is, they “paint their objects in more distinct colours.” “A man may indulge his fancy in feigning any past scene of adventures; now wou’d there be any possibility of distinguishing this from remembrance of a kind, were not the ideas of the imagination fainter and more obscure” (1739, pp. 9, 85). Since Hume holds that it is precisely the force and vivacity of ideas that constitutes belief, the difference between memory and imagination may also be expressed by saying that whereas memory involves belief (namely, in the previous occurrence of a train of impressions having an “order and form” corresponding to that of the ideas), imagination does not. In the “Appendix” to the Treatise of Human Nature Hume calls force and vivacity “feelings.” Were it not for these feelings, there would be no difference between “the mere reveries of the imagination” and “the most establish’d truths founded on history and experience. There is nothing but the feeling, or sentiment, to distinguish the one from the other” (1739, p. 624). And he continues: “We may venture to recommend to the consideration of philosophers…. Whether this feeling be any thing but a firmer conception, or a faster hold, that we take of the object” (p. 627). Though Hume took the phenomenal criterion to provide a basis for distinguishing between ideas of memory and ideas of the imagination, he was fully aware that this criterion is far from being infallible. He concedes that the memory ideas, “by losing its force and vivacity, may degenerate to such a degree, as to be taken for an
3 Hume laments that “vivacity” inadequately expresses the quality he refers to: “But when I wou’d explain this manner, I scarce find any word that fully answers the case, but am oblig’d to have recourse to every one’s feeling, in order to give him a perfect notion of this operation of the mind. An idea assented to feels different from a fictitious idea, that the fancy alone presents to us: And this different feeling I endeavour to explain by calling it a superior force, or vivacity, or solidity, or firmness, or steadiness” (1739, p. 628–629).
6.2 The Varieties of Memory Markers
89
idea of the imagination” and, in turn, ideas of imagination can grow in their vivacity (1739, p. 86). Indeed, dreams, reveries, and hallucinations can often be exceedingly vivid, far more than memory impressions. A second problem with the phenomenal criterion is that liveliness and vivacity are matters of degree and can be placed on a continuum from very faint to very distinct and detailed. There is no clear cut-off point on that continuum which would need to exist if Hume’s distinction were to be employed. Third, on Hume’s view, memories are more vivacious than imaginations and less vivacious than perceptions. While it might be plausible to suppose that the memory of a perception has less force and vivacity than the original perception it is highly questionable that the memory of an imagination surpasses the original imagination in liveliness. Fourth, vividness as such is no indication at all that the idea in question stands for some previous experience. Even if it were true that all memory ideas are particularly lively, vivacity could still not be used as a criterion for distinguishing memories from imaginations – unless it had been established that vivacity indicates pastness. To confirm that vividness is a sign for pastness we would need to show that genuine memories are more often vivid than phantasies. But how do we know that there is a relation between vivacity and pastness? Do we remember that such a relation holds? If so, how did we originally learn that it holds? And, indeed, how is it that we are able to remember the fact that vivacity indicates pastness? It seems that if we have any grounds for supposing that a given idea originates from memory rather than imagination, we need to have some means of authentication beyond the idea itself. Fifth, even if vivacity were a reliable memory marker and if we knew about it, the knowledge regarding this marker would itself need to be retained in memory in order for it to be applicable as a criterion for telling memories apart from imaginations. But now we have a problem of remembering what it is that marks the memory marker. The representative realist needs not only memory markers, but also memory marker markers, and memory marker marker markers, and so on. Given that neither the phenomenal nor the formal criterion amounts to a reliable tool for discriminating ideas of memory from ideas of imagination, what should we make of Hume’s theory of force and vivacity? Maybe Hume was not so much concerned with the question of what distinguishes memory from imagination, as with the pragmatic issue of how we go about differentiating ideas of memory from ideas of imagination. For the fact is that we do use, rightly or wrongly, considerations of vividness to make these kinds of differentiations. Following in the Humean tradition, Brian Smith has suggested memory markers not unlike force and vivacity. Smith claims that there are three memory markers: firmness, controllability, and expansion into context. As for the first, Smith asserts that memory-data are generally more “firm,” “self-sufficient,” “unitary,” and “constant” than imagination-data. Concerning the second property, controllability, Smith says that we are free to change our imagination-data. By contrast, when he tries to alter a memory-image Smith records the following impression: “I am immediately aware of having “taken charge of” my image. The [original image] will tend to reassert itself as soon as I discontinue my deliberate fantasy” (1966, p. 142). The
90
6 Against Representative Realism
third distinctive property of memory-images is that they naturally lead to a wider context; they grow compulsively to include additional background details. Imagination-images, on the other hand, do “not so much expand as simply change. We have no feeling of … compulsion; the development, if any, is haphazard and voluntary” (1966, p. 142). Smith cautiously frames his proposal as no more than a specification of the properties which prompt us to accept images as being of the memory rather than the imagination. He disavows any attempt to provide a “rigid distinction of kinds” (1966, p. 141). The Humean theory of memory markers more or less corresponds to empirical research regarding the detectable differences between ostensible memories derived from veridical perception versus ostensible memories generated by imagination. Marcia Johnson and Carol Raye (1981, see also Johnson 1988) list four differences. Unlike ostensible memories that originate from imagination, those originating from perception tend to have more perceptual information, more contextual information about time and place, and more meaningful detail. Ostensible memories originating from the imagination, on the other hand, tend to contain more information about the cognitive operations involved in the original imaginings. Differences in average value along these criteria guides us in our decisions whether the origin of a seeming memory is perceptual or fantastical. Unlike Hume who identifies the mark of memory as a property which inhabits the imagery, representative theorists like William James, Bertrand Russell, and Robert Audi maintain that the mark of memory is a property which invariably accompanies the imagery. James suggests that it is the feelings of “warmth and intimacy” and of “the past direction of time” which distinguishes memory from mere imagination (1890, I, p. 650). Russell’s candidates for memory markers are the feelings of familiarity and of pastness. And Audi (1995, pp. 35–36) maintains that a memory belief that p is accompanied by a feeling of having believed p. All three philosophers appeal to common experience to verify their observations, and suggest that if we examine our own memory-data we will find that they do, indeed, have these qualities. Since it is impossible to check the accuracy of a present memory image by comparing it with a past sensation Russell infers that “[o]ur confidence or lack of confidence in the accuracy of a memory-image must … be based upon a characteristic of the image itself” (1921, p. 161). According to Russell, memory images are distinguished from other kinds of images by two feelings that accompany them – the “feeling of familiarity” which leads us to trust the image and the “feeling of pastness” which leads us to refer the image to some incident in the past. Quoting Russell on these points: We may say, then, that images are regarded by us as more or less accurate copies of past occurrences because they come to us with two sorts of feelings: (1) Those that may be called feelings of familiarity; (2) those that may be collected together as feelings giving a sense of pastness (1921, p. 163; cf. 1927, pp. 198, 203–204).
Depending on the intensity of the feelings of pastness and familiarity, we refer the image to more or less remote times in the past. Russell also speaks of a “feeling of
6.2 The Varieties of Memory Markers
91
belief” as a constituent of memory; but this feeling is apparently nothing distinct from the feelings of pastness and familiarity.4 As in the case of Hume, it is not clear whether Russell conceived of the feelings of familiarity and pastness as phenomenological facts or as epistemic criteria. In other words, it is not clear whether the feelings of familiarity and pastness are supposed to be the mark of the images that, as a matter of fact, we take to be memory images or whether they are the mark of the images we are entitled to take to be memory images. Presumably it is the latter reading Russell has in mind. That the feeling of familiarity plays an important role in our assessment of the sources of our representations is a psychological fact. The feeling of familiarity belongs to the category known to psychologists as metacognitive signals. A metacognitive signal is a conscious state which makes information about unconscious mental processes available to the subject. The feeling of familiarity carries information about how swiftly a particular representation is retrieved. By and large, when a representation enjoys fluency of access it is deemed familiar and therefore considered a piece of memory rather than imagination. The psychologists Aaron Benjamin and Robert Bjork write: “How readily information ‘comes to mind’ is one index humans use to assess the accuracy of that information, and more generally, the adequacy of their own state of knowing in a given domain” (1996, p. 309). Though fallible, fluency and familiarity are indeed correlated with states of ostensible memory. “Such retrieval fluency provides, in fact, a useful heuristic: In general information that is better learned, more recent, and more strongly associated with the cues guiding recall (or any combination of the three) will tend to be more readily retrievable” (ibid.). It is an undeniable fact that some of our memory experiences have a recognizable feeling of familiarity about them, and that it is that which distinguishes them from other experiences. The trouble with Russell’s proposal, as with others of its kind, is that it does not offer a reliable mark. There are cases in which these memory markers are present, but in which there is no inclination to attribute memory and there are instances where genuine memories lack the feeling of familiarity. Suppose, for example, that you have imagined something so often before, wishing that it had happened, that the images themselves become familiar, and you forget that you only wish it had happened. In this case you might think you remember something, but you are wrong because you are only imagining it. C.D. Broad also takes the feeling of familiarity to be a memory marker. Familiarity is defined as the “peculiar feeling which seems to justify the judgement that a certain characteristic fits or fails to fit a certain past object” (1925, p. 249). But, unlike Russell, Broad explicitly says that the feeling of familiarity which
4 Russell writes: “Memory-images and imagination-images do not differ in their intrinsic qualities, so far as we can discover. They differ by the fact that the images that constitute memories, unlike those that constitute imagination, are accompanied by a feeling of belief which may be expressed in the words “this happened.” The mere occurrence of images, without this feeling of belief, constitutes imagination; it is the element of belief that is the distinctive thing in memory” (1921, p. 176).
92
6 Against Representative Realism
accompanies memory images is not an infallible sign for something’s being a memory image rather than, say, an imagination-image. In speaking of familiarity, Broad explains that “we are so constituted” that when a memory image “manifests the characteristic of familiarity, we inevitably apply the concept of pastness.” Familiarity “means” pastness “to such beings as we are” and this meaning is “primitive and unacquired” (1925, pp. 266–267). Another problem with Russell’s theory of memory markers is that familiarity and pastness are matters of degree. For these memory markers to be applicable we would need to know to what degree an image has to appear familiar for us to be entitled to conclude that it is a memory image. Yet, for obvious reasons, Russell is not in a position to specify the threshold of familiarity required for a memory claim to be epistemically justified.5 A third problem with Russell’s theory of memory markers is that familiarity as such – as with force and vivacity as such – is no indication for pastness. R.F. Holland phrases this point nicely when he asserts that “the crux of the matter is that [such a property] as familiarity does not, as it were, bear its own explanation upon its face” (1954, p. 468). What Holland is getting at is that the association between images that strike us as familiar and memory is not an epistemic reason for anything. For the mere fact, if it is one, that we are inclined to make this association does not imply that we are entitled to make it. To be so entitled, we would need independent evidence suggesting that genuine memories appear familiar more often than phantasies. If such evidence exists at all, it cannot be accessed by reflection alone. Finally it is far from obvious that the experience of familiarity is a sui generis feeling. After having studied Russell’s Analysis of Mind, Ludwig Wittgenstein recorded passages in his Brown Book suggesting that the experience of familiarity, rather than being any positive feeling, consists only in the awareness of absence of feeling, the absence of surprise or unfamiliarity, the absence of the confounding of expectations.6 Wittgenstein points out that the experience of familiarity does not recur whenever we are confronted with a familiar-seeming object. Pondering about the question whether there is such a feeling [of familiarity] or not, we are likely to gaze at some object and say, “Don’t I have a particular feeling when I look at my old coat and hat?” But to this we now answer: What feeling do you compare this with, or oppose it to? Should you say that your old coat gives you the same feeling as your old friend A with whose appearance too you are well acquainted, or that whenever you happened to look at your coat you get that feeling, say of intimacy and warmth (1960, pp. 180–181; cf. 1980, I, §120–121; 1967b, p. 596)?
5 Pears (1975b, p. 238) and Warnock (1987, p. 23) argued that memory cannot be explained in terms of familiarity because familiarity presupposes memory. One cannot explain what one means by saying something is familiar except by saying that one remembers or seems to remember it. Following Woozley (1949, pp. 46–47), I have doubts that this objection is valid, for there is no reason to suppose that the notion of memory is an indispensable part of the notion of familiarity. 6 Regarding Wittgenstein’s acquaintance with Russell’s Analysis of Mind see Hacker (1990, p. 232), Wittgenstein (1960, pp. 21–22), and Wittgenstein (1975, p. 64).
6.3 Remembering and Trying to Remember
93
According to Wittgenstein, the experience of familiarity, instead of being a sui generis feeling, is an awareness of unhesitating behavioral response to an environment. Wittgenstein’s favorite example is that one’s ability to handle objects and to produce the appropriate bodily movements depends on whether one is in a familiar room or an unfamiliar one. He suggests that the feeling of familiarity one experiences in a familiar room is nothing but the free and unhesitating manner in which one operates inside it: I know my way about in a room: that is, without needing a moment’s reflection, I can find the door, open and shut it, use any piece of furniture, I don’t have to look for the table, the books, the chest of drawers or think what can be done with them. That I know my way around will come out in the freedom with which I move about in the room. It will also be manifested in an absence of astonishment or doubt (Wittgenstein 1980, I, §295; cf. 1960, p. 182).
Henri Bergson also suggests that the sense of familiarity is nothing but a type of bodily recognition that consists in appropriate action in response to an object or environment: There is in the first place, … an instantaneous recognition, of which the body is capable by itself, without the help of any explicit memory-image. It consists in action and not in representation. For instance, I take a walk in a town seen for the first time. At every street corner I hesitate, uncertain where I am going. I am in doubt, I mean by this that alternatives are offered to my body, that my movement as a whole is discontinuous, that there is nothing in one attitude which foretells and prepares future attitudes. Later, after prolonged sojourn in the town, I shall go about it mechanically…. [M]ust we not assume that the consciousness of a well-regulated motor accompaniment, of an organized motor reaction, is here the foundation of the sense of familiarity? (1896, pp. 92–93).
Bergson suggests that this question should be answered in the affirmative. Organized motor reactions, he claims, are not only the “foundation” of the sense of familiarity but they are the sense of familiarity. Irrespective of whether the experience of familiarity is a sui generis feeling, it begins to show that nothing is going to provide the desired intrinsic feature that shows that a memory experience is indeed a memory experience. As long as one can be mistaken about whether one is (veridically or ostensibly) remembering, as opposed to imagining, it doesn’t seem possible to identify an intrinsic feature of memory experiences that stamps them as such.
6.3
Remembering and Trying to Remember
J.O. Urmson (1967, 1971) has proposed an account of memory markers that is different from Hume’s and Russell’s. Urmson suggests that memory images can be distinguished from imagination-images by one’s intentions, more precisely, by the criteria of success one has chosen to apply to one’s efforts. We are remembering if we think that agreement with the facts matters, but imagining if we think that only such criteria as general verisimilitude or interestingness are the relevant ones.
94
6 Against Representative Realism
Urmson (1967, p. 86) begins his account by drawing out two senses of the words “imagine” and “remember.” “Imagine” sometimes means “falsely believe” (imagine1) and sometimes means “creatively invent” (imagine2). Analogously, in addition to the agreed sense in which I can remember only what is the case (remember1 or genuine memory), there is another sense in which so-and-so can be “my genuine recollections of what happened, however imperfect they may be” (remember2 or ostensible memory). It is the second sense of each term that is of importance for Urmson’s purpose. To imagine2 is to make something up as a piece of fiction. In this sense of the term, truth and falsity have no application. Rather it is the general verisimilitude or interestingness of the fiction that serves as the criterion of success for imagining2. Similarly, the criterion for success for remembering2 is one’s intention rather than the satisfaction of the truth conditions of the memory content. If it matters to a person that a certain incidence occurred, then, whether or not it did occur, he is in a position to remember2 that it did occur. Regarding the first sense of the words “remember” and “imagine,” Urmson admits that it is impossible to know, by reflection alone, whether one is remembering1 or imagining1. “We cannot know from the description of an activity whether it is an activity of accurate recollection or of delusive imagination” (1967, p. 89). But, Urmson claims, we can know a priori whether we are remembering2 or imagining2 by virtue of knowing a priori our intentions. Suppose we are trying to decide whether we are remembering2 or imagining2 ourselves conducting the defense in a trial. To tell which is which [a]ll we have to do is to know what criteria of success are applicable, and that is a question which depends upon our own intentions. We are [remembering2] not if we did conduct the defense in the trial but if it matters whether we did. We are [imagining2] if some such criteria of success as general verisimilitude, or interestingness are the relevant ones (Urmson 1967, pp. 87–88).
The idea is that the difference between remembering and imagining can be known immediately, without collecting evidence. “Evidence is unnecessary because I know straight off whether I am, say, daydreaming or story-telling on the one hand or recollecting (well or ill) on the other” (1967, p. 86). This idea goes back to R.F. Holland who argued that it is logically impossible that one could imagine something, in the sense of “reigning some past scene of adventure” and at the same time think one was remembering (1954, p. 471). To further illustrate the criteria of verisimilitude and interestingness, Urmson provides the following example: If one produces “a mental picture of a three-legged, eleven-eyed monster” it will be a case of imagination[2] if the picture can be criticized on the ground that the eyes do not look very like eyes or that some of the legs look more like arms, but cannot be criticized on the ground that [one] has never seen anything like that (1971, p. 607).
Urmson maintains that the intention-based criterion for distinguishing remembering2 from imagining2 is “infallible” (1967, p. 89). Following Urmson, Stanley Munsat maintains that, in the case of imagining2, “a person cannot be mistaken about what he is imagining (unless this is a mistake about what that which he is imagining is called)” (1967, p. 95).
6.3 Remembering and Trying to Remember
95
Urmson’s proposal for distinguishing remembering2 from imagining2 is open to criticism. First, the proposal overestimates the extent to which remembering2 and imagining2 are voluntary and deliberate. Not every instance of remembering2 is preceded by an act of deliberation and intentional planning. Frequently remembering2 is not something we do, but something that happens to us. And when memories2 arise without any conscious decision, Urmson’s criterion for distinguishing imagining2 from remembering2 cannot get a foothold. Now Urmson might respond that in the case of involuntary memories2, we are in a position to detect the kind of intention that would underlie the episode of remembering2 if it were brought about voluntarily. Yet detection of possible intentions is certainly not infallible. Second, even if our intentions are transparent to us, they don’t tell us whether we are remembering2 or imagining2. At best, our intentions tell us whether we are trying to remember2 or trying to imagine2 (Furlong 1970, p. 137; White 1990, pp. 168–169). To drive this point home, suppose that I am taking myself to remember2 having climbed the Eiffel Tower but that, as a matter of fact, I have never been to Paris. Am I remembering2 or imagining2 having climbed the Eiffel Tower? According to Urmson, I can settle this question by recalling what criteria of success I have decided to adopt. But just because I cared about the truth rather than the interestingness of the representation of my climbing the Eiffel Tower does not mean that it is a memory2. A memory, however imperfect or faulty, is linked to an original experience in ways that imagination is not. I can no more choose what I remember2 than I can choose what I remember1. Just as with remembering1 and imagining1, the difference between remembering2 and imagining2 depends, in part, on the relationship between the present experience and what has happened in the past. In the words of E.J. Furlong, “remember is an achievement verb, not a trying verb; and achievement cannot be settled by prior decision” (1970, p. 138). Third, contrary to Urmson’s conviction, intentions are not always transparent to the mind and hence not infallibly knowable. It is possible that one should mistakenly take oneself to be remembering2 something when one is really imagining2 it. Experiments in social psychology show that subjects regularly fail to identify causally efficacious stimuli, and regularly appeal to causally inert stimuli in the explanation of their intentions. In one such experiment subjects were asked to rate the quality of four pairs of identical nylon pantihose. The rightmost pantihose was heavily preferred. Presumably what made the subjects choose this pantihose was that fact that it was placed on the right-hand side. Speakers of Indo-European languages read from left to right. Whatever is on the far right is that which is seen last and so has a preeminent status in regard to choice. Yet the subjects resolutely denied that their choices were influenced by the position of the objects. And when subjects were asked to give an account of the considerations which influenced them in their ranking of pantihose, and when the requests were made “within seconds of their occurrence,” they gave an account which can be shown to be at variance with what in fact must have been the thought process behind their behavior (Nisbett and Ross 1980, pp. 248 ff.). Given that we can err concerning why we prefer what we prefer, it is surely possible that we are also wrong about what it is we prefer or intend.
96
6 Against Representative Realism
Urmson might defend his theory of memory markers by adopting a constitutionalist account of our cognitive access to our intentions. Ordinarily we take a realist view towards our self-ascriptive judgements: when I claim to have a particular mental state, there is a fact of the matter about what the state is about, independent of my self-ascriptive judgment. Constitutionalism, however, is a version of antirealism: mental states are projections of one’s reports about one’s mental states. Constitutionalism explains self-ascriptive judgments not by appeal to some cognitive process by which we detect our own mental states but rather by appeal to the expressive nature of mental state avowals. My avowing that I am in a particular state of mind plays a constitutive role in determining the state of mind I am in. First-person authority (the thesis that the first-person introspective judgments that we make about our own occurrent mental states are on a qualitatively better epistemic footing than the third-person perceptual judgments we make about the external world) is due to the fact that what we think is often determined by what we take ourselves to think. Adapting constitutionalism, Urmson could argue that when I take myself to remember2 something I thereby make it the case that the attitude I am having towards that thing is one of remembering2. It was Ludwig Wittgenstein who first suggested we think about mental state avowals in terms of their expressive rather than their descriptive function. He remarked, for example, that when a child learns to speak, “the verbal expression of pain replaces crying” (1967b, §244). A baby’s cry is not the report of an inference or detection, but an expression of his discomfort. Likewise, if I say “I intend to climb the Eiffel Tower,” my words have a constitutive or stipulative force. My taking myself to be in a particular state of mind plays a constitutive role in determining the state of mind that I am in. First-person authority is due to the fact that what we think is often determined by what we take ourselves to think. Crispin Wright, a proponent of constitutionalism characterizes this position as follows: The authority which our self-ascriptions of meaning, intention, and decision assume is not based on any kind of cognitive advantage, expertise or achievement…. It is, so to speak, such a subject’s right to declare what he intends, what he intended, and what satisfies his intention; and his possession of this right consists in the conferral upon such declarations, other things being equal, of a constitutive rather than descriptive role.7
If constitutionalism were true, I would be responsible for my mental states in something like the same way in which I am responsible for my actions. Self-ascribing and avowing is, after all, the sort of activity for which we hold people responsible. If, say, my intention were constituted by my avowal, it would make sense to blame me for having the intention. The problem with this view is that not all of our mental
7 1987, pp. 400–401; cf. 1992, pp. 108–139; Bilgrami 1998. Wittgenstein also applies constitutionalism to self-ascriptions of past mental states. He says, for example, that a report of one’s past intention is a response and is not made “on grounds of self-observation” (1967b, §659). “What is the purpose of telling someone that a time ago I had such-and-such a wish? – Look on the language-game as the primary thing” (1967b, §656). For a detailed discussion of Wittgenstein in memory of past mental states see Child (2006).
6.4 The Validation of Ostensible Memory
97
states are under our control and that we are therefore not responsible for them in the way we are responsible for our actions. Call this the responsibility objection to constitutionalism. Even if we can attend, consider, intend and do other mental things at will, we cannot believe at will. For if we could believe at will we could control getting beliefs and we could stop ourselves from coming to beliefs. But this is impossible. It is the nature of belief to aim at truth, at correspondence to the world or the facts. And a belief’s being true has nothing to do with willing to believe it or willing that it be true. So, we are importantly passive in regard to getting or having a belief. And to the extent that we are passive in regard to something, we cannot be responsible in regard to it. If Urmson did resort to constitutionalism to defend his theory of memory markers he would have to find a way to dispel the responsibility objection. There have been many other attempts to distinguish memory from imagination, but it should be clear by now that nothing provides the desired a priori knowable feature that shows that a memory experience is a memory experience. There is no way to reliably tell, by reflection alone, the difference between (veridical or ostensible) memory and imagination, much less the difference between a veridical and ostensible memory. What makes a memory experience a memory experience is not some intrinsic and a priori knowable feature of the memory experience but rather an extrinsic or relational feature. The extrinsic feature in question is the fact that the memory experience stands in the right kind of a causal relation to some previous mental state and that the content of the memory experience is the same as, or sufficiently similar to the content of the past mental state. There is no way that this relational fact can be known by reflection alone.
6.4
The Validation of Ostensible Memory
How do we tell whether our seeming memory experiences are indeed representations of the past? What are the criteria that enable us to say that what we ostensibly remember is indeed remembered? We saw that there are no intrinsic features of memory experiences from which it can be “read off” that they are memory experiences rather than experiences of the act of imagining. Where does this leave us? Should we conclude that representative realism inevitably leads to skepticism about memory knowledge? Drawing this conclusion is premature, for there are still some anti-skeptical moves available to the representative realist that we haven’t considered yet. One (rather simple-minded) idea is that we can know that our memory-data represent past events by checking them against the originals, that is, the past events themselves. But this is impossible for two reasons. First, to discover whether a memory-datum represents a past event we would have to investigate it from an external point of view. But, according to the representative theory, the only way of finding out what happened is via memory-data. Second, the past events have ceased to exist, and hence are not available as objects of any kind of experience, apart from
98
6 Against Representative Realism
memory experience. This truism has been recognized both by critics and by proponents of the representative theory. Bertrand Russell (1921, p. 161), for example, asserts that “we cannot evoke the past bodily and compare it with the present image,” and R.F. Holland (1954, p. 478) speaks of the difficulty of trying to “resurrect the past bodily.” C.I. Lewis (1946, p. 354) says that “the past is irrecoverable,” and that we therefore “cannot go behind the epistemological present in any attempted justification of belief.” Granted that the past events cannot be used to confirm memory-data, the representative realist might look for other ways of verifying memory-data. Why should it not be possible to confirm the epistemic status of one’s memory-data by means of diaries, photographs, testimony and the like? The problem with this kind of evidence is that it begs the question at issue – whether one’s memory-data supply knowledge – in that the employment of this evidence assumes the trustworthiness of some memory-data (one’s own or someone else’s). To drive this point home suppose that you seem to remember putting a key in the drawer. You open the drawer, and there is the key. Does this not confirm your memory experience? Does it not show that what you took to be a memory experience was indeed a memory? No, for the key might be in the drawer without your having put it there. Then suppose that you try to confirm your memory experience of having put the key in the drawer by asking your friend whether he saw you put the key there. If your friend answers in the affirmative, does this justify your memory experience? No, for your friend’s (ostensible) memory experience might be just as much a product of the fancy as your own. The problem of verifying ostensible memory is only pushed from you to your friend. A possible way out of the present difficulty is to focus on memories about the present and the future rather than the past. Though memory knowledge and knowledge of the past are closely linked, it is a mistake to equate the two. I can remember things not only about the past (e.g., that Brutus stabbed Cesar), but also the present (e.g., that my friend is now shopping), the future (e.g., that I have a dentist’s appointment tomorrow morning) as well as timeless truths (e.g., E = mc2). In the case of ostensible memories about the present, we seem to be able to confirm their accuracy without having to rely on memory: we can use a present perception to show that a certain ostensible memory about the present amounts to a genuine memory. Yet this strategy does not establish the reliability of memories in general – not even of memories about the present. The reason is that to infer that memories (about the present) are in general reliable we would have to rely once again on our memory. H.H. Price nicely summarizes the reasoning in the following passage: [N]o one memory can be validated or invalidated without relying on other memories…. It is often supposed that we can validate or invalidate a memory-judgement by means of a present perception, for example by consulting documents or records. Again, it is supposed that we can do it by appealing to the established laws of nature…. But in both cases we are using memory over again, because we are relying on inductive generalizations…. [H]owever great the probability of an inductive generalization may be, its probability is derived from past observations. We have only memory to assure us that those past observations existed, or what sort of observations they were (1969, pp. 78–79; cf. Furlong 1951, pp. 64–66; Russell 1940, p. 157).
6.4 The Validation of Ostensible Memory
99
What Price points out is that any inductive argument for the trustworthiness of memory experiences is open to the charge of vicious circularity. We cannot validate our ostensible rememberings without assuming memory. And since a circular justification is no justification at all, the problem of showing ostensible memory to be reliable seems to be insoluble. The inductive justification of memory is circular only if the memory presupposed by the justification is of the same kind as the memory whose validity is being scrutinized. But, it might be argued, the kind of memory that is involved in my knowing that, say, my friend is now shopping is very different from the kind of memory that is involved in my knowing that my ostensible memories (about the present) are usually reliable. Knowing that my friend is now shopping is based on memory of a particular event. Knowledge of generalizations usually do not depend on memories of particular events. However, even if there are circumstances where a non-circular inductive justification of ostensible memory is possible, the circularity issue remains to pose a problem. For when one tries to justify one’s ostensible memories in general, everything one can appeal to in giving such a justification is a subset of the kind of memories whose trustworthiness is being questioned (Shoemaker 1967, pp. 272–273). An ingenious attempt to prove that a non-circular justification of memory is possible is due to Roy Harrod. According to Harrod, it is indeed possible to establish the validity of memory experiences about the present without having to rely once again on one’s memory. Harrod invites us to suppose that the thesis according to which memory experiences are valid (or informative) is a mere hypothesis. What, if anything, could verify the hypothesis? Harrod’s point is that we don’t have to resort to the remembered past to verify the hypothesis but can instead use observable facts about our present environment. For instance, I may predict continuity for the ordinary objects around me and the prediction may be fulfilled within one present totality. And if a flash of lightning occurs, I may predict and experience its end within the same totality. Only if memory is informative should I have grounds for predicting continuity for the chairs and tables and death for the lightning flash; the success of the prediction within the specious present is therefore strong evidence of the informativeness of memory (1942, p. 65).
At first blush, Harrod’s argument has some plausibility. It is not obvious that the “specious present” is not long enough to accommodate both the prediction of the continuation of, say, a chair – where the prediction is based on memory experiences – and the fulfillment of such a prediction. Yet to see what is wrong with Harrod’s argument suppose that I wish to verify the hypothesis that I have been looking at a chair for the past 10 minutes. The crux is that my present perceptual experience of looking at the chair may actually disconfirm the validity of my ostensible memory. For precisely the same perceptual experience can be looked upon as lending (even stronger) support to the hypothesis that I have been looking at the chair for the past 11 minutes. But if the 11-minute hypothesis is true, my ostensible memory is delusive. Thus, from the fact that the chair continues to be there now it does not follow that the things happened before as I remember their happening. Moreover, since perceptual facts about my present environment do not
100
6 Against Representative Realism
constitute, by themselves, evidence for the assertion that my ostensible memories are generally true, Harrod’s proposal is of no use in refuting skepticism about memory knowledge. It seems that any attempt to confirm the validity of memory experiences must rely on memory. We don’t seem to be able to put our reliance on the memory in question and then demonstrate the reliability of ostensible memory. Where does this leave us? If it were established that we cannot validate memory without assuming memory, it would be appropriate to counter skepticism about memory knowledge with skepticism. A.J Ayer suggests such a skeptical solution to skepticism when he writes: Our only resource is to point out … that the proof that [the skeptic] requires of us is one that he makes it logically impossible for us to give. It is, therefore, not surprising that we cannot furnish it: it is no discredit to the proofs which we do rely on that they do not imply that we can achieve the impossible; it would be a discredit to them, rather, if they did (1956, p. 164).
But maybe it is premature to give in to skepticism about memory knowledge. Maybe there is still a way to validate ostensible memory. The problem of verifying our ostensible remembering has striking similarities to Hume’s problem of induction, that is, the problem of justifying our tacit belief in the principle of the uniformity of nature. Both problems result from attempts to justify epistemic methods which apparently can be justified only by appeal to their own principles. It is not surprising then that the various putative solutions to the problem of induction have also been applied to the problem of validating memory experiences. The three main putative solutions to the Humean problem of induction are the postulational, the pragmatic, and the analytic approach. Let us discuss these putative solutions in turn. Some philosophers have tried to handle the problem of verifying our ostensible memories by maintaining that the reliability of ostensible memory is a postulate that cannot be proven or disproven but must be assumed. For example, John T. Saunders holds that “the reliability of ostensible memory is a fundamental assumption of the knowledge enterprise as we in fact pursue it. Such an enterprise must have its unprovable first principles (that is, principles which are unprovable within that enterprise)” (1963, p. 486). Similarly, Bertrand Russell declares that the proposition according to which “memory is in the main veridical is, in my opinion, one of the premises of knowledge” (1948, p. 288). Saunders’ and Russell’s postulational approach to the problem of justifying the trustworthiness of memory experiences is unsatisfactory since it amounts to an admission of defeat. The postulation of an unjustified or unjustifiable principle makes our reliance on memory experiences a matter of faith. If memory knowledge is basically a matter of faith, then this faith exists on a par with other faiths. Reliance on memory experiences has no ground on which to maintain its cognitive superiority to any other form of irrationalism, such as voodoo or crystal gazing. Another approach to the problem of justifying our trust in our ostensible memories is to develop a pragmatic justification. A pragmatic justification of our reliance
6.4 The Validation of Ostensible Memory
101
on ostensible memories points out that trusting one’s memory experiences, though ultimately unjustified, is better suited to the goal of uncovering the past than any other method that might be adopted. Something like this strategy is what Richard Brandt has in mind: Any satisfactory theory of nature obviously has to be based on, and to explain, the unquestionable facts: the facts of present experience, and the fact of our memory beliefs…. [W]hen we begin looking about for a theory which will do what is wanted, we find that there is only one type of theory which comes near to success: a theory which postulates an historical sequence of events and human beings with a capacity to recall …, and which hence explains memory beliefs as the precipitate of a process of interaction between human brains-minds and the world. In other words, the only acceptable theory is one which asserts that a large proportion of our memory beliefs are veridical. No alternative to such a theory has been proposed; nor can one imagine what one would be like (1955, pp. 92–93).
As was noted by Don Locke (1971, pp. 113–114), the crux with Brandt’s justification of ostensible memory is that such an alternative theory has indeed been proposed as a logical possibility that has not yet been ruled out. The alternative is Russell’s suggestion that there has actually been no past at all for us to remember since the world sprang into being five minutes ago. Until Russell’s proposal has been ruled out, Brandt is not entitled to conclude that ostensible memory is reliable. Both the postulational and pragmatic approaches to the justification of our trust in ostensible memories would be redundant if it could be shown that those memories are necessarily reliable in representing the truth for the most part. Norman Malcolm and Sydney Shoemaker have in fact argued that the general reliability of ostensible memories is an analytic truth. In his own words, Shoemaker wants to show “that it is a necessary (logical or conceptual) truth, not a contingent one, that when perceptual and memory statements are sincerely and confidently asserted, i.e., express confident beliefs, they are generally true” (1963, p. 229). (The expression “perceptual and memory statements,” as Shoemaker uses it, refers not only to statements explicitly referring to memories but also to statements that are “directly based on … memory, i.e., are putative reports of what the speaker … remembers.”) Malcolm makes essentially the same point when he declares: “Knowing how to use the past tense cannot be completely separated from making many true statements with it” (1963, p. 196). There are two main arguments to the effect that ostensible memories are necessarily reliable. This is the first argument as Shoemaker presents it: A primary criterion for determining whether a person understands the meaning of such terms as “see” and “remember” is whether under optimum conditions the confident claims that he makes by the use of these words are generally true. If most of a person’s apparent perceptual and memory claims turned out to be false, this would show, not that the person had exceptionally poor eyesight or an exceptionally bad memory, but that he did not understand, had not correctly grasped, the meanings of the words he was uttering, or was not using them with their established meanings, i.e., was not using them to express the perceptual and memory claims they appear to express (1963, p. 231; cf. Malcolm 1963, pp. 193–194).
102
6 Against Representative Realism
Shoemaker claims that if someone were to consistently make wildly inaccurate claims about the past, and instead seemed to remember things that never happened, we would have to say not that he was misremembering, but that he had lost his understanding of “to remember.” He declares: We would not be satisfied that one of our own children had learned the correct use of the word “remember” and of the expressions that indicate a past tense unless the sincere statements he made by the use of these expressions were normally true – just as we would not allow that someone knew the meaning of the word “blue” if he typically applied it to such things as grass and trees. In this case, as in many others, using an expression correctly necessarily goes together with using it to make statements that are (for the most part) true (1967, p. 273).
Shoemaker is certainly right in that habitual mistakes regarding one’s memory claims can be due to a mistake of terminology. The question is, however, whether someone whose memory claims are habitually wrong necessarily misunderstands “to remember” and the past tense. Could the habitual mistakes in question not be due to mistakes of fact rather than meaning? To see that this question should be answered in the affirmative, consider the cognitive verb “to predict” and future tense statements. Given Shoemaker’s position, someone can be said to understand the meaning of “to predict” only if his claims about the future and the unobserved past and present, respectively, are generally true. But it is surely possible for someone to understand the concept to predict while being mostly wrong in his predictions. Past tense statements seem relevantly analogous to future tense statements. So why should it be impossible that someone understands “to remember” and uses it to construct wellformed sentences that are mainly false? In support of the claim that if someone understands “to remember” and past tense speech, then the statements he makes in the past tense employing “to remember” are mainly true Malcolm (1963, p. 194) claims that the language game is like the game of chess: if someone makes mainly incorrect uses of sentences of a language, it can be inferred that he is not playing the game involving those sentences; for someone who makes illegal chess moves is not playing chess. Following James Cornman (1965, p. 163), I doubt that legal (illegal) chess moves correspond to true (false) uses of sentences. Illegal chess moves correspond not to uses of sentences that violate the rule of truth, but to uses of sentences that violate the rules of grammar. Using the chess king to move like the bishop is not analogous to saying something false like “Britain was once an American colony.” An illegal move in the language game is one which brings about an ungrammatical sentence such as “Britain will once an American colony.” Knowing what a term means and how to use it does not necessarily go together with using it to form true statements. Whether someone understands an expression is determined by what he thinks he is saying when he uses that expression to form a statement and what he thinks would make the statement true – not by the truthvalue of the statement. Even if someone’s memory claims were consistently wrong, he could still have a correct understanding of the verb “to remember.” That he correctly understands “to remember” could be established by the fact that he uses it
6.4 The Validation of Ostensible Memory
103
only to talk about things that, he believes, did happen and not about things that, he believes, he imagined. Shoemaker’s second argument to the effect that ostensible memory is necessarily reliable rests on the claim that one cannot question one’s confident perceptual and memory beliefs. He writes: It is precisely one’s confident beliefs, and especially one’s confident perceptual and memory beliefs, that one expresses by saying “I know…;” it is not a psychological fact, but rather a logical fact, that one cannot help regarding one’s confident perceptual and memory beliefs as constituting knowledge (1963, p. 234).
Shoemaker goes on to argue that what is true of myself is also true of other people. Each one of us has to claim that his confident perceptual and memory beliefs are generally true. Shoemaker concludes that it is necessarily true of confident perceptual and memory beliefs in general that they are generally true. But this does not seem right. Just because I cannot question my own confident memory beliefs doesn’t mean that I cannot question someone else’s claim concerning his confident memory beliefs. Recognizing that the other person cannot question his own confident memory beliefs is not the same as accepting them as true, or even generally true (O’Connor and Carr 1982, p. 140). A further problem with Shoemaker’s second argument is that even if it were incoherent to question one’s confident memory beliefs, this does not mean that one’s memory beliefs could not be consistently false. Thus the skeptical problem actually gets worse because not only is it possible that one’s memory beliefs are consistently false but also one is not even in a position to coherently entertain this possibility. In the end, none of the putative solutions to the problem of verifying our ostensible memories work. That is why Russell insists that “no memory proposition is, strictly speaking, verifiable, since nothing in the present or future makes any proposition about the past necessary” (1940, p. 154). In my view, the only promising strategy available to the representative realist for fending off skeptical worries is to reject epistemic internalism, that is, the view that all of the factors required for a belief to be justified must be cognitively accessible to the subject and thus internal to his mind. The skeptic about memory knowledge asks how one can know (or justifiably believe) that the memory experience on which one bases one’s memory-claim is in fact an accurate, a reliable, guide to the past. As was explained in section 6.1, the skeptic’s question is committed to internalism about justification. For unless it is deemed reasonable to expect an epistemic agent to have insight into his justifying reasons, the skeptic’s question is irrelevant. The representative realist (or the direct realist, for that matter), can undermine the skeptic’s position simply by denying that an epistemic agent must have cognitive access to his justifying reasons. This is the position of epistemic externalism. The externalist holds that some of the justifying factors may be external to the subject’s cognitive perspective. Given externalism, we need not be able to respond to the skeptical query. The fact that we don’t know (or justifiably believe) that our memory-claims amount to knowledge does not mean that they do not amount to knowledge. A person who knows something does not have to know that what he has in his evidential base
104
6 Against Representative Realism
amounts to knowledge. As long as he in fact satisfies the conditions of knowing something, there is nothing more he has to do in order to know. No skeptical worry gets started. Using epistemic externalism to neutralize skepticism about memory knowledge is neither legitimate nor promising unless there are reasons for adopting epistemic externalism that are independent of the goal of neutralizing skepticism. The goal of the next chapter is to look for skepticism-independent reasons for adopting externalism about memory knowledge.
Chapter 7
Skepticism, Externalism, and Closure
The skeptic challenges our ability to know (or justifiably believe) that the experience of seeming to remember on which we base our memory-claims is in fact a reliable guide to the past. How can we be sure that seeming memory experiences are representations of the past rather than, say, products of our imagination? As was shown in the previous chapter, skepticism about memory knowledge assumes internalism about justification. At the end of section 6.4, I suggested rejecting internalism and proposed an externalist response to skepticism about memory knowledge. The main aim of this chapter is to elaborate on the externalist response. As was explained in section 6.1, internalism about justification is the view that all of the factors required for a belief to be justified must be accessible to the subject merely by reflecting and thus internal to his mind. Externalism about justification, on the other hand, is simply the denial of internalism, holding that some of the justifying factors may be external to the subject’s cognitive perspective. A belief is justified if it has the property of being truth-effective. No more than this is necessary for justification. Whether the subject takes his belief to be truth-effective doesn’t add anything to the belief’s epistemic status. Since externalism is the denial of internalism and since there is no third position besides internalism and externalism, any argument against internalism is an indirect argument for externalism. My arguments in favor of externalism about memory knowledge are, for the most part, indirect ones. The most prominent recent externalist theories have been versions of reliabilism. While externalism is only a negative thesis consisting in the denial that justification and knowledge are completely internal, reliabilism is a positive thesis maintaining that what qualifies a belief as knowledge or as justified is its reliable linkage to the facts that make the belief true. The reliable linkage, which is commonly expressed by means of subjunctive conditionals, ensures that if the known proposition p were not true, the subject would not believe that p. Advocates of reliabilism argue that, a subject knows that p if he believes that p, p is true, and he would not believe that p unless p were true. The idea is that to know that p the belief must not be accidentally true. It must “track” or “indicate” the facts that make the belief true. The theory is externalist because it makes justification depend on factors that are not necessarily accessible to the knowing subject.
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
105
106
7 Skepticism, Externalism, and Closure
Given externalist reliabilism, a person can be justified in believing p without having to have cognitive access to any of the justificatory factors involved. The fact that he cannot answer the skeptical question as to what justifies him in thinking that his memory-claims amount to knowledge does not mean that they do not amount to knowledge. The skeptic fails to undermine our ordinary memory knowledge claims. Internalists argue that the externalist response to skepticism about memory knowledge is not satisfactory. Section 7.1 examines this line of criticism and rejects it. I argue that epistemic externalism does provide an escape from skepticism about memory knowledge; but I concede that the externalist should supply reasons for adopting his position – reasons that are independent of the issue of skepticism. When considering the plausibility of externalism as an account of memory knowledge two questions need to be distinguished. First, does externalism provide a plausible account of coming to justifiably believe something that is subsequently committed to memory? Second, how convincing is the externalist account of continuing to justifiably believe something one remembers? I criticize both internalism concerning the initial justification-acquisition (section 7.2) and internalism concerning justification-preservation (sections 7.3 and 7.4). Section 7.5 sets forth and defends an externalist account of justification-preservation. Finally section 7.6 tackles a related, yet distinct form of skepticism, namely skepticism regarding our ability to know on the basis of memory experiences that there is a past. After explaining this form of skepticism and rejecting attempts to declare it incoherent, I develop a solution based on the limitation of the scope of the epistemic closure principle.
7.1
Skepticism and Metajustification
The externalist account of justification promises to avoid the regress that seems to be an inevitable ingredient of skepticism. The external-world skeptic asks how one can be sure that sense experience is a reliable guide to what is going on in the external world. Analogously, the skeptic about memory knowledge inquires into one’s reasons for thinking that the experience of seeming to remember on which one bases one’s memory-claims are in fact a reliable guide to the past. The externalist acknowledges that he does not have such reasons, but this does not undermine his memory knowledge. To know something on the basis of the experience of seeming to remember it, a person does not have to know that the experience is a reliable indicator of the truth; it just has to be a reliable indicator of the truth. If the experience is reliable, it gives us knowledge. If it is not, then it does not. But the agent doesn’t himself have to know that what he has got in his prior evidential base amounts to justification. As long as he in fact satisfies the conditions of justifiably believing something in the domain in question, there is nothing more he has to do in order to know things in that domain. No regress gets started. Thus, second-level skepticism about memory knowledge may be in order (we cannot know whether we have memory knowledge), but arguments for first-order skepticism about memory
7.1 Skepticism and Metajustification
107
knowledge, arguments designed to show that we cannot know anything on the basis of memory experiences are refuted. Advocates of epistemic internalism argue that the externalist response to skepticism is not satisfactory. First, externalism shows only that it is possible that our memory beliefs amount to knowledge. A satisfactory response to skepticism, the internalist maintains, must establish that we in fact possess memory knowledge. Second, externalism is said to fail to supply reasons that make it rational to adopt this theory. The most detailed critique of the externalist response to skepticism is due to Barry Stroud and William Alston. What follows will focus on their case against externalism. Stroud’s main criticism of externalist reliabilism is that it does not and cannot provide a satisfactory explanation for why it is the correct theory of knowledge. Stroud writes: If “externalism” is correct, what [the externalist] would be saying here is true. His theory, if true, would explain his knowledge. The difficulty is that until he finds some reason to believe his theory rather than some other, he cannot be said to have explained how he knows the things he knows…: if his theory is true he will know the things he thinks he knows. But he is, in addition, a theorist of knowledge. He wants to understand how he knows the things he thinks he knows. And he cannot satisfy himself on that score unless he can see himself as having some reason to accept the theory that he (and the rest of us) can recognize would explain his knowledge if it were true…. But the scientific student of human knowledge must know or have some reason to believe his theory of knowledge if he is going to understand how knowledge is possible (1989, pp. 44–45).
Externalist reliabilism, Stroud argues, does not meet the metajustificatory requirement whereupon to understand one’s knowledge satisfactorily one must see oneself as having some reason to accept an epistemology that one can recognize would explain one’s knowledge if it were true. Externalism violates this requirement because it allows us to know something without understanding what it is that makes it a piece of knowledge rather than merely true belief. What would the justification of externalist reliabilism look like? What kind of reasons would convince us that externalist reliabilism is the correct analysis of knowledge? Prima facie externalist reliabilism can be defended either by means of an externalist argument or an internalist argument. The problem with an internalist justification of externalist reliabilism is that it will not convince the proponents of externalism. If you already believe in externalist reliabilism, being presented with an internalist justification of externalist reliabilism will not make you give more credence to externalist reliabilism. The only option is therefore an externalist justification of externalist reliabilism. The mark of an externalist justification is that it can obtain without the knower having cognitive access to it. One can be justified, in the externalist sense, in believing that externalist reliabilism is the correct account of knowledge, even though one is not in a position to cite any reason. Stroud thinks that such an externalist justification of externalist reliabilism is inadequate. [The externalist] is at best in the position of someone who has good reason to believe his theory if that theory is in fact true, but has no such reason to believe it if some other theory is true instead. He can see what he would have good reason to believe if the theory he believes were true, but he cannot see or understand himself as knowing or having good
108
7 Skepticism, Externalism, and Closure
reason to believe what his theory says…. But then [the externalist] is still in the same position two levels up that we found the imagined “externalist” Descartes to be in at the first level. If the imagined Descartes’s claim to self-understanding was inadequate there, any similar claim will be equally inadequate at any higher level of knowing that one knows or having reason to believe that one has reason to believe…. It is difficult to say precisely what is inadequate about that kind of response, especially in terms that would be acceptable to an “externalist.” Perhaps it is best to say that the theorist has to see himself as having good reason to believe his theory in some sense of “having good reason” that cannot be fully captured by an “externalist” account (1989, pp. 46–47).1
Presumably the externalist reasons for externalism are not of the right kind, according to Stroud, because they are not normative. Externalist justification is something that happens to a person rather than being something he achieves by his own efforts. An externalist reliabilist is well advised to dismiss Stroud’s metajustificatory requirement because it presupposes the truth of internalism and epistemic deontologism. Why, the externalist reliabilist asks, is it not sufficient that he has reasons to accept his theory but that these reasons need to be accessible by reflection? When internalism and deontologism are not taken for granted, it is hard to see why the externalist theorist of justification has to be in a position “to see himself as having good reasons to believe his theory.” In response to this objection, Stroud tries to put the burden of proof on the externalist reliabilist by claiming that the internalist conception of metajustification is the natural one: If we do recognize a certain ineliminable dissatisfaction in any such “externalist” attempt at self-understanding I do not think it is because of hidden attachment to an opposing “internalist” theory which requires that everything we know must be justified by reasonable inference from something else we believe…. I think the dissatisfaction, if we recognize it, is felt to come from the demands of the epistemological project itself …. What we seek, and what the theorists I have imagined appear to lack, is something that “externalism” alone seems unable to explain or to account for (1994, p. 304).
The externalist reliabilist will, of course, deny that the kind of justification whereupon the agent is in a position to “see himself as having good reasons to believe his theory” is part of our folk epistemology. The internalist conception of justification depends on theoretical commitments that are not forced on us by our ordinary ways of thinking about justification. Like Stroud, William Alston thinks that externalism in general, and reliabilism in particular are unable to neutralize the skeptical challenge. The reason is that externalist reliabilism, by itself, does not allow us to tell whether a particular linkage between beliefs and facts is reliable and when we possess knowledge. Alston writes: [W]hen we ask whether one or another source of belief is reliable, we are interested in discriminating those that can reasonably be trusted from those that cannot. Hence merely showing that if a given source is reliable it can be shown by its record to be reliable, does
1 Stroud’s fictional externalist whom he calls “Descartes” is not to be confused with the real René Descartes.
7.2 Internalism and the Acquisition of Justification
109
nothing to indicate that the source belongs with the sheep rather than with the goats. I have removed an allegedly crippling disability, but I have not given the argument a clean bill of health. Hence I shall disqualify epistemically circular arguments on the grounds that they do not serve to discriminate between reliable and unreliable doxastic practices (1991, p. 148; 1993, p. 17).
Alston maintains that a proponent of externalist reliabilism owes us a justification of externalist reliabilism that compares and contrasts the pros and cons of this epistemology vis-a-vis others. It is clear that the discriminatory justification Alston has in mind is internalist in nature. But an internalist justification of externalist reliabilism doesn’t stand a chance of convincing either internalists or externalists. Internalists will refuse to countenance that justification has anything to do with reliability and externalists will oppose the idea that justification is a matter of the rationality of belief. Ernest Sosa (1994, p. 286) notes that Stroud’s and Alston’s desire for a fully general, legitimating, philosophical understanding of one’s epistemology is unfulfillable. The quest for a presuppositionless understanding of one’s epistemology is unfulfillable because for something to give us real understanding we must know it. Real understanding is knowledgeable understanding, and not just “understanding by dumb luck” (Sosa 1994, p. 288). Since any attempt to really understand and to justify a theory of knowledge will have to be epistemically circular, there is no good reason why one should not be allowed to employ externalist reliabilism to argue that externalist reliabilism is the correct epistemology. Though I believe that externalist reliabilism provides an escape from skepticism in general and skepticism about memory knowledge in particular I concede to Stroud and Alston that the proponent of externalist reliabilism must supply reasons for adopting his position – reasons that are independent of the issue of skepticism. The following three sections will present skepticism-independent reasons in favor of externalist reliabilism and against internalism. Sections 7.2–7.4 criticize internalism concerning the initial justification-acquisition and internalism concerning justification-preservation, respectively. Section 7.5 explains and defends the externalist account of justification-preservation.
7.2
Internalism and the Acquisition of Justification
There are two kinds of counterexamples to internalism about coming to justifiably believe something that is subsequently committed to memory. Either the subject lacks awareness of his being justified in believing p or he lacks the relevant justification for believing p. In the former case – I refer to it as the absent justification case – the subject is simply unaware of the epistemic reasons he has, but in a sense might still possess. In the latter case – I call it the defeated justification case – the subject has counterevidence to his justifiably believing p. Let’s start with an absent justification case. The following example has been adapted from Robert Audi (1995, pp. 39–40). While taking a history exam you
110
7 Skepticism, Externalism, and Closure
provide a factual answer that seems right to you, though you cannot justify it. You believe it but feel you may be out on a limb. Later, when discussing the exam with a friend, she argues for a different answer, citing facts that you were well aware of but did not at the time bring to bear. Still later, however, it turns out that your answer was right, and you realize that you had written it in several places in your notebook. It seems that you did remember the answer when you took the exam, despite the fact that you were not justified, in the internalist sense, in believing the answer. Audi correctly notes that this example, if plausible, flies in the face of an internalist construal of the past justification condition of memory. He concludes that “perhaps remembering that p is, in at least some instances, a special case of reliably produced or sustained true belief, a kind that is memorically preserved” (1995, p. 40). The discussion of defeated justification counterexamples to internalism about coming to justifiably believe something that is subsequently committed to memory requires some stage setting. Despite the dazzling number of different conceptions of epistemic justification, virtually all epistemologists accept that knowledge is incompatible with undefeated defeaters. Defeaters come in at least two flavors – doxastic and normative ones. A doxastic defeater is a proposition D that is believed by S to be true, yet indicates that S’s belief that p is either false or unreliably formed or sustained. Defeaters in this sense function by virtue of being believed, regardless of their truth value or justificatory status. Apart from doxastic defeaters there are normative ones. A normative defeater is a proposition D that S ought to believe to be true, yet indicates that S’s belief that p is either false or unreliably formed or sustained.2 Defeaters in this sense function by virtue of being propositions that S should believe – whether or not S does believe them – given the presence of certain available evidence. The underlying thought here is that certain kinds of counterevidence contribute epistemically unacceptable irrationality to doxastic systems and, accordingly, that justification and knowledge can be defeated by their presence.3 A defeater may be either defeated or undefeated. When one has a defeater D for one’s belief that p that is not itself defeated, one has what is called an “undefeated defeater” for one’s belief that p. And it is the presence of undefeated defeaters that is thought to be incompatible with knowledge. In the case of epistemic internalism, it is obvious that the presence of undefeated defeaters undermines justification. Given that what justifies a belief is a mentally accessible item (something that the person can come to know whether it obtains just by reflecting on one’s mental states), being justified in believing p must exclude a person’s having sufficient
2 There are two different kinds of doxastic defeaters. Rebutting defeaters are propositions that are believed by S to be true yet indicate that the target belief is false while undercutting defeaters are propositions that are believed by S to be true yet indicate that the target belief is unreliably formed or sustained. 3 In addition to doxastic and normative defeaters there are factual defeaters. A factual defeater is a true proposition D such that if D were added to S’s belief system, then S would no longer be justified in believing that p. Defeaters in this sense function by virtue of being true.
7.2 Internalism and the Acquisition of Justification
111
reasons for supposing either that p is false or that the belief that p is not grounded or produced in a way that is sufficiently truth-indicating. Whether the presence of an undefeated defeater poses a threat to an externalist construal of justification depends on the particular version of externalism. Given an extreme form of epistemic externalism, the question of whether a belief is justified (truth-effective) is quite distinct from the question of whether the subject takes his belief to be justified (truth-effective). A belief is or isn’t justified regardless of what the subject thinks about the justificatory status of the belief. If one relies on what is, in point of fact, a conclusive reason for p, one is justified in believing that p, despite being convinced that p is false. Fred Dretske (2000, p. 595) calls such an austere form of epistemic externalism “mad-dog reliabilism.” For reasons I don’t have space to go into here I reject mad-dog reliabilism (see Bernecker 2007, pp. 143–144). Though a subject need not be aware of the factors that justify his belief, he may not be aware of evidence that undermines his belief. The negative coherence condition ensures that for a belief to become justified it must not be incoherent with the background information the subject possesses. Given this mitigated version of epistemic externalism (or “mitigated externalism” for short), defeated justification cases undermine justification. With these preliminary points in mind, we can see that it is possible to remember something one did not justifiably believe in the past. The following defeated justification case is adapted from Robert Shope (1973, pp. 308–309). Some time in the past, at t1, S stood at a lookout point and saw a rare bird, say, a bald eagle catch a fish. S came to believe that a bald eagle has passed by. As a matter of fact, a bald eagle had been there. Although S’s belief was correct, he had plausible but misleading reasons to suppose that he had just been given a strong hallucinogenic drug. Given the incompatibility of internalist justification with the presence of undefeated doxastic defeaters, S wasn’t justified at t1 in believing that a bald eagle had passed by and thus he didn’t know it. At t2, S learns that, in spite of his past evidence, he had not actually been given a hallucinogenic drug but only a placebo. He had really seen what he had irrationally believed to be seeing while falsely thinking to be hallucinating. So S remembers at t2 that there was a bald eagle at t1 even though, at the time, he wasn’t justified in believing it. The point of the example is that at t1 S was not justified in believing that there was a bald eagle even though at t2 he remembers that there was such a bird. Hence memory does not imply past internalist justification. The exam-example and eagle-example show that it is implausible to suppose that memory beliefs must meet the conditions for internalist justification. Even if the subject remembers everything from the time he first acquired the belief whose content he now remembers, he may not be in a position to give the reasons for the memory belief in question. His inability to give the reasons may be due either to the fact that the original belief was not justified or to the absence of awareness, at the time, of the justificatory factors for the indicated belief. The exam-example and eagle-example fly in the face of the epistemic theory of memory already discussed in section 2.5. According to the epistemic theory, remembering that p entails knowing that p, where this knowledge was previously acquired and preserved. On this view, memory is long-standing or continuing
112
7 Skepticism, Externalism, and Closure
knowledge. In addition to the claim that remembering that p entails knowing that p, the epistemic theory further maintains that memory is not capable of generating new justification and knowledge. Just as testimony transmits knowledge from one person to another, memory is said to preserve knowledge from one time to another. In this way, it is said to be true for both memorial knowledge and testimonial knowledge that the proposition in question must be known when it was originally acquired and, accordingly, that a source other than memory and testimony, respectively, must be responsible for its original acquisition. If I justifiedly believe that p on the basis of memory, then – so the claim goes– I must have acquired this justification in a non-memorial way at some earlier time. Memory cannot make a proposition acquire an epistemic status different from the one it had at the time it was originally acquired. For example, Michael Dummett writes: “Memory is not a source, still less a ground, of knowledge: it is the maintenance of knowledge formerly acquired by whatever means” (1993, pp. 420–421). Audi says that [j]ust as we cannot know that p from memory unless we have come to know it in another way, say through perception, we cannot know that p on the basis of testimony unless the attester … has come to know it (at least in part) in another way …. Memory and testimony … are not generative with respect to knowledge: characteristically, the former is preservative, the latter transmissive (1997, p. 410).
Alvin Plantinga makes the same point when he writes: [M]emory beliefs depend, for their warrant, upon the warrant of earlier beliefs. I have an orange for breakfast; if this belief has no warrant, then my later belief that I had an orange for breakfast will also have no warrant. Memory beliefs are like testimonial beliefs …: the warrant they have is dependent upon the warrant enjoyed by an earlier belief (1993, p. 61, fn. 22).
So according to the epistemic theory of memory, to remember that p is to know that p, where this knowledge was previously acquired and preserved. A source other than memory is responsible for the original acquisition of knowledge that p. Given that memory is preserved knowledge and given epistemic internalism, the exam-example and the eagle-example are incoherent. For if to remember that p is to know that p, if knowledge implies justification, and if memory cannot generate new justification and knowledge, then one can remember only what one justifiably believed in the past. Provided the both examples are coherent, at least one of the two assumptions of the epistemic theory of memory is incorrect: either memory is not a form of knowledge or memory can generate justification and knowledge. In fact, I think that both assumptions of the epistemic theory are incorrect. In section 7.4 I will argue that memory is indeed a generative source of justification. In the remainder of this section I will suggest that, contrary to what the epistemic theory says, remembering that p does not imply knowing that p. The eagle-example illustrated that one can remember something one did not justifiably believe in the past. Now consider an example of memory without present justification that was already discussed in section 2.6. In the past, at t1, S met his father and noticed that he was wearing a blue pullover. He knew this fact about his father. At t2 S’s friends play a practical joke on him. They tell him that his father wore a green pullover and present him with plausible but misleading evidence to the effect that his father wore a green pullover. Given the incompatibility of
7.3 The Problem of Forgotten Evidence
113
internalist justification with the presence of undefeated normative defeaters, S does not know at t2 that his father wore a blue pullover, for he is unable to rule out the relevant alternative that he wore a green pullover. He fails to know that his father wore a blue pullover, despite the fact that he still remembers it. What this example shows is that one can know at t1 that p, remember at t2 everything one knew at t1, and yet fail to know at t2 that p – even though one continues to believe that p and p continues to be true – for the reason that one isn’t justified in believing that p. One may attempt to undermine this example of memory without present justification by identifying the conditions for remembering with the conditions for saying that one remembers. If the conditions for memory were the same as the conditions for the self-attribution of memory, then, since S probably doesn’t feel confident at t2 to claim to remember that his father wore a blue pullover, it would follow that he doesn’t remember this fact about his father. This objection is not convincing. What makes a mental state a state of remembering is the fact that certain conditions are met; whether or not the subject takes himself to meet these conditions is irrelevant. Remembering is, in this respect, like seeing. Just as I can see something without being aware that I see it, I can remember something even if I think I am suffering a hallucination. Whether I remember p or whether it only seems to me that I remember p cannot be discriminable by means of introspection. The reason memories are not transparent to the mind is that they imply truth and that they presuppose a causal chain of memory traces which are inaccessible to consciousness. Hence just because S doesn’t feel confident to selfascribe the memory that his father wore a blue pullover does not mean that he does not indeed remember this. In sum, memory, unlike knowledge, does not imply internalist justification. Not only is it possible to remember something at t2 that one didn’t have internalist justification for at t1 but also one may acquire between t1 and t2 some plausible but misleading evidence that destroys the status as justified belief (in the internalist sense) of the once-genuine justified belief that one still remembers.
7.3
The Problem of Forgotten Evidence
After having discussed internalism about coming to justifiably believe something that is subsequently committed to memory, we can move on to investigate internalist accounts of continuing to justifiably believe something one remembers. Suppose at t1 you come to justifiably believe (and know) that p on the basis of a trustworthy friend’s having told you so. At t2, as is normal, you have forgotten a great deal of what you once knew.4 Among those things that you have forgotten is that it was
4 Presumably the reason we frequently forget the evidence for the beliefs we retain is that we thereby make better use of the finite memory space. Without constant condensation and summing up, we would clutter our minds with a lot of redundant pieces of information. See section 8.6 on Harman on “clutter avoidance.”
114
7 Skepticism, Externalism, and Closure
your trustworthy friend who told you that p. For all you know you could have acquired the belief that p by reading an article in the National Inquirer or from some other unreliable source. You no longer remember the original justification for believing p and you have not gained new evidence concerning the proposition in question. But since you remember p most of us would feel comfortable saying that you are still justified in believing p. Yet on what internalist basis can the retained belief that p be justified? As Alvin Goldman says, “[n]o perceptual experience, no conscious memory event, and no premises consciously entertained at the selected moment will be justificationally sufficient for such a belief” (1999, p. 278). So it appears that internalists are stuck with the unacceptable result that retained beliefs are unjustified (and hence don’t amount to knowledge) unless something internal that justified them can be found – unless the past evidence is recalled. This is what Gilbert Harman calls the problem of forgotten evidence (1986, p. 41). To illustrate the problem of forgotten evidence take, for example, the version of internalism developed by Earl Conee and Richard Feldman which goes by the name of evidentialism. The core idea of evidentialism is that a subject is justified in believing that p if and only if believing p fits the evidence the subject has (Feldman and Conee 2004, p. 83). Moreover evidentialism holds that, in the most fundamental sense of “justified” a belief can be justified for a person only by the person’s current and (dispositionally) conscious evidence (ibid., pp. 67–68). But if only conscious or dispositionally conscious beliefs are allowed as evidence, evidentialism doesn’t look appealing as a theory of the retention of justified belief. The problem for evidentialism arises most clearly when the original evidence is irretrievably lost and not part of any stored justification.5 Another internalist theory of justification that has problems accounting for continuous justification in the face of forgotten evidence is Laurence BonJour’s coherentism. In a nutshell, coherentism claims that S’s belief that p is justified only if p coheres with the rest of S’s (coherent) doxastic system. While BonJour is not specifically interested in the problem of continuous justification, he offers a general theory of justification that also applies to continuous justification. According to BonJour, S’s retained belief that p from t1 is continuously justified at t2 if it coheres with S’s (coherent) doxastic system at t2. The apparent problem with this view is that we have many retained beliefs that are not inferentially or evidentially related to other beliefs in our doxastic systems. BonJour attempts to fix the problem by introducing so-called “cognitively spontaneous beliefs,” that is, non-inferential beliefs that strike one “in a manner that is both involuntary and quite coercive” (1985, p. 117). Cognitively spontaneous beliefs don’t cohere with the other members in the doxastic system but they are nevertheless justified, provided the subject can justify their reliability. BonJour
5 It is not my intention to give a sustained argument against evidentialism. I am aware that evidentialists think that they don’t have any difficulty finding adequate justification in cases of forgotten evidence (see Feldman and Conee 2004, pp. 69–72). I use evidentialism for illustrative purposes to show that internalists have to be creative to account for the preservation of justification.
7.3 The Problem of Forgotten Evidence
115
gives the following example of a justificatory argument for a cognitively spontaneous perceptual belief. Suppose I look up and come to believe that there is a red book on the desk. The justificatory argument for this belief involves three premises: (1) I have a cognitively spontaneous belief of kind K1 that there is a red book on the desk. (2) Conditions C1 obtain [i.e., lighting is good, my visual system functions normally, etc.]. (3) Cognitively spontaneous visual beliefs of kind K1 in conditions C1 are very likely to be true. Therefore, my belief that there is a red book on the desk is likely to be true. Therefore, (probably) there is a red book on the desk (1985, p. 118). BonJour goes on to argue that since memory beliefs also can be cognitively spontaneous, they don’t have to cohere with the rest of one’s doxastic system to be justified; rather they are justified in much the same way that cognitively spontaneous perceptual beliefs are. A cognitively spontaneous memory belief is “a belief which is cognitively spontaneous and which is identifiable as a memory belief by reference to its content and very likely also to its phenomenological character and to various introspectively observable accompaniments” (1985, p. 155). The justification of memory beliefs, BonJour maintains, “depend[s] on the fact that certain specific kinds of cognitively spontaneous memory beliefs can, from within the system of beliefs, be identified and argued to be reliable, using arguments which are essentially parallel to those discussed above for observation” (1985, p. 155). So suppose that I am having the cognitively spontaneous memory belief that there was a red book on the desk yesterday. Following the master argument for perception, I am justified in believing that there was a red book on the desk yesterday only if I justifiably believe that it is a cognitively spontaneous belief I am having and that it is very likely to be true. Apart from the fact that cognitively spontaneous memory beliefs don’t necessarily present themselves under this description to the subject’s mind there is the further issue of how the subject can come to know that his seeming memory experiences are indeed reliable guides to the past. It was the point of chapter 6 that such knowledge is unobtainable. BonJour indirectly acknowledges the impossibility of validating one’s ostensible memories when he writes: “[T]he degree of reliability and hence the degree of justification attaching to a memory belief will typically be a good deal lower than that attaching to observation … – too low, in many or even most cases – to satisfy by itself the requirement for knowledge” (1985, p. 155). So in the end BonJour seems to give in to skepticism and accept that, given internalist coherentism, memory knowledge is a very rare commodity. Internalists could try to defend their position in light of the problem of forgotten evidence by pointing out the apparent absurdity of the first-person statement “I remember that p, though I am not justified in believing that p.” But surely the oddity of the locution in question doesn’t prove that remembering that p entails being internally justified in believing it. First, the analogous locution in the thirdperson mode isn’t odd at all. Second, as Robert Audi (1995, p. 39) notes, the oddity is
116
7 Skepticism, Externalism, and Closure
probably due to the difficulty of understanding how someone could have epistemic grounds for attributing the memory that p to himself while lacking epistemic grounds for attributing to himself justification for believing p. Another possible internalist response to the problem of forgotten evidence is to step up the requirements for memory beliefs to qualify as being justified. Matthias Steup, for example, holds that a subject must have adequate evidence for believing the following counterfactual: If I had acquired the belief that p from an untrustworthy source, I would not have formed the belief that p. The idea is that if you come to believe that p by, say, reading the National Inquirer you fail to justifiably believe that p because you fail to have adequate evidence for the indicated counterfactual. Following Alvin Goldman (1999, p. 281), I think that Steup’s counterfactual requirement on continuous justification is both too demanding and too weak. First, the counterfactual in question is too demanding, for it is difficult to get adequate evidence for it – let alone evidence that is accessible to the subject merely by reflection. Second, the counterfactual is too weak because one may have adequate evidence for it and hence be justified in believing it but not be justified in one’s particular memory belief since one fails to realize that one acquired the belief in question from an untrustworthy source. In other words, one may be justified in believing the counterfactual but fail to apply it appropriately. The upshot is that Steup’s proposal does not circumvent of the problem of forgotten evidence. The only way for an internalist to acknowledge that a retained belief can be justified even when the original justification is irretrievably lost and no new evidence is gained is by adopting something like the following principle: If a person is justified in a given belief, to the extent of knowing it, and if he simply retains it, then he continues to be justified in believing it, and accordingly, continues to know it – regardless of whether or not he has forgotten his initial grounds. Thus, one cannot use the fact that a person cannot recall the original evidence for his retained belief to argue that the belief doesn’t qualify as being justified. The intended principle may be called the principle of continuous justification: S’s memory belief that p from t1 is continuously justified at t2 if S continues to (dispositionally) believe that p throughout the interval – even if S lost his original knowledge-producing evidence and has acquired no fresh evidence in the meantime. (Something like the principle of continuous justification has already been encountered in the context of the evidential retention condition in section 2.6.) An early proponent of the principle of continuous justification is Sydney Shoemaker. In elaborating on the notion of retained knowledge, Shoemaker holds that “a person’s belief can be well-grounded (well-grounded enough to qualify it for knowledge if it is true) by virtue of its having been so grounded in the past” (1967, p. 274). Shoemaker elaborates his reason for advocating the principle in question as follows: If it be denied that knowledge and belief can be simply retained, and that a belief can be well grounded simply by virtue of having been acquired on the basis of good evidence or from a reliable source, the only alternative seems to be that we must be constantly re-acquiring, on the basis of a continuous supply of fresh evidence, every item of knowledge we possess and are said to remember. The latter seems a fantastic view (1967, pp. 271–272).
7.4 Internalism About Continuous Justification
117
Shoemaker is undeniably right in supposing that the view according to which memory knowledge is a matter of the constant reacquisition of knowledge is fantastic. It is quite common that we come to know some proposition on the basis of some evidence but lose our original evidence while retaining the knowledge. Rejection of the principle of continuous justification brings with it the rejection of much of what is ordinarily counted as knowledge. The literature contains two kinds of explanation for continuous justification: an internalist-foundationalist explanation (proposed by George Pappas and John Pollock) and an externalist-reliabilist explanation (proposed by Fred Dretske and Palle Yourgrau). Section 7.4 is a critical discussion of internalist explanations of continuous justification. In section 7.5 I will discuss and defend externalist explanations of continuous justification.
7.4
Internalism About Continuous Justification
How can someone be justified in believing that p, if he lacks the justificationproducing evidence he had when he first came to justifiably believe that p? Answering this question seems to be particularly hard for a proponent of epistemic internalism like George Pappas. For internalism requires that nothing which the agent fails to have cognitive access to is relevant for the epistemic appraisal of the target belief. Yet Pappas is not only an internalist but also a foundationalist. And it is his commitment to foundationalism that allows him to account for continuous justification. Continuous justification, he claims, is a kind of immediate or non-inferential justification, that is, “the justification that one has for a proposition and that one would have even if one were not to have any evidence that makes up or helps make up one’s justification for believing that proposition” (1980, pp. 132–133). Given this position, much of one’s knowledge of the past turns out to be non-inferential. Pappas’ chief motivation for his internalist-foundationalist account of continuous justification is the conviction that there is no viable alternative. He considers three alternative accounts. First, one might say that it is the past evidence which justifies S’s present belief that p. This, however, will not do, for, according to internalism, “a person is justified in believing a proposition on the basis of some evidence only if he is justified in believing the evidence he then has” (1980, p. 129). Since S has completely forgotten the original evidence, he is not justified (in the internalist sense) in believing p. Second, one might argue that the fact that S presently remembers that p justifies his belief that p. This proposal will not do, for, if remembering is a form of knowing, the proposal comes down to the explanatorily vacuous truism that knowing that p is a justification for believing that p. Given that “knowing that p just is remembering that p … [S’s] remembering that p is not the evidence on the basis of which S now knows that p” (1980, p. 129). Third, it might be thought that S’s seeming to remember provides him with the evidence that justifies his retained belief that p. The reason this proposal is not convincing is that there is no reason to insist that continuous justification requires the subject to seem to
118
7 Skepticism, Externalism, and Closure
remember that p. “Of course, S might seem to remember that p, now …. The point to emphasize, though, is that he would know that p, now, even if he did not now seem to remember that p. Thus, seeming to remember that p now is not the evidence on the basis of which S now knows that p” (1980, p. 130). Another internalist-foundationalist account of continuous justification has been developed by John Pollock (1974, ch. 7; 1986, ch. 2). Pollock distinguishes between “remembering that p” and “recalling that p,” the former being a success verb while the latter refers to having an experience just like remembering but without the implication of truth. “S recalls that p” is entailed by “S remembers that p,” but not vice versa. Contrary to Pappas, Pollock maintains that having an experience of seeming to remember (or recalling) that p makes one prima facie justified in believing that p. “‘S recalls that p’ is a prima facie reason for S to believe that it was true that p” (Pollock 1974, p. 193). Just as states of perceptual experience are said to confer justification to perceptual beliefs, so states of memorial experience, which are relevantly similar, are said to justify memory beliefs. Since recalling that p is only a prima facie reason for believing p it could be that, all things considered, S is not justified in his belief that p. S might, for example, have reason to believe that his memory is unreliable. This would undermine his prima facie justification because it would give him reason to deny that remembers that p.6 Pollock declares: This is not to say that my original warrant is totally irrelevant to the justification of my present claim to remember. Although I need not first ascertain that I originally had a good reason for believing-that-p before I can justifiably claim to remember-that-p, nevertheless, if I discover somehow that I did not have a good reason, this entails that I do not remember, and hence defeats the justification of my claim to remember. Thus my original warrant is relevant, but only negatively, as a defeater. If I recall-that-p, I have a prima facie reason for believing-that-p, but if I also know that my original belief-that-p, from which my present recollection stems, was unjustified, this defeats my prima facie reason (1974, p. 193).
But recalling and ruling out defeaters is not yet sufficient for continuous justification. On Pollock’s view, q is a reason for believing p if it is logically possible for a person to become justified in believing that p by believing it on the basis of q (1986, pp. 36–37). In other words, something’s being a reason for a belief entails that the belief is based on it. Thus for the belief that p to be continuously justified – in addition to recalling that p and not having any defeater for it – one may not base the belief on anything else but the recalling; one may not base it, for instance, on wishful thinking. Pollock’s final account of continuous justification can be summarized as follows: S’s memory belief that p from t1 is continuously justified at t2 if (1) S recalls that p, (2) S bases his belief that p on his recalling, and (3) S possesses no defeater for his belief that p.
6 Pollock’s term for positive counterevidence is “undercutting defeater” (1986, p. 39). He claims that only a proposition that is currently in working memory can be a candidate defeater. A potentially defeating proposition stored in long-term memory is irrelevant to the justificatory status of what I am now believing, even if I could retrieve this belief with relative ease (ibid., pp. 46–58).
7.4 Internalism About Continuous Justification
119
Pollock’s foundationalist account of continuous justification fails for at least three reasons. First, it cannot be seriously maintained that distinctive memorial phenomenology attends all of the memory beliefs we are justified in holding, even all our occurrent ones. And even if what Pollock labels “recalling” were a necessary component of occurrent memory beliefs, this doesn’t generalize to dispositional memory beliefs. Second, granted that memory beliefs are accompanied by states of recalling, it is not clear that having an experience of seeming to remember and having a memory belief are two distinct states so that the latter can be based on the former. Often we don’t notice a temporal succession between these two kinds of states (Senor 1993, pp. 460–461). Third, the experiential features of memory beliefs can’t do the epistemological work that Pollock assigns to them. For the experiential state of recalling to function as a justificatory factor it must either increase the likelihood of the memory belief in question being true, or make it rational for the subject to endorse the memory belief. The former interpretation is implausible, for the experiential state of seeming to remember is certainly not a reliable mark of truth-effectiveness. And since it is a well-known fact that the experience of seeming to remember something is frequently misleading, it does not seem rational to endorse a memory belief solely on the basis of such an experience. Assuming that the justification of retained beliefs is independent of the experiential features that may accompany such beliefs, one might try to rescue the gist of Pollock’s internalist-foundationalist account of continuous justification by dropping all reference to the experiential state of recalling. According to the new proposal, S’s retained belief that p from t1 is continuously justified at t2 if (1) S continues to believe that p throughout the interval from t1 to t2 and (2) S possesses no defeater for his belief that p. But this will not do. For whether S has a defeater for p depends in part on the breadth of his background knowledge and the care he puts into the investigation. According to the proposal at hand, S’s memory belief that p may be deemed justified simply because S can’t think of a defeater due to epistemic negligence. This view, I take it, is too counterintuitive to be convincing. To bring the last point home, consider Harman’s (1973, pp. 190–191) adoption of the famous library-detective example, derived in its basic scenario from Keith Lehrer and Thomas Paxson (1969): After having observed Tom stealing a book the library detective brings charges against him for theft. It was in fact Tom who stole the book and the detective knows it. The detective retains his belief and testifies before the court at Tom’s trial that Tom stole the book. However, before the trial, Tom’s mother tries to save her son by telling the detective that it was Tom’s identical twin who stole the book. The detective believes what she says, although her story is a lie. The detective, being somewhat slow, fails to see the connection between what Tom’s mother told him and continues to believe that Tom stole the book. The crucial point, for present purposes, is that the detective’s original evidence is undermined. He comes to know that Tom stole the book (p) on good enough evidence, and persists in believing that p. But once he accepts the lie Tom’s mother tells him, he is no longer justified in believing that p, and no longer knows that p, for “[o]ne does not continue to know something
120
7 Skepticism, Externalism, and Closure
if it becomes irrational to continue to believe it” (Harman 1973, p. 191). The detective stops knowing that p even though he does not stop believing that p. The reason he stops knowing is that no one who, purely out of epistemic negligence, does not consider a defeater to p can be said to justifiably believe that p. When this idea is incorporated in the above account of continuous justification, the following account emerges: S’s memory belief that p from t1 is continuously justified at t2 if (1) S continues to believe that p throughout the interval from t1 to t2, and (2) there is no further fact q such that if S came to believe q, his justification for believing that p would be defeated.7 The latest account of continuous justification is very promising but violates epistemic internalism. If continuous justification depends on the absence of counterevidence the subject may be unaware of, it is not the case, as internalists demand, that all of the factors that render a memory belief justified (or unjustified) are cognitively accessible to the subject merely by reflection. Thus it looks as if a convincing foundationalist account of continuous justification will have to be externalist in nature. Yet before examining externalist explanations of continuous justification I would like to investigate a further criticism of Pollock’s foundationalist account of continuous justification. According to Michael Huemer (1999, pp. 348–349), the problem with Pollock’s theory is that it has the consequence that remembering p increases one’s justification for believing p; memory can even convert an unjustified belief into a justified one. Suppose that S initially comes to believe p by means of an a priori proof. The next day S still remembers p and the proof of it. But since he also has the experience of seeming to remember that p, he now has two reasons for holding p true, an inferential and a foundational one. Thus S has more justification for p now than he had at the original learning. Huemer finds this counterintuitive. To see that Pollock’s foundationalism even allows memory to convert an unjustified belief into a justified one, suppose that yesterday S came to believe that p by wishful thinking. Today he remembers that p but not the way in which he formed the belief. Between yesterday and today S has acquired no new evidence for or against p. Yet since it seems to S that he remembers that p. Pollock will have to say that S’s belief that p is justified to a greater degree today than it was yesterday. Compare S to his twin brother who acquired the belief that p in the same way as S but who wasn’t so fortunate as to forget how he acquired the belief in the first place. The twin brother’s belief is justified to a lesser extent than his forgetful twin brother S. Huemer reckons that this is an implausible position. Interestingly, Huemer does not explain why he thinks that memory cannot make a belief take on an epistemic status different from the one it had at the time it was originally acquired. Contrary to Huemer, I think that there are indeed cases where memory makes an unknown proposition known, an unjustified belief justified, and an irrational belief rational. One such case has been described in section 7.2. The eagle-example is meant to show that the justification for the memorial belief can
7
Harman (1973, p. 192) ultimately rejects (a version of) this principle.
7.4 Internalism About Continuous Justification
121
exceed the justification for the original belief. In the eagle-example it is the presence of a doxastic defeater that prevents S from justifiably believing that p. Once the doxastic defeater is removed S is in a position to justifiably believe that p simply in virtue of remembering p. The eagle-example worked with doxastic defeaters. In the remainder of this section I will make the case for memory’s being a generative source of justification by employing normative defeaters. The following example is adapted from Jennifer Lackey (2005, pp. 640–641). The mayor of a big city has been assassinated. At t1 S learned about the assassination by reading a report in a generally reliable newspaper. The report was written by someone who was an eyewitness. Two days later, at t2, the mayor’s associates, wishing to prevent panic, have issued television announcements and newspaper reports saying falsely that the assassination attempt had failed and that the mayor is alive and well. Most people see the television announcement and read the newspaper reports and therefore come to believe at t2 that the mayor is alive. However, by a fluke, S misses the denial of the assassination and continues to hold the true belief that the mayor had been assassinated. At t3, the scheme to cover up the mayor’s death is exposed, and all of the major newspapers and television stations report that the mayor has in fact been assassinated. Throughout t2 and t3, S remains blissfully ignorant of all of the relevant reports and continues to believe that the mayor has been assassinated solely on the basis of remembering the original newspaper report from t1. When S continues to believe at t2 that the mayor had been assassinated, he has an undefeated normative defeater for this belief and thus fails to have knowledge. Since every newspaper and television station was reporting that the mayor had not been assassinated, S should have believed that the mayor is alive and well. At t3, however, it is no longer the case that S should believe that the mayor is alive, because all of the newspapers and television stations have ceased reporting this false information. Hence, at t3 there is no undefeated normative defeater for S’s belief about the mayor. Given that the presence of the undefeated defeater was the only factor preventing S’s true belief from qualifying as knowledge at t2, its absence enables him to know at t3. Thus S knows at t3 that the mayor is dead without having known this proposition at t2. What this example shows is that a subject’s relation to normative defeaters can change over time as a result of changes in the external environment, thereby enabling memory to generate knowledge. Now there are two ways in which one may attempt to deny that the assassination-example undermines the epistemic theory of memory. First, one may attribute the generation of the epistemic status of knowledge in this example to the external environment rather than to the memory. After all, it is changes in the environment that are responsible for the absence of defeaters; and it is because of the absence of defeaters that the belief in question qualifies as knowledge. So shouldn’t we credit the external environment, rather than memory, with the ability to generate knowledge? Certainly not. For a change in an environment has epistemic significance only relative to some particular faculty of a given subject. The only plausible candidate for a faculty operative in this case is memory. Hence proper credit for the generation of the epistemic properties should be understood to accrue to memory.
122
7 Skepticism, Externalism, and Closure
Now, of course, in the example at hand memory can generate knowledge only because the environment cooperates by removing defeaters. But this shouldn’t worry us. For as Wittgenstein says in On Certainty (§505): “It is always by favor of Nature that one knows something.” Another way of undermining the assassination-example is to deny that S does not know at t2 of the mayor’s death. Why, one might ask, is the unreliable evidence against S’s belief in the mayor’s death capable of preventing an otherwise epistemically impeccable belief from being an instance of knowledge? The suggestion is that because S is unaware of the misleading evidence against his belief about the mayor, he does indeed know at t2 that the mayor is dead. I disagree. There are certain pieces of information that one is expected to be aware of by virtue of being a member of society, and the mere fact that one happens to be ignorant of such information does not enable one to have knowledge that everyone else lacks. Had S turned on any news channel or picked up any newspaper, he would have come to believe that the mayor is alive. Unless this sort of ignorance of “common knowledge” is appropriately grounded, it is incompatible with having the knowledge in question. For otherwise one would know more by knowing less. Given that memory can function as a generative source of justification Huemer’s criticism of Pollock’s internalist-foundationalist account of continuous justification crumbles. This does mean, however, that the internalist-foundationalist account is unproblematic. As was shown before, the combination of foundationalism and internalism fails to yield a plausible account of continuous justification. A convincing foundationalist account of continuous justification will have to be externalist in nature.
7.5
Externalism About Continuous Justification
Given that epistemic internalism has problems accounting for continuous justification, there are two possible conclusions one can draw. Either one rejects the idea of continuous justification and maintains that for a retained belief to continue to be justified the subject must remember the justification-producing evidence; or one settles for an externalist-reliabilist account of continuous justification. I reckon that the former is not a viable option, for it threatens a drastic diminution in the stock of beliefs ordinarily deemed justified, and hence in the stock of knowledge. I attempt to show that externalist reliabilism squares well with many of our intuitions regarding the continuous justification of memory beliefs. Three distinct externalist accounts of continuous justification can be found in the literature – Tyler Burge’s analytic account, Alvin Goldman’s process reliabilist account, and the dualistic account proposed by Fred Dretske and Palle Yourgrau. Tyler Burge argues that when a belief stems from what he calls preservative memory it gets its positive epistemic status from the conditions of the initial acquisition of this belief by the original sources. Preservative memory (which he contrasts with substantive memory) is “memory that preserves beliefs with their
7.5 Externalism About Continuous Justification
123
justifications, but contributes no independent source of justification” (1997, p. 37). A person who is engaged in preservative memory is a priori entitled to accept a proposition that he seems to remember, unless there are strong reasons not to do so. Entitlements are defined as “epistemic rights or warrants that need not be understood or even accessible to the subject” (1993, p. 458). In other words, an entitlement is a warrant that is external, in that the subject need not realize that the beliefs stemming from his preservative memory are so warranted. Entitlements to trust memory need not be accessible to a subject. But any conditions which overturn the default entitlement must be accessible to the subject. Burge writes: The [acceptance] principle says that the entitlement holds unless there are stronger reasons (available to the person) that override it …. It is enough for the individual’s being warranted that there are no defeaters; defeaters of the entitlement must be available to him (1997, p. 45, fn. 4; 1993, p. 469).
So there is an asymmetry between the entitlement to trust in ostensible memory, which need not be accessible to the subject, and any defeaters, which must be accessible to him. The awareness of a defeater is not the only instance when the a priori entitlement to trust one’s ostensible memories is suspended. When the original belief is unjustified, we are also not entitled to accept a proposition that we seem to remember. We would not be entitled to the belief if it were preserved from unwarranted acquisitions that we had forgotten: we cannot, I think, become warranted by forgetting the poor grounds we originally had, and then relying on the remembered belief (1997, pp. 39–40).
The reason a belief does not become justified simply in virtue of being remembered is that, according to Burge, memory can only preserve but not generate justification. But as was explained in section 7.2, there are good reasons to think that memory is capable of rendering an unjustified belief justified. Instead of explaining continuous justification in terms of an a priori entitlement some externalists such as Alvin Goldman (1999, pp. 280–281) hold that for a retained belief to be justified, the belief must be sustained by a reliable cognitive process. According to this account, what allows remembered beliefs to inherit their positive epistemic status is the fact that memory is by and large a reliable cognitive process. Yet for a retained belief to be justified it must not only originate from the reliable process of memory but it must also be the case that the agent does not have significant reason to distrust his memory. Like Burge, Goldman adds a no-defeater clause to his account of continuous justification. In sum, S’s memory belief that p from t1 is continuously justified at t2 only if (1) S continues to believe that p throughout the interval between t1 and t2, (2) S was justified at t1 in coming to believe that p, (3) S’s belief that p has been sustained by a reliable cognitive mechanism from t1 to t2, and (4) for every time ti (1 < i > 2) there is no epistemically relevant proposition for S that would have defeated the justification he had at t1 nor is there an epistemically relevant proposition for S to the effect that his memory is unreliable from t1 to t2. While Burge and Goldman require only the absence of background beliefs providing reasons to question the truth of one’s retained belief or the reliability of one’s
124
7 Skepticism, Externalism, and Closure
memory, Fred Dretske and Palle Yourgrau (1983) argue that, in special circumstances, continuous justification also involves the presence of positive reasons to the effect that the original belief-formation process was a reliable one. It is the following example that convinces them that an externalist account of continuous justification à la Burge and Goldman is too weak: Clyde has two friends, R (Reliable) and D (Deceiver) both of whom he considers reliable informants. He knows that R is a reliable informant and falsely believes that D is as well. Clyde wants to find out whether the enemy plans to attack and both R and D have this information. R tells Clyde that the enemy does plan to attack (=p). Since R is a reliable informant and is known as such by Clyde, we can conclude that Clyde comes to know that p. He makes a mental note of this piece of information and falls asleep. When he wakes up, Clyde can remember only that the enemy plans to attack but he is unable to remember whether it was R or D who told him. However, since he regards both R and D reliable informants, this does not shake his confidence in the belief that p. He knows that he got the information p from one and therefore believes that p. The question is whether Clyde, upon awakening, is justified in believing p? Dretske and Yourgrau think that Clyde’s inability, upon awakening, to remember from whom he got p (R or D) leaves his retained belief that p without justification. It is not sufficient, they claim, that a belief be acquired through a completely reliable process; in addition the subject must be able to tell that this is in fact the process by means of which he acquired the belief. [T]he process by means of which Clyde acquired his belief (about the enemy plans) is either to be described as: (P1) “Told by a trusted informant” in which case the process, not being reliable (D is also a trusted informant), does not yield knowledge, or it is to be described as: (P2) “Told by R” in which case the process, though perfectly reliable, is one that Clyde, having forgotten who told him, is unable to tell is the process by means of which he acquired his belief (1983, p. 359).
For Clyde, upon awakening, to justifiably believe that p, he must remember not only p but also that it was R who told him this. Unless Clyde can reconstruct the original belief-formation process in a way that brings out its reliability, his continuous belief that the enemy plans to attack does not qualify as being justified. The point is not that Clyde falsely believes D to be reliable. Rather, the point is that Clyde cannot remember the crucial justificatory factor of his original belief – that the person who told him p is someone he knows to be reliable. Since it was the known reliability of the informant of p that rendered the original belief that p justified, the loss of this piece of information robs the belief that p of its justification; or so Dretske and Yourgrau argue. Whether Clyde justifiably believes that the enemy plans to attack depends on whether he remembers that it was R who told him this. Moreover, Dretske and Yourgrau argue that the memory that it was R who told him p must itself be justified in the sense of being reliably correlated with R’s having told him. “Clyde’s morning belief that R told him about the enemy plan will be a reliable index to the enemy’s actual plan if Clyde’s morning belief that it was R that told him this is itself a reliable index to R’s having told him” (1983, p. 364). And should Clyde’s belief that it was R who told him that p be unreliable (he often confused R and D in his
7.5 Externalism About Continuous Justification
125
memory), then he needs some more specific belief. He may need to, for example, believe not only that R told him that p but that he told him this while waiting in line at the grocery store. This more specific belief may be reliable after all. Clyde wouldn’t believe having met R while waiting in line at the grocery store if this is not what has happened. The upshot is that “[m]emory, in so far as it involves a retention of knowledge, is not simply the preservation of a suitably acquired belief, it is the suitable preservation of a suitably acquired belief” (1983, p. 363). In Clyde’s case, the belief that p is “suitably” preserved if he can reliably tell that it was acquired by means of a reliable belief-formation process. Continuous justification presupposes the ability to reconstruct the justificatory factors of the original belief. This position can be called the dualistic theory because it holds that the question “What justifies S in believing that p on the basis of seeming to remember p?” requires a twopart answer: “Was S justified in adopting the belief that p at some point in the past?” and “Is S justified in believing that he was justified in adopting the belief that p?”8 Dretske and Yourgrau don’t claim that whenever one forgets the source of one’s past belief that p, one thereby ceases to justifiably believe that p. They acknowledge that there are incidences where the epistemic status of a retained belief is hereditary. “Ordinarily we don’t have to remember what justified us in believing in order to retain knowledge of what we thereby came to believe” (Dretske 1982, p. 111n; Dretske and Yourgrau 1983, p. 359n). Yet they think that the case of Clyde is different. Unfortunately they don’t say what needs to be the case for justification to be preserved simply by virtue of the belief’s having been retained by the subject. Dretske and Yourgrau go on to draw some general conclusions regarding memory. First, the richer and more interconnected one’s system of beliefs about the past, the more likely that an individual belief qualifies as knowledge. Someone who remembers only what he is told but not who told him and in which context knows less than someone whose retained belief is accompanied by a lot of episodic “texture.” Second, beliefs about public events are, by and large, more likely to qualify as knowledge than beliefs about private ones. The reason is that beliefs about public events tend to be better supported by other beliefs one holds than beliefs about private events. Third, there is a division of labor between the acquisition and preservation process. If one is very cautious about the beliefs one acquires, then one’s memory has an easier time preserving the knowledge so acquired. Though Clyde’s brother learns exactly what Clyde learns from his conversation with R (namely, that the enemy plans to attack), his brother, being a much more cautious fellow (he doesn’t trust D) needn’t, like Clyde, remember who told him about the enemy plans (R or D) in order to remember that the enemy plans to attack. If he is aware of his own caution
8 The label “dualistic theory” is used by Huemer (1999, p. 351) to refer to a very different conception. According to Huemer’s dualistic theory, a memory belief is justified “if and only if one had an adequate justification for adopting it at some point, and thenceforward one was justified in retaining it.”
126
7 Skepticism, Externalism, and Closure
in this regard, he can reconstruct this fact, but the point is that the belief that R told him need not be preserved in memory for him to continue knowing something that Clyde can only continue knowing by remembering who told him. The conclusion seems to be that the gullible may learn as much as their more critical colleagues, but they need a much better memory to keep it (1983, pp. 366–367).
My objection to Dretske’s and Yourgrau’s dualistic theory of continuous justification is the same as my objection to internalist theories of continuous justification – it threatens skeptical outcomes. A large class of what is commonly counted as memory knowledge is made up of beliefs for which we have forgotten the original evidence. Adopting the dualistic theory has the consequence of rendering memory knowledge a very rare commodity indeed. Unless we are willing to give in to skepticism about memory knowledge, we should settle either for Burge’s analytic account or Goldman’s process reliabilist account of continuous justification.
7.6
Russell’s Hypothesis and Closure
So far I have discussed the skeptical question of how we can know (justifiably believe) that the experiences of seeming to remember on which we base our memory-claims are veridical. In this section I wish to examine a related yet distinct form of skepticism, namely skepticism regarding our ability to know on the basis of memory experiences that there is a past. After explaining this form of skepticism and rejecting attempts to declare it incoherent, I will set forth an anti-skeptical strategy based on the limitation of the scope of the epistemic closure principle. Memories of the past imply that there is a past. To remember, for example, that J.F. Kennedy was assassinated in 1963 it must be the case that there is a past of at least 45 years. If I know that memories of the past imply that there is a past and if I know that Kennedy was assassinated in 1963, then, given that knowledge is closed under known entailment, I know that there is a past of at least 45 years. The closure principle states that if S knows p, and if S knows that p entails q, then S knows q. In other words, for S to know that p, S must know those propositions which he knows are entailed by p. Usually the closure principle is used to back skepticism regarding our ability to know about the external world: If I know that this is a chair, I also have to know that I am not dreaming. But I don’t know that I am not dreaming, since if I were, things would seem exactly as they seem now and there might be no chair. It follows that I don’t know that this is a chair. Generally speaking, if knowing p entails knowing that the skeptical hypothesis is false, then, if you don’t know that the skeptical hypothesis is false, then you don’t know that p. In the case of memory knowledge, the skeptic argues that if remembering a proposition about the past entails that there is a past and if you know about this entailment, then you must know that there is a past to know the proposition in question. The skeptic then goes on to show that you cannot know that there is a past. The reason you don’t know that there is a past is that there is a counterfactual situation where there is no past but where your memory
7.6 Russell’s Hypothesis and Closure
127
experiences are the same as they are now. The impossibility of ruling out such a counterfactual situation in combination with the closure principle yield the result that memory knowledge regarding the past is unobtainable. The significance of the skeptical problem under consideration crucially depends on the plausibility of the skeptical hypothesis according to which, for all we know, there may be no past at all. How plausible is the suggestion that we cannot tell whether the world came into existence 5 minutes ago, complete with ostensible memories, records, fossils, and so on? Bertrand Russell took the suggestion seriously: It is not logically necessary to the existence of a memory-belief that the event remembered should have occurred, or even that the past should have existed at all. There is no logical impossibility in the hypothesis that the world sprang into being five minutes ago, exactly as it then was, with a population that “remembered” a wholly unreal past. There is no logically necessary connection between events at different times; therefore, nothing that is happening nor or will ever happen in the future can disprove the hypothesis that the world began five minutes ago…. I am not suggesting that the non-existence of the past should be entertained as a serious hypothesis. Like all skeptical hypotheses, it is logically tenable, but uninteresting (Russell 1921, pp. 159–60; cf. 1927, p. 7; 1948, p. 228).
Though Russell’s skeptical hypothesis seems anything but plausible, the fundamentalist biologist Philip Henry Gosse (1857) promoted an analogous view. Gosse argued that God had created the world in 4004 B.C., with its fossils and lava and other “records” of a distant past, for the sole purpose of trying men’s faith. Norman Malcolm develops two (ultimately unsuccessful) arguments to the effect that Russell’s skeptical hypothesis is incoherent. Malcolm’s first argument is meant to show that the inhabitants of Russell’s imaginary world possess genuine memories. Given the skeptical hypothesis, even though there is no past, the ostensible memories are in agreement with each other and with the so-called “records.” But if there were this kind of agreement then the apparent memories would be verified as true. This is what the verification of apparent memories means…. And if the apparent memories were verified it would not be intelligible to hold that, nevertheless, the past they describe may not have existed…. The supposed unreal past has turned out to be real. Russell said that his skeptical hypothesis was logically tenable but uninteresting. I believe the exact opposite has proved to be the case (Malcolm 1963, pp. 198–199).
Malcolm’s argument can be parsed into two steps: (1) If ostensible memories are in agreement with one another and with the “records,” then the ostensible memories are verified as true. (2) If ostensible memories are verified as true, then it is impossible that the past they refer to did not exist. Malcolm concludes that the allegedly unreal past turns out to be real after all. The problem with this argument is that it rests on an equivocation of the verb “to verify” (Cornman 1965, p. 166; Locke 1971, pp. 125–126) and of the term “record” (Naylor 1966, pp. 125–126). “Verification” can mean establishing that something is true or it can mean trying to establish that something is true. According to the first reading, “to verify” is a success verb, according to the second it is not. The way “to verify” is used in (1), it refers to the method one uses in trying to discover whether one’s ostensible memories are true. According to this sense of “verification”
128
7 Skepticism, Externalism, and Closure
it remains an open question whether one’s ostensible memories are indeed veridical. And this is the only kind of verification of which the inhabitants of Russell’s imaginary world are capable. In (2), however, “to verify” is used in an altogether different sense. According to (2), something’s being verified as true entails that it is true; otherwise it wouldn’t follow that it is impossible that ostensible memories be verified as true and yet not be true. Given that (1) and (2) operate with different senses of “to verify,” Malcolm’s argument is invalid. If the sense of “to verify” used in (2) were applied to (1), the argument would beg the question against the skeptic, rather than proving him wrong. And if the sense of “to verify” used in (1) were applied to (2), premise (2) would be plainly false. By “records” we usually mean things which are evidence for past events. But if we suppose that the past is unreal, then whatever seeming evidence there is for an ostensible memory is not a genuine record but a spurious one. Agreement with spurious records is not sufficient to verify anything, where “to verify” means to establish that something is true. Thus, if we interpret “to verify” in (1) as a success verb, the “records” have to be taken as genuine even though they are, of course, spurious. Malcolm’s second (unsuccessful) argument to the effect that Russell’s skeptical hypothesis is incoherent starts from the observation that the hypothesis can be neither verified nor falsified. Neither geology, nor chemistry or radar astronomy can supply evidence that the world came into existence 5 minutes ago (1963, p. 200). Some hypotheses concerning the past cannot be verified or falsified because the relevant evidence is lost or destroyed or is otherwise unobtainable. Those hypotheses don’t pose a philosophical problem, for we can specify what would count as evidence for them, whether such evidence is available or not. Russell’s hypothesis is different though. In the case of Russell’s hypothesis, it is logically impossible to produce evidence either way, since any possible evidence would itself fall within the scope of the hypothesis. Marcus Singer makes the same point as Malcolm when he writes: [I]t is impossible to conceive of anything that might possibly serve as evidence for an answer that would not itself come within the terms of the question [How do you know that the world did not come into existence five minutes ago?]. Thus for anything selected as evidence for an answer to this question, the very same question demands evidence that this is evidence, and then of course one would have to supply evidence for this further evidence, and so on. This is why any proposed answer to this question of Russell’s would be question-begging, and this is why it is logically impossible for the question to be answered (1963, p. 190).
According to some, the meaninglessness of Russell’s hypothesis is a direct consequence of the hypothesis’ not being verifiable or falsifiable. Ludwig Wittgenstein, for example, writes: By examining Russell’s hypothesis that the world was created five minutes ago I shall try to explain what I mean in saying that it is meaningless. Russell’s hypothesis was so arranged that nothing could bear it out or refute it. Whatever our experience might be, it would be in agreement with it. The point of saying that something has happened derives from there being a criterion for its truth. To lay down the evidence for what happened five minutes ago is like laying down rules for making measurements (1979, pp. 25–26).
7.6 Russell’s Hypothesis and Closure
129
A number of philosophers have followed Wittgenstein in arguing that the reason Russell’s hypothesis is meaningless is that it is logically impossible to produce evidence either way (cf. Butler 1959, p. 17; Lewis 1946, p. 358; Singer 1963, p. 194; Waismann 1965, pp. 20–21). But as Don Locke (1971, p. 127) has pointed out, this argument assumes a widely rejected theory of meaning – verificationism. Verificationism has it that a statement has meaning only to the extent that it is capable of verification. Verificationism, of course, begs the very question at issue. If we stipulate that a theory’s being meaningful is nothing more than its being optimally rational according to our methodology, it follows that Russell’s hypothesis is meaningless. But this is really no more than a stipulation. It does not capture our straightforward notion of meaning which is entirely non-epistemic, that is, not relativized to any methodology. To say about a hypothesis that it is meaningful is to say quite simply that the world could be the way the hypothesis says it is. Of course, the operation of producing evidence for the hypothesis is bound to our epistemic principles. But it does not follow that a hypothesis’ being meaningful depends in any way on epistemic procedures. All it depends on is the way the world is or could be. The skeptic about memory knowledge is a realist about meaning, and his realism is incompatible with verificationism. According to Malcolm, the reason Russell’s hypothesis is meaningless is not so much that it rules out the possibility of evidence for or against that hypothesis, but that it destroys any possibility of our having evidence for or against anything. This amounts to a transcendental argument against Russell’s hypothesis, that is, an argument that shows that the hypothesis has to be accepted as false if our epistemic practice is to be possible. Malcolm declares: The five-minute hypothesis is incompatible with the very concept of evidence. Consider what would be implied by our “believing” that the earth and mankind have just come into existence. If one of us were to “believe” this he would have to renounce not only his previous conception of his own identity, but his entire store of common knowledge – his knowledge of natural processes; his knowledge of the normal properties of anything, so that he would no longer know what wood, water, and fire are; his knowledge of how the words of his language are used or even that he has a language; of how people live, act, and react; of what interests them and what kinds of inquiries they make; of how measurements, experiments and arguments are conducted, and when something is held to be proved…. If he thought out the consequences of this hypothesis he would realize that it is not anything he could rationally believe, because “believing” it would mean that he no longer understood anything at all…. To accept this “hypothesis” as true would mean the destruction of all our thinking.9
Malcolm’s point is that without a past and without memory we would not only know nothing at all but we would have not acquired any reason for believing
9 1963, pp. 200–201. Malcolm’s idea that the presentation of one’s skeptical doubts presupposes, among others, the certainty of one’s mastery of the language through which the doubts are expressed is borrowed from Wittgenstein. Wittgenstein declares: “If you are not certain of any fact, you cannot be certain of the meanings of your words either. If you tried to doubt everything you would not get as far as doubting anything. The game of doubting itself presupposes certainty” (1969, §114, §115). Naturally, if one is to assume the mastery of language, one has to rely on memory.
130
7 Skepticism, Externalism, and Closure
anything, not even about our present environment. If Russell’s hypothesis were correct, we would have no reason for regarding the hypothesis itself as correct. Making sense of the truth of the hypothesis presupposes that “there is a framework of logical possibilities which is independent of our concepts and our application of language.” Since Malcolm thinks that such a framework is “a mere piece of imagery” (1963, p. 202) he concludes that Russell’s hypothesis is meaningless or incoherent. Malcolm is neither the only one nor the first one to claim that Russell’s hypothesis is meaningless. This charge goes back to Ronald Butler who thought that skepticism concerning the reality and knowability of the past is “queer” because “once the hypothesis is formulated there is nothing which members of the population might do either to confirm it or disprove it…. With no conceivable way of establishing its truth or falsehood, one is inclined to balk at admitting it as a genuine hypothesis at all” (1959, p. 18). Marcus Singer argues that Russell’s hypothesis is meaningless “because all of the evidence for it would equally be evidence for its denial, and there could not possibly be any evidence for either, and consequently there could not possibly be any evidence either for it or against it” (1963, p. 197). C.I. Lewis (1946, p. 359) thinks that the supposition that we are victims of a “systematic delusion of memory” is meaningless because unverifiable in principle. And following Lewis, Don Locke claims that Russell’s hypothesis is not “logically tenable” since “rationality itself requires that we accept our evidence for the past as evidence for the past, even though we cannot … prove that there has been the past that this evidence points to” (1971, p. 131). Finally Shoemaker maintains that “it is not intelligible to say (and hence not logically possible) that memory beliefs are always or generally false” (1967, p. 274). (As was explained in section 6.3, Shoemaker thinks that it is a necessary truth that memory beliefs are for the most part correct.) Despite the arguments put forward by Malcolm and others, Russell’s skeptical hypothesis seems clearly coherent. For anyone who finds Russell’s skeptical hypothesis at all persuasive its very persuasiveness provides at least as strong an argument against accepting the incoherence-thesis as the incoherence-thesis can provide against the coherence of Russell’s skeptical conclusion. What is more, those who accuse Russell’s skepticism of meaninglessness misjudge the status of the skeptical hypothesis. They assume that the skeptic is someone who takes his hypothesis to be correct and who attempts to provide evidence for it. Given this assumption, skepticism is incoherent, for if skepticism were correct, we would have no reason for holding anything true, let alone skepticism itself. But the skeptic is not committed to the truth of skepticism and does not try to justify skepticism. Skeptical hypotheses are nothing but a trick to illustrate the suggestion that our goal of obtaining knowledge about the world and its past cannot ever be reached. Hence the fact that it is logically impossible to justify skepticism shouldn’t be held against the skeptic. If anything, the critique of Russell’s hypothesis brought forward by Malcolm and others shows that if one takes skepticism about memory knowledge to be true, one gets tangled up in a pragmatic contradiction. But from this it does not follow that skepticism about memory knowledge cannot be true. This would follow only if the
7.6 Russell’s Hypothesis and Closure
131
skeptic’s position contained a logical contradiction; but a logical contradiction is out of the question. Malcolm himself acknowledges that “the sentence ‘The earth came into existence five minutes ago’ is not self-contradictory” (1963, p. 201). For the sentence to be self-contradictory the age of the world would have to be determinable a priori. The fact that skepticism about memory knowledge may involve a pragmatic contradiction doesn’t disqualify the skeptical position; rather it aggravates the skeptical problem: if the suggestion that we lack memory knowledge cannot even be coherently accepted, then, if this suggestion were true, we would be incapable of acknowledging its truth. After having criticized attempts to declare Russell’s skeptical hypothesis incoherent, I will develop my own response to skepticism regarding our ability to know on the basis of memory experiences that there is a past. As was explained at the beginning of this section, remembering (and knowing) what occurred in the past entails that there is a past. Assuming closure of knowledge under known entailment, one needs to know that there is a past to know what occurred in the past. Granted that Russell has established that we cannot know that there is a past, it follows that we cannot know what occurred in the past. I propose to disarm the skepticism under consideration by limiting the scope of the epistemic closure principle. Regarding knowledge of the outside world, most epistemologists agree that the closure principle is too strict to be convincing. The fact that we know ordinary empirical propositions and the fact that we do not know the falsity of skeptical alternatives is used as an argument against the truth of the closure principle. If knowing that p would require the elimination of every known alternative to p, as suggested by the closure principle, we could never know anything about the world around us. A more plausible view is that knowledge requires the elimination of relevant alternatives only and that skeptical alternatives are normally not relevant. In an ordinary case of claiming to know that some animals in the zoo are zebras, to use Fred Dretske’s (1981b) example, the alternative that they are cleverly painted mules is not relevant. Thus, one can truthfully claim to know that they are zebras despite one’s inability to rule out this fanciful alternative. But in some extraordinary cases, the painted mules hypothesis is a relevant alternative; and then we have to eliminate this alternative to know that what we are seeing are really zebras. Since skeptical alternatives are usually not relevant, the fact that our evidence cannot eliminate them doesn’t force us to abandon our ordinary knowledge claims. The relevant alternatives account of knowledge can be motivated by noting that other concepts exhibit the same logical structure. Two examples of this are the concepts empty and flat (Dretske 1981b). Although both concepts are absolute, just as knowledge is, they are absolute relative to a standard. In the case of flat, there is a standard for what counts as a bump and in the case of empty there is a standard for what counts as containing something. No one would deny that a table is flat because, if observed under a microscope, it looks bumpy. Nor would we deny that a refrigerator is empty because it contains shelves. To be flat means to be free of relevant bumps and to be empty means to be devoid of relevant objects. Similarly, the relevant alternatives account of knowledge states that to know a proposition one’s evidence needs to rule out only relevant alternatives.
132
7 Skepticism, Externalism, and Closure
What makes an alternative relevant? What standard do the alternatives raised by the skeptic fail to meet? It has proven to be difficult to answer these questions with any degree of precision and generality. Roughly speaking, there are an objective and a subjective reading of the concept of relevance. In the objective reading, relevant alternatives consist in certain features of the subject’s circumstances beyond the evidence she has. Among such features are the availability of unpossessed misleading counter-evidence and the objective probability of being wrong. In this sense, whether a subject knows that p on the basis of certain evidence is determined relative to the extra-evidential circumstances of the subject. Two subjects could possess the same evidence for the truth of a certain proposition, and one of them know the proposition while the other fails to know the proposition. In the subjective reading, relevant alternatives consist in certain features of the conversational context. Among such features are the assumptions, purposes and intentions of the attributors of knowledge. Given this sense of context-relativity, two speakers can say of a subject, “S knows that p” and it can be the case that one speaker thereby says something true while the other says something false. It is this subjective or conversational notion of relevance that is suggested by the above mentioned analogy with the concepts flat and empty. I will assume an objective reading of relevant alternatives according to which relevance is concerned with the kind of possibilities that actually exist in the objective situation. Let’s return to skepticism regarding our ability to know on the basis of memory experiences that there is a past. Regarding knowledge about the external world, most philosophers agree that the closure principle is too strict and should be replaced by a version of the relevant alternatives account. What I am suggesting now is that there is no reason to believe that the conditions for memory knowledge are any stricter than the conditions for knowledge about the external world. If knowledge about the external world isn’t closed under known entailment, why should memory knowledge be? And if memory knowledge isn’t closed under known entailment, one need rule out only relevant counter-possibilities to know one’s specific thought contents. Ordinarily, the hypothesis that the world sprang into existence 5 minutes ago is irrelevant and thus doesn’t need to be eliminated – not even if one is aware of the fact that knowledge of what occurred in the past entails that there is a past. Hence, one can know what occurred in the past without knowing that there is a past. Since ordinarily, knowing what occurred in the past doesn’t require knowledge that there is a past, Russell’s skeptical hypothesis does not automatically undermine the possibility of memory knowledge. But if the hypothesis turns out to be a relevant alternative, we would have to eliminate it to know what occurred in the past. To avoid misunderstanding, I should note that the relevant alternative analysis of knowledge is not committed to denying closure tout court. Just because one cannot know that there is a past by way of knowing (on the basis of remembering) what occurred in the past doesn’t mean that one cannot, for example, know that one is petting a mammal by means of knowing that one is petting a cat. If one couldn’t know any of the known consequences of the things one knows, reasoning in general would be in jeopardy. One of the difficulties facing the relevant alternative account
7.6 Russell’s Hypothesis and Closure
133
of knowledge is coming up with a criterion by which to determine the number and kind of counter-possibilities that are ruled out by a particular knowledge claim. To complete my argument I would have to come up with a criterion to differentiate between benign and incorrect applications of the closure principle. This, however, must wait for another occasion. In sum, Russell’s skeptical hypothesis undermines our ordinary memory claims only if it is assumed that knowing what occurred in the past (on the basis of memory experiences) entails knowing that there is a past. I have argued against this entailment. Our memory provides us only with knowledge about past events, if there is a past. But that there is a past is not something we can know on the basis of remembering. It is something we know to be entailed by the events we remember, but it is not itself something we can remember.
Part III
Truth in Memory
Chapter 8
The Factivity Constraint
Memory implies truth. In the case of fact memory this is obvious. “S remembers that Fido was on the sofa” implies that Fido was the sofa. And generally, “S remembers that p” entails that p is the case. Object-, property-, and event memory are also factive since I cannot remember Fido’s flap-ear and his sitting on the sofa without Fido having a flap-ear and without it being the case that he sat on the sofa. “I remember Fido” implies that there is something (Fido) that I remember. Yet in what follows I will focus on the factivity of fact memory (propositional memory). Sections 8.1–8.3 motivate and explain the thesis that propositional remembering is factive. Many philosophers hold that one of the consequences of the factivity constraint on memory is that memory is supposed to work like a photocopier producing duplicates of past contents. On the xerox model of memory the recalled thought must be type-identical with the original thought. The content of a reproductive memory (and the embedded content of a meta-representational memory content, respectively) must be a token of the same type of content as the original thoughts from which it causally derives. Section 8.4, characterizes the xerox model and gives examples of its employment. Section 8.5 argues that the xerox model flies in the face of our actual practice of attributing memory. Frequently, our memory not only stores but also processes the encoded information. Three such information processes are discussed: cognitive dynamics, condensation and schematic processing. Finally section 8.6 argues that the fact that our memory not only retains but also processes the incoming information should not be regarded as an abnormal lapse of an otherwise reliable cognitive faculty, but as part of the very function of memory.
8.1
Motivation of the Truth Condition
Suppose your elementary school teacher played a trick on you by telling you that Columbus discovered America in 1392 (rather than in 1492). Later you are asked “When did Columbus discover America?” and you answer “1392.” Do you remember that America was discovered in 1392? Following the everyday notion of “to remember,” we have to answer in the negative. You cannot remember that Columbus S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
137
138
8 The Factivity Constraint
discovered America in 1392 since it wasn’t in 1392 that he discovered America. You can only remember having (falsely) believed that Columbus discovered America in 1392. What are arguments in favor of the thesis that propositional remembering is factive? Though virtually everyone writing in this field agrees that propositional remembering is factive, hardly anyone has tried to come up with an argument for this claim. The reason is probably that the factivity of memory is thought to be evident. Yet hardly anything is evident. Why should we not regard the non-factive notion of memory (called “ostensible memory”) which does not imply truth to be more basic than the factive notion of memory?1 One argument for the factivity of “to remember” makes use of the undeniable fact that statements of the form “S remembers that p, but p is false” and “S remembers event E, but E never happened” have a paradoxical ring to them. Just as G.E. Moore’s famous dictum “It’s raining outside, but I don’t believe it,” though not literally contradictory, cannot be used to make coherent assertions, the statement “I remember that p, but p is false” is pragmatically incoherent. The problem with this argument for the factivity of remembering is that it is possible to explain the pragmatic incoherence of the statement “I remember that p, but p is false” while maintaining that memory does not imply truth. When I claim to remember that p, I am convinced that p is the case. This is what the first part of the statement expresses. Yet the second part of the statement denies that p is the case. Thus “I remember that p, but p is false” is incoherent not because one cannot remember that p without p being the case but because one cannot claim to remember that p while claiming that p is false. And since the conditions for claiming to remember are distinct from the conditions for remembering it does not follow that remembering that p implies the truth of p because claiming to remember that p implies the truth of p. A more convincing argument for the factivity of propositional memory rests on a grammatical point. A paradigm example of a factive verb is “to know” and the paradigm example of a non-factive verb is “to believe.” Both verbs can be combined with that-clauses: “S knows that Bert killed Oscar” and “S believes that Bert killed Oscar.” But there is a grammatical difference between both sentences. Zeno Vendler (1972, ch. 5; 1980) argues that the former but not the latter phrase can be transformed into sentences with wh-clauses, that is, clauses beginning with “who,” “whom,” “what,” “where,” “when” and “why.” “S knows that Bert killed Oscar” can be transformed into “S knows whom Bert killed,” “S knows who killed Oscar,” “S knows why Bert killed Oscar,” “S knows where Oscar was killed,” “S knows when Bert killed Oscar.” However, the non-factive verb “to believe” cannot take whclauses: “S believes whom Bert killed,” “S believes who killed Oscar,” “S believes why Bert killed Oscar” are ungrammatical sentences. 1 The Columbus-example is borrowed from Malcolm (1963, p. 189; cf. 1977, p. 194). Apart from Malcolm, the factive sense of “to remember” is explicitly embraced by Moore (1959, pp. 213– 214), Chisholm (1989, p. 68) and Shoemaker (1967, p. 266). Among the few who use “to remember” in the ostensible sense are Goldman (1986, p. 199) and Siebel (2000, p. 256).
8.1 Motivation of the Truth Condition
139
Given Vendler’s criterion for distinguishing factive from non-factive verbs, it can be proven that the cognitive verb “to remember” is factive. “To remember” is factive because it can take wh-clauses: “Oscar remembers that Bert killed Oscar” can be transformed into “Oscar remembers whom Bert killed,” “Oscar remembers when Bert killed Oscar,” “Oscar remembers where Oscar was killed,” “Oscar remembers why Bert killed Oscar,” etc. But, one might object, doesn’t the factivity thesis regarding “to remember” imply that memory claims are, by definition, infallible? If memory entails truth, it seems to follow that we can never claim to remember things which are false. But given that memory claims are frequently erroneous, it would follow that memory is not factive. This critique of the factivity of memory is mistaken. Though memory entails truth, we can be mistaken in thinking that we remember something. “What I remember is false” is shorthand for “What I thought I remember is false.” It is, in part, due to the entailment of truth that memory states are not transparent to the subject. In fact, all factive cognitive activities are opaque from a first-person perspective (see section 7.2). Knowledge, for example, is not transparent because the subject cannot discriminate knowledge from belief just on the basis of introspection. To identify one’s mental state as a state of knowing one needs to know that the proposition in question is true. The same applies to seeing. To know that I am seeing that p, I need information beyond what is available through introspection, for I need to know that it is in fact p I am claiming to see. If I mistake q for p, I don’t see p but only think that I see p. Memory isn’t transparent to the mind because one may not be able to tell, on the basis of reflection alone, whether the proposition one allegedly remembers is in fact true. And if one cannot introspectively know this, one cannot tell, by reflection alone, whether the activity one is engaged in qualifies as remembering or whether it is merely imagining. Since self-attributions of both memory and perception may be false and unremediable by introspection, we run a fourfold risk when we appeal to memory of perceptual experiences. Suppose S claims to remember having seen Fido on the sofa. First, S may correctly remember having thought he saw Fido on the sofa; but what he saw was not Fido but, say, a cat. In this case it is the past perception and not the memory which is to blame. Second, S may incorrectly remember having thought he saw Fido on the sofa; but what he really thought he saw, at that time, was a cat on the sofa and there was in fact a cat on the sofa. In this case the fault lies with the memory and not the past perception. Third, both kinds of mistakes can be combined: suppose S incorrectly remembers having seen Fido on the sofa; but what he thought he saw, at that time, was a cat and the perception was false for there was no cat but, say, a squirrel. In this case the fault lies both with the memory and the perception. Fourth, the perceptual mistake and the memory mistake may balance each other out. Suppose S claims to remember having seen Fido on the sofa; but what he thought he saw, at that time, was a cat, and the perception was false for it was in fact Fido he saw. In this case, though the memory claim is veridical, it doesn’t qualify as memory, for the causal condition of memory is violated. Let me add a proviso. Sensory experiences are identified and distinguished from one another in part by their qualia which are independent of the denotation of the
140
8 The Factivity Constraint
experiences. Smelling a certain food may differ markedly from tasting the food even if these experiences are of the same thing – the food. Moreover, sensory experiences differ from the propositional attitudes they can give rise to. In seeing or hearing Mozart’s symphony Jupiter being performed, one normally comes to believe that Mozart’s Jupiter is being performed. Compared to the experience, though, the belief that Mozart’s Jupiter is being performed is rather boring. After all, one can believe Mozart’s Jupiter is being played without hearing it being played, and vice versa. Now, the question arises whether the qualia-free belief that I heard Mozart’s Jupiter can count as a memory of my qualia-laden auditory experience of Mozart’s Jupiter. Since the topic of qualia is connected with a whole host of problems, none of which are specific to memory, I will confine myself to memories of qualia-free representational states. For the discussion of the factivity of memory it is important to differentiate between reproductive and meta-representational memory, for the truth conditions of both kinds of propositional memory are quite different. Consider an instance of metarepresentational memory: suppose I remember having believed that John F. Kennedy was assassinated. What I remember is true if I did believe that Kennedy was assassinated. Whether the content of the belief is true is irrelevant to the truth of the memory claim. While the truthmaker of meta-representational memory (of a past non-factive attitude such as belief) is our mind, the truthmaker of reproductive memory is our mind and the world. My claim to remember that Kennedy was assassinated is veridical only if I thought that Kennedy was assassinated and if Kennedy was indeed assassinated. The goal of the following two sections is to elaborate on the truth conditions for reproductive and meta-representational memory claims.
8.2
Truth in Reproductive Memory
The truth of reproductive memory claims is not fully determined by the truth of their contents. Apart from the truth of its content, the truth of a reproductive memory claim depends on whether one represented the content in the past. If I have only just learned that Kennedy was assassinated, my claim that he was assassinated is true but doesn’t qualify as a memory. In itself the claim is true but qua memory claim it is false. Hence, the truth value of reproductive memories is co-determined by one’s mental history, even though one’s mental history is not part of the content of reproductive memories. My claim to remember that Kennedy was assassinated may be false (i) because, contrary to my contention, Kennedy wasn’t assassinated, or (ii) because at no time in the past did it occur to me that Kennedy was assassinated, or (iii) because of a combination of (i) and (ii). Let’s begin with case (i). For my claim to remember that Kennedy was assassinated to be true, it must be the case that Kennedy was assassinated. Only if Kennedy was assassinated can I remember that he was assassinated. If it wasn’t Kennedy (but some look-alike) who was assassinated, I don’t remember that Kennedy was assassinated. And again if Kennedy wasn’t assassinated but died of, say, a stroke shortly
8.2 Truth in Reproductive Memory
141
before Lee Harvey Oswald’s bullet reached him, I don’t remember that Kennedy was assassinated. Let’s turn to case (ii). Remembering that Kennedy was assassinated presupposes not only that Kennedy was assassinated but also that, at some point in the past, one thought that he was assassinated. If I never thought that Kennedy was assassinated, my claim to remember that Kennedy was assassinated is false in that I don’t remember this fact. There are, of course, numerous reasons for why I may not have thought that Kennedy was assassinated. The most straightforward reason is that I had other things on my mind. Another reason is that my present “me” is not numerically identical with the person who used to believe that Kennedy was assassinated. Maybe I only quasi-remember the content of a previous thought but are not numerically identical with the person who hosted that thought (cf. section 2.3). But there is a general problem with (ii)-type errors of reproductive memory claims. In section 1.2, I suggested that reproductive remembering calls for a previous representation condition: For S’s representing at t2 that p to qualify as remembering, he must, among other things, have represented that p at t1. (ii)-Type errors occur when the previous representation condition is not met. Yet there are cases of reproductive memory – so-called negative memories – that seem to contradict the previous representation condition. Suppose that after you have left your house you remember that you didn’t lock the door. You walk back to your house, confirm that the door is unlocked, lock it, and leave again. Given the previous representation condition, for you to remember that you didn’t lock the door, you must have been aware of the fact that you didn’t lock the door, when you left the house the first time. But that’s absurd, one might argue, for if you were aware that you didn’t lock the door, why didn’t you lock it right away? Now one could try to defend the previous representation condition by arguing that when you initially left the house you were subconsciously aware of the fact that you didn’t lock the door. Yet I don’t think this strategy is convincing. Instead, I think negative memories are instances of inferential remembering. It is not quite right to say that you remember that you didn’t lock the door. What you do remember is that you closed the door behind you and walked away. It is only by means of an inference that you realize that you didn’t lock the door after closing it. Since I focus on non-inferential reproductive memory, I hold on to the previous representation condition. After having identified two distinct components of the truth-condition of reproductive memory claims – the mind and the world – it needs to be emphasized that these two components do not work independently of each other but that they are interrelated. For a reproductive memory claim that p to be veridical, it is not enough that the present thought that p be true simpliciter; it has to be true because the past thought that p was true. If my past thought-token that p had not been true, my present reproduction of this thought-token wouldn’t be true either and therefore couldn’t qualify as memory. And analogously, if my present thought-token is true and if it is a reproduction of a previous thought-token, the previous thought-token must have been true as well. Thus, neither the reproduction of a false thought-token nor the defective reproduction of a true thought-token gives rise to reproductive memory.
142
8 The Factivity Constraint
I have claimed that in order to remember that Kennedy was assassinated I must, in the past, have thought that Kennedy was assassinated. Reproductive memory requires the reproduction of thought-tokens. But in what respect and to what degree must past and present thought-tokens match for the truth-condition of reproductive memory to be fulfilled? May, for example, my present thought that Kennedy was murdered (or simply that Kennedy died) qualify as a reproduction of my previous thought that Kennedy was assassinated? And what about my thought that Bob Kennedy’s brother was assassinated? May this thought qualify as a memory of my previous belief that John F. Kennedy was assassinated? According to the standard view, remembering requires that the contents of the past and the present thought-tokens are of the same type. Since, on this view, memory is conceived of as a photocopier of propositional attitudes I label this view the xerox model of memory. In sections 8.5 and 8.6 I will argue that the xerox model is mistaken, for it reduces memory to a purely passive device of registering, storing, and regulating information. Sometimes memory is an active device for reconstructing information. And as a result of such information processing, the content of a memory state may differ from the content of the original representation it causally derives from. Depending on the context, the contents of our memories may be similar to, but not identical with, the contents of the thoughts they causally derive from. In chapter 9, I will define criteria for determining whether content-tokens are sufficiently similar to meet the truth condition of memory.
8.3
Truth in Meta-representational Memory
Mental states represent the union of an attitude (character or force) and a content. The propositional content of “I believe that p” is expressed by “p.” The attitude that is taken towards this content is that of believing. Other examples of attitudes are affirming, denying, wanting, fearing, etc. Frequently the attitudinal component of a mental state is not made explicit. For example, the normal expression of a belief doesn’t mention the fact that it is a belief. When I believe that Kennedy was assassinated, ordinarily I simply say “Kennedy was assassinated” rather than “I believe that Kennedy was assassinated” (or “I know that Kennedy was assassinated”). Likewise when I remember having thought that Kennedy was assassinated, frequently I don’t say “I remember having thought that Kennedy was assassinated” but simply “I thought that Kennedy was assassinated.” For a meta-representational memory claim to be true its content must be true. The content of meta-representational memory states has the form “I believed that p” and thus consists of three elements: the subject (“I”), the attitudinal component (“believed”), and the embedded first-order content (“p”). Only if each one of these elements is correctly identified is the meta-representational memory claim as a whole veridical. Thus there are three aspects to the factivity constraint on metarepresentational memory.
8.3 Truth in Meta-representational Memory
143
Let’s begin with errors due to mistaken self-identification. Assuming a bodily criterion of personal identity, S’s meta-representational memory claim to the effect that S thought that p may be false because the person-stage at t1 which thought that p and the person-stage at t2 which thinks that he thought that p don’t belong to the numerically same person. Since throughout this study I have avoided dealing with the relation between personal identity and memory, I will abstain from errors due to mistaken self-identification. Another reason why a memory claim of the form “I thought that p” may be false is because it misrepresents the attitudinal component of one’s past mental state. Suppose I claim to remember having believed that Kennedy was assassinated. If instead of having believed that Kennedy was assassinated, I supposed or considered that Kennedy was assassinated, the memory claim is incorrect and may not count as memory. Common sources of error concerning the identification of one’s past attitude types are inattention, self-deception, and forgetting. The attitude-identification which is part of meta-representational remembering is particularly error-prone when the past thought was framed in an attitude expressed by a factive verb. Wolfgang von Leyden (1961, p. 60) provides the following instructive example2: S sees a panic in Trafalgar Square and thinks erroneously that it is a setting for a film. He takes himself to be seeing a setting for a film. Later, by memory, he reports having seen a stage setting for a film at that time. Since there was no film setting, and since no one can see what is not the case, S cannot have seen a film setting and, consequently, cannot remember having seen a film setting. Strictly speaking, his meta-representational memory claim “I saw a setting for a film” is false. But it is not his memory but the original perception which is at fault. It is, as Leyden (1961, p. 62) elegantly phrases, “the inheritance of a mistake, not a mistake of inheritance.” Now suppose we inform S that his memory claim is false. If he wanted to account for the fact that his memory was working all right, he would have to reformulate his memory claim. Instead of claiming to have seen a setting for a film, he would have to assert something like this: “It seemed to me as if I saw a setting for a film” or “I thought that I saw a setting for a film.” The crucial difference between these memory claims, on the one hand, and the original one, on the other, is that “seeming” and “thinking” are non-factive. Rephrased in this way, S’s memory claim comes out true and – provided the other three memory conditions are fulfilled – can constitute an instance of meta-representational remembering. A meta-representational memory claim may also be deceptive because the embedded content doesn’t match the content of the original representation. My claim to remember having believed that Kennedy was assassinated is false if what
2 Anscombe (1981a, pp. 105–106) develops a similar example. Suppose someone is mistaking a wax dummy for a man. Later on he claims to remember having seen a man. Again, this way of expressing his alleged personal factual memory is unfortunate because it makes the memory claim come out false. Instead of saying “I remember having seen a man” S should express his memory claim by saying “I remember having thought that I saw a man” or “I remember it seeming to me that I saw a man.”
144
8 The Factivity Constraint
I in fact believed was, say, that a Kennedy-look-alike was assassinated while Kennedy collected his life insurance and bought a condo in downtown Havana. The most common reason for misrepresentations of embedded contents is lack of attention, self-deception, or forgetting. In the remainder of this chapter I will focus on misidentifications of embedded propositional contents and abstract from other sources of errors in meta-representational memory claims. The aim is to say what constitutes a misrepresentation of a past thought content.
8.4
The Xerox Model of Memory
We saw that regardless of whether I remember that p or whether I remember having thought that p, it has to be the case that, at some point in the past, I thought that p. In both cases the (embedded) memory content must have been entertained before. The truth of meta-representational and of reproductive memory claims depends, in part, on the match between the (embedded) memory content and some past content. Since this match can be more or less perfect, the question arises to what extent past and present thought contents have to be alike for the latter to meet the truth condition of memory. Many philosophers hold that our memory is purely a passive device for registering, storing, and reproducing information. The characteristic feature of this xerox model is that the recalled thought is required to be type-identical with the original thought. On the xerox model, the content of a reproductive memory claim (and the embedded content of a meta-representational memory content, respectively) must be a token of the same type of content as the original thoughts from which it causally derives. The xerox model is incomplete; it captures only one aspect of what our memory is supposed to do. Our memory is not only a passive device for reproducing information but also an active device for reconstructing information. Of course, when the reconstructive nature of memory gains the upper hand the distinction between memory and confabulation becomes blurred. But sometimes remembering allows for the moderate transformation of the informational content. There are cases where the content of a memory claim is veridical even though it is not an exact duplicate of the relevant past content but contains some reconstructive elements. But before we can dismiss the xerox model we need to examine it. The xerox model comes in two flavors: either recall is conceived of as the revitalization of an existing trace bringing it once more into active thought – the persistence model – or recall is construed as the reproduction of the original experience or thought – the reproduction model.3 The persistence model has gained more widespread support than the reproduction model. Typically, the persistence
3
What I call the “xerox model of memory” has been given various names: “the copy theory of memory” (Brewer 1988, p. 26), “the reappearance hypothesis” (Neisser 1967, pp. 280–284), and “passivism” (Casey 2000, p. 15).
8.4 The Xerox Model of Memory
145
model is illustrated by comparing our memory to a storehouse or repository. Two features are characteristic of all versions of the xerox model. First, memory is exclusively portrayed as a passive process or device for registering, storing, and regurgitating immutable items of information. Second, discrete, elementary stimuli are used as memory input, allowing for the quantification of memory. Given the quantitative approach, measures of memory effectiveness are based on the number of recovered elements and forgetting is conceived of as information loss. One would expect that cognitive psychologists would have always been aware of the fact that it is part of the job of our memory to reconstruct and process the stored information. But this is not the case. In fact it took until the 1930s, when Bartlett published his seminal book Remembering, for psychologists to acknowledge the reconstructive nature of memory. The xerox model is a tacit presumption of associationism, behaviorism, gestalt psychology, and Freudian psychoanalysis alike. Legions of cognitive psychologists have used memory metaphors which rest on the xerox model. Donald Broadbent (1971), for example, compares the storage of memories to storing books in a library, Elizabeth Loftus (1977) to storing information in a dictionary, and Michael Posner (1972) to storing sounds on a tape recorder. In the field of philosophy, the xerox model of memory is even more deeply entrenched than in cognitive psychology. Apart from a few exceptions, the xerox model is assumed by virtually everyone, though explicit endorsement is rare. An early explicit advocate of the persistence model was Augustine, who called the memory the “belly of the mind” and compared it to “a large and boundless inner hall,” a “storehouse,” and a “vast cave” within which “the images of things perceived” are laid away, to be “brought forth when there is need for them.”4 Moreover, the British Empiricists assumed that our memory retains “ideas” or “conceptions” which are nothing but slightly faded copies of sensory experiences. Hume, for example, claimed that memory stores ideas in their “original form.” Generally, ideas in memory are accurate, vivid, and can be automatically called up. Any modification of ideas in memory can be attributed to “some defect or imperfection in that faculty” (1739, p. 9). Likewise, John Locke called our memory a “store-house” in which ideas are “as it were laid aside out of sight,” and he spoke of the “repository of the memory” (1694, pp. 149–150, 152–153).
4
400, p. 191. Augustine wasn’t entirely happy with the persistence model of memory, for he realized that it yields a counterintuitive result: in order to know that I have forgotten something, I need to remember it. Suppose that a particular event, C, has been forgotten. If I want to dig up the memory of C I have to browse various other memories – A, B, D, E, etc. – until I find C. The problem is that this procedure seems to require that I have not forgotten C. For, first, if I had completely forgotten C, I would not even be able to attempt to remember C. “We do not say we have found the thing which was lost unless we recognize it, and we cannot recognize it if we do not remember it” (ibid.). Second, if I had truly forgotten C, I would not be able to recognize C, even if I do recall it: “What when the memory itself loses something? This happens when we forget and attempt to recall. The only place to search is in the memory itself. If something other than what we want is offered us, we reject it until the thing we are looking for turns up. And when it comes, we say “That is it.” We would not say this unless we recognized it, and would not recognize it unless we remembered. It seems certain, then, that we had forgotten” (ibid., p. 195).
146
8 The Factivity Constraint
One of the main goals of Locke’s Essay is the rejection of the doctrine of innate ideas. Yet there is a tension between the rejection of innate ideas, on the one hand, and the characterization of memory as a storehouse of ideas, on the other hand. The problem is that the storehouse model of memory presupposes the existence of unconscious ideas; and from countenancing unconscious ideas there is only a small step to acknowledging innate ideas. This may explain why, when Locke prepared the second edition of his Essay he gave up the persistence model of memory and adopted the reproduction model instead: But our Ideas being nothing, but actual Perceptions in the Mind, which cease to be any thing, when there is no perception of them, this laying up of our Ideas in the Repository of Memory, signifies no more but this, that the Mind has a Power, in many cases, to revive Perceptions, which it has once had …. And in this Sense it is, that our Ideas are said to be in our Memories, when indeed, they are actually no where, but only there is an ability in the Mind, when it will, to revive them again; and as it were paint them anew on itself (1694, p. 150).
The claim is that ideas of perception merely get “revived” as qualitatively identical memory ideas, and that in the interim, when they are unconscious, they exist “nowhere.”5
8.5
Adaptive Variation
Sometimes our memory is supposed to work like a tape recorder giving a verbatim rerun of some stored information. Examples are the learning of poems and songs, dramatic acting, and particularly vivid memories of important events. Yet precise repetition of movements, words, or sequences of thought is not the only function of memory. The error of the xerox model is to reduce the job of memory to literal repetition. The relation of our memories to the past can be much more complicated than simple reproduction. Sometimes our memory processes and is supposed to process the encoded information. As a result of such information processing the content of a memory state differs from the content of the thought it causally derives from. When the stored information is transformed radically, the resulting state cannot classify as remembering for it violates the truth condition. Minor information 5 Locke’s reproduction model elicited a famous objection from Reid in his Essay on the Intellectual Powers of Man: “But it seems to me as difficult to revive things which have ceased to be anything, as it is to lay them up in a repository, or to bring them out of it. When a thing is once annihilated, the same thing cannot be again produced, though another thing similar to it may. Mr Locke acknowledges in another place that the same thing cannot have two beginnings of existence; and that things that have different beginnings are not the same, but diverse. From this it follows that an ability to revive our ideas or perceptions, after they have ceased to be, can signify no more but an ability to create new ideas or perceptions similar to those we had before” (1785, pp. 284–285). I am puzzled by Reid’s critique. Of course, if unconscious ideas are impossible, the recalled ideas and the original ideas cannot be numerically identical but only qualitatively identical. But since there is no indication that Locke confused qualitative for numerical identity, Reid’s objection seems to come to nothing.
8.5 Adaptive Variation
147
processing, however, is compatible with memory. An example of moderate information processing is when the thought that Kennedy was assassinated in Texas is remembered as the thought that Kennedy died in Texas. Of the numerous modes by which our memory may edit the encoded information, I will discuss three: cognitive dynamics, condensation, and schematic processing. These kinds of information processes modify the encoded information to different degrees. Memory condensation is a more invasive form of information processing than cognitive dynamics; and schematic processing is more invasive than condensation.
Cognitive Dynamics It is a precondition of rationality that information acquired at one time should be available to the subject later and that he should be capable of continued believing. In the case of beliefs involving demonstratives, one needs to change the judgements in order to continue to believe the same thing. Suppose that S points to a man at a party, saying “That fellow is Bert.” I believe him, saying the same sentence to myself. Later on, S is near the bar, the man is next to me, and I say “So, you are Bert.” A week later I see him in the street and think to myself, “He is Bert.” I came to believe something and continued to believe it. Thus, when I keep track of someone or something as one (or both) of us move or as time passes by, I must vary the judgments so as still to express the same belief as before. David Kaplan (1989) calls the ability to vary one’s judgements so as to manifest a single persisting belief “cognitive dynamics.” When recalling propositional contents entailing demonstratives, the content of the memory state must differ from the content of the original state for it to express the same belief. Suppose, for example, you arrange to meet S next Tuesday. Next Tuesday, you remember having arranged to meet S today. Or suppose that, while visiting London, you come to believe that this city is pretty. Back in your home town you remember having believed that London is pretty. Using Kaplan’s character/content distinction we may say that in order for two statements uttered in different circumstances to express the same content they may have to employ different characters.6 Moreover, many contents of memory states are not exact replicas of the contents of the original thoughts for the simple reason that the former are formulated in the past tense while the latter are framed in the present tense. Suppose you
6 It is not just in the case of propositions involving genuine demonstratives that the recalled thought must differ from the original thought for it to express the same belief. Suppose that before the German reunification in 1989, you came to believe that Karl-Marx-Stadt is beautiful. After the reunification the city’s name changed (back) to Chemnitz. After the reunification you may recall your pre-1989 belief that Karl-Marx-Stadt is beautiful by thinking that you believed that Chemnitz is beautiful. Karl-Marx-Stadt and Chemnitz have an indexical component, for they refer to the same city at different times. Remembering a Karl-Marx-Stadt-thought as a Chemnitz-thought is an instance of inferential memory. And as was explained in section 8.2, this study concentrates on non-inferential memory.
148
8 The Factivity Constraint
watch S’s behavior at a party and come to believe that he is drunk. What you remember the next day is not that S is drunk but that he was drunk. (It might be that S is still drunk the next morning, yet you don’t remember this but learn it for the first time upon watching or smelling S during breakfast.) With the passing of time, a propositional content changes like this: “S is F now” (original thought); “S was F a moment ago” (memory shortly after the original thought); “S was F a while back” (memory long after the original thought).7 If memory were a purely passive repository – as the xerox model maintains – it would not be allowed to adapt the tenses of verbs and the demonstratives included in the original thought contents; an absurd conclusion.
Condensation Many of our memories do not refer to particular events but to periods of our lives, ways we used to feel or act, or things we used to believe. For example, I remember that when I first moved to San Francisco I used to work out a lot, that for a while I was meeting a lot of new people, that there was a stretch of time when I was always traveling, etc. It is not uncommon to have these kinds of condensed memories without being able to remember in detail any particular instance of what is remembered, or remembering some instances distinctly, some vaguely, and some not at all.8 Condensed memories present a difficulty for the xerox model of memory which reduces remembering to the passive regurgitation of discrete items of encoded information. According to the xerox model, memory is a simple one-toone connection between two discrete diachronic moments of consciousness. In the case of condensed memories, however, it is not only a one-to-many relationship but also the content of a summarized memory state may be different from any of the contents it comprises. To drive this point home, consider the following example. I remember having disliked eating vegetables as a child. Though it is true that I used to dislike vegetables, at no time in my childhood did I (dispositionally) believe that I didn’t like vegetables. Rather than entertaining this general idea, what I thought was that I don’t like Brussel sprouts, that mangel is too bitter, carrots too sweet, that spinach has a disgusting texture, etc. The content of my comprehensive meta-representational memory, though no doubt correct, has no single counterpart in the past and therefore cannot count as a replica of a previous thought content of mine. It stems from the amalgamation of a whole host of past thought contents.
7
The concept of content employed here is not Kaplan’s. Cf. Schechtman (1994, pp. 7–11) and (1996, pp. 124–125). Conway (2001) distinguishes between two types of condensed memories: knowledge of lifetime periods, such as when I was at school X, worked for company, lived in Z, etc. and general event knowledge, such as holiday in A, eating out at B, working project C, etc. 8
8.5 Adaptive Variation
149
There are numerous experiments demonstrating the prevalence of condensed memories. Laurence Barsalou, for example, stopped college students at the beginning of the academic year and asked them to list all the events they could remember from their summer vacations. Barsalou reports: We originally believed that the primary content of autobiographical memory was supposed to be memories of specific events. Yet subjects, when asked to describe “events” from their summer vacation, spent only 21% of their time recalling specific events… The retrieval of summarized and extended events, along with other kinds of information, appears to play an important role in accessing information about periods of one’s life (Barsalou 1988, p. 201).
On the basis of this study Barsalou concludes that the ratio of summarized eventmemories to specific event-memories comprises around 60–40%.
Schematic Processing This form of information processing is even more invasive than condensation. Frequently our memory edits the encoded information in accordance with schemas, i.e., cultural stereotypes used to structure information (cf. Brewer and Nakamura 1984). A schema is an organized bundle of default features, so that, for example, when one thinks of a bird, one tends to think of something that looks like a robin (the prototype) rather than an ostrich. The schema of a room implies that it has walls, a door, a ceiling and window. When recall is filtered through schemata, we tend to omit things that are not part of or incompatible with our schemata and we tend to remember those things better which match with our preferences or biases. In the case of schematically processed information, what is reported about the past sometimes has very little in common with what really happened. Frequently schematic processing yields ostensible memories. A classic experiment illustrating schematic processing is due to Frederick Bartlett. Bartlett’s experimental method was that of a serial reproduction, in which a subject would attempt to recall the same material on several occasions. As stimuli he used passages of prose, pictures, and stories, the most famous of which being a folktale from the Canadian-Indian Kwakiutl people entitled The War of the Ghosts. Bartlett was interested in how subjects would remember a story that fit in poorly with their cultural background. The test subjects were Cambridge students. This is the original version of the Indian folktale which Bartlett presented to the students: The War of the Ghosts. One night two young men from Egulac went to the river to hunt seals, and while they were there it became foggy and calm. Then they heard war-cries, and they thought: “Maybe this is a war-party.” They escaped to the shore, and hid behind a log. Now canoes came up, and they heard the noise of paddles, and saw one canoe coming up to them. There were five men in the canoe, and they said: “What do you think? We wish to take you along. We are going up the river to make war on the people.” One of the young men said: “I have no arrows.” “Arrows are in the canoe,” they said. “I will not go along. I might be killed. My relatives do not know where I have gone. But you,” he said, turning to the other, “may go with them.” So one of the young men went, but the other returned home. And the warriors went on up the river to a town on the other side of Kalama. The people came down to the water, and they began to fight, and many were killed. But presently the young man heard one of the warriors say: “Quick, let us go home: that Indian has been hit.” Now he thought: “Oh, they are
150
8 The Factivity Constraint
ghosts.” He did not feel sick, but they said he had been shot. So the canoes went back to Egulac, and the young man went ashore to his house, and made a fire. And he told everybody and said: “Behold I accompanied the ghosts, and we went to fight. Many of our fellows were killed, and many of those who attacked us were killed. They said I was hit, and I did not feel sick.” He told it all, and then he became quiet. When the sun rose he fell down. Something black came out of his mouth. His face became contorted. The people jumped up and cried. He was dead (1932, p. 65).
He had his subjects recall the story after various delays, from immediately after study to years later. This story was produced by one subject on the first recall after an interval of about 15 minutes: War Ghost Story. Two young men from Egulac went out to hunt seals. They thought they heard war-cries, and little later they heard the noise of the paddling of canoes. One of these canoes, in which there were five natives, came forward towards them. One of the natives shouted out: “Come with us: we are going to make war on some natives up the river.” The two young men answered: “We have no arrows.” “There are arrows in our canoes,” came the reply. One of the young men then said: “My folk will not know where I have gone;” but, turning to the other, he said: “But you could go.” So the one returned whilst the other joined the natives. The party went up the river as far as a town opposite Kalama, where they got on land. The natives of that part came down to the river to meet them. There was some severe fighting, and many on both sides were slain. Then one of the natives that had made the expedition up the river shouted: “Let us return: The Indian has fallen.” Then they endeavored to persuade the young man to return, telling him that he was sick, but he did not feel as if he were. Then he thought he saw ghosts all round him. When they returned, the young man told all his friends of what had happened. He described how many had been slain on both sides. It was nearly dawn when the young man became very ill; and at sunrise a black substance rushed out of his mouth, and the natives said one to another: “He is dead.” (1932, p. 70).
After an interval of 20 hours another subject gave the following first reproduction of the story: The War of the Ghosts. Two men from Edulac went fishing. While thus occupied by the river they heard a noise in the distance. “It sounds like a cry,” said one, and presently there appeared some men in canoes who invited them to join the party on their adventure. One of the young men refused to go, on the ground of family ties, but the other offered to go. “But there are no arrows,” he said. “The arrows are in the boat,” was the reply. He thereupon took his place, while his friend returned home. The party paddled up the river to Kaloma, and began to land on the banks of the river. The enemy came rushing upon them, and some sharp fighting ensued. Presently some one was injured, and the cry was raised that the enemy were ghosts. The party returned down the stream, and the young man arrived home feeling none the worse for his experience. The next morning at dawn he endeavored to recount his adventures. While he was talking something black issued from his mouth. Suddenly he uttered a cry and fell down. His friends gathered round him. But he was dead (1932, p. 66).
In a second recall after 4 months, another subject produced the following story: I have no idea of the title. There were two men in a boat, sailing towards an island. When they approached the island, some natives came rushing towards them, and informed them that there was fighting going on on the island, and invited them to join. One said to the other: “You had better go. I cannot very well, because I have relatives expecting me, and they will not know what has become of me. But you have no one to expect you.” So one accompanied the natives, but the other returned. Here there is a part that I can’t remember.
8.5 Adaptive Variation
151
What I don’t know is how the man got to the fight. However, anyhow the man was in the midst of the fighting, and was wounded. The natives endeavored to persuade the man to return, but he assured them that he had not been wounded. I have an idea that his fighting won the admiration of the natives. The wounded man ultimately fell unconscious. He was taken from the fighting by the natives. Then, I think it is, the natives describe what happened, and they seem to have imagined seeing a ghost coming out of his mouth. Really it was a kind of materialization of his breath. I know this phrase was not in the story, but that is the idea I have. Ultimately the man died at dawn the next day (1932, pp. 70–171).
Yet another subject recalled after 6 months the following story: (No title was given). Four men came down to the water. They were told to get into a boat and to take arms with them. They inquired “What arms?” and were answered “Arms for battle.” When they came to the battle-field they heard a great noise and shouting, and a voice said: “The black man is dead.” And he was brought to the place where they were, and laid on the ground. And he foamed at the mouth (1932, pp. 71–72).
It is hardly surprising that, as time passes by, subjects omit more and more of the folktale, change many of the facts, and import new information. Bartlett’s important observation is that these inaccuracies are systematic and that there are a number of consistent tendencies: the remembered story is always shorter, more coherent, and tends to fit in more closely with the subject’s own viewpoint and cultural stereotypes. For example, “something black came from his mouth” in the original story becomes “he foamed at the mouth” or, in some stories, “he vomited.” “Hunting seals” is changed to “fishing” and “canoe” changed to “boat.” The puzzling (because supernatural) aspects of the story – such as the connection between the Indian’s injury and the termination of the battle – are often omitted. Furthermore, some subjects introduced inferences in an attempt to explain incongruous features of the story. For example, when one of the men declines to go off to battle he states “ ‘I will not go along. I might be killed. My relatives do not know where I have gone. But you’, he said, turning to the other, ‘may go with them’.” One subject recalled this last passage as “But you have no one to expect you” and another subject recalled it as “You have no parents” (1932, pp. 71, 120). In each case the subject has provided an explanation not given in the original text for why it would be all right for the second man to go off to battle. Frequently, the rememberer selects certain features and uses these to anchor the whole story. In the case of The War of the Ghosts the death scene often serves this role. Remembering, Bartlett concludes, is essentially active, with the subject showing a constant “effort after meaning.” Instead of being only reproductive, remembering is also a reconstructive process. Sometimes “remembering is not the re-excitation of innumerable fixed, lifeless and fragmentary traces” but “an imaginative reconstruction, or construction” (1932, p. 213).9
9 For a review of literature on recent replications of Bartlett’s experiments see Brewer and Nakamura (1984) and Roediger et al. (2000). On memory reconstruction in general see Ross and Buehler (1994).
152
8.6
8 The Factivity Constraint
The Function of Adaptive Variation
We have examined three kinds of information processes and have established the prevalence of adaptive variation of information stored in (long-term) memory. But none of this proves that the xerox model of memory is incomplete or mistaken. For a proponent of the xerox model could concede that encoded information is frequently condensed and edited, but claim that these information processes give rise to ostensible rather than genuine memories. Assuming that memory has the function to reproduce information, when the retrieved information differs from what has been encoded only in the slightest, the memory malfunctions. To neutralize this rejoinder and to demonstrate that the xerox model is indeed incomplete, it has to be shown that moderate forms of information processing such as cognitive dynamics and condensation are part of the function of remembering – that they are compatible with the truth condition of memory. As far as I can see, there are two arguments for the thesis that remembering involves the condensing of information, an argument from narrative self-definition and an argument from clutter avoidance.
Narrative Self-Definition Each one of us possesses a more or less coherent story of his life. These lifestories are derived to a large extent from one’s memories. But to form a coherent autobiographical story, memories have to be condensed (and filtered through schemata). For life-stories stress the relevant aspects of one’s past experiences, depict behavioral patterns, and identify representative judgements and characteristic preferences. In the process of editing, anomalous events may be omitted, typical ones underlined, and other changes undertaken to increase the smoothness and intelligibility of the narrative. The narrative is expressive of the overall contours of one’s character and behavior but it doesn’t need to be faithful to each and every specific experiences and events. Marya Schechtman declares: Like a scientist who creates a continuous graph by drawing a line close to, but not necessarily in contact with, all the data points, our autobiographical memories draw a smooth storyline among the various experiences we have had – a storyline which is constrained by the bulk of those experiences, but which need not contain them exactly, and which gives our lives a narrative unity (1994, p. 13).
The thesis that condensation (and schematic processing) of stored information is a necessary condition for our narrative self-constitution fits nicely with a conclusion that Craig Barclay and Peggy DeCooke (1988) have drawn on the basis of their empirical evidence. Barclay and DeCooke asked subjects to keep a journal recording events and experiences. Later these subjects had to distinguish their journal entries from entries added by the experimenter. Subjects were frequently wrong when they were presented with fictitious journal entries which were written in a style unlike theirs but reported experiences the subjects actually had had. Barclay and DeCooke conclude: Autobiographical recollections are not necessarily accurate, nor should they be; they are, however, mostly congruent with one’s self-knowledge, life themes, or sense of self …. As
8.6 The Function of Adaptive Variation
153
with allegory, autobiographical memory often is a constructive and reconstructive process used to condense everyday memories of events and activities, extracting those features that embrace and maintain meaning in one’s self-knowledge system …. In turn, seemingly unconnected episodic recollections become allegorical in that particular events can be remembered and used as instances of generalized life experiences to convey one’s sense of self to an audience (1988, p. 92).
Clutter Avoidance Every day we receive an overwhelming amount of information. Since there is a limit to what one can remember and a limit to what one can retrieve from memory, it is vital to condense and compress the incoming information. Without constant condensation and summing up, we would clutter our minds with a lot of redundant pieces of information. Gilbert Harman develops the following clutter avoidance principle: “It is important to save room for important things and not to clutter one’s mind with a lot of unimportant matters” (1986, pp. 41–42). In chapter 3, it was argued that the stipulation of memory traces is indispensable for understanding the causal process underlying remembering. Given that information processing is part of the function of memory, the format of information stored in traces cannot be that of finished mental events, like images, sentences, or propositional attitudes, but has to be that of fragments of those complex entities. Depending on whether one endorses the language-of-thought thesis or connectionism, these fragments are either syntactic structures of a language of thought or connections between neural units. Before terminating this chapter I need to set my position apart from constructivism. There is a difference between saying, as I do, that memory need not always amount to the exact reproduction of previously recorded information and saying, as constructivists do, that, as a matter of principle, memory constructs (rather than reproduces) previously encountered information. Two of the earliest exponents of an experimentally-based psychology, Wilhelm Wundt and William James, seem to have played the reconstructive nature of remembering against its reproductive nature. writes: It is found that the reproduction of ideas in the strict sense of a renewal in its unchanged form of an earlier idea, never takes place at all, but that what really does happen in an act of memory is the rise of a new idea in consciousness, always differing from the earlier idea to which it is referred, and deriving its elements as a rule from various preceding ideas (1896, p. 225).
And James declares that “a permanently existing ‘idea’ or ‘Vorstellung’ which makes its appearance before the footlights of consciousness at periodical intervals, is as mythological an entity as the Jack of Spades” (1890, I, p. 236). Among today’s cognitive psychologists, Barclay, William Brewer, and Ulric Neisser are the leading proponents of constructivism. Like Wundt and James, Barclay highlights the constructive and downplays the reproductive nature of memory. He states that “memories for most everyday life events are … transformed, distorted, or forgotten” (1986, p. 89). Brewer, on the other hand, is more careful than Barclay when he proposes that recent personal memories retain a relatively large amount of specific information from the original phenomenal experience (e.g., location, point of view) but that with time, or under
154
8 The Factivity Constraint
strong schema-based processes, the original experience can be reconstructed to produce a new non-veridical personal memory that retains most of the phenomenal characteristics of other personal memories (1986, p. 44).
Neisser likens the process of remembering to a paleontologist reconstructing a dinosaur. Just as a paleontologist constructs a complete model of a dinosaur from a few pieces of bone and some knowledge of how they should be put together, we reconstruct the memory of a complex event from a few details and the general theme. Neisser argues that our original perception of the world is a construction and that what we remember is the constructive act of the initial perception and comprehension: The present proposal is, therefore, that we store traces of earlier cognitive acts, not of the products of those acts. The traces are not simply “revived” or “reactivated” in recall; instead, the stored fragments are used as information to support a new construction. It is as if the bone fragments used by the paleontologist did not appear in the model he builds at all – as indeed they need not, if it is to represent a fully fleshed-out, skin-covered dinosaur. The bones can be thought of, somewhat loosely, as remnants of the structure which created and supported the original dinosaur, and thus as sources of information about how to reconstruct it (1967, pp. 285–286).
Both austere constructivism and the xerox model strike me as overreactions. The problem with the xerox model is that there are numerous instances where our memory does more than regurgitate the encoded information, namely process the information. But sometimes our memory does have the function of giving a verbatim re-run of stored information. Constructivists are right on the mark when they claim that the retrieved information does not have to be identical with the encoded information and that information processing is an essential aspect of remembering. However, constructivists seem to forget that the factivity constraint on memory demands that the memory output be true and sufficiently similar to the input.
Chapter 9
Diachronic Content Similarity
Memory implies truth, but it does not imply that the memory content is an exact duplicate of the past thought content. Sometimes memory allows for moderate transformations of the informational content. In the previous chapter it was shown that it depends on the context whether memory requires literal reproduction of contents or whether there is room for minor variations of the previously learned information. Sometimes the content (or the that-clause embedded in the content) of a memory state has to be type-identical with the thought content it causally derives from, sometimes it suffices if the two content-tokens are of similar types. The aim of this chapter is to investigate the extent to which two content tokens may differ from one another and one of them still count as a memory of the other. Given that the context is such that remembering does not require the regurgitation of contents, what is the permissible range of aberration between the original content and the memory content? When are two diachronic content-tokens sufficiently similar for the later to qualify as a memory of the earlier? What is the semantic tolerance of remembering? What is the margin of error of the content-identification of propositional remembering? I will set forth principles which specify the degree to which two thought-content-tokens may differ from one another and the later one still be able to be classified as a memory of the earlier one. Section 9.1 discusses various context-dependencies of “similarity.” In response to recent attacks on the coherence of the notion of similarity, I argue that, at least with respect to the analysis of memory, talk of semantic similarity presents a useful abbreviation. In section 9.2, I develop some principles of semantic similarity. The idea behind these principles is that, in the process of remembering, the informational content stored in traces may stay the same or decrease (to a certain degree); but it may not increase. In section 9.3, I back up this idea by pointing out that we frequently correct our memory claims by reducing their informational content. And just as type-identity of contents is not a necessary condition for remembering, it might be argued that sometimes the attitude-type mentioned in the content of a meta-representational memory does not have to be identical with, but only similar to, the attitude-type of the original state. This is the topic of section 9.4.
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
155
156
9.1
9 Diachronic Content Similarity
The Relativity of Similarity
What does it mean to say of two objects that they are similar? Intuitively, we say that objects are similar because they have most or many properties in common. But this intuition doesn’t take us very far, because all objects have infinite sets of properties in common. A peach and a bicycle, for example, both share the properties of weighing less than 50 kg (and less than 51 kg, etc.). This seems to imply that all objects are similar to all others! And, by the same token, all objects will have infinite sets of properties that are not in common. A peach weighs less than 1 kg, while a bicycle weighs more than 1 kg (and similarly for 1.1 kg and 1.11 kg, etc.). Perhaps, then, all objects are dissimilar to all others! “Similar” is a meaningful notion only as “similar in a certain respect.” It is only within a context that claims concerning the similarity of things are meaningful. Independently of a standard, everything is similar or dissimilar to everything else. The analysis of criteria of similarity has to take place against the background of some context. Although similarity superficially appears to be a two-place relation, it is really a three-place relation. A is similar to B with respect to C, where C stands for a perspective or point of view. So-and-so is like so-and-so for he also has three pounds excess weight or because both of them are married and have two children. Apples are like pears because they both belong to the species of pomes. Brazil is like Australia in that they are both located in the southern hemisphere. Depending on what one takes as the measure of comparison, different things turn out to be similar to one another. Even when the frame of reference is fixed, similarity is a graded notion, a matter of degree. We can say things like, regarding the degree of latitude, New York is more similar to Rome than to Munich. The truth of similarity statements is not only relative to the frame of reference but also to the threshold value of shared properties. Establishing the threshold value of shared properties is contextually relative. Whether or not New York’s latitude counts as similar to Rome’s latitude depends, among other things, on which pairs of latitudes one uses as a contrast. In comparison to Minneapolis/Saint Paul, New York/Rome are not on similar latitudes, yet in comparison to Buenos Aires-Toronto they are. It is uncontroversial that similarity statements are relative to the frame of reference (qualitative relativity) and to the threshold value of shared properties (quantitative relativity). It is controversial, however, whether, despite its twofold relativity, the notion of similarity possesses explanatory value. W.V.O. Quine and Nelson Goodman deny this. Quine’s and Goodman’s main objection to the notion of similarity concerns the specification of the frame of reference. With a bit of imagination, for any two objects or events, one can find a standard with respect to which they are alike. Any two objects share some property (just as any two objects are joint members of any number of sets). For example, water and fire are alike in that they both belong to the four elements; Gandhi and Stalin have in common that they were both born in the 19th century, etc. If “similarity” isn’t an empty notion, it cannot be any property which can constitute similarity. “It must be,” as Quine notes, “that properties are
9.1 The Relativity of Similarity
157
shared only by things that are significantly similar” (1969, p. 118). Following this proposal, the notion of similarity turns out to be circular: the notion of similarity presupposes the notion of a frame of reference which itself presupposes the notion of similarity. (Structurally the same problem faces the specification of the threshold value of similarity.) And in “Seven Strictures of Similarity” Goodman poses a further problem for the notion of similarity. He invites us to compare similarity and motion. Both notions require a frame of reference. To say of something that it moves is incomplete. We have to say what it moves in relation to. Analogously, talk of similarity between two things must specify some respect in which they are similar. But, Goodman notes, there is an important difference between the relativity of motion and the relativity of similarity. In the case of statements about motion, when we add a specification of the frame of reference, the ambiguity is removed and the statement is completed. In the case of statements about similarity, however, once the respects are specified, similarity has no role to play; the respects do all the work. For example, if we say of two objects that they are similar with respect to their shared property of being green, what we say is nothing more than that they have the property “green” in common – the notion of similarity can be removed without loss. Statements of the form “A is similar to B” are reducible to statements of the form “A and B share property C.” What this suggests is that, unlike motion, the notion of similarity cannot be rescued merely by recognizing its relativity. Given the problems associated with the notion of similarity, Quine speaks of the “dubious scientific standing” of this notion, and remarks that “it is a mark of maturity in a branch of science that the notion of similarity or kind finally dissolves, so far as it is relevant to that branch of science” (1969, p. 121). Goodman uses even harsher terms. He declares that “similarity … is a pretender, an imposter, a quack. It has, indeed, its place and its uses, but is more often found where it does not belong, professing powers it does not possess” (1972, p. 437). And he concludes that while statements of similarity “are still serviceable in the streets” they “cannot be trusted in the philosopher’s study.” Quine’s and Goodman’s strictures on similarity are correct. What I question, though, is the reasonability of their recommendation that philosophy, tout court, should refrain from employing the notion of similarity.1 For example, the notion of similarity plays an important role in the explanation of counterfactual conditionals.
1 Interestingly, neither Goodman nor Quine follow their own recommendation of expelling the notion of similarity in the field of philosophy. Quine stresses that an innate sense or standard of similarity is something we share with animals; it is, as he says, “part of our animal birthright” (1969, p. 123). And despite the low regard in which Goodman holds the notion of similarity he expresses the suspicion that the notion of similarity can be explicated, in part, in terms of the notion of inductive projectibility. He writes that “rather than similarity providing any guidelines for inductive practice, inductive practice may provide the basis of some canons of similarity” (p. 441). The problem with this suggestion, according to Shoemaker (1975, p. 12–17), is that the projectibility of a property cannot be directly experienced, whereas the similarity or dissimilarity of two properties can be directly given in experience.
158
9 Diachronic Content Similarity
Robert Stalnaker (1968) proposes an approach to counterfactual conditionals in terms of comparative similarity between possible worlds. The idea is that a counterfactual conditional is true if and only if its consequent is true at the antecedent satisfying worlds most similar to ours. Likewise, with respect to the analysis of memory, the notion of similarity does indeed possess explanatory value. Talk about thought-state-tokens A and B being similar is a useful shorthand for a lengthy explanation of the kinds of properties which, according to one’s favorite semantic theory, A and B must share in order to be similar.
9.2
Principles of Semantic Similarity
Given that type-identity of contents is not a necessary condition for remembering and given that diachronic thought-state-tokens connected by memory traces may belong to (slightly) different types of contents, the question arises: when are content-tokens sufficiently similar for the later one to qualify as a memory of the earlier one? To what extent and in what manner may the information processes inherent in the faculty of memory edit a propositional content for the resulting thought-statetoken to still count as a memory representation? What is the permissible deviation from the type-identity of the contents of diachronic thought-state-types? In this section I will set forth individually sufficient principles of semantic similarity, that is, principles indicating when two thought-state-tokens are sufficiently similar to still meet the truth condition of remembering. As far as I can see, the philosophical literature contains no attempt to explicate the notion of semantic similarity.2 This isn’t surprising, since the concept of semantic similarity is encumbered with some vexing problems. I will mention just three. First, criteria of semantic similarity are relative to the individuation conditions of content which, in turn, are relative to the presumed semantics. Second, as was shown in the previous section, the concept of similarity is relative with respect to the standard of comparison and to the number of shared properties. Even if we fix the standards of comparison, we are left with the quantitative relativity of similarity. And it seems as if the threshold value of semantic similarity is itself contextually relative. Depending on the context, a memory state has to share more or less semantic
2
An exception to the rule is holistic functionalism. Functionalism holds that the content of a mental representation is grounded in its causal (computational or inferential) relation to other mental representations. Versions of functionalism differ on whether relata include all other mental representations (holism) or only some of them (localism). Holistic functionalism has the consequence that it would be a cosmic coincidence if two people ever shared a mental content. Given this problem, some holistic functionalists (e.g. Harman 1993) have argued that even if holism precludes the notion of content identity, it permits a notion of content similarity. Yet Fodor and Lepore (1992, pp. 17–22) have convincingly argued that this move is hopeless, for the notion of belief similarity presupposes a notion of belief identity.
9.2 Principles of Semantic Similarity
159
properties with the past thought-state it is causally connected with by memory traces. Since limitations of space prevent me from examining the effects of the conversational context on the notion of semantic similarity, I will try to concentrate on cases that classify as semantic similarity in most, if not all contexts. Third, depending of whether one is talking about memory de re or memory de dicto, the criteria for semantic similarity are very different. The difference between memory de re and memory de dicto is that the former but not the latter allows for the substitution salva veritate of coreferential expressions. On a de dicto reading, you can remember, say, that Marilyn Monroe is blond without remembering that Norma Jean Mortenson is blond. According to the de re reading, however, when we say of S that he remembers that Marilyn Monroe is blond, we are indicating a connection between S and a fact in the world – that Marilyn Monroe is blond. Like attributions of object-, property-, and event memory, de re attributions of propositional memory are not sensitive to the choice of words and allow for the substitution of coreferential expressions. To make things easier, I will concentrate on de re attributions of remembering.3 In the remainder of this section I will sketch some principles of semantic similarity. Let’s begin with the Principle of Tense Adjustment. In section 8.5, we saw that frequently memory contents need not be exact replicas of the original thought contents for the simple reason that the latter are framed in the present tense and the former have to be formulated in the past tense. Suppose you come to believe that S is drinking beer. What you remember the next day is not that S is drinking beer but that he drank beer. (It might be that S is still drinking beer but you cannot know this via memory). Although S is drinking beer and S drank beer have different truth conditions, the latter content can classify as a memory of the former. Just as our memory may replace statements in the (continuous) present tense for statements in the (continuous) present perfect, it may transform statements in the present perfect into statements in the past perfect and statements in the future tense into statements in the present tense. To drive home this last point, suppose that S tells you that he will have beer in his favorite bar tomorrow after work. The next day at 6 pm you remember that right now, S is having beer in his favorite bar. A principle closely related to that of tense adjustment is the Principle of Article Substitution. It states (i) that demonstrative and possessive pronouns may be replaced by indefinite and/or definite articles, (ii) that definite articles may be replaced by indefinite articles, and (iii) that definite and indefinite articles may be omitted altogether. To drive home this principle, suppose you come to believe that S drank this beer, where “this beer” refers to the only bottle of beer in your sight. If, later on, you
3
A number of philosophers hold that there is an everyday linguistic device that allows us to tell apart propositional attitudes de re from attitudes de dicto. The de re style of believing is said to be captured by “S believes of Marilyn Monroe that she is blond” and the de dicto style by “S believes that Marilyn Monroe is blond.” Yet it seems to me that in ordinary language the usage of “thinks of” and “thinks that” constructions isn’t a reliable indicator of whether the subject expresses a belief de re or de dicto (cf. Guttenplan 1994, p. 39). “Remembering that,” it seems to me, can be used to express either memory de re or memory de dicto.
160
9 Diachronic Content Similarity
recall this incident by entertaining the thought that S drank the beer (or a beer) – and if the previous thought causes the present thought in the right kind of way – then, from an intuitive point of view, the present thought classifies as an accurate representation of the past thought. Despite the fact that S drank this beer and S drank the/a beer are different types of propositions, they are similar enough to (most of the time) fulfil the factivity constraint of memory. These two principles have a special status. For while the application of the other principles of semantic similarity to be discussed is optional, there are circumstances where we have no choice but to employ the Principle of Tense Adjustment and the Principle of Article Substitution. Due to the nature of indexicals, “now” has to be changed to “then” as times passes by and “this” has to be changed to “a” as objects are moved. In the same way, if we want to refer to temporally well-defined events it may be necessary to modify the verb forms. The next principle of semantic similarity is the Principle of Universal Instantiation: statements about individuals (F(k) ) may be substituted for universally quantified statements (∀x F(x) ), as long as the rememberer knows that the individuals in question (k) belong to the domain of discourse. Sticking to the above example, consider the following scenario: you come to believe that all Bavarians drink beer. Some time later, when you are asked whether S, whom you know to be Bavarian, drinks alcohol, you recall that S drinks beer. The past thought that Bavarians drink beer [(∀x) (x is Bavarian ⊃ x drinks beer)] gives rise to the thought that A Bavarian drinks beer [(x) (x is Bavarian ⊃ x drinks beer)]. Intuitively, the recalled thought is an instance of memory, notwithstanding the fact that the two propositional contents are of different types. But couldn’t one object that this example is invalid, for it deals with inferential memory, the very type of memory which was excluded from this study (cf. sections 1.1 and 8.2)? I disagree. Granted that the transition from the proposition everyone drank beer to S drank beer rests on inferential reasoning, it is reasoning of a very basic sort and cannot be identified with genuine cases of inferential memory. To compare, here is an genuine case of inferential memory: you remember that S drank either beer or wine; you are told that S didn’t consume wine and from this you infer that S drank beer. The fourth principle of semantic similarity is a relative of the third principle. The Principle of Existential Generalization states that existentially qualified statements ( (∃ x F(x) ) may be substituted for statements about individuals ( (F(k) ), as long as the rememberer knows that the individual in question (k) belong to the domain of discourse. Suppose you watch S drinking beer at a party. Later on you are asked about the kinds of beverages consumed at the party and you recall that someone drank beer. The thought that S drank beer is remembered as the thought that someone drank beer. Even though the past and present propositional contents are of different types, the present thought-state qualifies as a memory of the past thoughtstate, provided the other conditions of remembering are fulfilled. The fifth principle is the Principle of Conjunction Elimination: conjunctive statements (p ∧ q) may be replaced by statements mentioning only one of the conjuncts (p or q). You might, for example, remember that S drank beer, even if, at no
9.2 Principles of Semantic Similarity
161
time in the past, you entertained this very thought but instead entertained the conjunctive thought that S drank beer and wine. Next, there is the Principle of Conceptual Universalization which states that, up to a point, particular concepts may be replaced by implied universal concepts, as long as the rememberer knows of the implication. Just as our memory has the tendency to condense and to generalize information, it frequently replaces particular concepts by universal ones. You might, for example, witness S drinking beer and, later on, refer to this event by believing that S drank some alcoholic beverage – provided you know that beer is a kind of alcoholic beverage. In the same vein, thoughts about poodles may be recalled as thoughts about dogs, and lemonade-thoughts may turn into generic juice-thoughts. It goes without saying that there is a limit to the permissible degree of conceptual generalization. For if all you can remember from your having witnessed S drinking beer is that S drank something (or, even worse, that someone consumed something), no one (including yourself) will be willing to grant you memory of the incident in question. The range of allowable generalization of concepts is contextually sensitive. To avoid misunderstandings, let me emphasize once again that I am only dealing with memory de re. De dicto beliefs are sensitive to the choice of words used in attributions; they are beliefs about linguistic items (dicta) rather than about the things (re). Consequently, the thought that S drank some alcoholic beverage may not count as a memory de dicto of the thought that S drank beer. The Principle of Omission is particularly interesting, since it does not rest on a logical or grammatical rule. This principle claims that, up to a point, our memory may omit contextually irrelevant information. In section 8.5, I have argued that, apart from storing, our memory sometimes has the function of processing the encoded information. The function of memory is to bring past experience to bear on present circumstances. Carrying out this function sometimes requires simplifying and abridging the stored information. To see this consider, once again, our beerloving S. Suppose you come to believe that S drinks a medium-sized bottle of ice-cold wheat beer. A few days later, all you remember is that S drank beer. Provided you are not interested in the amount, the temperature and the brand of beer consumed by S (and provided the other memory conditions are met), you (and we) would, no doubt, classify your belief that S drank beer as an instance of memory. Hence, it may not matter that a memory content is informationally impoverished in comparison to the thought content it causally derives from. It is time to step back and to examine what these principles of semantic similarity have in common. A little thought reveals that the common denominator among these principles is the loss of information. With the exception of the first principle, each one of the above principles brings it about that the encoded information is richer than the retrieved information. The retrieved information is part of, or an implication of, the encoded information. To see this, let us briefly review the examples. Going from the belief that S drank this beer to believing that he drank a beer consists of the omission of the information contained in the demonstrative pronoun “this” (Principle of Article Substitution). The proposition Bavarians drink beer is more informative than the proposition A Bavarian drinks beer
162
9 Diachronic Content Similarity
(Principle of Universal Instantiation). Similarly, the proposition Someone drank beer contains less information than S drank beer (Principle of Existential Generalization). The same is true for the proposition S drank beer vis-à-vis the proposition S drank beer and wine (Principle of Conjunction Elimination). Going from S drank beer to S drank some alcoholic beverage also involves a loss of information for it consists of the elimination of the information that the beverage in question was beer rather than wine or gin (Principle of Conceptual Universalization). And, finally, knowing that it was ice-cold wheat beer which S had is richer in information than knowing that S had beer (Principle of Omission). The amount of information that may be lost in the course of remembering is relative to the context of remembering. Given certain contexts, even the austere information that someone consumed something may count as a memory of the profuse and more specific information that S drank wheat beer. Yet there is probably an upper limit on how much information may be lost in the course of remembering. For not any fraction of the original information, no matter how small, and not any implication of the original information, no matter how far removed and irrelevant, can constitute a memory state. Otherwise, given the appropriate context, Air molecules moved could classify as a memory of S drank beer, since the former proposition is implied by the latter. While the truth condition of memory is compatible with the loss of information, it does not tolerate the increase or the enrichment of information. Whenever the retrieved information is richer than the encoded information, psychologists speak of “confabulation.” In the remainder of this section, I want to propose three additional principles of semantic similarity which, like the Principle of Tense Adjustment, bring about neither a loss nor an increase of information but cause the amount of information to stay (more or less) the same. According to the Principle of Negation Elimination a doubly negated statement (~~p) may be replaced by a positive statement (p). Suppose you believe that it is not the case that S doesn’t drink beer. If what you recall, some time later, is that S drank beer, this qualifies as a memory even though the memory content is not identical to the original content. This principle also works the other way round: if you originally thought that S drank beer, you may remember that it is not the case that S didn’t drink beer. According to some semantic theories, the contrastive stress (i.e., the featured exclusion of certain possibilities) is part of the meaning of a statement (cf. Boër 1979; Dretske 1972). Consider the following examples of contrastive statements: S drank (rather than spilled) beer; S drank beer (rather than wine); S (rather than Bert) drank beer. Assuming that contrastive stress is a semantically relevant contextual feature on a par with the identity of its utterer, the time of its utterance, and the like, the following question arises concerning memory: may a belief which embodies a certain dominant contrast count as an accurate representation of an earlier belief which does not contain this particular or any dominant contrast? Consider an example. Suppose you witness S drinking beer. A few weeks later,
9.2 Principles of Semantic Similarity
163
you overhear someone claiming that S hasn’t touched alcohol in a long time. You correct the person and assure him that S drank beer. Is the present thought that S drank beer an accurate representation of the previous thought that S drank beer? I think the answer should be affirmative, for otherwise memory becomes an exceedingly rare commodity. Hence, I propose the Principle of Contrastive Stress which states that, by and large, diachronic thought-content-tokens which differ only with respect to their contrastive stress are sufficiently similar to fulfil the truth condition of memory. Let us take up the tenth principle. Assuming that synonymy and literal meaning are not just a philosopher’s chimeras, our memory may substitute known synonyms for the words and phrases used in the original experiences and thoughts. Your current thought that prestidigitators are a nuisance may count as a memory of your previous thought that experts at sleight of hand are annoying – provided you know that “nuisance” and “annoyance,” “prestidigitator” and “expert at sleight of hand” mean the same thing. Or, to stick to the example used throughout this section, the thought that S drank beer may be recalled as the thought that, say, the husband of so-and-so imbibed an alcoholic beverage made by brewing and fermentation from cereals and flavored with hops – provided you know that S is married to so-and-so, that “drink” and “imbibe” mean the same thing, and that beer is made from fermented cereals. Thus, I propose the Principle of Synonym Substitution which states that, by and large, the faculty of memory may replace terms with their known synonyms without thereby violating the truth condition of memory.4 Like all other principles of semantic similarity, this principle is restricted to memory de re. I should also stress that this list of principles of semantic similarity does not lay claim to completeness. These principles form only the smallest common denominator of semantic similarity. Though these ten principles of semantic similarity carry over to most contexts of remembering, there are some contexts in which they do not apply. In the case of each one of these principles, one can imagine a conversational context that makes the threshold value of semantic similarity increase so that, even if two thoughtstate-tokens meet the criterion, intuitively we wouldn’t attribute memory. For example, when I recite a poem I have learned by heart, I am not entitled to omit irrelevant information, or to replace synonyms, or to change the contrastive stress. The same goes when one is repeating marriage vows in church. These contexts call for the verbatim reproduction of the learned material. And while there are some contexts in which the proposed principles of semantic similarity are too broad, there may be other contexts in which additional principles may be permitted. I will return to this issue in chapter 10.
4
The principle of synonym substitution is vital not only for propositional memory but also for memory of visual symbols. Frequently we remember the visual experience of a symbol (e.g., a white flag, a lion) by translating it into prose (truce, St. Mark).
164
9.3
9 Diachronic Content Similarity
Rectifying Memory Claims
Remembering demands that the memory content be true, but it does not require that the past and the present thought contents be identical. In the process of remembering, the informational content stored in traces may stay the same or decrease (to a certain degree); but it may not increase. In this section, I want to back up this idea by showing that we often rectify our memory claims, once their falsity has been pointed out to us, by reducing their informational content. Suppose I claim to remember that p. Shortly afterwards I find out that p is false. Instead of admitting that I did not remember that p, I may substitute a metarepresentational memory claim for a reproductive one: instead of claiming to remember that p I retreat to the claim that I thought or believed that p. This is a very common move and there is a good reason for it. By replacing a reproductive memory claim for a meta-representational one we change the informational content and thereby reduce the chance of being wrong. For as was shown in section 8.1, the truthmaker of reproductive memory is our mind and the world while the truthmaker of meta-representational memory (of non-factive attitudes) is just our mind. Errors come in different degrees. A memory claim can be inaccurate regarding specific facts and yet accurate in terms of general themes it conveys, or it can be inaccurate even in these. Examples of gross inaccuracies are people who take themselves to be Napoleon or who claim to remember having visited Mars. The only way to rectify these memory claims is by turning them into meta-representational ones: instead of claiming to remember having visited Mars (which is impossible), they may claim to remember having had the impression of visiting Mars. But very few of the inaccuracies of our memory claims are so dramatic. Most of the time we get the general themes but misrepresent some specific fact. Take, for instance, the case of someone who insists that the liquor store in his old neighborhood was on Twenty-Fifth Street when it was in fact on Twenty-Second Street. In the case of minor inaccurancies, memory claims can be corrected by omitting specific details and by moving to a more general level of description. After finding out that the liquor store is on Twenty-Second Street, I may say: “See, I told you I remember it is somewhere in this area!” “Somewhere in this area” is obviously less informative than “On Twenty-Fifth Street.” A good example of how the omission of specific facts can rectify memory claims is Ulric Neisser’s study of John Dean’s memory. When John Dean, Counsel to President Nixon, testified against Nixon before the Senate’s Watergate Committee, he was unaware that the conversations he had with Nixon had been tape-recorded. In his testimony Dean provided an incredibly detailed account of these conversations, giving almost verbatim accounts of conversations held much earlier. Later, when the tapes of these conversations were discovered, Neisser compared Dean’s testimony with the tapes and discovered that Dean’s alleged memories were frequently wrong. Interesting for the present context is Neisser’s emphasis that Dean’s alleged recollections, though literally wrong, were in some sense correct. Though Dean was mostly wrong about the
9.4 Attitudinal Similarity
165
details of the conversations with Nixon, he did recall the important and relevant information about these conversations. Dean’s testimony had much truth in it, but not at the level of “gist.” It was true at a deeper level. Nixon was the kind of man Dean described, he had the knowledge Dean attributed to him, there was a cover-up. Dean remembered all of that; he just didn’t recall the actual conversation he was testifying about…. In memory experiments, subjects often recall the gist of a sentence but express it in different words. Dean’s consistency was deeper, he recalled the theme of a whole series of conversations and expressed it in different events (1982, pp. 151, 159).5
Whether or not John Dean remembers his conversations with Nixon depends on how we formulate the proposition he is said to remember. If the proposition refers only to the general themes of the conversations, Dean does remember, for the proposition is true. Thus, once it is pointed out to John Dean that many of the specific facts he reported are inaccurate, he could retreat to the level of gist and thereby render his reproductive memory claims true. And there is another method of diluting the information content of a memory claim and thereby rendering it true: sometimes our memory claims are basically accurate as to observable facts about the world while they are mistaken because of their bizarre interpretation of those facts. Consider the following example adopted from Marya Schechtman (1996, pp. 126): someone who constantly sees in clearly innocent actions and gestures evidence of a sinister conspiracy directed against him gives the following memory report: “Yesterday, while shopping, there was a group of men in black suits watching me and taking notes. This is no coincidence; the CIA is once again after me.” If the paranoiac is not suffering from hallucinations, the observable facts mentioned in him memory report – men in black suits watched him and took notes – may be correct. It is just his interpretation of these facts which is absurd. The men were not CIA agents but market researchers counting the number of shoppers. When this is pointed out to the paranoiac, he may correct his memory claim by omitting some of its informational content, namely the part about the CIA.
9.4
Attitudinal Similarity
Meta-representational remembering consists of a two-fold classification. Remembering that one used to be in mental state X implies remembering that X had a certain kind of attitude and content. For example, remembering that one believed Kennedy was assassinated involves remembering that the first-order state is about what one expresses by “Kennedy was assassinated” as well as remembering that the content was framed in the attitude of believing.
5 Cf. Hirst and Gluck (1999). It is not entirely clear whether Neisser thinks that one can remember falsehoods as long as they belong to the periphery or embroidery of a narrative rather than to its core. If Neisser does sign up to this claim, I disagree.
166
9 Diachronic Content Similarity
Throughout this and the previous chapter I have been focusing on contentidentification because it is this aspect that reproductive and meta-representational memory have in common. Now I will say a few words about the margin of error regarding attitude-identification. We saw that the (embedded) content of memories are not necessarily type-identical with the original contents. But what about attitude-identification? Are there cases where the attitude-type identified by a metarepresentational memory does not have to be identical with, but only similar to, the attitude-type of the past state it is causally related with? And if this is so, when are attitude-types sufficiently similar for the truth condition of memory to be met? What is the permissible range of aberration between the original attitude-type and the attitude-type identified in a meta-representational memory claim? The level of scrutiny regarding the identification of past attitudes is dependent on the context. Sometimes when we recall our past mental states we are primarily interested in what we were thinking about, not in the mode in which the content was realized. Sometimes, however, the attitude does play an important role. Suppose you believe seeing an UFO landing in your backyard. Later on, when you report this incident to a skeptical police officer, you insist on remembering having seen a UFO while he might try to convince you that you merely remember having imagined seeing a UFO. Here the identification of the attitudinal component is vital for the informational content of the memory claim. Focusing on cases of meta-representational memory where the stress is not on attitude-identification, are there attitude types which are similar enough so that if I had a token of one type but remembered having had a token of another type, the memory claim would still turn out true according to most contextual standards? The answer to this question crucially depends on how attitudes are defined. Broadly speaking there are two (related) ways of defining attitudes. Using John Searle’s notion of direction of fit, attitudes can be divided into two groups (1983, pp. 7–8). On the one hand are cognitive attitudes which are supposed, in some way, to match an independently existing world and thus have a mind-to-world direction of fit. The paradigm example is believing; other examples are supposing, doubting, suggesting. On the other hand are conative attitudes that are supposed to bring about changes in the world and thus have a world-to-mind direction of fit. The paradigm example is wanting; other examples are choosing, intending, trying. Believing and wanting are paradigm examples for the mind-to-world direction of fit and the world-to-mind direction of fit, respectively, because they are simple attitudes. Complex attitudes, by contrast, presuppose other (simple) attitudes and involve concepts or dispositions. “To expect,” for example, requires the concept of future, “to be certain” involves a certain doxastic commitment, and “to intend” not only implies having a desire but also a belief of how to fulfil the desire. Given the distinction between simple and complex attitudes, one can differentiate between a grave and a benign misrepresentation of the attitudinal component of one’s past mental state. In certain contexts, meta-representational memory needs to identify a past attitude only as of either the cognitive or the conative type. Suppose that yesterday, when I was offered chocolate and vanilla ice cream,
9.4 Attitudinal Similarity
167
I decide in favor of vanilla. Today, I recall my decision by claiming to remember that I wanted vanilla ice cream. Following the proposal at hand, the memory claim would be correct because deciding in favor of something is a complex form of wanting. Functionalism provides a more fine-grained taxonomy of attitudes than the distinction of directions of fit and the simple/complex distinction. In this view, to understand attitudes is to understand their functional (causal or conceptual) roles.6 Functional roles are characterized by their inputs and outputs. A pain, for example, is a state that is often caused by damage to the body, it tends to produce a desire to get rid of that state, and tends to produce wincing, groaning, and so on. A belief state, on the other hand, normally results from perception or inference and gives rise to other beliefs or to behavior. The actual or potential actions caused by a belief state depend on the presence of other states. The belief that something is food may or may not lead to the attempt to eat it, depending on whether one has the desire to eat.7 Assuming functionalism about attitudes, if functional roles allow for similarity relations so do attitudes. Then we can say things like this: believing is more like supposing and less like doubting since the behavioral output of a belief resembles that of a supposition but not that of a doubt. And if there are similarity relations among attitudes, one could argue that an original attitude-type and the one identified in a meta-representational memory claim do not always have to be the same. Sometimes it may suffice to identify a sufficiently similar attitude.
6 Cf. Fodor (1987, pp. 68–70), Harman (1973, pp. 45–46), Lewis (1972, p. 256n) and Shoemaker (1990, pp. 200–205). In recent literature one can find the metaphor of “boxes” in the brain. To believe something is to store a sentence of the language of the mind in one’s “belief box;” to desire something is to store a sentence of the language of the mind in one’s “desire box;” and so on. Box-talk is a shorthand for individuating attitude types in functional terms. 7 According to short-arm functionalism the inputs and outputs of mental states are peripheral stimuli and bodily movements. Since narrowly specified functional roles are transparent to the mind one can, via introspection, acquire knowledge about the attitudinal components of one’s first-order states. Long-arm functionalism, on the other hand, individuates functional roles (among other things) by reference to external states of affairs. To see the difference between these versions of functionalism, consider the propositional attitudes “to see” and “to hallucinate.” From the point of view of short-arm functionalism the input conditions of both attitude types are the same, namely the presence of sensory stimuli. Long-arm functionalism, however, defines “to see” in terms of distal inputs (light waves) and therefore distinguishes between the functional properties of seeing and hallucinating.
Chapter 10
The Pragmatic Dimension of Memory
In chapter 8 it was argued that the xerox model of memory is false in that it acknowledges only the reproductive nature of memory. Our memory is not only a passive device for reproducing contents but also an active device for processing stored contents. As a result of such information processing, the content of a memory state may differ, to some degree, from the content of the original representation. Two propositional attitudes may stand in a memory relation to one another even though they are not type-identical but only sufficiently similar. Chapter 9 tried to specify to what extent propositional attitude tokens may differ from one another and one of them still count as a memory of the other. The present chapter addresses the question what determines whether the factivity constraint on memory requires that diachronic propositional attitudes must be identical or whether it allows for them to be only similar. I will propose a contextualist answer to the question at hand: whether a memory state must be identical with the representational state it causally derives from or whether it suffices that the two states are merely similar depends on the conversational context of the rememberer and of the attributor, i.e., the person describing the subject as a rememberer. The truth condition for remembering is pragmatically sensitive in that it depends on the context whether memory requires literal reproduction of previously encoded representations or whether it allows for some moderate reconstruction. Consider an analogy. Imagine two pictures of Fido, a needle-sharp photograph and a cubistic watercolor. Obviously, the former is a more accurate representation of Fido than the latter. Whether the watercolor is a good enough representation of Fido depends on what it is used for. If Fido is missing and I want to put up posters with his picture hoping that someone will find and return him, then I would better use the needle-sharp photograph. If, on the other hand, I want to use the picture to decorate my office and to be reminded of Fido while at work, then I might as well take the cubistic watercolor. In short, the required accuracy of a representation is, to a certain extent, dependent on the context in which it is used. The point of this chapter is that the same thing is true of memories. The level of authenticity with which memories must reproduce (or represent) past contents is, to some extent, dependent on the conversational context in which the memory state is employed. There are contexts where remembering calls for the exact reproduction of learned
S. Bernecker, The Metaphysics of Memory. © Springer Science + Business Media B.V. 2008
169
170
10 The Pragmatic Dimension of Memory
material. For example, when one is reciting a poem one has learned by heart or when one is repeating marriage vows in church one may not transform the learned contents. But while our memory is sometimes supposed to work like a tape recorder by giving a verbatim re-run of the stored information it may, at other times, process the encoded information. Section 10.1 gives a brief overview of different kinds of contextualism about knowledge and justification. Section 10.2 sets forth a contextualist interpretation of the factivity constraint on memory. The idea is that, depending on the conversational context, remembering requires either verbatim reproduction of the encoded information or allows for some degree of reconstruction. Section 10.3 briefly summarizes this investigation.
10.1
Epistemic Contextualism
To explain the kind of contextualism about memory that I have in mind it is useful to compare it to, and distinguish it from, contextualism about knowledge and justification. Generally speaking, epistemic contextualism maintains that whether one knows may (but need not) be relative to the conversational context. The truth conditions of knowledge ascription of the form “S knows that p” vary in certain ways according to the context in which they are uttered or thought. As long as error-possibilities are ignored, the standards for knowledge remain low, and the concept expressed by the term “knowledge” is that of low-standard knowledge. But when error possibilities become salient, then the standards for knowledge rise. Epistemic contextualism has been proposed as a way to resolve scepticism. Given contextualism, when the sceptic presents his argument to the effect that knowledge is impossible, he raises the standards for knowledge, thereby creating a conversational context in which he can truthfully say that we are unable to know anything. Yet the fact that the sceptic introduces very high standards for knowledge which we don’t satisfy does not mean that we fail to meet the lower standards that are ordinarily in place. The sceptic mistakenly assumes that because we don’t know in the high standard context we don’t know in all contexts. So contextualism allows for the skeptical conclusion to be true, but limits the damage of this concession by arguing that this conclusion only holds in certain non-everyday contexts. What are the sorts of conversational factors which the contextualist invoke as affecting epistemic standards? Most contextualists maintain that standards for knowledge are governed by a salience constraint: If the person ascribing knowledge to S is worried about the possibility that S has made a certain kind of mistake, then that renders the error-possibility a relevant alternative. Unless S’s evidence is good enough to rule out the error-possibility to which he is paying attention, he cannot be said to know. There is little agreement on what else, apart from salience considerations, affects knowledge standards. Stewart Cohen (1999, p. 61) thinks that knowledge standards are determined “by some complicated function of speaker intentions, listener expectations, presuppositions of the conversation, salience
10.1 Epistemic Contextualism
171
relations, etc.” And David Lewis (1996, p. 556) thinks that high stakes can make a difference: “When error would be especially disastrous, few possibilities may be properly ignored.” In recent years, a number of different positions have been developed under the heading of epistemic contextualism. The most basic distinction is that between attributor contextualism (or simply contextualism), as advanced by Cohen (1986), Keith DeRose (1992) and Lewis (1996), and subject contextualism (or subject sensitive invariantism), as advanced by John Hawthorne (2004, ch. 4) and Timothy Williamson (2000, ch. 7). The issue is whether knowledge is relative to the conversational context of the epistemic subject, or rather to the conversational context of the attributor, i.e., the person describing the subject as a knower or a non-knower. As the name suggests, subject contextualism is the view that whether someone knows depends on the conversational context of the subject rather than of the attributor. As the stakes go up for the subject of a knowledge attribution, more evidence is needed for that person to have knowledge and to be justified in his beliefs. According to attributor contextualism, however, the varying standards that the epistemic subject must live up to in order to count as knowing are relative to the conversational context of the attributor. One attributor can truthfully say “S knows that p,” while another attributor, in a different context where higher standards are in place, can truthfully say “S doesn’t know that p,” even though both speakers are talking about the same subject knowing the same proposition at the same time. DeRose (1992, pp. 921–922) explains the distinction between subject and attributor contextualism as follows: Attributor factors set a certain standard the putative subject of knowledge must live up to in order to make the knowledge attribution true: They affect how good an epistemic position the putative knower must be in to count as knowing. They thereby affect the truth conditions … of the attributions. Subject factors, on the other hand, determine whether or not the putative subject lives up to the standards that have been set, and thereby can affect the truth value of the attribution without affecting its [truth conditions]: They affect how good an epistemic position the putative knower actually is in.
To illustrate the difference between subject and attributor contextualism consider the following example: I am engaged in a discussion with Oscar in which we are employing very high epistemic standards. For us it is of a matter of live and death whether we know that p. Although we have good enough evidence for the belief that p that we would claim to know it if we were in a more ordinary context, we each deny that we know that p. Suppose Oscar raises the question of whether an absent friend of ours, Bert, knows that p. We both know that Bert has the same evidence for the belief that p that we have, but that since the issue isn’t important to him his epistemic standards are much lower than ours. Should we describe Bert as knowing that p? According to subject contextualism, it is Bert’s own conversational context that determines the truth conditions of “Bert knows that p.” And since we are aware that Bert’s epistemic standards are lax, we should indeed attribute to Bert knowledge that p. Given attributor contextualism, however, if we are not willing to ascribe knowledge to us, the attributors, then we shouldn’t ascribe knowledge to Bert, who possesses the same evidence we do.
172
10 The Pragmatic Dimension of Memory
What seems to speak in favor of attributor contextualism is that, in high standard contexts, attributors tend to apply to an absent subject the same high standards that they apply to themselves, even if they are aware of the subject’s low-standard context. And what seems to speak in favor of subject contextualism is that in situations, where the standards for knowledge are more relaxed, it is natural for attributors to apply to absent subjects standards that are appropriate to the context of those subjects. Obviously neither of these considerations amounts to a knock-down argument for or against subject contextualism and/or attributor contextualism. But it is hardly surprising that there are no decisive arguments either way because attributor contextualism can incorporate subject contextualism: There is nothing in attributor contextualism to prevent an attributor from selecting epistemic standards appropriate to the subject’s context (cf. DeRose 2004, p. 349).
10.2
Memorial Contextualism
After having outlined epistemic contextualism, it is time to set forth the contextualist account of memorial authenticity. I do not want to say that the truth condition of memory ascriptions of the form “S remembers that p” vary according to the conversational context in which they are uttered or thought. “S remembers that p” has context-insensitive or invariantist truth conditions. For it is context independent whether a past and a present content token are sufficiently similar for the later one to count as a memory of the earlier one. Instead of proposing context sensitive conditions for content and attitude similarity, I suggest a contextualist criterion for determining which of the two disjuncts of the content and the attitude conditions is meant in a particular memory ascription. Depending on the context, diachronic content tokens and attitude tokens must be identical or they need to be only sufficiently similar. To drive home the difference between epistemic contextualism and my contextualism about memorial authenticity compare indexical expressions such as “here” and “tomorrow” with ambiguous expressions such as “light” and “bank.” Ambiguous expressions have more than one meaning. The word “light” can mean not very heavy or not very dark. Indexicals, on the other hand, have fixed meaning but variable reference. The meaning of the word “tomorrow” doesn’t change from a day to the next, though, of course, its reference does. Indexicals in a sentence contribute to the sentence’s having different truth conditions in different contexts. Epistemic contextualists hold that “to know” is an indexical that refers to different epistemic positions in different conversational contexts (Cohen 1988, p. 97; DeRose 1992, p. 920–921). To use David Kaplan’s terminology, when “to know” is used in different conversational contexts it has the same character, but different content (1989, pp. 500–507). On my account of memorial authenticity, “to remember” is, in a certain respect, an ambiguous expression: depending on the conversational context, “to remember” can mean that the present and past thought tokens are identical or that they are only sufficiently similar.
10.2 Memorial Contextualism
173
Notwithstanding the differences between epistemic contextualism and contextualism about memorial authenticity both positions have something in common: Something like the distinction between subject and attributor contextualism carries over to contextualism about memorial authenticity. Subject contextualism about memorial authenticity maintains that which of the two disjuncts of the content and the attitude conditions of remembering is meant in a particular memory ascription depends on the conversational context of the rememberer himself. Attributor contextualism about memorial authenticity, on the other hand, holds that it is the context of the memory ascriber that determines which of the two disjuncts of the content and the attitude conditions is meant in a particular situation. To get a feeling for the difference between subject and attributor contextualism about memorial authenticity consider the following example. Unbeknownst to you, Bert is an industrial spy working for a company which is competing with the company you work for. Bert asks you to recount a conversation you had with your boss concerning the company’s strategy. Since the conversation in question concerns a topic of great interest to Bert’s company, Bert is hoping you provide him with a verbatim report of what your boss said on this occasion. Yet, in order not to arouse your suspicion and not to be uncovered as a spy, Bert conceals his burning interest in the details of the conversation and instead asks you in a casual manner about the conversation. Unaware that Bert would like a full report, you provide him with a brief summary of the conversation which conveys the general theme but omits many particulars. Since Bert’s standards for remembering are more rigid than your own standards, the question arises of who calls the shots. Depending on whether one subscribes to subject or attributor contextualism about memorial authenticity your summary of the conversation with your boss does or does not meet the content and attitude conditions for remembering. Are memory ascriptions dependent on the conversational context of the remembering subject or of the person describing the subject as a rememberer? In my view, it is both the subject context and the attributor context that determine the authenticity standard of a given memory claim. One of the factors that figures in the subject context of remembering is the agent’s intention. If a rememberer is trying to give a verbatim account of what he thought or experienced, he thereby raises the standards for fulfilling the truth condition of memory. If, however, the rememberer is trying only for a rough account of what he thought or experienced, he thereby lowers the standards for meeting the truth condition. Suppose Oscar tries to memorize a role in a stage play. After having studied the text, Oscar asks Bert to read the other parts while he recites his part. In this context, neither Oscar nor Bert would say that Oscar remembers his role, unless he can recite it perfectly (i.e. verbatim). The context is such that remembering demands type-identity of thought contents. But now suppose that shortly after having memorized the theater play, Oscar has a car accident and gets a severe blow on the head. Suspecting that he is suffering from retrograde amnesia, a physician asks Oscar to recall events and experiences prior to the accident. Among other things, Oscar is able to recite his role in the stage play. Now I reckon that even if he makes a few small mistakes and cannot recall the entire role, both the physician and Oscar
174
10 The Pragmatic Dimension of Memory
himself will say that he does remember his role in the stage play and they will dismiss the suggestion that he is suffering from retrograde amnesia. The context is such that remembering does not demand type-identity of thought contents. It suffices if the encoded information and the retrieved information are similar. The idea that the truth conditions of one’s memory claims are also sensitive to the attributor context might arouse disconcertment. One might think that the function of memory is determined exclusively by the rememberer himself, not by the interests and intentions of the bystanders. However, I think this is false. Suppose some time ago Oscar has witnessed fiso biting the mailman in the left leg. Further suppose that, unbeknownst to Oscar, within minutes of Fido having bitten the mailman in the left leg another dog, Spot, bit the mailman in the right leg. The wound on the left leg was relatively harmless but the one on the right leg was severe and brought about a high fever and loss of consciousness. Since the right leg required elaborate and expensive medical treatment the mailman’s health insurance company seeks to determine the culprit. Unfortunately, due to the blackout, the mailman has permanently forgotten which dog bit which leg. And since the bite marks of both dogs are the same, the entire case hangs on Oscar’s eyewitness testimony. Oscar tells the attorney everything he remembers which is that that Fido bit the mailman; Oscar has forgotten which part of the mailman’s body Fido bit Once the attorneys realize that Oscar’s memory is not detailed enough to decide the case one way or the other, one of them says to another: “Too bad, Oscar doesn’t really remember the circumstances of the offence.” Given the analysis of remembering, as it stands, the attorney’s comment is plainly false. Oscar does remember what he claims to remember. The reason is that, apart from meeting all the other memory conditions he satisfies the second disjunct of the content condition, that is, the disjunct demanding content similarity. But I am reluctant to simply dismiss the attorney’s comment. Given his high-stakes context, the attorney does have a point. The function of memory is to bring past representations to bear on present circumstances. Thus, if, like in the present case, the content of a memory state is too unspecific to be of any use, it is quite natural to deny that it is a memory state at all. Before terminating this chapter let’s consider an objection a critic might launch against memorial contextualism. A critic might argue that my analysis of the factivity of memory is contradictory. In chapter 8, I started out by maintaining that memory implies truth and that one can remember only what is/has been the fact. Saying that remembering implies truth implies, among other things, that the content of the memory belief is an accurate representation of the content of the original belief it causally derives from and refers to. In the present chapter I hold that what counts as “accurate” may, to a certain extent, depend on the intentions and purposes of the rememberer and the attributor. A critic might see the following problem: truth is an absolute matter. A proposition is either true or false; it cannot be neither true nor false; nor can it be true to a certain degree or more or less true than some other proposition. Given that memory implies truth and that truth is an absolute matter, it follows that memory too is an absolute concept. Either I remember or I don’t. There is no such thing as remembering an event better or worse than someone else. (But
10.2 Memorial Contextualism
175
one can remember more about an event than someone else.) Unlike truth, semantic similarity is a matter of degree. A can be more similar to B than to C. If there are contexts where memory is bound by similarity constraints, it allows for gradation; yet if memory implies truth, it is an absolute matter and does not allow for degrees. The critic concludes that memorial contextualism is an attempt to have my cake and eat it too. The problem at hand evaporates as soon as one realizes that the concept of memory, like the concept of knowledge, is relationally absolute. It is absolute, relative to a certain standard. The concept of knowledge is also relationally absolute in that in order to know a proposition to be true, our evidence need not eliminate all the alternatives to that proposition but only the relevant alternatives (where the set of relevant alternatives is determined by some standard). Other examples of relationally absolute concepts are “empty” and “flat.” Both concepts are absolute, for a space is empty only if it does not contain anything and a surface if flat only if it does not have any bumps. However, this absolute character is relative to a standard. The standard for what counts as flat is such that we would not deny that a table is flat because a microscope reveals irregularities on its surface. And we would not deny that a football stadium is empty because it contains particles of dust. To be flat is to be free of relevant bumps and to be empty is to be devoid of relevant objects. To remember that p means to accurately represent a previously had thought or experience. In this respect memory is an absolute concept. Yet what counts as an accurate or veridical representation may be to some degree relative to the interests and purposes of the rememberer. In this respect memory is a context-dependent concept.
Bibliography
Aho, Tuomo and Niiniluoto, Ilkka (1990), “On the Logic of Memory,” Acta Philosophical Fennica 49, pp. 408–429. Alexander, Samuel (1920), Space, Time, and Deity, 2 volumes, London: Macmillan. Alston, William (1991), Perceiving God: The Epistemology of Religious Experience, Ithaca, NY: Cornell University Press. Alston, William (1993), The Reliability of Sense Perception, Ithaca, NY: Cornell University Press. Anderson, John R. (1995), Cognitive Psychology and Its Implications, New York: W.H. Freeman. Annis, David B. (1980), “Memory and Justification,” Philosophy and Phenomenological Research 40, pp. 324–333. Anscombe, G.E.M. (1981a), “The Reality of the Past,” in her Collected Philosophical Papers 2: Metaphysics and the Philosophy of Mind, Oxford: Oxford University Press, pp. 103–119. Anscombe, G.E.M. (1981b), “Memory, ‘Experience’ and Causation,” in G.E.M. Anscombe (ed.), Collected Philosophical Papers II: Metaphysics and the Philosophy of Mind, Oxford: Blackwell, pp. 120–130. Armstrong, David M. (1979), “Perception, Sense Data and Causality,” in G.F. MacDonald (ed.), Perception and Identity, London: Macmillan, 1979, pp. 84–98. Armstrong, David M. (1987), “Mental Concepts: The Causal Analysis,” in R.L. Gregory (ed.), The Oxford Companion to the Mind, Oxford: Oxford University Press, pp. 464–465. Audi, Robert (1995), “Memorial Justification,” Philosophical Topics 23, pp. 31–45. Audi, Robert (1997), “The Place of Testimony in the Fabric of Knowledge and Justification,” American Philosophical Quarterly 34, pp. 405–422. Audi, Robert (1998), Epistemology, London: Routledge. Augustine, St. (400), Confessions, in H. Chadwick (ed.), 1991, Oxford: Oxford University Press. Austin, J.L. (1962), Sense and Sensibilia, Oxford: Oxford University Press. Ayer, A.J. (1956), The Problem of Knowledge, Harmondsworth, UK: Penguin. Ayer, A.J. (1972), Probability and Evidence, London: Macmillan. Barclay, Craig R. (1986), “Schematization of Autobiographical Memory,” in D.C. Rubin (ed.), Autobiographical Memory, Cambridge: Cambridge University Press, pp. 82–99. Barclay, Craig R. and DeCooke, Peggy A. (1988), “Ordinary Everyday Memories: Some of the Things of Which Selves Are Made,” in U. Neisser and E. Winograd (eds.), Remembering Reconsidered, Cambridge: Cambridge University Press, pp. 91–125. Barsalou, Laurence (1988), “The Content and Organization of Autobiographical Memories,” in U. Neisser and E. Winograd (eds.), Remembering Reconsidered: Ecological and Traditional Approaches to the Study of Memory, Cambridge: Cambridge University Press, pp. 193–244. Bartlett, Frederick C. (1932), Remembering: A Study in Experimental and Social Psychology, Cambridge: Cambridge University Press. Beauchamp, Tom L. and Rosenberg, Alexander (1981), Hume and the Problem of Causation, New York: Oxford University Press.
177
178
Bibliography
Benjamin, Aaron and Bjork, Robert (1996), “Retrieval Fluency as a Metacognitive Index,” in L.M. Reder (ed.), Implicit Memory and Metacognition, Mahwah, NJ: Erlbaum, pp. 309–338. Benjamin, B.S. (1956), “Remembering,” Mind 65, pp. 312–331. Bergson, Henri (1896), Matter and Memory, trans. by N.M. Paul and W.S. Palmer, 1994, New York: Zone Books. Bernecker, Sven (2001), “Russell on Mnemic Causation,” Principia 5, pp. 149–185. Bernecker, Sven (2007), “Remembering without Knowing,” Australasian Journal of Philosophy 85, pp. 137–156. Bilgrami, Akeel (1998), “Self-Knowledge and Resentment,” in C. Wright, B. Smith and C. MacDonald (eds.), Knowing Our Own Minds, Oxford: Oxford University Press, pp. 207–243. Boër, Steven E. (1979), “Meaning and Contrastive Stress,” Philosophical Review 78, pp. 263–298. BonJour, Laurence (1985), The Structure of Empirical Knowledge, Cambridge, MA: Harvard University Press. BonJour, Laurence (2007), “Epistemological Problems of Perception,” in E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (May 2007 edition), http://plato.stanford.edu/entries/ perception-episprob/ Braddon-Mitchell, David and Jackson, Frank (1996), The Philosophy of Mind and Cognition, Oxford: Blackwell. Brandt, Richard (1955), “The Epistemological Status of Memory Beliefs,” Philosophical Review 64, pp. 78–95. Brewer, William F. (1986), “What is Autobiographical Memory?,” in D.C. Rubin (ed.), Autobiographical Memory, Cambridge: Cambridge University Press, pp. 25–49. Brewer, William F. (1988), “Memory for Randomly Sampled Autobiographical Events,” in U. Neisser and E. Winograd (eds.), Remembering Reconsidered, Cambridge: Cambridge University Press, pp. 21–90. Brewer, William F. and Nakamura, G.V. (1984), “The Nature and Function of Schemas,” in R.S. Wyer and T.K. Srull (eds.), Handbook of Social Cognition, Volume 1, Hillsdale, MI: Earlbaum, pp. 119–160. Broad, C.D. (1925), The Mind and its Place in Nature, London: Kegan Paul, Trench, Trubner. Broadbent, Donald E. (1971), Decision and Stress, New York: Academic Press. Burge, Tyler (1993), “Content Preservation,” Philosophical Review 102, pp. 457–488. Burge, Tyler (1997), “Interlocution, Perception, and Memory,” Philosophical Studies 86, pp. 21–47. Butler, Ronald (1959), “Other Dates,” Mind 68, pp. 16–33. Cacioppo, John T., Berntson, Gary G., Lorig, Tyler S., et al. (2003), “Just Because You’re Imaging the Brain Doesn’t Mean You Can Stop Using Your Head: A Primer and Set of First Principles,” Journal of Personality and Social Psychology 85, pp. 650–661. Casey, Edward S. (2000), Remembering: A Phenomenological Study, 2nd edition, Bloomington, IN: Indiana University Press. Child, William (1994), Causality, Interpretation and the Mind, Oxford: Clarendon. Child, William (2006), “Memory, Expression, and Past-Tense Self-Knowledge,” Philosophy and Phenomenological Research 73, pp. 54–76. Chisholm, Rodrick M. (1989), Theory of Knowledge, 3rd edition, Englewood Cliffs, NJ: Prentice-Hall. Cohen, Stewart (1986), “Knowledge and Context,” Journal of Philosophy 83, pp. 574–585. Cohen, Stewart (1988), “How to be a Fallibilist,” Philosophical Perspectives 2, pp. 91–123. Cohen, Stewart (1999), “Contextualism, Skepticism, and the Structure of Reason,” Philosophical Perspectives 13, pp. 57–89. Conway, Martin (2001), “Sensory-Perceptual Episodic Memory and its Context: Autobiographical Memory,” Philosophical Transactions of the Royal Society of London 356, pp. 1375–1384. Cornman, James (1965), “Malcolm’s Mistaken Memory,” Analysis 25, pp. 161–167. Dancy, Jonathan (1985), An Introduction to Contemporary Epistemology, Oxford: Blackwell.
Bibliography
179
Davidson, Donald (1980), “Actions, Reasons, and Causes,” in his Essays on Actions and Events, Oxford: Oxford University Press, pp. 3–19. DeRose, Keith (1992), “Contextualism and Knowledge Attributions,” Philosophy and Phenomenological Research 52, pp. 913–929. DeRose, Keith (2004), “The Problem with Subject-Sensitive Invariantism,” Philosophy and Phenomenological Research 68, pp. 346–350. Deutscher, Max (1969), “A Causal Account of Inferring,” in R. Brown and C.D. Rollins (eds.), Contemporary Philosophy in Australia, London: Allen & Unwin, pp. 97–118. Deutscher, Max (1989), “Remembering ‘Remembering’,” in J. Heil (ed.), Cause, Mind, and Reality: Essays Honoring C.B. Martin, Dordrecht, The Netherlands: Reidel, pp. 53–72. Dretske, Fred (1969), Seeing and Knowing, Chicago, IL: University of Chicago Press. Dretske, Fred (1972), “Contrastive Statements,” Philosophical Review 81, pp. 411–437. Dretske, Fred (1979), “Simple Seeing,” in D.F. Gustafson and B.L. Tapscott (eds.), Body, Mind and Method: Essays in Honor of Virgil Aldrich, Dordrecht, The Netherlands: Reidel, pp. 1–15. Dretske, Fred (1981a), Knowledge and the Flow of Information, Cambridge, MA: MIT. Dretske, Fred (1981b), “The Pragmatic Dimension of Knowledge,” Philosophical Studies 40, pp. 363–378. Dretske, Fred (1982), “A Cognitive Cul-de-Sac,” Mind 91, pp. 109–111. Dretske, Fred (2000), “Entitlement: Epistemic Rights Without Epistemic Duties,” Philosophy and Phenomenological Research 60, pp. 591–606. Dretske, Fred and Yourgrau, Palle (1983), “Lost Knowledge,” Journal of Philosophy 80, pp. 356–367. Ducasse, C.J. (1951), Nature, Mind, and Death, La Salle, IL: Open Court. Dudai, Yadin (1989), The Neurobiology of Memory, Oxford: Oxford University Press. Dummett, Michael (1993), “Testimony and Memory,” in his Seas of Language, Oxford: Oxford University Press, pp. 411–428. Earle, William (1956), “Memory,” Review of Metaphysics 10, pp. 3–27. Ebbinghaus, Hermann (1885), Über das Gedächtnis: Untersuchungen zur experimentellen Psychologie, Leipzig, Germany: Duncker & Humblot. Evans, Gareth (1982), The Varieties of Reference, Oxford: Clarendon. Feldman, Richard and Conee, Earl (2004), Evidentialism: Essays in Epistemology, Oxford: Clarendon. Flage, Daniel E. (1985a), “Hume on Memory and Causation,” Hume Studies 10, suppl., pp. 168–188. Flage, Daniel E. (1985b), “Perchance to Dream: A Reply to Traiger,” Hume Studies 11, pp. 173–181. Fodor, Jerry A. (1987), Psychosemantics: The Problem of Meaning in the Philosophy of Mind, Cambridge, MA: MIT. Fodor, Jerry A. (2000), “A Science of Tuesdays: Review of Hilary Putnam The Threefold Cord,” London Review of Books 22, no. 14. Fodor, Jerry A. and LePore, Ernest (1992), Holism: A Shopper’s Guide, Oxford: Blackwell. Friedman, Lesley (1993), “Another Look at Flage’s Hume,” Hume Studies 19, pp. 177–186. Furlong, E.J. (1948), “Memory,” Mind 57, pp. 16–44. Furlong, E.J. (1951), A Study in Memory: A Philosophical Essay, London: Thomas Nelson. Furlong, E.J. (1954), “Memory and the Argument from Illusion,” Proceedings of the Aristotelian Society, pp. 131–144. Furlong, E.J. (1970), “Mr. Urmson on Memory and Imagination,” Mind 79, pp. 137–138. Ginet, Carl (1975), Knowledge, Perception, and Memory, Dordrecht, The Netherlands: Reidel. Ginet, Carl (1988), “Memory Knowledge,” in G.H.R. Parkinson (ed.), The Handbook to Western Philosophy, New York: Macmillan, pp. 159–178. Gloor, Pierre, Olivier, A., Quesney, L.F. et al. (1982), “The Role of the Limbic System in Experimental Phenomena of Temporal Lobe Epilepsy,” Annals of Neurology 12, pp. 129–144.
180
Bibliography
Goldman, Alvin I. (1967), “A Causal Theory of Knowing,” Journal of Philosophy 64, pp. 357–372. Goldman, Alvin I. (1986), Epistemology and Cognition, Cambridge, MA: Harvard University Press. Goldman, Alvin I. (1999), “Internalism Exposed,” Journal of Philosophy 96, pp. 271–293. Goodman, Nelson (1972), “Seven Strictures on Similarity,” in his Problems and Projects, Indianapolis, IN: Bobb-Merrill, pp. 437–446. Gosse, Philip Henry (1857), Omphalos: An Attempt to Untie the Geological Knot, intro. by D. Knight, 2003, London: Routledge. Guttenplan, Samuel (1994), “An Essay on Mind,” in S. Guttenplan (ed.), A Companion to the Philosophy of Mind, Oxford: Blackwell, pp. 1–107. Hacker, P.M.S. (1990), An Analytical Commentary on the “Philosophical Investigations,” Volume 3: Wittgenstein: Meaning and Mind, Oxford: Blackwell. Hamilton, Andy (1998), “False Memory Syndrome and the Authority of Personal Memory Claims: A Philosophical Perspective,” Philosophy, Psychiatry, and Psychology 5, pp. 283–297. Harman, Gilbert (1973), Thought, Princeton, NJ: Princeton University Press. Harman, Gilbert (1986), Change in View: Principles of Reasoning, Cambridge, MA: MIT. Harman, Gilbert (1993), “Meaning Holism Defended,” in J.A. Fodor and E. LePore (eds.), Holism: A Consumer Update, Amsterdam: Rodopi, pp. 163–171. Harrod, R.F. (1942), “Memory,” Mind 51, pp. 47–68. Harvey, J.W. (1941), “Knowledge of the Past,” Proceedings of the Aristotelian Society 41, pp. 149–166. Hawthorne, John (2004), Knowledge and Lotteries, Oxford: Clarendon. Heil, John (1978), “Traces of Things Past,” Philosophy of Science 45, pp. 60–72. Hirst, William and Gluck, David (1999), “Revisiting John Dean’s Memory,” in E. Winograd, R. Fivush, and W. Hirst (eds.), Ecological Approaches to Cognition: Essays in Honor of Ulric Neisser, Mahwah, NJ: Erlbaum, pp. 253–281. Holland, Alan (1974), “Retained Knowledge,” Mind 83, pp. 355–371. Holland, R.F. (1954), “The Empiricist Theory of Memory,” Mind 63, pp. 464–486. Huemer, Michael (1999), “The Problem of Memory Knowledge,” Pacific Philosophical Quarterly 80, pp. 346–357. Hume, David (1739), A Treatise of Human Nature, in L.A. Selby-Bigge (ed.), 2nd edition by P.H. Nidditch, 1978, Oxford: Oxford University Press. Hume, David (1748), An Enquiry Concerning Human Understanding, in L.A. Selby-Bigge (ed.), 3rd edition by P.H. Nidditch, 1975, Oxford: Oxford University Press. Jackson, Frank (1977), Perception: A Representative Theory, Cambridge: Cambridge University Press. Jacoby, L.L. and Whitehouse, K. (1989), “An Illusion of Memory: False Recognition Influenced by Unconscious Perception,” Journal of Experimental Psychology: General 118, pp. 126–135. James, William (1890), The Principles of Psychology, 2 volumes, London: Macmillan. Johnson, Marcia, Foley, M.A., Suengas, Aurora and Raye, Carol (1988), “Phenomenal Characteristics of Memory for Perceived and Imagined Autobiographical Events,” Journal of Experimental Psychology: General 117, pp. 371–376. Johnson, D.M. (1983), “Memory and Knowledge: The Epistemological Significance of Biology,” American Philosophical Quarterly 20, pp. 375–382. Johnson, Marcia and Raye, Carol (1981), “Reality Monitoring,” Psychological Review 88, pp. 67–85. Johnson, Marcia (1988), “Reality Monitoring: An Experimental Phenomenological Approach,” Journal of Experimental Psychology: General 117, pp. 390–394. Kaplan, David (1989), “Demonstratives,” in J. Almog, H. Wettstein and J. Perry (eds.), Themes from Kaplan, New York: Oxford University Press, pp. 481–563. Kim, Jaegwon (1973), “Causes and Counterfactuals,” Journal of Philosophy 70, pp. 570–572. Kneale, Martha (1972), “Our Knowledge of the Past and Future,” Proceedings of the Aristotelian Society 72, pp. 1–12.
Bibliography
181
Kripke, Saul A. (1980), Naming and Necessity, Cambridge, MA: Harvard University Press. Lackey, Jennifer (2005), “Memory as a Generative Epistemic Source,” Philosophy and Phenomenological Research 70, pp. 636–658. Laird, John (1920), A Study in Realism. Cambridge: Cambridge University Press. Laurence, Jean-Roch and Perry, Campbell (1983), “Hypnotically Created Memory Among Highly Hypnotizable Subjects,” Science 222, pp. 523–524. Lehrer, Keith and Paxson, Thomas D. (1969), “Knowledge: Undefeated True Justified Belief,” Journal of Philosophy 66, pp. 215–234. Lewis, C.I. (1946), An Analysis of Knowledge and Valuation, La Salle, IL: Open Court. Lewis, David (1972), “Psychological and Theoretical Identifications,” Australasian Journal of Philosophy 50, pp. 249–258. Lewis, David (1996), “Elusive Knowledge,” Australasian Journal of Philosophy 74, pp. 549–567. Lewis, Delmas (1983), “Dualism and the Causal Theory of Memory,” Philosophy and Phenomenological Research 44, pp. 21–30. Leyden, Wolfgang von (1961), Remembering: A Philosophical Problem, London: Duckworth. Locke, Don (1971), Memory, London: Macmillan. Locke, John (1694), An Essay Concerning Human Understanding, in P.H. Nidditch (ed.), 1975, Oxford: Clarendon. Loeb, Louis E. (1974), “Causal Theories and Causal Overdetermination,” Journal of Philosophy 71, pp. 525–544. Loftus, Elizabeth F. (1977), “How to Catch a Zebra in Semantic Memory,” in R. Shaw and J. Bransford (eds.), Perceiving, Acting and Knowing, Hillsdale, MI: Erlbaum, pp. 393–411. Lowe, Jonathan (1992), “Experience and its Objects,” in T. Crane (ed.), The Contents of Experience, Cambridge: Cambridge University Press, pp. 79–104. Lukasiewicz, Jan (1967), “Determinism,” in S. McCall (ed.), Polish Logic 1920–1939, Oxford: Oxford University Press, pp. 19–39. Mackie, John L. (1965), “Causes and Conditions,” American Philosophical Quarterly 2, pp. 245–264. Malcolm, Norman (1963), Knowledge and Certainty, Ithaca, NY: Cornell University Press. Malcolm, Norman (1976), “Memory as Direct Awareness of the Past,” in G. Vesey (ed.), Impressions of Empiricism, London: Macmillan, pp. 1–22. Malcolm, Norman (1977), Memory and Mind, Ithaca, NY: Cornell University Press. Margolis, Joseph (1977), “Remembering,” Mind 86, pp. 186–205. Martin, C.B. and Deutscher, Max (1966), “Remembering,” Philosophical Review 75, pp. 161–196. Mill, John Stuart (1843), A System of Logic, London: Longmans. Moore, G.E. (1959), “Four Forms of Scepticism,” in his Philosophical Papers, London: Muirhead Library, pp. 196–226. Munsat, Stanley (1967), The Concept of Memory, New York: Random House. Nagel, Ernest (1961), The Structure of Science, New York: Harcourt, Brace & World. Naylor, Andrew (1966), “On ‘Remembering’ an Unreal Past,” Analysis 26, pp. 122–128. Naylor, Andrew (1971), “B Remembers that P from Time T,” Journal of Philosophy 68, pp. 29–41. Neisser, Ulric (1967), Cognitive Psychology, New York: Meredith. Neisser, Ulric (1982), “John Dean’s Memory: A Case Study,” in his Memory Observed, San Francisco, CA: W.H. Freeman, pp. 139–162. Nigro, Georgia and Neisser, Ulric (1983), “Point of View in Personal Memories,” Cognitive Psychology 15, pp. 467–482. Nisbett, Richard and Ross, Lee (1980), Human Inference: Strategies and Shortcomings of Social Judgment, Englewood Cliffs, NJ: Prentice-Hall. Noonan, Harold W. (1989), Personal Identity, London: Routledge. Noxon, James (1976), “Remembering and Imagining the Past,” in D.W. Livingston and J.T. King (eds.), Hume: A Re-Evaluation, New York: Fordham University Press, pp. 270–295. O’Connor, D.J. and Carr, Brian (1982), Introduction to the Theory of Knowledge, Minneapolis, MN: University of Minnesota Press.
182
Bibliography
O’Shaughnessy, Brian (1980), The Will, 2 volumes, Cambridge: Cambridge University Press. Pappas, George (1980), “Lost Justification,” Midwest Studies in Philosophy 5, pp. 127–34. Pears, David (1975a), “Causation and Memory,” Philosophic Exchange 2, pp. 29–40. Pears, David (1975b), “Russell’s Theories of Memory,” in his Questions in the Philosophy of Mind, London: Duckworth, pp. 224–250. Pears, David (1990), Hume’s System: An Examination of the First Book of his Treatise, Oxford: Oxford University Press. Pincock, Christopher (2006), “Richard Semon and Russell’s Analysis of Memory,” Russell 26, pp. 101–125. Pitcher, George (1971), A Theory of Perception, Princeton, NJ: Princeton University Press. Plantinga, Alvin (1993), Warrant and Proper Function, New York: Oxford University Press. Pojman, Louis P. (1995), What Can We Know?, Belmont, CA: Wadsworth. Pollock, John L. (1974), Knowledge and Justification, Princeton, NJ: Princeton University Press. Pollock, John L. (1986), Contemporary Theories of Knowledge, Savage, MD: Rowman & Littlefield. Posner, Michael I. and Warren, Robert E. (1972), “Traces, Concepts, and Conscious Constructions,” in A.W. Melton and E. Martin (eds.), Coding Processes in Human Memory, Washington, DC: V.H. Winston, pp. 25–43. Price, H.H. (1932), Perception, London: Methuen. Price, H.H. (1936), “Memory-Knowledge,” Proceedings of the Aristotelian Society, suppl. vol. 15, pp. 16–33. Price, H.H. (1969), Thinking and Experience, 2nd edition, London: Hutchinson. Prior, Arthur (1970), “The Notion of the Present,” Studium Generale 23, pp. 245–248. Putnam, Hilary (1999), The Threefold Cord: Mind, Body and World, New York: Columbia University Press. Quine, W.V.O. (1969), “Natural Kinds,” in his Ontological Relativity and Other Essays, New York: Columbia University Press, pp. 114–138. Reason, James and Mycielska, Klara (1982), Absent Minded? The Psychology of Mental Lapses and Everyday Errors, Englewood Cliffs, NJ: Prentice-Hall. Reid, Thomas (1764), An Inquiry into the Human Mind, in D.R. Brookes (ed.), 1997, Edinburgh: University of Edinburgh Press. Reid, Thomas (1785), Essays on the Intellectual Powers of Man, in D.R. Brookes (ed.), 2002, Edinburgh: University of Edinburgh Press. Reid, Thomas (1895), Philosophical Works, with notes by Sir William Hamilton, 8th edition, 2 volumes, Edinburgh: MacLachlan and Stewart. Reprint Hildesheim: Georg Olms 1967. Robinson, Howard (1994), Perception, London: Routledge. Roediger, Henry L., Bergman, Erik T. and Meade, Michelle L. (2000), “Repeated Reproduction from Memory,” in A. Saito (ed.), Bartlett, Culture and Cognition, Guildford, UK: Psychology Press, pp. 115–134. Rose, Steven (1992), The Making of Memory, London: Bantam. Rosen, Deborah A. (1975), “An Argument for the Logical Notion of a Memory Trace,” Philosophy of Science 42, pp. 1–10. Rosenfield, Israel (1988), The Invention of Memory, New York: Basic Books. Ross, Michael and Buehler, Roger (1994), “Creative Remembering,” in U. Neisser and R. Fivush (eds.), The Remembering Self, Cambridge: Cambridge University Press, pp. 205–235. Rumelhart, David E. and Norman, Donald A. (1981), “A Comparison of Models,” in G.E. Hinton and J.A. Anderson (eds.), Parallel Models of Associative Memory, Hillsdale, MI: Erlbaum, pp. 1–7. Rundle, Bede (1986), “Memory and Causation,” Philosophical Investigations 9, pp. 302–307. Russell, Bertrand (1912a), “On the Notion of Cause,” in his Mysticism and Logic, London: Allen & Unwin, pp. 172–199. Russell, Bertrand (1912b), The Problems of Philosophy, Oxford: Home University Library. Russell, Bertrand (1921), The Analysis of Mind, intro. by T. Baldwin, 1995, London: Routledge. Russell, Bertrand (1927), An Outline of Philosophy, New York: W.W Norton.
Bibliography
183
Russell, Bertrand (1940), An Inquiry into Meaning and Truth, London: Allen & Unwin. Russell, Bertrand (1948), Human Knowledge: Its Scope and Limits, New York: Simon & Schuster. Ryle, Gilbert (1949), The Concept of Mind, London: Hutchinson. Saunders, John T. (1963), “Skepticism and Memory,” Philosophical Review 72, pp. 477–486. Saunders, John T. (1965a), “Professor Malcolm’s Definition of ‘Factual Memory’,” Theoria 31, pp. 282–288. Saunders, John T. (1965b), “Does All Memory Imply Factual Memory?,” Analysis 25, suppl., pp. 109–115. Semon, Richard (1909), Die mnemischen Empfindungen in ihren Beziehungen zu den Originalempfindungen, Leipzig, Germany: Engelmann. Schechtman, Marya (1990), “Personhood and Personal Identity,” Journal of Philosophy 87, pp. 71–92. Schechtman, Marya (1994), “The Truth about Memory,” Philosophical Psychology 7, pp. 3–18. Schechtman, Marya (1996), The Constitution of Selves, Ithaca, NY: Cornell University Press. Schumacher, John A. (1976), “Memory Unchained Again,” Analysis 36, pp. 101–104. Searle, John R. (1983), Intentionality, Cambridge: Cambridge University Press. Sellars, Wilfried (1963), Science, Perception and Reality, London: Routledge/Kegan Paul. Senor, Thomas (1993), “Internalistic Foundationalism and the Justification of Memory Belief,” Synthese 94, pp. 453–476. Sherouse, Mark A. (1979), Memory and Knowledge: An Examination of the Epistemic Theory of Memory, Dissertation, Ohio State University. Shoemaker, Sydney (1963), Self-Knowledge and Self-Identity, Ithaca, NY: Cornell University Press. Shoemaker, Sydney (1967), “Memory,” in P. Edwards (ed.), The Encyclopedia of Philosophy, Volume 5, New York: Macmillan, pp. 265–274. Shoemaker, Sydney (1970), “Persons and Their Pasts,” American Philosophical Quarterly 7, pp. 269–285. Shoemaker, Sydney (1975), “Phenomenal Similarity,” Critica 7, pp. 3–37. Shoemaker, Sydney (1990), “First-Person Access,” Philosophical Perspectives 4, pp. 187–214. Shope, Robert (1973), “Remembering, Knowledge, and Memory Traces,” Philosophy and Phenomenological Research 33, pp. 303–322. Siebel, Mark (2000), Erinnerung, Wahrnehmung, Wissen, Paderborn, Germany: Mentis. Singer, Marcus (1963), “Meaning, Memory, and the Moment of Creation,” Proceedings of the Aristotelian Society 63, pp. 187–202. Smart, J.J.C. (1959), “Sensations and Brain Processes,” Philosophical Review 68, pp. 169–176. Smith, Brian (1966), Memory, London: Allen & Unwin. Sosa, Ernest (1994), “Philosophical Scepticism and Epistemic Circularity,” Proceedings of the Aristotelian Society suppl. vol. 68, pp. 263–290. Spiegel, David (1995), “Hypnosis and Suggestion,” in D.L. Schacter (ed.), Memory Distortions: How Minds, Brains, and Societies Reconstruct the Past, Cambridge, MA: Harvard University Press, pp. 129–149. Squire, Larry R. (1987), Memory and Brain, New York: Oxford University Press. Squires, Roger (1969), “Memory Unchained,” Philosophical Review 78, pp. 178–196. Stalnaker, Robert (1968), “A Theory of Conditionals,” in N. Rescher (ed.), Studies in Logical Theory, Oxford: Blackwell, pp. 98–112. Strawson, Peter (1979), “Perception and its Objects,” in G.F. MacDonald (ed.), Perception and Identity, London: Macmillan, pp. 41–60. Stroud, Barry (1977), Hume, London: Routledge. Stroud, Barry (1989), “Understanding Human Knowledge in General,” in M. Clay and K. Lehrer (eds.), Knowledge and Skepticism, Boulder, CO: Westview, pp. 31–50. Stroud, Barry (1994), “Scepticism, ‘Externalism’, and the Goal of Epistemology,” Proceedings of the Aristotelian Society suppl. vol. 68, pp. 291–307.
184
Bibliography
Sutton, John (1998), Philosophy and Memory Traces: Descartes to Connectionism, Cambridge: Cambridge University Press. Sutton, John (2004), “Memory,” in E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Spring 2004 edition), http://plato.stanford.edu/entries/memory/ Szubka, Tadeusz (2002), “The Causal Theory of Perception and Direct Realism,” in J. Conant and U.M. Zeglen (eds.), Hilary Putnam: Pragmatism and Realism, London: Routledge, pp. 109–127. Traiger, Saul (1985), “Flage on Hume’s Account of Memory,” Hume Studies 11, pp. 166–172. Tulving, Endel (1972), “Episodic and Semantic Memory,” in E. Tulving and W. Donaldson (eds.), Organization of Memory, New York: Academic Press, pp. 381–403. Tulving, Endel (1983), Elements of Episodic Memory, Oxford: Clarendon. Urmson, J.O. (1967), “Memory and Imagination,” Mind 76, pp. 83–91. Urmson, J.O. (1971), “Memory and Imagination,” Mind 80, p. 607. Vendler, Zeno (1972), Res Cogitans: An Essay in Rational Psychology, Ithaca, NY: Cornell University Press. Vendler, Zeno (1980), “Telling the Facts,” in J.R. Searle, F. Kiefer and M. Bierwich (eds.), Speech Act Theory and Pragmatics, Dordrecht, The Netherlands: Reidel, pp. 273–290. Vision, Gerald (1993), “Animadversions on the Causal Theory of Perception,” Philosophical Quarterly 43, pp. 344–357. Vision, Gerald (1997), Problems of Vision: Rethinking the Causal Theory of Perception, New York: Oxford University Press. Waismann, Friedrich (1965), The Principles of Linguistic Philosophy, in R. Harre (ed.), New York: St. Martin’s Press. Warnock, Mary (1987), Memory, London: Faber & Faber. Warnock, G.J. (1965), “Seeing,” in R.J. Swartz (ed.), Perceiving, Sensing, and Knowing, Garden City, NY: Doubleday, pp. 49–67. White, Alan R. (1990), The Language of Imagination, Oxford: Blackwell. Wiggins, David (1967), Identity and Spacio-Temporal Continuity, Oxford: Blackwell. Wiggins, David (1980), Sameness and Substance, Cambridge: Cambridge University Press. Williams, Bernard (1973), Problems of the Self, Cambridge: Cambridge University Press. Williamson, Timothy (2000), Knowledge and its Limits, Oxford: Oxford University Press. Wittgenstein, Ludwig (1960), The Blue and Brown Books, Oxford: Blackwell. Wittgenstein, Ludwig (1967a), Zettel, in G.E.M. Anscombe and G.H. von Wright (eds.), Berkeley, CA: University of California Press. Wittgenstein, Ludwig (1967b), Philosophical Investigations, in G.E.M. Anscombe and R. Rhees (eds.), 3rd edition, Oxford: Blackwell. Wittgenstein, Ludwig (1969), On Certainty, in G.E.M. Anscombe and G.H. von Wright (eds.), New York: Harper & Row. Wittgenstein, Ludwig (1975), Philosophical Remarks, in R. Rhees (ed.), Oxford: Blackwell. Wittgenstein, Ludwig (1979), Wittgenstein Lectures, Cambridge 1932–1935, in A. Ambrose (ed.), Totowa, NJ: Rowman & Littlefield. Wittgenstein, Ludwig (1980), Remarks on the Philosophy of Psychology, in G.E.M. Anscombe and G.H. von Wright (eds.), 2 volumes, Oxford: Blackwell. Wittgenstein, Ludwig (1993), Philosophical Occasions: 1912–1951, in J.C. Klagge and A. Nordmann (eds.), Indianapolis, IN: Hackett. Woozley, A.D. (1949), Theory of Knowledge, London: Hutchinson University Library. Wright, Crispin (1987), “On Making Up One’s Mind: Wittgenstein on Intention,” in P. Weingartner and G. Schurz (eds.), Logic, Philosophy of Science and Epistemology: Proceedings of the 11th International Wittgenstein Symposium, Vienna: Holder-Pichler-Tempsky, pp. 391–404. Wright, Crispin (1992), Truth and Objectivity, Cambridge, MA: Harvard University Press. Wundt, Wilhelm (1896), Outline of Psychology, trans. by C.H. Judd, New York: G.E. Stechert, 1897. Zemach, E.M. (1968), “A Definition of Memory,” Mind 77, pp. 526–536.
Index
Page numbers in italics denotes a footnote A a priori knowability 41, 76, 78–9, 86 absent justification case 109–10 activity theory of causation 44 adaptive variation 146–51 function of 152–4 Alexander, Samuel 68 Alston, William 107, 108–9 ambiguous expressions 172 analytic approach and continuous justification 122–3, 126 and validation of ostensible memory 101–2 Annis, David 28 Anscombe, G.E.M. 17, 143 Aquinas 62 Aristotle 62 Armstrong, David 17, 18, 64 article substitution, principle of 159–60, 161 assassination-example 121–2 associative learning 7 attitude-identification 155, 165, 166–7 attitudinal similarity 165–7 attributor contextualism 171–2, 173 Audi, Robert 26, 90, 109–10, 112, 115–16 Augustine 145 Austin, J.L. 62, 72–3 autobiographical memory 22, 23, 152–3 Ayer, A.J. 33, 100
B Barclay, Craig 152–4 Barsalou, Laurence 149 Bartlett, Frederick 149–51 Remembering 145 Beauchamp, Tom L. 32
Benjamin, Aaron 91 Benjamin, B.S. 17 Bergson, Henri 68, 93 Berkeley 82 Bjork, Robert 91 BonJour, Laurence 82, 114–15 brain studying of processes of 39–40 brain-injured patients 39 Brandt, Richard 101 Brewer, William 153, 154 Broad, C.D. 44, 91–2 The Mind and Its Place in Nature 42–3 Broadbent, Donald 145 Burge, Tyler 122–3, 126 Butler, Joseph 23 Butler, Ronald 130
C Cacioppo, John T. et al 40 car accident example 25, 52–3 causal overdetermination 48, 52–3 causal reinforcement 52, 53–54, 57 causal supplementation 5, 52, 53, 57 causal theory of memory 1, 9–10, 17–29, 49 alternatives to 9–10, 26–9 arguments against 18–20 conditional argument 20–2 and direct realism 11, 74–6 idea behind 17 and memory traces 1, 9, 10, 31–46 personal identity argument 22–4 proponents of 17–18 temporal forgetting argument 24–6 causal theory of perception 74–5 185
186 causality contiguous 32–4, 42, 44 nomological character of 19 cause and effect 18–19, 32–4, 42, 44, 76–8 Child, William 78 classical computationalism 10, 36 closure principle 126–7, 131–2 clutter avoidance 152, 153 co-temporality objection and direct realism 11, 68–71 cognitive attitudes 166 cognitive dynamics 137, 147–8, 152 cognitive psychology and xerox model 145 cognitively spontaneous beliefs 114–15 Cohen, Stewart 170–1 coherentism 114–15 complete prompting 50, 51 conative attitudes 166 concepts 2 conceptual universalization, principle of 161, 162 condensation 137, 147, 148–9, 152, 153 Conee, Earl 114 conjunction elimination, principle of 160–1, 162 connectionist models of memory 10, 36–8 constitutionalism 96–7 constructivism 153–4 content addressable memory 37–8 content condition 8 contextualism 1, 13, 169–70 distinction between attributor and subject 171–2, 173 epistemic 170–2 and factivity constraint 1, 172–5 memorial 172–5 and skepticism 170 contiguous causation 32–4, 42, 44 continuous justification 113–26 analytic account 122–3, 126 dualistic theory of 122, 123, 124–6 and externalism 122–6 and internalism 117–22 principle of 116–17 and problem of forgotten evidence 113–17 process reliabilist account 122, 123, 126 contrastive stress, principle of 162–3 Cornman, James 102 counterfactuals 21–2, 48, 116, 126–7 similarity and conditionals of 157–8 cross-talk 38 cued recall 50
Index D de dicto memory 4, 159, 161 de re memory 4, 159, 161 Dean, John 164–5 DeCooke, Peggy 152–3 defeated justification 109, 110–11 defeaters 110–11, 123 doxastic 110, 111, 121 factual 110 normative 110, 113, 121 demonstratives 147 DeRose, Keith 171 Descartes, René 62 Deutscher, Max 6, 20–1, 28, 50, 51, 55–7, 66 diachronic content similarity 155–67 direct awareness 68, 69 direct realism 1, 10–12, 61–80 and causal theory 11, 74–6 co-temporality objection 11, 68–71 distinction between representative realism and 11, 62–8 hallucination argument 72–4 Humean argument 76–80 objections to 11 and perception 62–4, 69–70, 73, 74 dispositional memory 3, 24, 35, 36, 45–6 distinct existences 78–9 doxastic defeater 110, 111, 121 Dretske, Fred 111, 122, 124–6, 131 dualistic theory and continuous justification 122, 123, 124–6 Dummett, Michael 112
E eagle-example 111–12, 120–1 Earle, William 61, 68 Ebbinghaus, Hermann 20 eidetikers 6 Empiricists 145 empty concept 131, 132, 175 engrams see memory traces episodic memory 4 epistemic contextualism 170–2 epistemic deontology 84 epistemic externalism see externalism epistemic internalism see internalism epistemic theory of memory 8, 10, 26, 29, 111–12, 121 essentialism 79 eternalism 71 event memory 2, 3, 7, 35, 137 evidential retention condition 26–7, 28–9 evidentialism 114
Index exam-example 109–10, 111–12 existential generalization, principle of 160, 162 explicit memory 7 externalism 1, 11, 12, 76, 84, 85, 103–4, 105–33 and continuous justification 122–6 and justification-preservation 106, 113–22 and skepticism 85, 103–4, 106–9 see also internalism externalist reliabilism 105–9, 122 extrinsic properties of memory-data 86
F fact memory 2, 3, 5, 9, 26, 27, 35, 137 distinction between personal memory and 6–7 factivity of 137–40 kinds of 4 and memory-images 65 factivity constraint 12, 137–54, 169 contextualist interpretation of 1, 172–5 motivation of the truth condition 137–40 truth in meta-representational memory 140, 142–4, 164 truth in reproductive memory 140–2, 164 and xerox model see xerox model factual defeaters 110 familiarity as a memory marker 90–3 Feldman, Richard 114 first-person perspective 96, 115–16, 139 flat concept 131, 132, 175 Fodor, Jerry A. 74, 158, 167 forgetting 27, 143 forgotten evidence 113–17 formal criterion 87–8, 89 foundationalism 117 and continuous justification 117–20 free recall 50 functional magnetic resonance imaging (fMRI) 40 functionalism 158, 167 Furlong, E.J. 65, 72 gerund clauses 7
G Ginet, Carl 19–20 Goldman, Alvin 17, 114, 116, 122, 123 Goodman, Nelson 156–7 ‘Seven Strictures of Similarity’ 157 Gosse, Philip Henry 127 graceful degradation 37
187 H hallucination 11, 63, 72–4 Hamilton, William 69 Harman, Gilbert 114, 119, 153, 167 Harrod, Roy 99–100 Hawthorne, John 171 holistic functionalism 158 Holland, Alan 28 Holland, R.F. 84, 92, 94, 98 ‘honest liars’ 54 how-clauses 2 Huemer, Michael 120, 122, 125 Hume, David 33, 63, 64, 65, 85, 100, 145 and cause and effect 18, 19, 32, 76–7 Enquiry Concerning Human Understanding 32 and memory markers 87–9, 90 Treatise of Human Nature 32, 77, 88 Humean approach 76–80 hypnosis and suggestibility 54–5
I identity theory 41 imagination 11, 87, 88 differences between memory and 85, 87–90, 93–7 implicit memory 7 inattention 143 indexical expressions 172 indirect realism see representative realism induction 100 infallibility 63–4 inferential memory 4, 141 innate ideas 146 intentional aspect of memory traces 10, 35–6 intentions distinguishing between memory and imagination 93–6 internalism 11, 84, 86, 103, 105, 106, 107, 185 about continuous justification 117–22 and justification-acquisition 106, 109–13 and justification-preservation 106, 113–22 and problem of forgotten evidence 113–17 and skepticism 11, 84–5, 103, 105, 107 see also externalism internalist-foundationalist account and continuous justification 117–22 intrinsic properties of memory-data 86 introspection and memory traces 40–1 inus condition 48
188 J Jackson, Frank 64 James, William 65, 90, 153 justification-acquisition and internalism 106, 109–13 justification-preservation and internalism 106, 113–22
K Kaplan, David 147, 172 Kennedy assassination case 140–3 Kim, Jaegwon 22 Kneale, Martha 41 Kripke, Saul 79
L Lackey, Jennifer 121 Laird, John 67, 68 Laurence, Jean-Roch 55 laws of memory 20 learning and practice 20 Lehrer, Keith 119 Lepore, Ernest 158 Lewis, C.J. 98, 130 Lewis, David 171 Leyden, Wolfgang von 17, 143 library-detective example 119–20 life-stories 152 linguistic prompts 50, 51 Locke, Don 5–6, 26, 101, 130 Memory 26 Locke, John 62, 63, 65, 81, 82, 83, 145 Essay Concerning Human Understanding 146 Loftus, Elizabeth 145 long-term memory 7 Lukasiewicz, Jan 71
M Mackie, John 48 mad-dog reliabilism 111 Malcolm, Norman 7, 22, 41, 69, 101, 102, 127, 129–30, 131 ‘A Definition of Factual Memory’ 18–19 Martin, Charles 6, 20–1, 50, 51, 55–7, 66 memorial contextualism 172–5 memory as an absolute concept 131, 175 analysis of 7–9 classification of 2–7 as a context-dependent concept 13
Index differences between imagination and 85, 87–90, 93–7 function of 13, 161, 174 psychological taxonomy of 7 reconstructive nature of 37 memory causation approaches to 47–9 counterfactual analysis 48 necessary/sufficient conditions analysis 47–8 probabilistic analysis of 49 memory claims, rectifying 164–5 memory-data 7–8, 66–7, 68, 75, 79, 83, 84, 86, 97–8 memory-images 6, 65–6, 90–2 memory markers 11, 67, 84, 85–6 and familiarity 90–3 and formal criterion 87–8, 89 Humean theory of 87–9, 90 and phenomenal criterion 88–9 Smith on 89–90 Urmson theory of 93–6 varieties of 87–93 and vivacity 88–9, 92 memory traces 1, 9, 10, 31–46, 49, 153 computational aspect 10, 36–8 cooperation between retrieval cues and 49, 50–1, 52–5, 57 intentional aspect of 10, 35–6 and introspection 40–1 as knowable a priori 41 and mnemic causation 39, 42–6 nature of 10, 34–8 and neurobiology 39–40 Russell’s critique of 38–9 verifiability of 38–41 meta-representational memory 8, 9, 12, 137 attitude-identification 165, 166–7 distinction between reproductive memory and 4–5 preliminary analysis of 8 truth in 140, 142–4, 164 metacognitive signals 91 Mill, John Stuart 63 mind concept of as immaterial substance 35 mistaken self-identification 143 mnemic causation 10, 31, 39, 42–6 Moore, G.E. 138 motion and similarity 157 Munsat, Stanley 94
N Nagel, Ernest 33 naive direct realism 62
Index narrative self-definition 152–3 Naylor, Andrew 28–9 necessary/sufficient conditions 47–8, 49 negation elimination, principle of 162 negative memories 8, 141 Neisser, Ulric 153, 154, 164 neurobiology and memory traces 39–40 neurology 37, 54 Nixon, Richard 164 non-associative learning 7 non-inferentiality 4, 63, 64, 68 Norman, Donald 37 normative defeaters 110, 113, 121
O object memory 2, 3, 7, 35, 137 occurrent memory 3 omission, principle of 161, 162 operative conditions 55–7 Oscar example 51, 52–3, 138–9 ostensible memories 3, 5, 11–12, 72, 74, 78, 85, 86, 123, 127, 138 differences between originating from imagination and from perception 90 validation of 97–104, 115
P Pappas, George 117–18 pattern completion 37 Paxson, Thomas 119 Pears, David 51, 54, 85, 92 Penfield, Wilder 39–40 perception 139 disjunctive theories of 65 distinction between representative and direct realism about 62–3, 69–70, 73, 74–5, 81–3 and hallucination 72–3 perceptual memories 6 Perry, Campbell 55 persistence model of memory 144–5, 146 personal identity 22–4 and causal theory of memory 22–4 psychological continuity theory of 23 relation between memory and 9 personal memory 5–6, 65 distinction between factual memory and 6–7 phenomenal criterion 88 phenomenal principle 73 Pincock, Christopher 39 Pitcher, George 69 Plantinga, Alvin 112
189 Pollock, John 6–7, 28, 117, 118–19, 120, 122 positron emission topography (PET) 40 Posner, Michael 145 postulational approach and validation of ostensible memory 100, 101 practical memory 2, 5 pragmatic dimension of memory 169–75 pragmatic justification 100–1 preemptive causation 52, 53 presentism 70–1 preservative memory 122–3 previous representation condition 8, 141 Price, H.H. 64, 71, 73, 98 probabilistic causation 49 process reliabilist account 122, 123, 126 prompts see retrieval cues property memory 2, 3, 7, 35, 137 propositional memory see fact memory Proust, Marcel Remembrance of Things Past 50 pure retention condition 9–10, 26–8 Putnam, Hilary 74–5 The Threefold Cord 74
Q quantum mechanics 33 quasi-memory 22, 23 quasi-retention 23–4 Quine, W.V.O. 156–7 re-learning counterexamples 29 rectifying of memory claims 164–5 Reid, Thomas 67–8, 69, 82 Essay on the Intellectual Powers of Man 146
R relationally absolute concepts 175 relevant alternative account of knowledge 131–3 reliabilism 105–9, 122 representative (indirect) realism 10–11, 61–8, 81–104 distinction between direct realism and 11, 62–8 and Humean distinct-existences requirement 79 and memory 65–7, 68 and perception 62–4 and skepticism about memory knowledge 11, 81, 83–6, 103 validation of ostensible memory 97–104 reproduction model and xerox model 144, 146
190 reproductive memory 12, 137, 164 distinction between meta-representational memory and 4–5 preliminary analysis of 7–8 presupposition of personal identity 9 previous representation condition 8, 141 truth in 140–2, 164 retrieval cues (prompts) 10, 47–8, 52–5 cooperation between memory traces and 49, 50–1, 52–5, 57 distinction between strict and complete prompting 50–1 kinds of 50 and operative conditions 55–7 Robinson, Howard 70, 75 Rosen, Deborah 49 Rosenberg, Alexander 32 Rumelhart, David 37 Rundle, Bede 47 Russell, Bertrand 6, 68, 100, 101, 103 The Analysis of Mind 10, 31, 38–9, 92 critique of memory traces 38–9 ‘On the Notion of Cause’ 33–4 and representative realism 65, 66, 83, 98 skeptical hypothesis 83, 127–31, 132, 133 theory of memory markers 90–1, 92, 93 theory of mnemic causation 42–6 Ryle, Gilbert The Concept of Memory 27
S Saunders, John T. 100 Schechtman, Marya 152, 165 schemas 149 schematic processing 137, 147, 149–51, 152 scientific direct realism 62 Searle, John 67, 79, 166 self-deception 143 Sellars, Wilfried 64 semantic memory 4 semantic similarity 155, 158–63 article substitution principle 159–60, 161 conceptual universalization principle 161, 162 conjunction elimination principle 160–1 contrastive stress principle 162–3 existential guidelines principle 160, 162 negation elimination principle 162 omission principle 161, 162 synonym substitution principle 163 tense adjustment principle 159, 162 universal instantiation principle 160 Semon, Richard 38 sense-data 62–4, 65, 66, 72, 73, 75, 82
Index sensory cues 50, 51 sensory experiences 139–40 Sherouse, Mark 29 Shoemaker, Sydney 8, 9, 17, 21, 28, 101–2, 103, 116–17, 130, 167 ‘Persons and their Past’ 22–3 Shope, Robert 111 short-term (working) memory 7 Siebel, Mark 36 similarity 155 attitudinal 165–7 and motion 157 and philosophy 157–8 relativity of 156–8 semantic 155, 158–63 simultaneity paradox 33, 34 Singer, Marcus 128, 130 skeptical hypothesis 83, 127–31, 132, 133 skepticism 11, 84–6, 100, 126–7, 130–1 and closure principle 126–7 and contextualism 170 and externalism 85, 103–4, 106–9 and internalism 11, 84–5, 103, 105, 107 and memory markers 84, 85 and metajustification 106–9 and representative realism 11, 81, 83–6, 103 validation of ostensible memory 97–104 smell as retrieval cue 50 Smith, Brian 89–90 Sosa, Ernest 109 Spiegel, David 54–5 Squires, Roger ‘Memory Unchained’ 24–6 Stalnaker, Robert 158 Steup, Matthias 116 storehouse model of memory 145, 146 Strawson, Peter 75, 78, 79 strict prompting 50–1 Stroud, Barry 107–9, 109 subdoxastic states 36 subject contextualism 171–2, 173 subjunctive conditionals 105 substantive memory 122 suggestibility 54–5 Sutton, John 61 synonym substitution, principle of 163 Szubka, Tadeusz 74
T tastes as retrieval cues 50 temporal forgetting 24–6 temporal and spacial contiguity 32–3
Index tense adjustment, principle of 159, 162 trace hypothesis see memory traces truth in meta-representational memory 140, 142–4, 164 in reproductive memory 140–2, 164 truth conditions 8, 12–13, 162, 172, 173–4 motivation of 137–40 Tulving, Endel 4
U undercutting defeater 118 universal instantiation, principle of 160, 161–2 Urmson, J.O. 93–6, 97
V Vendler, Zeno 138 verbal reminders 50 veridical memories 3, 85, 86 verification 127–8, 129 verificationism 129 Vision, Gerald 78 visual prompts 50, 51 vivacity 88–9, 92
W Waismann, Friedrich 129 War of the Ghosts, The 149–51
191 Warnock, Mary 92 Watergate 164–5 wh-clauses 2, 138, 139 Williamson, Timothy 171 witness condition see previous representation condition Wittgenstein, Ludwig 43–4, 92, 96, 128–9 Brown Book 92–3 On Certainty 122 Woozley, A.D. 68, 92 working memory (short-term) 7 Wright, Crispin 96 Wundt, Wilhelm 153
X xerox model of memory 9, 12, 137, 142, 144–6, 148, 154 and attributing memory 137, 146–51 and cognitive psychology 145 features of 145 incompleteness of 152–3, 154 persistence model 144–5 and philosophy 145
Y Yourgrau, Palle 122, 124–6
Z Zemach, E.M. 27
PHILOSOPHICAL STUDIES SERIES 1. 2. 3.
4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28.
Jay F. Rosenberg: Linguistic Representation. 1974 ISBN 90-277-0533-X Wilfrid Sellars: Essays in Philosophy and Its History. 1974 ISBN 90-277-0526-7 Dickinson S. Miller: Philosophical Analysis and Human Welfare. Selected Essays and Chapters from Six Decades. Edited with an Introduction by Lloyd D. Easton. 1975 ISBN 90-277-0566-6 Keith Lehrer (ed.): Analysis and Metaphysics. Essays in Honor of R. M. Chisholm. 1975 ISBN 90-277-0571-2 Carl Ginet: Knowledge, Perception, and Memory. 1975 ISBN 90-277-0574-7 Peter H. Hare and Edward H. Madden: Causing, Perceiving and Believing. An Examination of the Philosophy of C. J. Ducasse. 1975 ISBN 90-277-0563-1 Hector-Neri Castañeda: Thinking and Doing. The Philosophical Foundations of Institutions. 1975 ISBN 90-277-0610-7 John L. Pollock: Subjunctive Reasoning. 1976 ISBN 90-277-0701-4 Bruce Aune: Reason and Action. 1977 ISBN 90-277-0805-3 George Schlesinger: Religion and Scientific Method. 1977 ISBN 90-277-0815-0 Yirmiahu Yovel (ed.): Philosophy of History and Action. Papers presented at the First Jerusalem Philosophical Encounter (December 1974). 1978 ISBN 90-277-0890-8 Joseph C. Pitt (ed.): The Philosophy of Wilfrid Sellars: Queries and Extensions. 1978 ISBN 90-277-0903-3 Alvin I. Goldman and Jaegwon Kim (eds.): Values and Morals. Essays in Honor of William Frankena, Charles Stevenson, and Richard Brandt. 1978 ISBN 90-277-0914-9 Michael J. Loux: Substance and Attribute. A Study in Ontology. 1978 ISBN 90-277-0926-2 Ernest Sosa (ed.): The Philosophy of Nicholas Rescher. Discussion and Replies. 1979 ISBN 90-277-0962-9 Jeffrie G. Murphy: Retribution, Justice, and Therapy. Essays in the Philosophy of Law. 1979 ISBN 90-277-0998-X George S. Pappas (ed.): Justification and Knowledge. New Studies in Epistemology. 1979 ISBN 90-277-1023-6 James W. Cornman: Skepticism, Justification, and Explanation. With a Bibliographic Essay by Walter N. Gregory. 1980 ISBN 90-277-1041-4 Peter van Inwagen (ed.): Time and Cause. Essays presented to Richard Taylor. 1980 ISBN 90-277-1048-1 Donald Nute: Topics in Conditional Logic. 1980 ISBN 90-277-1049-X Risto Hilpinen (ed.): Rationality in Science. Studies in the Foundations of Science and Ethics. 1980 ISBN 90-277-1112-7 Georges Dicker: Perceptual Knowledge. An Analytical and Historical Study. 1980 ISBN 90-277-1130-5 Jay F. Rosenberg: One World and Our Knowledge of It. The Problematic of Realismin PostKantian Perspective. 1980 ISBN 90-277-1136-4 Keith Lehrer and Carl Wagner: Rational Consensus in Science and Society. A Philosophical and Mathematical Study. 1981 ISBN 90-277-1306-5 David O’Connor: The Metaphysics of G. E. Moore. 1982 ISBN 90-277-1352-9 John D. Hodson: The Ethics of Legal Coercion. 1983 ISBN 90-277-1494-0 Robert J. Richman: God, Free Will, and Morality. Prolegomena to a Theory of Practical Reasoning. 1983 ISBN 90-277-1548-3 Terence Penelhum: God and Skepticism. A Study in Skepticism and Fideism. 1983 ISBN 90-277-1550-5
PHILOSOPHICAL STUDIES SERIES 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.
45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56.
James Bogen and James E. McGuire (eds.): How Things Are. Studies in Predication and the History of Philosophy of Science. 1985 ISBN 90-277-1583-1 Clement Dore: Theism. 1984 ISBN 90-277-1683-8 Thomas L. Carson: The Status of Morality. 1984 ISBN 90-277-1619-9 Michael J. White: Agency and Integrality. Philosophical Themes in the Ancient Discussions of Determinism and Responsibility. 1985 ISBN 90-277-1968-3 Donald F. Gustafson: Intention and Agency. 1986 ISBN 90-277-2009-6 Paul K. Moser: Empirical Justification. 1985 ISBN 90-277-2041-X Fred Feldman: Doing the Best We Can. An Essay in Informal Deontic Logic. 1986 ISBN 90-277-2164-5 G. W. Fitch: Naming and Believing. 1987 ISBN 90-277-2349-4 Terry Penner: The Ascent from Nominalism. Some Existence Arguments in Plato’s Middle Dialogues. 1987 ISBN 90-277-2427-X Robert G. Meyers: The Likelihood of Knowledge. 1988 ISBN 90-277-2671-X David F. Austin (ed.): Philosophical Analysis. A Defense by Example. 1988 ISBN 90-277-2674-4 Stuart Silvers (ed.): Rerepresentation. Essays in the Philosophy of Mental Representation. 1988 ISBN 0-7923-0045-9 Michael P. Levine: Hume and the Problem of Miracles. A Solution. 1989 ISBN 0-7923-0043-2 Melvin Dalgarno and Eric Matthews (eds.): The Philosophy of Thomas Reid. 1989 ISBN 0-7923-0190-0 Kenneth R. Westphal: Hegel’s Epistemological Realism. A Study of the Aim and Method of Hegel’s Phenomenology of Spirit. 1989 ISBN 0-7923-0193-5 John W. Bender (ed.): The Current State of the Coherence Theory. Critical Essays on the Epistemic Theories of Keith Lehrer and Laurence BonJour, with Replies. 1989 ISBN 0-7923-0220-6 Roger D. Gallie: Thomas Reid and ‘The Way of Ideas’. 1989 ISBN 0-7923-0390-3 J-C. Smith (ed.): Historical Foundations of Cognitive Science. 1990 ISBN 0-7923-0451-9 John Heil (ed.): Cause, Mind, and Reality. Essays Honoring C. B. Martin. 1989 ISBN 0-7923-0462-4 Michael D. Roth and Glenn Ross (eds.):Doubting. Contemporary Perspectives on Skepticism. 1990 ISBN 0-7923-0576-0 Rod Bertolet: What is Said. A Theory of Indirect Speech Reports. 1990 ISBN 0-7923-0792-5 Bruce Russell (ed.): Freedom, Rights and Pornography. A Collection of Papers by Fred R. Berger. 1991 ISBN 0-7923-1034-9 Kevin Mulligan (ed.): Language, Truth and Ontology. 1992 ISBN 0-7923-1509-X Jesús Ezquerro and Jesús M. Larrazabal (eds.): Cognition, Semantics and Philosophy. Proceedings of the First International Colloquium on Cognitive Science. 1992 ISBN 0-7923-1538-3 O.H. Green: The Emotions. A Philosophical Theory. 1992 ISBN 0-7923-1549-9 Jeffrie G. Murphy: Retribution Reconsidered. More Essays in the Philosophy of Law. 1992 ISBN 0-7923-1815-3 Phillip Montague: In the Interests of Others. An Essay in Moral Philosophy. 1992 ISBN 0-7923-1856-0 Jacques-Paul Dubucs (ed.): Philosophy of Probability. 1993 ISBN 0-7923-2385-8
PHILOSOPHICAL STUDIES SERIES 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69.
70. 71. 72.
73. 74. 75. 76. 77. 78. 79. 80. 81. 82.
Gary S. Rosenkrantz: Haecceity. An Ontological Essay. 1993 ISBN 0-7923-2438-2 Charles Landesman: The Eye and the Mind. Reflections on Perception and the Problem of Knowledge. 1994 ISBN 0-7923-2586-9 Paul Weingartner (ed.): Scientific and Religious Belief. 1994 ISBN 0-7923-2595-8 Michaelis Michael and John O’Leary-Hawthorne (eds.): Philosophy in Mind. The Place of Philosophy in the Study of Mind. 1994 ISBN 0-7923-3143-5 William H. Shaw: Moore on Right and Wrong. The Normative Ethics of G.E. Moore. 1995 ISBN 0-7923-3223-7 T.A. Blackson: Inquiry, Forms, and Substances. A Study in Plato’s Metaphysics and Epistemology. 1995 ISBN 0-7923-3275-X Debra Nails: Agora, Academy, and the Conduct of Philosophy. 1995 ISBN 0-7923-3543-0 Warren Shibles: Emotion in Aesthetics. 1995 ISBN 0-7923-3618-6 John Biro and Petr Kotatko (eds.): Frege: Sense and Reference One Hundred Years Later. 1995 ISBN 0-7923-3795-6 Mary Gore Forrester: Persons, Animals, and Fetuses. An Essay in Practical Ethics. 1996 ISBN 0-7923-3918-5 K. Lehrer, B.J. Lum, B.A. Slichta and N.D. Smith (eds.): Knowledge, Teaching and Wisdom. 1996 ISBN 0-7923-3980-0 Herbert Granger: Aristotle’s Idea of the Soul. 1996 ISBN 0-7923-4033-7 Andy Clark, Jesus Ezquerro and Jesús M. Larrazabal (eds.): Philosophy and Cognitive Science: Categories, Consciousness, and Reasoning. Proceedings of the Second International national Colloquium on Cogitive Science. 1996 ISBN 0-7923-4068-X J. Mendola: Human Thought. 1997 ISBN 0-7923-4401-4 J. Wright: Realism and Explanatory Priority. 1997 ISBN 0-7923-4484-7 X. Arrazola, K. Kortaand F.J. Pelletier (eds.): Discourse, Interaction and Communicationion. Proceedings of the Fourth International Colloquium on Cognitive Science. 1998 ISBN 0-7923-4952-0 E. Morscher, O. Neumaier and P. Simons (eds.): Applied Ethics in a Troubled World. 1998 ISBN0-7923-4965-2 R.O. Savage: Real Alternatives, Leibniz’s Metaphysics of Choice. 1998 ISBN 0-7923-5057-X Q. Gibson: The Existence Principle. 1998 ISBN 0-7923-5188-6 F. Orilia and W.J. Rapaport (eds.): Thought, Language, and Ontology. 1998 ISBN 0-7923-5197-5 J. Bransen and S.E. Cuypers (eds.): Human Action, Deliberation and Causation. 1998 ISBN 0-7923-5204-1 R.D. Gallie: Thomas Reid: Ethics, Aesthetics and the Anatomy of the Self. 1998 ISBN 0-7923-5241-6 K. Korta, E. Sosa and X. Arrazola (eds.): Cognition, Agency and Rationality. Proceedings of the Fifth International Colloquium on Cognitive Science. 1999 ISBN 0-7923-5973-9 M. Paul: Success in Referential Communication. 1999 ISBN 0-7923-5974-7 E. Fischer: Linguistic Creativity. Exercises in ‘Philosophical Therapy’. 2000 ISBN 0-7923-6124-5 R. Tuomela: Cooperation. A Philosophical Study. 2000 ISBN 0-7923-6201-2
PHILOSOPHICAL STUDIES SERIES 83. 84. 85. 86. 87. 88. 89. 90. 91. 92.
93. 94. 95. 96. 97. 98. 99.
100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111.
P. Engel (ed.): Believing and Accepting. 2000 ISBN 0-7923-6238-1 W.L. Craig: Time and the Metaphysics of Relativity. 2000 ISBN 0-7923-6668-9 D.A. Habibi: John Stuart Mill and the Ethic of Human Growth. 2001 ISBN 0-7923-6854-1 M. Slors: The Diachronic Mind. An Essay on Personal Identity, Psychological Continuity and the Mind-Body Problem. 2001 ISBN 0-7923-6978-5 L.N. Oaklander (ed.): The Importance of Time. Proceedings of the Philosophy of Time Society, 1995–2000. 2001 ISBN 1-4020-0062-6 M. Watkins: Rediscovering Colors. A Study in Pollyanna Realism. 2002 ISBN 1-4020-0737-X W.F. Vallicella: A Paradigm Theory of Existence. Onto–Theology Vindicated. 2002 ISBN 1-4020-0887-2 M. Hulswit: From Cause to Causation. A Peircean Perspective. 2002 ISBN 1-4020-0976-3; Pb 1-4020-0977-1 D. Jacquette (ed.): Philosophy, Psychology, and Psychologism. Critical and Historical Readings on the Psychological Turn in Philosophy. 2003 ISBN 1-4020-1337-X G. Preyer, G. Peterand, M. Ulkan (eds.): Concepts of Meaning. Framing an Integrated Theory of Linguistic Behavior. 2003 ISBN 1-4020-1329-9 W. de Muijnck: Dependencies, Connections, and Other Relations. A Theory of Mental Causation. 2003 ISBN 1-4020-1391-4 N. Milkov: A Hundred Years of English Philosophy. 2003 ISBN 1-4020-1432-5 E.J. Olsson (ed.): The Epistomology of Keith Lehrer. 2003 ISBN 1-4020-1605-0 D.S. Clarke: Sign Levels.Language and Its Evolutionary Antecedents. 2003 ISBN 1-4020-1650-6 A. Meirav: Wholes, Sums and Unities. 2003 ISBN 1-4020-1660-3 C.H. Conn: Locke on Esence and Identity. 2003 ISBN 1-4020-1670-0 J.M. Larrazabal and L.A. Pérez Miranda (eds.): Language, Knowledge, and Representation. Proceedings of the Sixth International Colloquium on Cognitive Science (ICCS-99). 2004 ISBN 1-4020-2057-0 P. Ziff: Moralities. A Diachronic Evolution Approach. 2004 ISBN 1-4020-1891-6 J.A. Corlett: Terrorism: A Philosophical Analysis. 2003 ISBN 1-4020-1694-8; Pb 1-4020-1695-6 K. Korta and J.M. Larrazabal (eds.): Truth, Rationality, Cognition, and Music. Proceedings of the Seventh International Colloquiumon Cognitive Science. 2004 ISBN 1-4020-1912-2 T.M. Crisp, M. Davidson and D. Vander Laan (eds.): Knowledge and Reality. 2006 ISBN 1-4020-4732-0 S.E. Boër: Thought-Contents. On the Ontology of Belief and the Semantics of Belief Attribution. 2006 ISBN 1-4020-5084-4 A. Voltolini: How Ficta Follow Fiction. A Syncretistic Account of Fictional Entities. 2006 ISBN 1-4020-5146-8 K.L. Miller: Issues in Theoretical Diversity. Persistence, Composition, and Time. 2006 ISBN 1-4020-5255-3 M. Kieran and D.M. Lopes (eds.): Knowing Art. Essays in Aesthetics and Epistemology. 2006 /2007 ISBN 978-1-4020-5264-4; Pb 978-1-4020-6785-3 M. Murray: The Moral Wager. Evolution and Contract. 2007 ISBN 978-1-4020-5854-7 P. Saka: How to Think about Meaning. 2007 ISBN 978-1-4020-5856-1 D.K. Chan: Moral Psychology: Essays on Values, Rational Choice, and the Will. 2008 ISBN 978-1-4020-6871-3 S. Bernecker: The Metaphysics of Memory. 2008 ISBN 978-1-4020-8219-1