Declarative and Procedural Determinants of Second Languages (Studies in Bilingualism)

Declarative and Procedural Determinants of Second Languages

Michel Paradis

John Benjamins Publishing Company

Declarative and Procedural Determinants of Second Languages

Studies in Bilingualism (SiBil) The focus of this series is on psycholinguistic and sociolinguistic aspects of bilingualism. This entails topics such as childhood bilingualism, psychological models of bilingual language users, language contact and bilingualism, maintenance and shift of minority languages, and sociopolitical aspects of bilingualism.

Editors Kees de Bot

University of Groningen

Dalila Ayoun

University of Arizona

Editorial Board Michael Clyne

University of Melbourne

Kathryn A. Davis

University of Hawaii at Manoa

Joshua A. Fishman Yeshiva University

Francois Grosjean

Université de Neuchâtel

Thom Huebner

San José State University

Georges Luedi

University of Basel

Christina Bratt Paulston University of Pittsburgh

Suzanne Romaine

Merton College, Oxford

Merrill Swain

Ontario Institute for Studies in Education

G. Richard Tucker

Carnegie Mellon University

Wolfgang Klein

Max Planck Institut für Psycholinguistik

Volume 40 Declarative and Procedural Determinants of Second Languages by Michel Paradis

Declarative and Procedural Determinants of Second Languages Michel Paradis McGill University & Cognitive Neuroscience Center, UQÀM

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Paradis, Michel. Declarative and procedural determinants of second languages / Michel Paradis.        p. cm. (Studies in Bilingualism, issn 0928-1533 ; v. 40) Includes bibliographical references and index. 1.  Bilingualism--Psychological aspects. 2.  Bilingualism--Physiological aspects. 3.  Second language acquisition. 4.  Neurolinguistics.  I. Title. P115.4.P38   2009 404'.2019--dc22 2008046405 isbn 978 90 272 4176 4 (Hb; alk. paper); isbn 978 90 272 4177 1 (Pb; alk. paper)

© 2009 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Preface

ix

chapter 1 Key concepts, framework, and clarifications 1 1. Definition of key concepts  1 1.1 Definitions of “implicit”  3 1.2 Automaticity  5 1.3 Proficiency, accuracy, fluency, and other measures  6 2. Clarifications about the framework  9 2.1 Serendipity or the birth of the application of the declarative/procedural distinction to language representation and processing  9 2.2 Declarative/procedural models  12 2.3 Why vocabulary and lexicon differ  16 2.4 Degree of availability of procedural memory  22 2.5 There is no continuum from automatic to controlled processing  26 2.6 The content of metalinguistic knowledge and implicit competence  28 2.7 Interference, variability, and other indicators of explicitness  30 2.8 Macro-anatomical and micro-anatomical levels of representation  32 3. Conclusion  34 chapter 2 Consciousness in L2 appropriation 1. Only specific types of representations can become conscious – others cannot  41 1.1 Only a subset of explicit representations is active at any given time  45 1.2 The threshold of consciousness  45 1.3 Consciousness of input and output but not of implicit processes in between  47 1.4 Consciousness and working memory  49

37

 Declarative and procedural determinants of second languages

2. 3. 4. 5.

Perception, attention and noticing  50 2.1 Attention in second language acquisition and learning  51 Explicit input is not implicit intake  53 3.1 The double implicitness of intake  56 Neurobiological and neurochemical bases of consciousness  58 Conclusion  59

chapter 3 The disintegration of the explicit/implicit interface debate (or interface newspeak?) 1. The meaning of interface  61 1.1 The premises: Learning and acquisition are distinct; explicit knowledge is not transformed into implicit competence  64 2. The so-called “dynamic interface” is no interface  65 2.1 No interface but switching from one to the other  68 3. Consciousness cannot possibly be the interface  71 4. An indirect influence is not an interface  76 5. None of the proposed characterizations are compatible with an interface  79 6. Illusory and untenable would-be evidence  82 6.1 From seeds to trees  83 6.2 Tuning  84 6.3 Proceduralization  85 6.4 Ambiguities  87 6.5 Inapplicable analogies and metaphors  88 7. Description of explicit phenomena contributing to metalinguistic knowledge  93 8. Why adults should need explicit metalinguistic knowledge  96 9. Indirect influence of metalinguistic knowledge on acquisition not denied  97 10. How explicit knowledge benefits implicit acquisition – indirectly  99 11. The contexts of learning and acquisition  101 12. Conclusion  103 chapter 4 Ultimate attainment in L2 proficiency 1. Ultimate attainment in L1 and L2  110 2. The optimal period   113

61

109



3. 4.

5. 6. 7.

8.

Table of contents 

Optimal window of opportunity  114 The optimal period is restricted to implicit linguistic competence  117 4.1 Inter-individual variability in attainment  118 4.2 The impact of working memory and level of education  120 4.3 The success in semantics relative to syntax and phonology  121 4.4 The decline in L2 performance with increasing age  122 4.5 The ease of appropriation and use of L1 vs. L2  123 4.6 You don’t learn L2 the way you acquired L1, do you? How come?  124 Optimal period and the right hemisphere  126 Evidence adduced against a critical period  128 Factors invoked in lieu of a neurobiological critical period to account for poor performance in L2 are actually the consequences of an optimal period  129 7.1 Effects due to age are a consequence of brain processes  130 7.2 Native language entrenchment  133 Conclusion   134

chapter 5 The pervasive relevance of the distinction between implicit competence and explicit knowledge 1. Implications of the declarative/procedural distinction for laterality studies  137 2. Implications of the declarative/procedural distinction for imaging studies  140 2.1 Words of caution about the interpretation of neuroimaging studies  141 2.2 Consequences of not distinguishing word studies from sentence studies  143 2.3 The nature of the additional cortical resources reported to be recruited for L2  151 3. Procedural and declarative language switching and mixing  155 3.1 Types of switches and consequences  155 3.2 Switching data from neuroimaging studies  157 3.3 Switching data from clinical studies  160 3.4 Conscious and automatic control mechanisms in language switching  163 4. Data from clinical studies  169 4.1 Data from bilingual neuropsychiatric disorders  169 4.2 Data from bilingual aphasia  171 4.3 Data from other cerebral accidents/conditions  174

137

 Declarative and procedural determinants of second languages

5. 6. 7.

8.

 e declarative/procedural distinction and the subsystems hypothesis  177 Th Declarative and procedural translation strategies  180 Further indications of declarative/procedural relevance  182 7.1 Variability in appropriation in L2 vs. systematicity in L1  183 7.2 L2 accent changes faster than L1 accent when speakers relocate to an area where a different variety is spoken  183 7.3 Additional evidence for L1 implicit procedural memory and L2 explicit declarative memory  184 Conclusion  184

Summary of key proposals

187

References

191

Subject index

217

Preface

Rather than confidently aiming at absolute truth, scientific research strives to reduce ignorance. Sustainable knowledge is backed by healthy skepticism and constant willingness to critically reconsider even the best entrenched assumptions.  Paul Bouissac, Semiotix, 9: May 2007

In Paradis (2004), I proposed a neurolinguistic theory of bilingualism. It integrates a number of hypotheses – namely the three-store hypothesis, the direct access hypothesis, the activation threshold hypothesis, and the subsystems hypothesis – within the framework of a neurocognitive megasystem that comprises a number of independent neurofunctional systems that collaborate in the representation and processing of verbal communication. These independent systems include a common conceptual system; motivation/affect; and, for each language, implicit linguistic competence, explicit metalinguistic knowledge, and linguistic pragmatics. The present volume takes matters up where the previous one left off. Except for a brief recapitulation in this preface of the main relevant points, none of the contents of the earlier book are repeated. Readers are referred to the latter for further background information in places where it could be of advantage. Updated information on the various topics that were covered and additional evidence for the proposed theoretical constructs, including evidence from previously unexplored research domains, are provided, and new issues that have emerged since are discussed, as they relate to the framework of declarative and procedural memory. The current volume will explore further implications of these constructs for the appropriation, representation and processing of a second language. This will require careful consideration of a number of concepts associated with current issues pertaining to second language research, namely consciousness, interface, modularity, automaticity, proficiency, accuracy, fluency, intake, and ultimate attainment – informed by data from a variety of domains including language pathology and neuroimaging. A list of constructs not presented in the 2004 volume is provided in the Summary of key proposals appendix (pp. 187–190).



Declarative and procedural determinants of second languages

Components of verbal communication Verbal communicative capacity comprises linguistic competence (phonology, morphology, syntax and the lexicon), metalinguistic knowledge (the conscious knowledge of language facts used to keep track of the output when sentences are long and complex, particularly in a formal context), pragmatic competence (the ability to infer meaning from discourse and situational contexts, paralinguistic phenomena and general knowledge, and to use language effectively in order to achieve a specific purpose), and motivation (the desire to acquire a language in order to communicate, modulated by a range of affective factors that result in great variability among second-language learners). Each of these systems is necessary but not sufficient for normal verbal communication and relies on its own specific neural substrate, which is susceptible to selective impairment. Implicit linguistic competence is sustained by procedural memory, metalinguistic knowledge by declarative memory. Pragmatics relies mainly on areas of the right hemisphere. Speakers who have learned a second language after acquiring their native language will compensate for gaps in their implicit competence by relying more extensively on the other components of verbal communication, namely metalinguistic knowledge and pragmatics. The type and degree of motivation may influence the level of success in both the appropriation and use of a second language. To the extent that the teaching of L2 is formal, it will involve the learner’s declarative memory (and result in metalinguistic knowledge); to the extent that it provides motivation, it will engage the dopaminergic system (and improve performance in both learning and acquisition); to the extent that it is communicative, it may involve procedural memory (and result in some implicit linguistic competence). Practice will either speed up controlled processing or promote implicit competence (or both, to different extents and at different times). A language needs to be used in order to keep its activation threshold sufficiently low to prevent accessibility problems. Within each language, the ease of access to various items is proportionate to the recency and frequency of their use. Both affective factors and pathology may modify this correlation.

Implicit and explicit language processes The procedural/declarative dimension is a critical element in the appropriation, use and loss of languages. Implicit linguistic competence is acquired incidentally, stored implicitly, and used automatically. In the context of this neurolinguistic study of bilingualism, implicit linguistic competence refers to the cerebral representation of a set of computational procedures (the form of which is not overtly known). These procedures are implemented automatically. We cannot consciously control their



Preface 

use since we are not aware of their structure. Competence (“knowing-how”) is subserved by procedural memory, as opposed to knowledge (“knowing that”), which is subserved by declarative memory. Implicit linguistic competence and metalinguistic knowledge are distinct, as suggested by neurofunctional, neurophysiological, and neuroanatomical evidence, and recently confirmed by a number of neuroimaging studies of bilinguals. They have different memory sources (declarative vs. procedural), each subserved by neuroanatomical structures and neurophysiological mechanisms that differ from those subserving the other (hippocampal system and extensive areas of tertiary associative cortex vs. cerebellum, striatum, and focalized cortical areas). Moreover, they bear on different entities (e.g., surface form versus underlying structure; acoustic properties versus proprioception). The former is consciously controlled; the latter is used automatically. As will be discussed at some length, implicit competence and explicit knowledge coexist. Neither one can become the other. Second-language learners may gradually shift from the almost exclusive use of metalinguistic knowledge to more extensive use of implicit linguistic competence. The output of L2 speakers is not evidence that a given structure has been incorporated in their implicit linguistic competence. It may be the result of controlled use of explicit knowledge (albeit relatively fast). When controlled processes are speeded-up, they can give the illusion of automaticity. But a task component cannot be more or less automatized. It either is automatized or it is not. Conscious production can be more or less speeded-up, that is, more or less efficiently controlled. Control admits of degrees of velocity in the performance of a task. But we cannot have more or less control over computational procedures that we are unaware of. Hence, automaticity does not admit of degrees. It is systematic whereas a speeded-up process is variable. Speakers can only notice and pay attention to what they can perceive. What is internalized as implicit linguistic competence cannot be noticed. Speakers are aware of the output of the computational procedures that underlie implicit linguistic competence, not of the procedures themselves. One can only observe what has been produced, not how it was produced.

About the contents of this volume Kathryn Kohnert (2008) proposes to view language from a dynamic interactive processing perspective, which has its roots in five complementary theoretical classes: social construction (Vygotsky, 1978); interactive processing (“top-down” and “bottom-up” processes); functionalism (competition, usage-based, and pragmaticbased models); connectionism (the brain as a network of connected neurons); and dynamic systems theory (de Bot, 2007; De Bot, Lowie, & Verspoor, 2007).

 Declarative and procedural determinants of second languages

Here, I recognize the relevance of all five and focus on the contributions of declarative and procedural memory systems involved in each. A chapter on bilingualism and neuropsychiatric disorders, first conceived of as part of this volume, was published separately (Paradis, 2008a) so as to narrow our focus on the appropriation, representation and processing of second languages. In that paper, though it had not been the initial purpose of the research, patterns were detected in an assortment of data that had hitherto been dispersed as unrelated items in the literature (individually uninterpretable, as pieces of a puzzle in isolation tend to be), which turned out to fit a number of hypotheses discussed in Paradis (2004). Several of the hypotheses integrated into a neurolinguistic theory of bilingualism have been shown to be relevant in the neuropsychiatric domain: (1) The activation threshold in the differential abilities to understand and to produce language; (2) the selective impairment of L1, L2, or both, indicating subsystems rather than a single system or two independent systems; (3) the reversibility of symptoms, pointing to the inhibition and disinhibition of subsystems rather than their physical destruction; (4) the fact that the poorer the L2 is, the greater the reliance on declarative memory (metalinguistic knowledge and pragmatics), irrespective of the type of pathology; and (5) the role of affect in sustaining normal and pathological language. Supported by experimental studies and clinical observations, these hypotheses will be considered insofar as they are applicable to the issues treated in this volume, namely the contributions of declarative and procedural memory to second language appropriation and processing. Throughout in the text, redundancy has been favored over ambiguity. I hope the reader will forgive occasional repetitions that may seem tedious at times, as they generally serve to clarify the meanings of terms in particular contexts. Too many barren controversies stem from different interpretations of statements that lend themselves to more than one possible reading. If something can be misinterpreted, believe me, it will be.

A philosophical note When I say that thought (or any other neural function) is subserved by the brain, I do not imply that thought is independent of the brain. Thinking is what the brain does. Thought has no existence outside neural activation; it is not detachable from its organ. Cerebral functions (such as vision, consciousness, or feelings) are emergent properties of the brain (Bunge, 1980; Paradis, 2004). They emerge from the activation of particular neural circuits. Thinking is to the brain what walking is to legs or rotation to the wheel – an abstraction (Bunge, 2007).

chapter 1

Key concepts, framework, and clarifications God showed it to them at the Tower of Babel! In order to sow confusion among humans, He gave words new meanings – different meanings for different people – so that His creatures would get all confused and would not be able to communicate among themselves: Burro came to mean butter to Italians and donkey to Spaniards. Hence, the people were not able to continue the construction of the tower. To avoid such a situation, in scientific English, one must use words with their conventional denotations or define their special technical meanings if a new concept is introduced. Thus, in order to avoid fruitless protracted hollow polemics caused by confusion over the exact intended meaning of propositions related to the various issues considered in this volume, the words that refer to key concepts will be defined below (not that it is claimed that these are the true meanings of these words, but only that, for the sake of clarity, these are the meanings they have in the context of the present discussion). In addition, the background against which some of these notions are debated will be provided.

1.  Definition of key concepts First of all, once the difference between explicit knowledge and implicit competence and between declarative and procedural memory has been acknowledged, there is nothing to be gained by using one term (either learn or acquire) to mean both to learn and to acquire, or by using both terms interchangeably to refer to either learning or acquisition. There is an inherent ambiguity in the term SLA itself, since the A stands for Acquisition but this acronym is commonly used unsystematically to refer to conscious second language learning. Therefore, the term acquire (and its derivatives acquisition, acquirer) will be used to refer only to implicit (non-conscious) items and processes, whereas learn (and its derived forms learning, learner) will refer only to explicit (conscious) items and processes (except in direct quotations). When a statement applies to either or both processes, the term appropriation will be used. The outcome of acquisition is (implicit) competence and the outcome of learning is (explicit) knowledge. Implicit computational procedures refer to whatever mechanism constitutes what is inferred to be an implicit grammar (the speaker’s



Declarative and procedural determinants of second languages

implicit linguistic competence), whether a set of rules (as assumed by generativist linguists) or a network of weighted connections (as assumed by connectionist psychologists). Implicit linguistic competence refers to the neurofunctional system that allows an individual to speak automatically, without conscious control. Its content is acquired incidentally while focusing on something other than what is internalized. It is stored implicitly (i.e., it remains opaque to introspection). As pointed out in Paradis (2004), none of the current competing descriptions of linguistic structure is likely to correspond to the actual implicit computational procedures that are activated when people understand and produce sentences; we can only infer that such procedures exist. The neuroimaging techniques that have been used so far to identify the locus and modus operandi of the relevant neural substrates are not adequate for capturing the ultra-rapid and complex phenomena involved in the generation of a sentence from the initial intention to communicate a message to its phonetic realization. We can only hope that one day linguistic theory and brain theory will be unified, but as Beretta (2006) laments, “at present, far from these theories being compatible, any possibility of unification seems utterly remote” (p. 526). This, however, does not mean that we cannot keep trying. Let us just be aware that we are not there yet. One major problem involves treating current hypotheses as though they were scientific evidence and inferring from them how the brain actually works (and worse, make practical recommendations about language teaching or rehabilitation). There is nothing wrong with (1) formulating hypotheses in theoretical linguistics in an attempt to determine the set of coherent rules that are most compatible with the observed systematic use of language by speakers; or with (2) attempting to verify by all available means (e.g., ERP, neuroimaging, clinical data, etc.) whether there are neural correlates for these theoretical constructs. Nor is there anything wrong with modifying the theory as one goes along. But it should give one cause to tolerate diversity in approaches, to temper one’s conviction of having reached the ultimate truth, and to refrain from treating one’s current theoretical linguistic description as though it corresponds to the way the brain actually processes language. Hypotheses that are considered invalid should be formally refuted; that is, it should be shown why they are flawed. We can be certain, on logical and empirical grounds, of what could not possibly be the case; it is much more difficult to determine what is the case. Theoretical linguistic constructs tend to have a short lifespan. There is no point in claiming that the construct hatched last week is precisely how the brain encodes underlying procedures. We simply have not attained this degree of knowledge. Not only are we not reasonably sure that the underlying procedures are rules of any particular type (which is why Chomsky called them implicit), but we do not even know that they are coded as rules at all, rather than, for instance, as statistically driven



Chapter 1.  Key concepts, framework, and clarifications

connections modulated by the context of each utterance. Let us retain a modicum of humility and a healthy grain of skepticism.1 At this point we can only infer that procedures (whatever their form) sustain what allows speakers to generate sentences. 1.1  Definitions of “implicit” Knowledge is said to be represented implicitly when it is inferred to exist from individuals’ systematic behavior (their competence), though these individuals are unaware of the content of their knowledge. Implicit is said of something that is not observable but inferred. Implicit linguistic competence is a functional system capable of generating sentences, which is inferred from speakers’ systematic verbal behavior. It is inferred that they must have stored in their brains some entity that can be considered as a grammar that allows them to speak the way they do by combining elements in a consistent fashion; they keep this grammar in memory so as to speak in the same way from day to day. Implicit memory is a memory system whose existence is inferred from individuals’ verbal behavior and whose contents are not available for description. It contains various constituents according to the domain of application or relevance (procedural memory, priming, conditioned reflexes). The component of implicit memory that sustains skills (including cognitive skills, such as implicit linguistic competence) is called procedural memory. It is inferred to contain computational

1.  What I caution the reader against is precisely the arrogant attitude of those who are convinced that they hold the Truth and who treat anyone with diverging views as muddleheaded, using intimidation to impose their views – only to discover a decade or less later that they were wrong, at which point they go on to defend and try to impose the new Truth with the same determination and contempt for diverging views. (The disdain displayed for stratificational linguistics by Beretta (2006) on the grounds that it “has very few adherents” (p. 527) is a case in point; so is his attitude towards the study of affect – wondering whether “Schumann and his amygdala are matters that could ever engage” (p. 526) SLA researchers.) This prompts me to point out in passing that scientific truth is not a matter settled by opinion polls, and that if it were, Beretta might find himself in the minority, worldwide. Anyone has a right to focus on a specific component of normal verbal communication. I have not read anything by Schumann that would deny the role of syntax in language use. He does not cast a shadow on the work of theoretical linguists, any more than those who focus on the pragmatic aspects of language, or who stress the imbeddedness of the language module within a set of collaborating complementary cognitive systems. None of these researchers call for a moratorium on the search for whether the brain cares about morphological roots. In fact, Paradis (1998b) argued that, even though the language system (qua implicit linguistic competence) needs to be supplemented by metalinguistic knowledge, pragmatics and motivation/affect to fully account for verbal communication, there is nevertheless a good theory-external justification for the study of context-independent sentence grammar.





Declarative and procedural determinants of second languages

procedures or action schemas. Implicit linguistic competence (i.e., what is inferred to support automatic language comprehension and production) is acquired incidentally, by focusing one’s attention on something other than what is internalized (meaning and surface form vs. the computational procedures that generate sentences). Thus, Hulstijn (2003:  360) is right in assuming that implicit and incidental are not synonymous and in recommending that the distinction be maintained. Implicit refers to a property (i.e., that which is not directly observed but inferred) and incidental refers to a manner of appropriation (i.e., by focusing attention on something other than what is internalized and eventually stored implicitly). It would be equally beneficial for the sake of clarity to maintain the distinction between (implicit) acquisition and (explicit) learning, as discussed above. According to Roehr (2008a), implicit linguistic competence is stored in and retrieved from an associative network during parallel distributed processing, whereas explicit knowledge is processed sequentially with the help of rule-based algorithms. The difference in kind between these two processes results in phonology, morphology, syntax, and lexical retrieval being processed in parallel (hence simultaneously) by linguistic competence, while metalinguistic knowledge is processed only one item at a time; metalinguistic knowledge requires attention, whereas linguistic competence does not. Implicit linguistic competence refers to the generation of novel sentences (propositionizing) by combining and recombining linguistic units (words, phrases, syntactic frames) into linguistic sequences, which, like Heraclitus’ river2 are never the same sentences twice although made up of the same type of material and the same structure. The initiation of an utterance is deliberate, but its elaboration is automatic (from the intention to communicate to its articulatory realization in production and from acoustic analysis to the decoding of its conceptual meaning in comprehension). Acquisition is appropriation of information without awareness on the part of the acquirer of what is acquired and stored in implicit memory. The fact that it is incidental is a characteristic of acquisition as a particular manner of appropriation. Schmidt (1994: 16) defines incidental acquisition as learning without intention to learn; learning one aspect of a stimulus while paying attention to another; learning one thing when the learner’s primary objective is to do something else, for example, learn formal features by focusing one’s attention on semantic features. Granted that we are always paying attention to something, this does not mean that attention is involved in acquisition. Attention is not directed to the element that is acquired. Practice refers to repeated use (involving both comprehension and production) in interactive communicative situations. Structure generation refers to the

2.  You cannot step twice in the same river because the water that flows is never the same.



Chapter 1.  Key concepts, framework, and clarifications

activation of implicit procedures in both comprehension and production to understand or produce sentences; this definition does not prejudge the actual nature of these procedures (whether a generative grammar or a statistically based associative network). Intake denotes the implicit components of the input that contribute to the establishment of implicit linguistic competence. To be internalized means to be or to become part of implicit linguistic competence. The term grammar covers all parts of the implicit language system (whatever corresponds to phonology, morphology, syntax and the implicit grammatical properties of the lexicon). 1.2  Automaticity It is necessary to specify what it means to say that implicit linguistic competence is automatic in the context of the present discussion; the term has been used in neuropsychology in the past with very different (sometimes opposite) meanings. One source of the problem is that not all language material is part of implicit linguistic competence (e.g., vocabulary is not). Another is that implicit linguistic competence is not the only thing said to be automatic; the term is also used in clinical tests to refer to memorized sequences, such as counting, prayers, poems, slogans, nursery rhymes and multiplication tables. First, as it is used here, automaticity is not to be confused with Hughlings Jackson’s (1878) notion of automatic speech (Figure 1.1), which he opposed to propositional speech. Propositionizing, in his vocabulary, refers to the generation of sentences from scratch, the production of novel sentences in order to convey information. Automatic speech refers to the result of rote learning (often produced without really thinking about the meaning of what is said), which global aphasic and echolalic patients are sometimes able to produce in the absence of any ability to propositionize. Global aphasics are also able to produce emotional expressions and interjections, also called automatic speech in the Jacksonian sense. It is interesting to note that automatic speech (e.g., repeating a memorized passage) typically does not engage the language cortex (Bookheimer et al., 2000). (Hughlings Jackson) automatic ––––––→ propositional | | (Paradis) | | ↓ controlled Figure 1.1.  Two meanings of “automatic” Jackson: from automatic to propositional speech; Paradis: from automatic use of implicit linguistic competence to controlled use of explicit meta-linguistic knowledge.





Declarative and procedural determinants of second languages

According to Paradis (2004), on the other hand, the use of propositional speech is automatic, as opposed to being under conscious voluntary control. In other words, what Hughlings Jackson called propositionizing is run by an automatic process (in fact, a set of integrated automatic processes, since those that run phonology are not the same as the ones that generate syntax or sustain lexical searches). So is Hughlings Jackson’s automatic speech, for that matter, except that it is not part of the language system (as implicit linguistic competence) and is not deliberately set in motion, in the case of emotional expressions and interjections. 1.3  Proficiency, accuracy, fluency, and other measures Proficiency is usually measured in terms of accuracy and fluency. Neither accuracy nor fluency, nor both combined, guarantee that the speaker has acquired and uses implicit linguistic competence. Thus, accuracy and fluency alone cannot be used to distinguish between automatic (implicit) performance and controlled (explicit) performance, for the reasons outlined below. Fluency refers to the absence of pauses and other indices of word-finding (or grammatical) difficulty. Clinically, it is often measured by counting the number of words within a specified semantic field (e.g., animals, tools) or words starting with a specified letter (or phoneme) a speaker is able to produce within a specified period of time (usually one minute). In itself, this is not a very informative measure of implicit competence. The fluency test may assess either access to vocabulary (an implicit function) or the richness of the vocabulary (a measure of explicit knowledge). For non-clinical purposes, such as the evaluation of second language appropriation, it is preferable to use natural language performance tasks such as authentic speech obtained in interviews. So-called formal tests of competence assess knowledge, and do not necessarily tap implicit competence. Because off-line tasks allow speakers to consciously control their performance, they cannot be used to measure implicit competence. Fluency in conversation (speed of delivery) is not necessarily an indication of implicit linguistic competence because it may result from speeded-up controlled performance. Accuracy refers to the similarity to native speakers’ grammar in the case of L2, or to prescriptive grammar in the case of L1. Speakers may be quite fluent in L2 but quite inaccurate at the same time (which means that they have internalized (automatized) a grammar containing some items that are deviant with respect to the L2 norm). “Immersion” students3 are notoriously fluent but inaccurate (Lyster, 2004). 3.  This term designates students who attended so-called immersion schools (in particular in the Montreal area in the 1960s), in which native speakers of English were taught all subjects



Chapter 1.  Key concepts, framework, and clarifications

Experimental behavioral measures could be obtained with a wug-type test of syntactic constructions or morphological forms to determine reaction times and variations in their standard deviations (Segalowitz & Segalowitz, 1993; Segalowitz, Segalowitz, & Wood, 1998). This could reveal whether performance is not just fast (possibly speeded-up) but also automatic, given that a prime characteristic of automaticity in speech is systematicity (hence, negligible variability). Under brain imaging conditions, one can assess accuracy and fluency and measure interindividual variation in reaction times. Imaging should show heightened activation of the basal ganglia, and Brodmann’s areas 22, 40, and 44, reflecting implicit language processing. The hippocampal system should also be activated to some extent, reflecting the activation of the content of speech (concepts and memories talked about) and monitoring of performance, in addition to right-hemisphere areas reflecting the use of pragmatics. Lack of automaticity would be reflected in longer reaction times and, more importantly, greater variation; greater activation of the hippocampal system, prefrontal attentional systems, and, in production tasks, the anterior cingulate gyrus. There is little evidence that L2 learners actually do acquire (part of) the L2 grammar. On the one hand, accuracy of performance is no assurance that speakers have internalized the procedures; on the other hand, they may, through extensive practice, have internalized a defective grammar. Moreover, while accuracy is not a sign of implicit linguistic competence, inaccuracy is not a sign that some grammar (albeit inaccurate from the L2 normative point of view4) has not been internalized. Roehr (2008b) found a strong positive relationship between L2 proficiency and L2 metalinguistic knowledge: Advanced learners’ proficient use of L2 structures and vocabulary and their explicit knowledge of these features co-vary strongly and significantly. But neither the tasks nor the conditions of test administration guaranteed that the learners’ proficiency reflected implicit competence rather than

in French, as opposed to being taught French as a second language for one hour a day or to being truly immersed (“submerged”) in a school for French-speaking students where they would hear only native French and would converse with native French-speakers all day. In an immersion school, students hear mostly the interlanguage of their peers since the ratio is 1 (French) teacher to 25 (English-speaking) students. In the case of French Immersion students, “it is clear that the incorrect use of some aspects of syntax and lexis has become part of automatized routines” (Lyster, 1990: 170). 4.  Some deviant features may have been incorporated into a speaker’s L2 implicit system, either features of L1 or wrongly analyzed features attributed in error to the L2 system and repeatedly practiced.





Declarative and procedural determinants of second languages

speeded-up metalinguistic knowledge. The strong correlation would in fact suggest the latter. Incidental acquisition through practice is the only way to internalize implicit linguistic competence. But it is not the only way to become a proficient, fluent speaker of L2: Explicit learning may lead to speeded-up controlled use of a second language (and may even, with repeated practice, indirectly lead to the internalization of some components of L2). As Hulstijn (2007) puts it, “although explicit knowledge cannot be transformed into implicit knowledge neurophysiologically, explicit grammar instruction may indirectly be beneficial to the establishment of implicit knowledge” (p. 701). This issue will be discussed at length in Chapter 3. Late second language acquisition is hard: “L1 is entrenched and plasticity is low” (Seidenberg & Zevin, 2006: 601). In order to overcome this hurdle, learners make use of declarative memory. Of all components of their grammar, native speakers are least aware of their prosody and many other features of their phonology which they nevertheless implement constantly and consistently, such as devoicing a consonant in certain contexts (of which they are not aware, and are in fact quite surprised to learn in Linguistics 101). These are the forms that individuals with genetic dysphasia are unable to acquire, as tests have shown across languages (e.g., morphological operations in Greek (Dalalakis, 1999), Japanese compounding (Fukuda & Fukuda, 1999), or English stress assignment (Piggott & Kessler Robb, 1999)), whereas individuals without dysphasia perform these operations successfully without being aware of them. These same types of implicit features (prosody, segmental phonology, etc.) are those that most readily betray speakers of a second language. Thus, it seems that the features least noticed by native speakers are the ones that individuals with specific language impairment and second language learners have the most difficulty with. This suggests that (1) L2 learners have the greatest difficulty acquiring the features that native speakers acquire incidentally and of which they remain unaware for the rest of their lives unless they take a course in Linguistics. Hence, (2) L2 speakers are aware of and learn what can be fairly easily noticed, and they consciously control their performance both in learning and practice. (In other words, they deliberately learn the rules when these are provided in a formal setting, or deduce a set of rules by conscious analogy, induction or deduction, in addition to memorizing chunks and phrases; they then use these structures consciously.) All of these processes are sustained by declarative memory and are under deliberate control. Automatic comprehension is “ballistic”: You cannot choose not to understand a sentence you have perceived in your (non-attrited) native language. In contrast, when you try to understand a sentence in L2, in order to figure out who does what



Chapter 1.  Key concepts, framework, and clarifications

to whom, you may have to consciously look for cues to identify the subject (such as word order or case inflection), and then select the verb by its position or inflection, even when you understood the meanings of the words from the start. This process may be speeded up over time, and eventually the implicit decoding process may be automatized.

2.  Clarifications about the framework The framework for a neurolinguistic theory of bilingualism was proposed in Paradis (2004). Some of the assumptions and hypotheses set out therein will be clarified below. Following a brief history of the emergence of the declarative/ procedural distinction in the study of second languages, a number of constructs will be elucidated: the crucial distinction between lexicon and vocabulary; what it means to rely more or less on declarative memory; the nature of the continuum from controlled to automatic processing; the nature of the differing structures of metalinguistic knowledge and implicit linguistic competence; various indicators of metalinguistic explicitness in second language performance; and the macro- and micro-anatomical levels of cerebral representation. 2.1  S erendipity or the birth of the application of the declarative/procedural distinction to language representation and processing On November 6, 1991, Neil Cohen gave a lecture titled “Memory, amnesia and the hippocampal system” at the McGill University Cognitive Neuroscience Colloquium. It occurred to me that all the characteristics of procedural memory that sustain motor and cognitive skills, as outlined in Cohen’s talk, were compatible with the processing of syntax. When, at the end of the talk, I made a comment to that effect, the speaker replied: “Do you have a problem with that?” No, indeed I did not. But I remarked that some of my colleagues of the generative grammar persuasion might not like the idea. After all, language was supposed to be completely different from any other cognitive function. (And indeed, they pooh-poohed the idea for a while; it took a few years for the notion to become acceptable.) Cohen also mentioned that, though H.M., the famous anterograde amnesic individual, was able to acquire new motor and cognitive skills, he was never able to learn new words, whether the names of the attending staff or words he did not know before. There thus appeared to be a dissociation between vocabulary and morphosyntax, with the former seeming to be sustained by explicit memory and the latter by implicit memory:  Speakers explicitly know the forms of words and can describe their meanings. I immediately incorporated this insight into the paper



 Declarative and procedural determinants of second languages

on “the involvement of different cerebral memory mechanisms depending on learning methods” I had been invited to give at the First International Congress on Memory and Memorization in Acquiring and Learning Languages (Brussels, Belgium, November 1991), the proceedings of which were published in 1993. The fact was also emphasized in Paradis (1994a). At that conference, Martial Van der Linden of the Cognitive Neuropsychology Unit at the University of Louvain objected that some amnesic patients were able to acquire some new words; it depended on how the words were presented. At the time, I thought that the severity of the amnesia might be responsible for the inter-individual variation in performance, but I soon came to realize that it probably had to do indeed with “how the words were presented”: Although new words could not be consciously learned, no matter how many times one tried to teach them, they might eventually be acquired incidentally, through frequent use in context, without having one’s attention drawn to their form or meaning. As a matter of fact, in studies that have reported on amnesic individual’s ability to acquire new words, the use of concepts in meaningful and varied contexts (as opposed to deliberate, rote learning) has had a positive influence on the extent of acquisition (Van der Linden, Meulemans, & Lorrain, 1994). Note that in all cases, amnesic patients took considerably longer than normal controls (Van der Linden, 1993). It appears that anterograde amnesiacs can learn novel words, but only when assessed by tests of implicit memory (Williams, 2005). At around the same time, Myrna Gopnik of the Linguistics Department at McGill University, who had been investigating a bilingual teenager in Montreal with specific language impairment (SLI), saw a British TV program about a London family in which several members presented with the same symptoms as her examinee. She got in contact with the family and investigated the nature of their language impairment. From the description of their language behavior, it became obvious to me that these individuals were exhibiting a specific impairment of procedural memory for language and that they were learning their native language explicitly, through the use of declarative memory, as one learns facts about chemistry or geography; I proposed the term genetic dysphasia to characterize the deficit (Paradis & Gopnik, 1994). The term was censured by the editor and replaced by familial impairment; however, this paper was subsequently published in a special issue of the Journal of Neurolinguistics explicitly devoted to genetic dysphasia in 1997. The chromosome involved (7q31) was identified soon afterwards (Fisher et al., 1998) and the FOXP2 gene5 responsible for this

5.  FOXP2 is not “a grammar gene” (as only journalists of the popular press would print, in blatant contradiction to what had actually been said in the (misquoted) researcher’s



Chapter 1.  Key concepts, framework, and clarifications

condition was pinpointed three years after that (Lai et al., 2001). The concept of implicit and explicit memory had previously been used in the SLA literature. We now had empirical evidence of the involvement of cerebral mechanisms in these processes and of their genetic underpinnings. Michael Ullman, who was working on inflectional morphology at MIT at the time, heard about the research project on this familial language impairment, came to visit and contributed with Myrna Gopnik to the special issue of the McGill Working Papers in Linguistics devoted to the topic. In it he argued that the deficit was specific to grammatical rules rather than a general cognitive deficit (Ullman & Gopnik, 1994) and referred the reader (p. 83) to Paradis and Gopnik (1994) for a neurolinguistic hypothesis (namely, that the deficit is “a dysfunction of the left-hemisphere-based procedural memory system that subserves the acquisition and processing of certain aspects of morphology” and that the impaired individuals compensate for their deficit by using “metalinguistic knowledge stored in declarative memory” (p. 142)). The declarative/procedural framework for language processes, born in 1991, was coming of age. Five years later, Ullman and Gopnik (1999) specifically referred to declarative and procedural memory to explain the deficit. Michael Ullman has continued to develop the model, incorporating various neurophysiological and biochemical dimensions.

presentation). It is a gene whose transcription factor protein regulates many genes. Many of the functions of FOXP2 are not related in any direct way to language; these include the development of the lungs, intestine and heart. As a result, some individuals with mutated FOXP2 may develop other abnormalities, some of which may affect speech to various degrees, such as abnormal development of the lower jaw (affecting articulation). But, whereas the gene’s effects are wide-ranging, the acquisition of implicit linguistic competence can nevertheless be disrupted in specific and systematic ways when it presents a minute mutation at the forkhead DNA-binding domain (MacDermot et al., 2005; Vargha-Khadem et al., 2005). Individuals with a mutated FOXP2 have few or no cognitive handicaps and the nature of these handicaps differs among individuals; there is therefore no causal link between a particular cognitive deficit and the language impairment. Neuroimaging indicates anomalies in Broca’s area and other cortical language-related regions, the basal ganglia and cerebellum, also known to be language-related (Watkins et al., 2002; Liegeois et al., 2003; Marcus & Fisher, 2003; White et al., 2006). Given that gene expression generally involves external factors (e.g., biochemical environment or sensory input), the mutation of FOXP2 responsible for specific language impairment does not necessarily affect all the genes whose expression the normal human FOXP2 regulates. However, when it does affect language, the nature of the deficits is systematic and predictable, ensuing from an inability to develop (among other things) procedures that permit rule-governed generalizations.



 Declarative and procedural determinants of second languages

2.2  Declarative/procedural models Ullman (2006a) insists that “the DP model differs from Paradis’ perspective” (p. 100). However, “Paradis’ perspective” (1991, 1994a, 2004) is a Declarative/ Procedural model. Ullman’s (2001, 2005, 2006a) DP model (“The DP model”) can only be said to differ from “the view espoused by Paradis” (Ullman, 2006a: 100) in that it provides (very welcome) additional evidence from other perspectives, such as the differential impact of sex hormones and a more detailed exploration of the neuroanatomical underpinnings of the distinction between implicit linguistic competence and explicit metalinguistic knowledge, as subserved by procedural and declarative memory respectively. Note, by the way, that it is not the case that “for Paradis, all that is conscious is declarative and all that is nonconscious is procedural” (Ullmann 2006: 100). Indeed, what is conscious is sustained by declarative memory, but not all that is nonconscious is procedural: in the context of language, only the computational procedures that yield the generation (automatic comprehension and production) of grammatical structures (phonological, morphological and syntactic) are. The access to any representation (whether a conscious or a nonconscious one) is also implicit (i.e., nonconscious) (Paradis, 1991, 2004). As an instance of divergence, Ullman refers to the increased reliance on procedural memory, in both L1 and high-proficiency L2, in terms of greater automatization and implicitness across various domains of language, including both lexicon and grammar. First, it must be clear that “lexicon,” as defined in Paradis (2004), refers to the implicit grammatical properties of lexical items, and explicitly not the form-meaning relations that represent what is called the vocabulary, namely the conscious aspects of words, such as their form-meaning relationships, which are subserved by declarative memory (two entries in the glossary underscore this distinction, which is applied throughout). Once this semantic oversight has been cleared up, I do not see the mark of a different model here. Unless Ullman denies the distinction between the declarative (conscious) and procedural (nonconscious syntactic) properties of words, I see nothing incompatible with his account of the DP Model in this regard. A possible misunderstanding with respect to the lexicon/vocabulary distinction (Ullman, 2001; note 2) is that when components of the L2 lexicon have not been internalized, learners may rely on explicit knowledge about the grammatical properties of these lexical items and process them declaratively in the same way they process explicit syntactic rules, namely in a conscious, controlled manner. In fact, non-automatized L2 function words may also be processed as vocabulary items (Weber-Fox & Neville, 1996). Vocabulary, as defined, is subserved by declarative memory in L1 and L2 alike (to the extent that speakers are aware of the sound-meaning relationships of words).



Chapter 1.  Key concepts, framework, and clarifications 

Even if it were true (which it probably is) that Ullman and Paradis focus on different aspects, this would not make their positions divergent, unless one were to deny any of the other’s proposals. Focusing on different components or properties of the same construct does not make different models; focusing on some areas does not deny the relevance of other areas. The main claim is that items that are sustained by procedural memory for language are subserved by neural substrates different from those that are sustained by declarative memory. If Ullman disagrees with this, then there is a substantive difference between our views – but I do not think he does. We may focus on different aspects of the same phenomenon, but the view I espouse is compatible with everything that Ullman has proposed (and vice versa). One may say that, in comparison, my proposal is underspecified, but not that it is a different model until one finds a proposal in one model that is not compatible with the other. I consider the fact that sex hormones differentially affect which items are subserved by the declarative and which by the procedural systems (if confirmed) to be further evidence for the declarative/procedural distinction. To that extent, Ullman has enriched the model by adding support to the distinction between implicit linguistic competence (and its acquisition) and explicit metalinguistic knowledge (and its learning). This makes the model more elaborate, further specified, and to this extent, of course, one could say that it is different – but only different in the sense of more worked-out, not of proposing anything that would make opposite predictions. (It simply makes additional predictions that verify the model with evidence from other domains, e.g., hormonal.) Ullman’s DP model is a (valuable) elaboration of the declarative/procedural distinction in language appropriation, representation and use, as proposed in 1991, made more explicit in 1994 and expanded in 2004. So far, Ullman’s significant additional contributions contain nothing that makes his model incompatible with the original proposal. As Ullman (2001) put it, the two accounts “share many assumptions, and are perhaps best thought of as complementary rather than competing models” (p. 110). When authors refer to the Paradis/Ullman perspective (Libben, 2006), Ullman and Paradis’ proposal (Sabourin, 2006), or the Ullman/Paradis account (Clahsen & Felser, 2006a), they mean the parts relevant to the L1 and L2 processing differences that Paradis’ and Ullman’s models have in common, namely the distinction between implicit grammar and explicit knowledge, and the differential degree of reliance on each in L1 and L2. Ullman (2006a: 101) does state that the DP model is inconsistent with aspects of the claims of Paradis but does not specify which ones. The reader is referred to Ullman (2005), where the same “differences” are referred to, and one rather paradoxical claim is made: that his DP model assumes that “all lexical knowledge

 Declarative and procedural determinants of second languages

resides in declarative memory (whether or not the knowledge is available to conscious awareness),” whereas for Paradis (2004), “lexicalized knowledge of grammatical properties, such as argument structure, is not declarative” (p. 162). First, Paradis (2004) does not speak of lexicalized knowledge, which is an expression that usually means that a particular phrase has been stored in declarative memory as an unanalyzed chunk (i.e., not generated by a grammatical process, as it could be), but of the implicit grammatical features of lexical items which, as such, are part of linguistic competence, and hence, like the rest of the grammar, sustained by procedural memory. To claim that implicit grammatical features of lexical items (which are unavailable to conscious awareness) reside in declarative memory would be a contradiction in terms in Paradis’ framework.6 It would negate the implicit/explicit, declarative/procedural distinction. Since the grammatical properties of the lexicon are part of the grammar (more specifically morphosyntax), it would amount to saying that portions of the implicit computational procedures that subserve the grammar reside in declarative memory. One would then have to ask which procedures the claim applies to, why, and how something of which one is not aware can be consciously controlled (the way the use of the explicit knowledge of a rule is); and what the criteria for assigning a grammar rule to procedural or declarative memory are. This is unlikely to be what Ullman intends to say. There must be some misunderstanding stemming from words being used with different meanings in Paradis (2004) and Ullman (2005). This is precisely the reason why Paradis (2004) placed so much emphasis on the need to define terms and use specific words for specific concepts; and why, for instance, acquisition or learning, competence or knowledge, lexicon or vocabulary, should be used with their specific meanings to avoid possible ambiguity. The distinction between lexicon and vocabulary was prompted by the need to distinguish between the mere sound-meaning pairings of words, as memorized in traditional second language vocabulary learning, which is a portion of the declarative memory system, and the grammatical properties of words which, independently of their lexical semantic meaning constraints and phonological form, may differ from one language to the next. Because these grammatical properties are usually not explicit, they are not learned as part of the word but are acquired as elements of the implicit grammar of each language subsystem. And because these features are grammatical, they are part of the implicit linguistic competence of

6.  As discussed in the next chapter, words reside in declarative memory whether or not this knowledge is available to conscious awareness at a given time, but (unlike implicit grammatical procedures) they are capable of reaching awareness or else, by definition, they could not be said to be declarative.



Chapter 1.  Key concepts, framework, and clarifications

each language subsystem, not of the explicitly known properties of words, and thus are not represented in declarative memory. (Note that individuals with genetic dysphasia and L2 learners may be taught some of the declarative grammatical properties of words and store that knowledge in declarative memory.) This is why the three-store hypothesis postulates that the lexicon is part of each language subsystem whereas the vocabulary is part of the nonlinguistic cognitive system. Thus, what is called vocabulary in this context corresponds to what Ullman (Pinker & Ullman, 2003; Ullman, 2001) calls the lexicon. The assumptions about the declarative/procedural word/rule model are the same; the distinction between lexicon and vocabulary is relevant to other aspects of the proposed neurolinguistic theory of bilingualism, in particular the subsystems and three-store hypotheses. Note that it is possible, by observing the sentences produced by speakers, to become aware of the fact that a particular verb in English takes two arguments (whereas the equivalent verb in French takes three), and that observation then becomes metalinguistic knowledge that is indeed sustained by declarative memory. Before this conscious observation, though, the fact was part of the implicit competence of the speaker, since it was used without this knowledge.7 Any memorized wordspecific knowledge is sustained by declarative memory. Whereas Ullman (2006a) posits that experience with L2 eventually leads to proceduralization8 of grammar, resulting in L1-like grammatical processing,

7.  For example, whether a verb is intransitive, reflexive, requires a direct or an indirect object; whether a noun is a count or a mass noun (and hence whether it can take the plural form). These features are specific-language-dependent:

English: He laughs Russian: Он смеётся (‘he laughs himself ’) English: He faints French: Il s’évanouit (‘he faints himself ’) English:  He obeys her French: Il lui obéit (‘he obeys to her’)

These grammatical properties may be inferred from examples, but are not part of the average speaker’s metalinguistic knowledge. Even the gender of some noun may not be explicitly known and native speakers, when asked the question, will need to use an adjective with it to deduce its gender from the (automatically generated) agreement (as when you do not remember a certain string of numbers to activate the alarm, but can activate it by pressing the appropriate digits on the keyboard in rapid succession). This is even more obvious with the mass/count distinction. In my experience, native French speakers who come across the concept when studying English do not even realize that the distinction exists in their native language and consider this to be idiosyncratic to English grammar. Yet they will automatically use French mass nouns only in the singular, and with a partitive article. 8.  Note another useful technical distinction: The grammar (qua metalinguistic knowledge) is not proceduralized; rather, corresponding procedures are incidentally acquired, and their use will replace the use of metalinguistic knowledge, as will be argued in the following chapters.



 Declarative and procedural determinants of second languages

Paradis (2004) reckons that it is not theoretically impossible, but that in practice, it is at best very rare that the L2 grammar in its entirety will be internalized and hence subserved by procedural memory. To the extent that L2 is internalized, it is indeed processed like L1, but most L2 speakers retain lacunae in their L2 competence. At the same time, some of the procedures incorporated in the L2 implicit grammar may deviate from their native counterparts (in the form of L1 static interference or other types of erroneous features) but will nevertheless be processed as if they were bona fide elements of L2 grammar. Here again, it is important to specify the meanings of the words we use. Proceduralization could not refer to the transformation of particular explicitly known rules into implicit computational procedures, but only to the gradual replacement of the use of explicit knowledge in constructing sentences by the use of the implicit competence newly (and independently) acquired through repeated use, allowing the speaker to automatically generate sentences. Acquisition is not, and cannot be, the proceduralization of explicit knowledge because an implicit procedure does not constitute the faster and faster application of an explicit rule; it operates on the basis of principles of a different type. 2.3  Why vocabulary and lexicon differ The lexicon is a network of interrelated elements with their inherent morphosyntactic properties. In addition to their specific lexical semantic constraints, both the interrelationships between lexical items and their syntactic properties differ from one language to the next (and are implicit). Anything that is not explicitly known until it has been taught (or observed) is not subserved by declarative memory. Once learned, an item becomes part of knowledge and is stored in declarative memory, though its implicit counterpart continues to be used automatically on-line. For example, one may be taught, or discover by observation, that the verb ‘to obey’ takes a direct (or an indirect) object in one’s native language. This information was not known before but was part of implicit linguistic competence; one automatically used the correct form, without knowing why. This knowledge now exists independently of, and is not used by, the implicit computational system that continues to automatically implement on-line the procedure that corresponds to this explicit knowledge. Similarly, bound morphemes, though they can be made explicitly known, are automatically processed within implicit competence. Only native speakers with an impairment of implicit linguistic competence will form the English past tense of regular verbs by consciously adding -ed. Items stored in the lexicon (as opposed to the explicit vocabulary stored in declarative memory) are subserved by procedural memory and their grammatical properties are part of the procedural system for implicit linguistic competence.



Chapter 1.  Key concepts, framework, and clarifications 

The implicit grammatical features of the lexicon are not part of the vocabulary. When learners become aware of a grammatical feature pertaining to a word (through instruction or observation), for instance the fact that a particular verb must take three arguments, what they learn is incorporated in their declarative metalinguistic knowledge. But this piece of knowledge is of a different nature from the corresponding implicit grammatical feature of the lexicon that is used when generating sentences. Because lexical items have language-specific grammatical properties, they are inextricably connected within a specific language subsystem network and would not fit within another language subsystem. The vocabulary is what you learn in school and find in dictionaries:  lists of words and their meanings. You see them, you hear them, you learn their form-meaning associations. The lexicon is what you acquire through use, from encountering words in different sentential contexts. Their implicit morphosyntactic properties are absorbed into linguistic competence in the same way any syntactic feature (such as long-distance agreement) is assimilated. Native speakers acquire a lexicon, and with time, in school, or by sheer observation, they learn new words with their associated meanings. Second language learners usually gain knowledge of a vocabulary before they acquire a lexicon, and often explicitly learn, at least partially, the syntactic properties of (some) words. As a result, most L2 learners have at their disposal a number of sound-meaning (and written word-meaning) associations but lack the competence related to (often also the conscious knowledge of) their morphosyntactic properties, which may vary from those of their native language. For example, the translation equivalents of direct transitive verbs can be indirect transitive (e.g., to phone/obey Paul → téléphoner/obéir à Paul) or intransitive (e.g., to start the car → *démarrer la voiture); masculine nouns can be feminine (e.g., ein Wagen → une voiture); count nouns can be mass nouns (with the consequent misuse in English of less instead of fewer, much instead of many, etc., – even if the rule stipulating the use of fewer and many before a plural count noun and less and much before a mass noun has been taught, L2 speakers may use the inappropriate adjective. Words are disambiguated by the linguistic context in which they are embedded. For instance, the sound [vεr] in French means ‘glass’, ‘worm’, ‘green’, ‘verse’, or ‘towards’, respectively spelled verre, ver, vert, and vers (more peripherally, it can also refer to Vaires (a city near Paris), and vair (Northern Russian squirrel; its fur)). If the immediate (i.e., sentential) linguistic context is not sufficient, as in “He went to look for his sister near the bank,” then additional context, either discourse or pragmatic, is required. Single words used in isolation as stimuli in psycholinguistic experiments are decontextualized. Only their explicit properties, subserved by declarative memory,

 Declarative and procedural determinants of second languages

are then accessed and the results can only speak to the representation of individuals’ vocabulary, not their lexicon. Consequently, they cannot be generalized to the representation of language (with its implicit phonological, morphological and syntactic structures). Because the form-meaning associations of words are sustained by declarative memory, they are not subject to the same processing as the implicit aspects of language (including the grammatical features of the lexicon). So-called lexical words,9 which have a declarative component (probably the only component tapped in single word tasks), are the very language items that are most likely to be similarly processed by L1 and by late and early L2 speakers, because they are of the same kind:  items in declarative memory – as illustrated by hippocampal activation (Halsband et al., 2002). Grammatical words, however, which are processed as part of the syntax by L1 speakers, tend to be treated as open-class words by late L2 speakers (Neville, Mills, & Lawson, 1992), who use declarative memory to process the morphosyntactic constructions they have not internalized (as do individuals with genetic dysphasia). In a lexical decision experiment, L2 speakers of German, unlike German native speakers, did not gain from a morphosyntactically congruent prime (adjectives that were correctly or incorrectly inflected and either semantically associated with the target word or not), whereas semantic priming effects were similar to those affecting the native speakers (Scherag et al., 2004). In an ERP study using a single group of subjects for a task involving words and a task involving sentences, native and non-native speakers showed similar semantic processing effects for words, but non-native speakers showed none of the effects associated with syntactic processing in native speakers (Sanders & Neville, 2003). It appears that adult L2 learners are guided less by syntactic information than by lexical-semantic clues during sentence parsing (Clahsen & Felser, 2006a). This is consonant with their treating function words declaratively, like lexical words, as previously reported (Neville, et al., 1992; Sanders & Neville, 2003). Non-native speakers lexicalize functional categories, for example attributing a specific meaning to prepositions, which are then processed like content words. To name a picture, individuals need not activate lexical nodes or lemmas (which contain morphosyntactic information) but only the explicit knowledge of the meaning (given by the picture) and the sound of the associated word in declarative memory. No part of the language system itself needs to be activated

9.  Also referred to as open-class words (i.e., nouns, verbs, adjectives, and most adverbs), as opposed to grammatical words (closed class, function words: articles, prepositions, conjunctions). Note that in the lexicon, open-class words too have grammatical properties that integrate them into each implicit language subsystem’s network of relationships.



Chapter 1.  Key concepts, framework, and clarifications 

in order to name the picture. The process is similar to any task requiring the retrieval of a piece of information from declarative memory. It is controlled by a mechanism outside the language system, in contrast to the activation of a word that occurs on-line in the course of producing (or comprehending) an utterance (in which case it is embedded in a syntactic unit, the sentence, and modulated by pragmatic considerations). In the production of an utterance, retrieval of a lexical item will be facilitated by a number of factors such as collocations, formulas, the semantics of what precedes, and morphosyntactic constraints (exemplified in a Cloze test). In the online generation of a sentence, words are not usually consciously selected; rather, the selection is accomplished by the implicit neurofunctional computational system, in the same way as the selection of tense, agreement, word order, and other automatic phenomena. In L2, lexical processing in the course of sentence production may not follow this pattern if a given word has not been incorporated as a lexical item within an implicit computational procedural system; in that case, the selection of the word is a declarative, controlled task, as is the explicit construction of the sentence within which it is inserted. But the picture naming task (confrontation naming in clinical assessment) is the same in L1 and L2. Note that the retrieval process itself remains implicit. Word selection is implicit and automatic: one is not aware of the access mechanisms that select the items during speech (Paradis, 1994a); both explicit and implicit knowledge are accessed automatically (Paradis, 2004). The control of the various components of language structure is automatic (like the control of your heartbeat but unlike the control of your selection of English to name a picture when the background is red and of Spanish when it is green). Automatic processing does not require the use of the anterior cingulate (see the neuroimaging studies that speak to this issue, i.e., those that, like Perani et al. (1996) and Dehaene et al. (1997), show the involvement of declarative memory structures (parahippocampal gyri) and conscious control (anterior cingulate cortex) for the (weaker) L2 only). In addition, the 28 neuroimaging studies reported in Paradis (2004) that used single words unanimously showed complete overlap between L1 and L2, irrespective of the degree of proficiency in L2. Therefore, considerable overlap in frontal activation is to be expected whenever a consciously controlled process is involved, because it is controlled in both L1 and L2, unlike tasks that are automatized in L1 but controlled in L2 (in proportion to the lack of implicit competence), in which one often observes differences in the volume, extent, and location of activation during L1 and L2 processing. “Common principles underlie the representation of words in the two languages” (Green & Abutalebi, 2008:  564) by virtue of being sustained by declarative memory, and “a common tissue underlies both” (ibid.), only at the

 Declarative and procedural determinants of second languages

conceptual representation level (Paradis, 2004, 2007b). The conceptual system is common not only across languages (Paradis, 2004, 2007b), but also across language modalities such as spoken and signed (Rönnberg, Rudner, & Ingvar, 2004). At the level of the lexicon, although they are located within a common gross anatomical area, the items are subserved at the micro-anatomical level by distinct circuits (Giussani et al., 2007) devoted to the respective language subsystem. In addition to the several ways mentioned in Paradis (2004) in which words stand apart from language structures, (1) with advanced age, there is an increase in word-finding difficulty (Obler & Albert, 1981) whereas syntactically complex sentences are preserved (Obler & Albert, 1984); indeed, the most common language complaint among healthy elderly people is the frequency of word retrieval failure (Nicholas et al., 1998). (2) When cultures come into contact with each other, borrowing takes place primarily in the realm of lexical items (Anderson, 1973). Syntax, morphology and phonology (in that order) are more resistant to foreign influence than words (Arabski, 2006). The Franks’ Germanic influence on the history of the French language is primarily found in the vocabulary (Walter, 1988: 52), and a recent study shows that many single words are currently being borrowed from English into Polish and adapted to Polish phonology, in the absence of syntactic or phonological borrowing (Arabski, 2007). (3) The idiot-savant described by Smith and Tsimpli (1995) was able to learn a considerable number of words in over a dozen languages, but did not acquire implicit linguistic competence, even in his native language. (4) Growth in vocabulary but considerable difficulty with syntactic structures is reported among congenitally deaf individuals who were not exposed to ASL from birth (Grimshaw et al., 1994). (5) Children who acquire two languages simultaneously from birth acquire the morphosyntax of both their languages at the same rate as unilingual speakers of each, but their vocabulary in each is less extensive than that of unilingual speakers (Genesee & Nicoladis, 2006); (6) Patients with Huntington’s disease are impaired in linguistic rule application but have spared word processing (Teichman et al., 2005). (7) Anterograde amnesics in general have great difficulty learning new vocabulary: they are unable to recognize or use words that they have been exposed to over long periods after the onset of amnesia (Williams, 2005); their overall linguistic capacity, however, remains unimpaired. (8) Neuroimaging studies show that early and late bilinguals, as well as unilinguals, treat single words in the same manner, unlike the rest of language (Paradis, 2004, Chapter 6, and references therein). Scherag et al. (2004) note that the data from their lexical decision experiment suggest that the full acquisition of at least some syntactic functions may be restricted to limited periods in life, whereas semantic functions are based on associative learning mechanisms that permit learning throughout life. Similarly, after reviewing 15 studies, Slabakova (2006) reports that, while the functional lexicon presents the most



Chapter 1.  Key concepts, framework, and clarifications 

formidable challenge to adult L2 appropriation, there is no apparent barrier to ultimate success in semantics. Why do words differ from the rest of language along so many parameters? One common underlying thread that may account for all of these disparate phenomena is that words (i.e., the conscious association of a default meaning with an oral or written form), unlike the rest of language (i.e., morphosyntax, phonology and prosody), are sustained not by procedural but by declarative memory. So are their corresponding conceptual representations (often referred to in the literature as semantics). Sentence processing is thus not only more complex than singleword processing, it is also supported by a different type of memory system and consequently involves different cerebral structures situated in different anatomical locations. Memorized words depend upon the temporal-lobe-based declarative memory system whereas the mental grammar, which underlies the computation of complex words, phrases and sentences, depends upon the frontal/basal-gangliabased procedural system, which subserves skills. “Explicit learning processes are essential for acquiring the semantic and conceptual aspects of vocabulary” (Ellis, 1994a:  49) in contrast to the implicit processes of the acquisition of the phonological and morphosyntactic features of the lexicon. Individuals acquire the pronunciation elements of the lexicon through a specialized module by implicit, incidental learning processes; this is done automatically (Ellis, 1994a: 51–52). Any lexical process that can be described in terms of a rule, for example the formation of compound words in Greek and Japanese, is subserved by procedural memory (and is therefore impaired in individuals with genetic dysphasia, who are able to correctly produce frequent (hence well-known) forms but cannot create novel compounds (Dalalakis, 1999; Fukuda & Fukuda, 1999)). Abstract representations such as word category are subserved by declarative memory only to the extent that they have been consciously learned. Children before school age use nouns, verbs, adjectives, and adverbs correctly, from which one can infer that they have an internal representation of such categories, but they have no explicit knowledge until this information is taught (or possibly deduced from conscious observation and analysis). Similarly, as discussed in note 7, even the gender of each noun is not necessarily explicitly known, not even by adults. This is why the distinction between vocabulary and the lexicon was proposed (Paradis, 2004, 2007c): In this context, vocabulary refers to the set of explicit word sound-meaning (and word grapheme-meaning) pairings, that is, those aspects that are consciously observable or teachable. The conscious knowledge of the (default, dictionary) sound-meaning pairing of a word is demonstrated by the speaker’s ability to point to a cat when asked to point to a [kæt] and to say [kæt] when asked to name the animal. In contrast, the lexicon refers to the set of words represented in each

 Declarative and procedural determinants of second languages

neurofunctional language subsystem, including their implicit morphological and syntactic properties, such as whether a verb is intransitive and whether an adjective can modify a property that is inalienable, or not. Nonconscious aspects of the lexicon are part of implicit linguistic competence. Vocabulary items are not. While the proposed contrast between these terms is not systematically applied in the literature, the choice of words is nevertheless not unrelated to usage. The word lexicon has been chosen when implicit grammatical properties are incorporated because it is the term generally used in theoretical linguistics and in work on the Mental Lexicon. The term vocabulary is traditionally used in second language learning/teaching contexts. 2.4  Degree of availability of procedural memory Clahsen and Felser (2006a) ask: “What does it actually mean to say that the procedural memory is less available to L2 learners and that they are more dependent upon declarative memory?” (p. 30). Procedural memory is less available to L2 learners: They have fewer items in their implicit linguistic competence than native speakers; consequently, whereas items which they lack are available to native speakers, they are not available for use by L2 speakers. As stipulated in Paradis (2004), to the extent that there is a gap in their L2 implicit linguistic competence (the “rule” system), adult learners compensate by relying on their metalinguistic knowledge (concretely, if they cannot process the passive construction procedurally, they will consciously construct a passive sentence by applying the explicit rule they have learned); they therefore depend more than native speakers upon declarative memory. The gradients of more and less are to be understood in terms of the number of morphosyntactic and phonological features that adult L2 speakers have incorporated in their implicit competence and that are thus available for automatic use. The fewer features of implicit competence a speaker has internalized, the more that speaker will depend on declarative-memory-based knowledge. Clahsen and Felser find what they consider “the vagueness of notions such as ‘less available’ and ‘more dependent’ (p. 30) to be a problem. Second language speakers differ along a number of variables on each of several dimensions:  age of first contact with L2, method of language appropriation, type of exposure to L2, context and extent of use, etc. Given the specification that “to the extent that” some variable X is present, result Y will ensue, the proposed declarative/procedural model fits all possible types of L2 speakers, irrespective of the actual value of each variable. There is nothing vague about this notion. It can be quantified by taking into account the weight of each relevant variable (e.g., how many morphosyntactic features are missing). Here are just four of the numerous examples of this principle explicitly and repeatedly formulated in Paradis (2004):



Chapter 1.  Key concepts, framework, and clarifications 

To the extent that the teaching method is formal, it will involve declarative memory (and result in metalinguistic knowledge); to the extent that it provides motivation, it will engage the dopaminergic system (and improve performance in both learning and acquisition); to the extent that it is communicative, it may involve procedural memory (and result in some implicit linguistic competence) (Ch. 1, p. 31). To the extent that explicit knowledge has been learned through formal instruction and has not yet been forgotten (due to a lack of rehearsal over the years), patients must have access to their metalinguistic knowledge to help them substitute controlled production for the lack of automatic processing, in the same way as a foreign-language learner during the initial phases of formal language instruction (Ch. 2, p. 54). To the extent that X obtains for a given speaker-hearer, Y will result. This means that the more X is present, the more Y obtains. The less X is present, the less Y is used. If X is altogether absent, then there is no Y. But as long as there is one X, there will be some amount of Y. The mechanism is thus the same for all users, and there is only a difference in degree of use (not a difference in nature or in kind). This reasoning applies to any aspect of the proposed neurofunctional verbal communication system (Ch. 7, p. 216) . A concrete example is provided, citing Paradis, Hagiwara & Hildebrandt (1985): “to the extent that there are some regular graph-sound correspondences, and to the extent that there are exceptions to these correspondences, it will be possible to distinguish between surface and phonological dyslexia” (Ch. 7, p. 217). [This refers to the fact that in Japanese kana script there are very few exceptions to the graph-sound correspondence. Yet, the same principle still applies as in English, a language with a far more irregular orthography, and it is possible to demonstrate the same types of acquired dyslexia, provided the few exceptions are included in the experimental stimuli.]

Along this line of reasoning, whether we consider a so-called “perfect” bilingual or a late-onset second language learner at any stage of development, the same principles apply:  To the extent that a language has been internalized, its implicit grammatical competence is processed by procedural memory (and is available); to the extent that there are gaps in the implicit competence for one (or more) of the languages, the speaker will compensate for them by using explicit knowledge sustained by declarative memory (upon which the speaker is thus more dependent). Which components are and which are not available to L2 learners varies with each individual, given that no two individuals will have exactly the same learning experience and thus they will differ along the variables mentioned above (in addition to variables inherent in declarative memory, such as IQ, working memory capacity, attention span, etc.). Given the task-specificity of procedural

 Declarative and procedural determinants of second languages

memory,10 the nature of these components corresponds to the nature of what is internalized in the various modules of explicit competence, namely, phonological, morphological, and syntactic implicit rules. If a native speaker possesses X number of implicit rules, one L2 speaker may possess X minus 15, another X minus 27. These rules can be further subdivided into Xp, Xm, Xs corresponding respectively to the number of phonological, morphological and syntactic rules (each in its respective module). Procedural memory is not monolithic. There is procedural memory that sustains phonology, and procedural memory that sustains syntax, just as there is procedural memory dedicated to playing tennis and dedicated to playing the piano. Each procedural language module concerns a different set of objects, of a different nature, that engage a different type of implicit rule (procedures). In addition, it is also the case that the availability of procedural memory for acquiring language as a whole decreases with age (though with different optimal periods for the development of various components:  prosody, phonology, morphology, syntax, in that order). Because each module has a different optimal period, but also because of all the diversification in the relevant variables mentioned above, different learners will internalize different sets of implicit rules in each linguistic module. Some will speak with deplorable pronunciation but impeccable syntax, others the reverse. And the specific features missing in each module will also differ among speakers, some having internalized the rules A, B, C, others rules B, C, D. (Replace A, B, C, and D by actual rules and this is as precise as you can get.) There is no doubt that the declarative/procedural account is in need of theoretical elaboration, but what is meant by reduced availability should be clear from the preceding paragraphs. As Libben (2006: 72) notes, “the Paradis/Ullman perspective (Paradis, 1994a, 2004; Paradis & Gopnik, 1997; Ullman, 2001) … is indeed underspecified, but it must be admitted that their [Clahsen & Felser’s] alternative account is, at present, also a sketch.” So, what do “less available to L2 learners” and “more dependent upon declarative memory” actually mean? As stipulated in Paradis (2004), “Speakers who have learned a second language after the native language has been acquired will compensate for gaps in their implicit competence by relying more extensively on the other components of verbal communication, namely metalinguistic knowledge and pragmatics” (p. 30). “As a skill becomes more proficient, processing shifts from the use of one mechanism (controlled, declarative) to another (automatic, procedural)” (p. 36). “With practice, more and more tasks (or task components)

10.  The memory representations supporting implicit memory phenomena are inflexible and nonrelational, and are tied to specific processing modules (Cohen & Eichenbaum, 1997).



Chapter 1.  Key concepts, framework, and clarifications 

fall into the province of procedural competence (and hence automatic processing), replacing the use of controlled declarative processing by independently developed procedures” (p. 47). Finally, “[s]econd-language learners may gradually shift from the almost exclusive use of metalinguistic knowledge to more extensive use of implicit linguistic competence” (p. 61). In other words, every time speakers are unable to generate a sentence automatically, they resort to the use of metalinguistic knowledge to fill the gap. As implicit competence further develops (and only to the extent that it does), the speaker needs to rely on metalinguistic knowledge less and less because there are fewer instances where implicit linguistic competence is missing a required element. Clahsen and Felser’s (2006a, b) shallow structure hypothesis, if correct, provides further evidence for the dissociation between the declarative and procedural components of language:  What distinguishes nonnative comprehenders from native ones is that, in L2 processing, the shallow route predominates (Clahsen & Felser, 2006b: 117). In other words, L2 speakers use (declarative) lexical-semantic and pragmatic information whenever the system of symbolic rules and principles of structure building (i.e., procedural competence) fails. The application of the parsing mechanism is restricted because the L2 (implicit) grammar is incomplete (Clahsen & Felser, 2006b:  117). This is another way of saying that simple rules may be learned (noticed, practiced, and perhaps eventually acquired, along the lines discussed in Chapter 3) whereas more complex ones (such as long-distance dependencies) are much more difficult and require more explicit instruction. N.C. Ellis (2005) gives a good account of the important role of declarative memory in such appropriation. To the extent that learners’ ability to compute complex grammatical representations at the sentence level is reduced, they resort to declarative shallow processing. Note that expressions such as “the extent to which controlled processes can become automatized” (Clahsen & Felser, 2006a: 30) must be understood as a shortcut for “the extent to which controlled processes are replaced by the corresponding automatic processes,” because within the declarative/procedural framework explicit knowledge cannot become or be transformed into implicit competence, given the very nature of the two types of memory that sustain them, which bear on different types of entities (Paradis, 2004; N. Ellis, 2005). The fact that “even highly proficient L2 learners behave differently from native speakers” (Clahsen & Felser, 2006a: 30) is consonant with the DP framework. Paradis (2004) argues that complete acquisition of an L2 in adulthood is at best extremely rare and that many of the features that appear to have been automatized are in fact speeded-up but still controlled. The assumption that comprehension involves both the application of semantically based heuristics and full syntactic analysis is also entirely compatible with the neurolinguistic

 Declarative and procedural determinants of second languages

theory of bilingualism proposed in Paradis (2004). The two different routes for computing sentence interpretations, fed by the implicit grammar and explicit lexical-semantic and pragmatic information respectively, are assumed to be engaged in the processing of both L1 and L2, with the explicit route (Clahsen and Felser’s shallow processing route) predominating (i.e., used more often by necessity) in L2 processing. 2.5  There is no continuum from automatic to controlled processing It is often objected that some of the discussed entities susceptible of double dissociations are points on a continuum. In fact, there is no continuum between implicit competence and explicit knowledge, declarative memory and procedural memory, incidental acquisition and attentional learning, or automatic and controlled processing. Processing is either automatic or controlled. Controlled processing may be speeded-up but it remains qualitatively different from automatic processing (which admits of no degree of conscious control whatsoever and is subserved by different neural structures). Conscious control may be involved in the deliberate decision to initiate an automatic process, but it is not involved in the processing itself. What may, at some level of abstraction, be considered as a continuum is the gradual replacement of the conscious use of metalinguistic knowledge by the automatic use of implicit linguistic competence. If you gradually replace meat by vegetable protein in your diet, meat does not become vegetable protein, and there is no continuum between meat and vegetables (except, possibly, phylogenetically over millions of years of evolution – but not in the context of the period and situation that concern us). There is no continuum from automatic to controlled processing (i.e., no degrees of automaticity; a function is automatic or it is not); there is only a continuum ranging from predominant reliance on controlled processing to predominant reliance on automatic processing, or between the amount of controlled and automatic processing exerted on a particular function (e.g., on different types of language switching, as will be discussed in Chapter 5). To rely more on metalinguistic knowledge means using implicit competence when possible (i.e., when the procedure has been acquired and is available for use) and, where there is a gap, compensating by using metalinguistic knowledge to fill that gap. The fewer implicit linguistic competence procedures are available, the more the use of metalinguistic knowledge will be substituted. In the context of asking whether incidental and intentional learning should be thought of as two distinct learning processes or as poles on a continuum, Hulstijn (2003) quotes Paradis (1994a: 394) to the effect that implicit competence



Chapter 1.  Key concepts, framework, and clarifications 

“is acquired incidentally (i.e., by not focusing attention on what is being internalized), stored implicitly (i.e., not available to conscious awareness), and used automatically (i.e., without conscious control).” From this, Hulstijn unexpectedly deduces that “in the dimension of attention and noticing, incidental and intentional do not form two distinct categories” (p. 361). Generally, the distinction between separate entities and a continuum is that two entities (here, learning processes) refer to separate, distinct ways of functioning, as opposed to a grading between two poles from none to all. When one process is assumed to involve no focusing of attention on the item to be appropriated whereas the other one necessarily does, there is no room for degrees of attention to be focused on what is appropriated (here, acquired). Thus, in the dimension of attention and noticing, Paradis’ (1994a) claim is that incidental acquisition takes place without attention to, or noticing of, what is acquired; intentional, deliberate learning, on the other hand, necessarily requires attention to, and noticing of, what is learned. According to Hulstijn’s quoted passage from Paradis (1994a), there is no possible continuum: you either focus on what you learn (intentional learning) or you do not (incidental acquisition). There might possibly be degrees of attention to what is learned in different learning situations, but there can be none in incidental acquisition. What is acquired (as opposed to learned) is not amenable to being focused upon or noticed to any extent whatsoever. There are no degrees of focused attention on the item being acquired for the simple reason that what is acquired is often not there to be noticed in the first place. When children incidentally acquire what we infer to be a grammar (whatever the nature of its underlying computational procedures), they are not focusing somewhat less on the implicit underlying grammar than when a second language learner focuses on a grammatical rule to be learned. They do not (and cannot) focus at all on what they acquire, because it is not apparent. When children or L2 learners acquire the ability to produce new sounds, they do not focus their attention more or less on the voice onset time of a consonant or the precise tongue position (or the tonus of any of the 100 muscles involved) needed to produce it but on the acoustic properties of the target sound and of their own production. There is no (degree of) focus of attention on the proprioceptive data that are acquired. According to Paradis (1994a), and as reaffirmed more forcefully in Paradis (2004), incidental acquisition does not (cannot) require any attention or noticing. It is not just that attention is “not deliberately geared toward an articulated learning goal” (Hulstijn, 2003: 361); in the case of incidental acquisition, attention is focused neither on the intent to learn, nor on what is acquired. There is no point at which both automatic and controlled processing bear on the same object: each is sustained by a different memory system. The nature of the object sustained by each memory system differs: declarative memory sustains

 Declarative and procedural determinants of second languages

explicit rules, procedural memory sustains implicit procedures. Procedures do not have the same schemata; they are not just faster, more efficient applications of the explicit rules. A sentence is either produced consciously by the application of explicit rules, or generated automatically by the activation of implicit procedures. The two processes do not share any component. The processing of an explicit rule can be speeded up (Segalowitz, 2003), but it is not automatized, be it ever so gradually. As a heuristic comparison for the incompatibility between explicit rules and implicit procedures, let us consider the way speech sounds are produced when speaking a word. Speech sounds can be produced by a machine in the order in which the sequence of phonemes is heard. But because of the interdigitation11 of the muscular correlates of phonemes, the sequencing of impulses to the muscular events underlying the production of speech sounds does not correspond to the actual order of the produced phonemes. (This, by the way, is another illustration of the fact that, just because a machine can do something in a particular way, it does not mean that that is how the brain does it.) As a corollary to Paradis’ (1994a) now generally accepted proposal that metalinguistic knowledge does not gradually become implicit linguistic competence (N.C. Ellis, 2005; Ellis & Larsen-Freeman, 2006), there is no continuum between metalinguistic knowledge and implicit linguistic competence, but only between the degree of use of one system and of the other. The two may be used in sequence or in parallel, but knowledge is explicit and what underlies competence is implicit. As another corollary, there is no possible interface between implicit linguistic competence and metalinguistic knowledge. This will be discussed at length in Chapter 3. 2.6  The content of metalinguistic knowledge and implicit competence Explicit metalinguistic knowledge is not necessarily genuine knowledge, but a set of beliefs. In L2 explicit grammar (i.e., what individuals believe to be part of L2 grammar), some elements may be false (e.g., the belief that German Wagen ‘car’ is feminine). Individuals nevertheless treat the false beliefs as though they were bona fide elements of the grammar (e.g., Wagen is represented as being feminine

11.  Because there are considerable differences in the length and diameter of nerve fibers involved in articulation, the neuronal firing order for the production of the sequential phonemes in a word differs from the order of the phonemes produced. Innervation time for laryngeal muscles may be up to 30 msc longer than for muscles in and around the oral cavity. The firing order must at times be different from the order of events occurring at the periphery (See Lenneberg, 1967: 93–107).



Chapter 1.  Key concepts, framework, and clarifications 

in these persons’ explicit L2 grammar). Following very frequent use, these deviant features may be internalized.12 There may be (and for syntactic rules, often is) a discrepancy between the correct/incorrect status of elements in individuals’ metalinguistic knowledge and their implicit linguistic competence. Individuals may know the rule and be able to verbalize it and apply it in writing, in consciously controlled speech, or in a grammaticality judgment task, but nevertheless continue to systematically produce an incorrect form represented in their (inaccurate) implicit linguistic competence. Under monitoring conditions, that is, when speakers are paying attention to their output, they may correct the error. When an incorrect element was incorporated in speakers’ L2 implicit grammar before they were explicitly made aware of their error, eventually, after having repeated the correct form many times in the process of self-correction (and having noticed it now when it is spoken by native speakers), the incorrect element in implicit competence may be replaced by the correct one (thanks to the repeated use of the correct form in meaningful conversations). It is important to distinguish between what is represented (the content) and how it is represented (the mechanisms of representation). What is represented will vary considerably, not only across languages but also across individuals within the same native language (Paradis, 2007b, 2008c). However, the mechanisms involved are the same for all languages and all individuals; the only difference will be a quantitative one, depending on the degree of reliance on each neurofunctional system involved in verbal communication (implicit linguistic competence, metalinguistic knowledge, pragmatics and motivation). In one’s native language, the processing of heard speech is obligatory.13 During normal speech comprehension, the listener is aware of the overall meaning of the speaker’s utterance without any need to direct attention to individual linguistic and paralinguistic features contained within the speech signal (Crinion et al., 2003).

12.  Beginners have to think twice before they produce the proper article and adjective agreements. It takes a second language learner a long time to automatize gender (i.e., to use it systematically, applying all the possible consequences for agreement). This means that, if they wish to be accurate, L2 learners have to apply multiple pedagogic rules consciously in rapid succession when producing utterances until the process has been automatized. This is why beginners are notoriously slow. They may gradually speed up their control over their production, and some are better at it than others, resulting in considerable individual differences, as discussed in Chapter 4. 13.  It might be objected that sometimes one does not realize that one is hearing speech in one’s L1, (e.g., from a speaker with a strong accent). To the extent that the pronunciation does not conform to the sounds of your L1, it is not the language as you know it. In fact, you will have comprehension difficulties to the extent that a dialect differs from your own.

 Declarative and procedural determinants of second languages

To the extent that L2 features have not been internalized, L2 speakers may have to specifically direct attention to certain formal features to recover the meaning of utterances they hear; hence, to that extent, declarative memory is involved. 2.7  Interference, variability, and other indicators of explicitness Frequent “performance errors” in L2 speech are an indication that a rule has not been automatized. Variation is an index of speeded-up control. The use of implicit linguistic competence is automatic, which makes it resistant to interference from other sources requiring attention and allows production (and comprehension) despite a certain degree of “noise.” If one wishes to claim, for example, that 65% correct responses is evidence of an L2 item’s being part of implicit linguistic competence (or of UG) and that any errors are performance errors, one must stipulate what type of performance factors would induce a speaker to go wrong 35% of the time in L2 but not in L1. The correct use by native speakers of the rules of their grammar is not just “above chance” but systematic. To score above chance is no evidence that a rule has been internalized. If a rule has been internalized, it is systematically used correctly – barring the extremely rare occurrence of true performance errors, easily recognized by the speaker as such. These errors are caused by such factors as short-term memory (STM) constraints, attention span, and particular conditions such as stress, fatigue, or one too many Martinis. Under normal circumstances, they occur about one to two percent of the time in native speakers. Why would STM and/or attention fail 35% to 40% of the time when speaking a second language? The answer may be that these speakers of L2 are not actually using implicit linguistic competence but controlled metalinguistic knowledge, which is known to be vulnerable to STM constraints and requires attention. The implicit grammar cannot be modified by anything (e.g., facilitation or interference), implicit or explicit, that is external to it (i.e., that is of a nature other than that of the components of each module). In contrast, any conscious perception (irrespective of modality) may interfere with or facilitate a conscious task. The claim that language functions are neurofunctionally modular does not mean that the underlying substrate of linguistic competence is unique, in the sense of being of a nature, type and/or organization altogether different from any other function. It only means that, even though some of its substrates may be of the same general nature as those of certain other functions, implicit linguistic competence forms a modular system subserved by task-specific procedural memory, namely a set of computational procedures uniquely dedicated to it. This procedural memory system devoted to implicit competence is composed of a network of specific frontal, basal ganglia, parietal and cerebellar structures (Ullman, 2004). On the



Chapter 1.  Key concepts, framework, and clarifications 

other hand, vocabulary (i.e., the memorized arbitrary form-meaning associations) is subserved by declarative memory and largely depends on temporal-lobe substrates (Ullman, 2004). The learning and acquisition of grammar differ in important ways from learning and acquiring motor sequences. Implicit competence is not analogous to metalinguistic knowledge in the same sense that we speak of analogous knowledge of movement sequences. The analogy is possible only at a high level of abstraction. An implicit procedure is analogous to the corresponding explicit knowledge only in that it generates (albeit via means that do not correspond to the application of the corresponding explicit rule) an output that can be described by the explicit rule. But it is not the case that these explicit rules are automatized. The result of the conscious application of an explicit rule (e.g., a sentence of a particular type) is the same as that of the implicit generation of a sentence of the same type, though the computational procedures that served to generate it do not correspond to the steps that led to its explicit construction. The dynamic schema of the various phases of the production of an utterance differ: each follows a different type of flow chart. The explicit rule and the implicit procedure are not modeled on each other. What may be true of motor sequences (the declarative system initially learns sequences (as knowledge) while the procedural system acquires the analogous competence – the automatization of the same sequence of movements) cannot be true of the acquisition of language. Unlike a learned sequence of movements, to which the analogous competence – namely the automatic production of that sequence – corresponds, the acquired linguistic competence does not constitute the automatic application of explicit rules. The implicit computational procedure is not the automatized explicit rule. The automatic procedure that generates a sentence (or any other grammatical form) differs from the explicit rule that may be used to construct that sentence (or grammatical form). Both implicit and explicit processes produce the same output, but by means of different kinds of operations.14 The procedural memory system does not gradually acquire (i.e., automatize) an explicit metalinguistic rule but an analogous implicit procedure – analogous only in the sense that it will generate sentences of the same type as the ones explicitly constructed by the application of a metalinguistic rule. The procedure cannot be the automatization of an explicit rule, because there are many different explicit rules that correspond to the structural description of a given grammatical

14.  As a heuristic analogy, a cancerous tumor can be eliminated by chemotherapy, radiotherapy, or surgery – three different processes relying on different principles that produce the same outcome: the tumor is removed.

 Declarative and procedural determinants of second languages

construction, depending on the particular linguistic theory. A sequence of movements occurs in the same order, whether performed automatically or under conscious control. There is no such correspondence between an automatic implicit linguistic competence procedure and an explicitly known grammar rule. An implicit linguistic computational procedure is not the automatization of an explicit grammar rule. Adult native speakers do not consciously apply rules couched in terms of pedagogic grammars (Hulstijn, 2007: 705). Certainly, as R. Ellis (2005: 215) puts it, “it can be argued that explicit knowledge is used in both the process of formulating messages as well as in monitoring and that many learners are adroit in accessing their explicit memories for these purposes.” However, he goes on to add: “especially if the rules are, to some extent, automatized.” This could only happen if the explicit rules were identical to the automatized procedures that generate sentences implementing such rules. Yet automatic use does not involve the rapid, serial application of explicit rules (Hulstijn, 2007:  706). When adroit learners access their metalinguistic knowledge to formulate utterances, they cannot process this knowledge automatically but must consciously control it (no matter how speedily). Only what has been acquired can be used automatically. This is why learners’ grammatical accuracy improves significantly when they have time for “on-line planning” while performing a narrative task (Yuan & Ellis, 2003), which is indeed “a result most readily explained in terms of their accessing explicit knowledge” (R. Ellis, 2005: 215). Automatic processes are not only fast but systematic, subject to far fewer “performance errors” (i.e., external interference, such as noise, etc.) than controlled processes, however speeded up. 2.8  Macro-anatomical and micro-anatomical levels of representation One more point may need clarification, since Fabbro (2001) and Paradis (1989, 2004) have been claimed to assert “that the neural mechanisms underlying different languages in bilinguals are not distinct” (Hamberger, 2007: 483). What one actually finds in Paradis (1989) is that, whereas some authors propose that the two languages of a bilingual are represented in partly different anatomical areas in the dominant hemisphere, with some overlap, “Another possible hypothesis is that languages are subserved by different circuits intricately interwoven in the same language areas, so that both are represented in the same area at the gross anatomical level, while still being independently subserved by different neural circuits at the micro-anatomical level” (Paradis, 1989: 131; Paradis, 2004: 110). Paradis (2004) goes on to write that, because in earlier stimulation studies language-specific areas were reported only at the periphery of the language zone, it might have been the case that the stimulation interfered with language only some of the time, with



Chapter 1.  Key concepts, framework, and clarifications 

the electrode field falling just outside the language zone at others. However, an increasing number of cases of specific sites well within the classical language areas that respond to only one of the patients’ languages have been recently published (Roux & Trémoulet, 2002). This might be a first step toward empirical evidence that two languages are represented as independent subsystems intertwined within the same area” (p. 216). The evidence has increased since then (Lucas, McKhann, & Ojemann, 2004; Roux et al., 2004; Walker, Quiñones-Hinojosa, & Berger, 2004; Bello et al., 2006; Giussani et al., 2007; Aladdin, Snyder, & Nizam, 2008). The proposed hypothesis is clear: two languages are micro-anatomically distinct within the same macro-anatomical area. Neuroimaging studies are not adequate to measure these differences: “Gross anatomy (in terms of gyri, as is usually reported) is likely not to be an appropriate measure of functional localization if, as proposed, languages are represented as subsystems that differ micro-anatomically in their dedicated neuronal circuitry” (Paradis, 2004:  165). Roux and Lubrano (2006) concur:  “languages in bilinguals are represented, at least partially, in distinct microanatomical systems located in the same gross anatomical areas” (p. 129). The influence of the history of an individual’s bilingualism (including time of onset and manner of appropriation), pattern of use of each language, proficiency in each, and, just as importantly, methodological variations (including type of task and stimuli) cannot be ignored in the interpretation of the topological patterns of language organization inferred from neuroimaging: All of these factors contribute to the pattern of involvement of declarative or procedural memory in the appropriation of language. Time of onset is important for its influence on the choice of appropriation strategy – either incidental or attentive, with repercussions on the type of memory system involved (procedural or declarative). The manner of appropriation after a first language has been acquired may differ according to the circumstances of exposure, the motive for appropriating a second language, and the individual’s cognitive style. Note that a language acquired through extensive practice from total immersion in a second-language context may include numerous features at all levels of linguistic structure that deviate from the L2 norms, but to the extent that it is automatized, it will be processed as implicit linguistic competence and will be detected as such by neuroimaging procedures. These procedures speak to the organization and processing of representations, not to their contents. The pattern of use will not only influence the general degree of proficiency but will also impact independently on the development of each component skill, such as comprehending, speaking, reading, writing, switching, and translating. With respect to proficiency, the proportion of each language that has been automatized will play a role. Note that the native language will generally be fully acquired (with respect to the basic structures of the grammar at least – the size of the vocabulary

 Declarative and procedural determinants of second languages

may depend on sociocultural, educational and professional factors); however its fluency may eventually be reduced by the raising of its activation threshold following many years of disuse, while some form of the second language may have been largely automatized. What Fabbro (2001) asserts is that lexicons (i.e., vocabularies) of L1 and L2 are both stored in declarative memory, in agreement with the most recent hypotheses on the different memory systems involved in language acquisition and learning. He insists that the areas shared by both languages are at the macro-anatomical level, that syntax is macroscopically represented in the same brain areas, implying that the representation of each language may nevertheless be distinct microscopically, in individual circuits within a common area.

3.  Conclusion In this volume, for the sake of clarity, knowledge is considered to be explicit and competence implicit. What is implicit is inferred, not directly observable. Acquisition is incidental, while learning is conscious. Acquisition results in competence, and learning results in knowledge. The nature of what is represented in implicit and explicit representations differs: skills and competence are represented as procedures and schemata; knowledge is represented as facts. Implicit linguistic competence refers to the cerebral representation of the computational procedures that subserve the grammar of the language (phonology, morphology, syntax and lexicon). This implicit grammar is sustained by procedural memory, contrary to metalinguistic knowledge (i.e., what one is aware of about language), which is subserved by declarative memory. Performance involves implicit access, that is, on-line activation of representations, whether those representations themselves are implicit (linguistic competence) or explicit (declarative knowledge). Words, as form-meaning relationships, are known declaratively. When a word is introduced in context, inferences can be made from that context as to its possible meaning, thereby involving pragmatics. Words differ from the rest of language along many parameters. One common underlying thread that accounts for all of these disparate phenomena is that words (i.e., the conscious association of the default meaning and oral or written form), unlike the rest of language (i.e., morphosyntax, phonology and prosody), are sustained, not by procedural but by declarative memory. In fact, anything in a language that needs to be explicitly memorized (e.g., an irregular verb form not derivable by rule (Ullman, 2001)) is consciously learned and hence sustained by declarative memory. Items memorized by rote are not part of implicit memory either.



Chapter 1.  Key concepts, framework, and clarifications 

The most important characteristic of implicit linguistic competence is its systematicity (the virtual lack of variance in performance in all contexts and in all components of linguistic structure, from prosody to phonology, to morphology, to syntax and lexical retrieval). There is no continuum from automatic to controlled processing in the sense that the same material gradually becomes automatized over time. The observed continuum in performance reflects gradually reduced reliance on metalinguistic knowledge as implicit competence becomes available. The distinction between lexicon and vocabulary has a number of practical and theoretical advantages. It clarifies the distinction between explicit and implicit aspects of words. Only explicit aspects are listed in the vocabulary (notably the form-meaning relationship), just as they are usually listed in dictionaries. The distinction differentiates between these explicit properties (subserved by declarative memory) and those that are implicit, of which speakers are not generally aware (unless they are linguists or grammar teachers), such as the type of classifier a word takes in Chinese or Japanese, which are part of the implicit grammar. The implicit properties of the lexicon are thus hypothesized to be subserved by procedural memory. Word access is deliberate but not consciously controlled. This accounts for the increasing word retrieval difficulty with aging and the longer times experimentally measured in a second language. An early bilingual is a person who has acquired two languages in infancy. A late bilingual is a person who has learned a second language after the first had already been acquired. Even when exposed to a second language in a natural, informal, conversational environment, individuals who are appropriating a second language pay attention to the surface structure of what is being said, try to imitate what they hear, and consciously try to understand how the grammar works, “how one says things” in that language, thereby engaging declarative memory and gaining knowledge. They “operate in a self-initiated searching mode, trying to develop concepts and rules on their own” (Hulstijn, 2007: 706). This explicitly deduced knowledge may be accurate or not, as learners may make wrong assumptions, but this does not affect the manner in which it is represented, namely in declarative memory.

chapter 2

Consciousness in L2 appropriation Because consciousness is central to the implicit/explicit distinction debate, it will be discussed here in detail. Schmidt (1994) recognized the importance of the notion of consciousness in applied linguistic theory and pleaded for a clear definition of the term. He considered four basic senses of consciousness that are relevant to second language acquisition theory: consciousness as intentionality, consciousness as a subjective state arising from the allocation of attention, consciousness as awareness, and consciousness as control. The sense that is adopted here, trying to adhere to the ordinary meaning of the term as far as possible – its “natural meaning” (Dienes & Perner, 1999: 736) – is also the one that Schmidt (1994) considered the most common in colloquial use as well as in psychological and philosophical discussion, namely awareness. As applied to language appropriation, it is not ambiguous. We are aware (conscious) of what we learn: it is noticed before it is stored in declarative memory. But we are not aware of what we acquire:1 it is not noticed because it is not perceivable in the 1.  For example, (unconscious) proprioception for the articulation of (observable) speech sounds; underlying implicit procedures for the generation and production of (observable) surface syntactic structures. Note that the underlying structure is so opaque to introspection that linguists and psychologists still hotly debate whether it takes the form of symbolic rules or of weighted connections in a parallel distributed network. This does not mean that, in principle, we cannot ever learn the actual form and processes of implicit linguistic competence. Given the right technology, we might eventually find out. But (1) at this time (and probably for a long time to come) we are not able to know, and (2) even if we did, it could not help us understand or produce utterances on-line in L1 or L2. We would simply know what the mechanisms are (which would be one more piece of metalinguistic knowledge) but we would not be able to make use of this knowledge on-line any more than our knowledge of articulatory movements obtained from X-rays or our knowledge of the process by which vitamins and minerals are extracted during digestion from the food we eat allows us to control these processes. We may drink orange juice because we know that it will prevent scurvy, but that’s as far as it goes. We cannot influence the actual metabolism process or perceive the various vitamins and minerals in the food that we ingest. As we will see later in this chapter, we are aware of the input and of the results, but not of the intake process: how the organism selects, from the abundant input, what it needs; how the acquisition device abstracts the intake from the abundant verbal input.

 Declarative and procedural determinants of second languages

input stimuli before or after it is represented in procedural memory. Baars (1988) operationally defines consciousness as “the set of events that can be reported with verifiable accuracy and are claimed to be conscious under optimal reporting conditions” (p. 372). “Information that is clearly represented in the nervous system but cannot be retrieved, such as syntax and other highly automatic tasks” (p. 390) is unconscious. What is conscious is that of which we are aware, whether it is awareness of one’s feelings (affective, emotional consciousness, Panksepp, 2007) or of one’s perceptions, volitions, actions, and thoughts. “Consciousness is the perception of what passes in a Man’s own mind” (Locke, 1690: Book II, Chapter 1, section 20). That perception is explicit. Implicit knowledge refers to a competence, namely the ability to do something without knowledge of the actual underlying mechanisms that allow a particular performance. Theoreticians may infer the structure of such mechanisms but have no direct evidence. The only (logical) certainty is that some mechanism (whose structure is not known) must sustain the particular performance. Given that the structures of these mechanisms are unknown and not directly observable, we cannot be aware of them, and thus they are not conscious. Therefore, they cannot be under conscious control (we cannot consciously control something of which we are not aware). An action may be initiated with intent (i.e., deliberately triggered), but the operation underlying what is triggered is processed automatically (i.e., without conscious control). Control is either conscious or automatic. Only that which is observable, known, and accessible to consciousness can be consciously controlled. What is not susceptible to reaching consciousness (i.e., not open to introspection) is controlled automatically by autonomic, implicit systems. Known aspects of language can be consciously controlled; unknown (i.e., implicit) aspects must be controlled automatically. Similarly, language switching can be either consciously controlled or automatic (i.e., under autonomic cerebral control). Intentionality implies consciousness, but it is not synonymous with it. An intentional act is a conscious act, but a conscious act is not necessarily intentional. Consciousness is a subjective state arising from the allocation of attention, but it is not attention itself. Attention is the cognitive process whereby a person concentrates on some feature(s) of the environment to the (relative) exclusion of others; it is the ability to focus one’s awareness selectively on a particular stimulus among all the other aspects of current experience, sustaining that focus and shifting it at will. The information on which attention is focused is information that the person is aware of (Cowan, 1999). Consciousness is not control: an action may be consciously or automatically controlled. Not all phenomena are amenable to being consciously controlled (i.e., only those of which, by virtue of their nature, we can be aware).



Chapter 2.  Consciousness in L2 appropriation 

Whether or not we intend to deliberately learn a second language or to acquire it incidentally through interactive verbal communication, we are aware of what we perceive and of what we produce, but not of how, in the process of interaction, the underlying grammar (prosody, phonology, morphology, and syntax) is acquired and represented for subsequent automatic use. The structures most often used in L2 are likely to be acquired first and become automatic when much of the rest of the grammar still remains under conscious control. Learning without attention to what is to be learned is impossible, but much of what is acquired is not observable, hence not noticeable, and attention cannot be focused on something that cannot be perceived. Attention can be focused on input (the surface form of what is perceived) but not on intake (what serves as material for the elaboration of the underlying structure, namely either the computational procedures underlying the generative grammar or the ascription of weights to the underlying connections on the basis of frequency of use); this will be considered in more detail below (p. 53). Implicit learning (i.e., acquisition) occurs when learners have no awareness of either the process or the outcome of learning (i.e., what is represented in their (implicit) memory). If learners become aware of generalizations, these become part of their metalinguistic knowledge. Metalinguistic knowledge refers to anything about language of which the person is aware, that is, what is consciously known. In other words, we may, thanks to the improvement in technology and other sources of knowledge, eventually become aware of the nature, structure and operations of implicit linguistic competence, but this knowledge will not be used in the acquisition and automatic use of language any more than our knowledge of the way the gallbladder produces bile allows us to consciously control this process. Metalinguistic knowledge and implicit competence will remain distinct systems sustained by different types of memory systems, just as the knowledge of the anatomy and physiology of bile production remains independent of the autonomic processes involved. Thus, what is acquired incidentally involves no awareness of what is acquired (i.e., implicit linguistic competence); it is represented implicitly (and remains inaccessible to introspection); and it is used automatically. There is no awareness of the process of acquiring, of the structure of the representation, or of the nature of the processes involved in its implementation. The terms implicit, unconscious, automatic, and procedural are not synonymous but all entail each other. What is implicit is not directly observable but inferred. What is unconscious is not within awareness. What is automatic is used without conscious control. Automatized skills and cognitive schemata are sustained by procedural memory. What is implicit cannot be consciously controlled. What is unconscious cannot be explicitly known. An automatic process

 Declarative and procedural determinants of second languages

cannot be consciously controlled, its procedures are implicit, not conscious, and hence impervious to introspection and sustained by procedural memory. The absence of consciousness is central to the characterization of implicit acquisition (Lewicki, Czyzewska, & Hill, 1997). Implicit knowledge cannot be defined in terms of context and mode of appropriation (e.g., picking up a second language in naturalistic situations, where target structures are not explicitly described), because adults often consciously infer (correctly or incorrectly) grammatical rules, which they subsequently apply. Learners have more awareness than they are often given credit for and are able to observe regularities and consciously mold their verbal behavior accordingly; they form conscious hypotheses about the target language without being told the rules (Schmidt, 1994). In the classroom, as in an experiment, “it is always possible that participants will enter an explicit, intentional, learning mode even if the task does not appear to demand it” (Williams, 2004: 206). Language as a set of explicit rules is the result of conscious linguistic analysis and does not describe our language ability in kind.2 Rather, it describes structural properties of the results produced by our language faculty. The unconscious workings of implicit linguistic competence (which correspond to a state of our neural networks) are quite different in nature from the conscious handling of its output and our interpretation and explicit analysis of it, namely metalinguistic knowledge. Implicit procedures are not describable in terms of algorithms or rules (Bartsch 2002). The data of our linguistic analysis (of which we are conscious) fundamentally differ from the unknown internal representations (of which we are not conscious); only the results of the unconscious processes of linguistic ability have an expression in consciousness (Bartsch, 2002). Statements such as “… non-conscious registration applies to well learned rather than new information. Syntactic categories may also be non-consciously activated once they are well established” (Schmidt, 2001: 31) seem to imply that there is a one-to-one correspondence between the explicit syntax (the syntax we are aware of, whether the observed surface form or the result of conscious analysis) and the computational procedures of implicit competence, as if the automatized syntax were a speeded-up processing of the explicit rules. Yet, (1) children acquiring their native language do not have an explicit syntax before they acquire the competence to speak syntactically; and (2) adult L2 learners surely do not end 2.  Bartsch (2002) considers that Chomsky’s and Fodor’s assumption that underlying linguistic rules preside over and guide our language behavior “does not make sense.” No introspection has ever led to such rules; they represent only the result of the analysis of the products of our language ability (p. 34). We never encounter these implicit rules in the data we experience. We can detect them only by systematic linguistic analysis of the perceived data (p. 35).



Chapter 2.  Consciousness in L2 appropriation 

up unconsciously activating the syntax they have learned (neither the generative grammar nor the connectionist network that are assumed to subserve competence resemble the speakers’ explicit knowledge of the agreement of the past participle in their L2 French). Nonetheless, if they have indeed automatized the implicit procedures corresponding to the L2 syntax, they are able to understand and produce such sentences at the rate of 14 phonemes per second (about 200 words a minute). The question of to what extent adult L2 learners actually do automatically activate syntactic categories remains unanswered – but as long as they do so, the unconscious mechanism does not activate the rule that is consciously known.

 nly specific types of representations can become 1.  O conscious – others cannot Over the last decade or so, studies have scrutinized the neurophysiological and neurochemical aspects of consciousness. These theories of consciousness explicitly exclude from their construct those phenomena that cannot be brought to consciousness. My reading of the data is that there are two kinds of mental representations: those that are inherently susceptible of becoming conscious (explicit representations) and those that are not (implicit representations, implicit procedures). The first, when they reach their activation threshold, give rise to conscious experience. The second do not; rather, they give rise to cerebral activities that are not consciously controlled (in fact, not consciously controllable: We cannot consciously control the generation of computations of which we are not aware). Much of the brain’s activity proceeds without giving rise to awareness (Zeman, 2005). We must distinguish between inherently implicit representations (which are not capable of reaching consciousness under any circumstance) and explicit representations that are temporarily not conscious because they are not currently the focus of attention and whose activation therefore does not cross the threshold of consciousness (though they are capable of doing so). Discussions of unconscious elements within the global workspace (Baars, 2003) and the conscious neuronal workspace model (Dehaene & Changeux, 2004) refer explicitly to the latter. Representations subserved by autonomous neurofunctional modules (e.g., implicit linguistic competence) are inherently implicit and therefore lie outside Baars’s (2005) global workspace and Dehaene & Changeux’s (2004) global neuronal workspace and are independent of Scott Kelso’s (2002) rhythmic synchrony of coordination (the relevance of this theory will become clear in Chapter 3). The bright spot (consciousness in Baars’s metaphor) on the stage of immediate memory, directed by a spotlight of attention under executive guidance, cannot shine on an inherently unconscious representation: That never emerges from the

 Declarative and procedural determinants of second languages

dark. In fact, it remains off stage. Attention, under executive guidance, can focus its spotlight only on what is temporarily in the dark on stage in the global theater (i.e., only what has the potential to become conscious). In Dehaene and Changeux’s (2004) conscious neuronal workspace model, it is clear that only what can enter consciousness will ever do so. The activity of many neurons is excluded from conscious mobilization (p. 1148). Implicit representations and processes are excluded from the conscious global workspace. Consciousness is enriched by two types of attention: bottom-up, caused by sensory input; and top-down, produced by the planning parts of the brain (Crick & Koch, 1998). A piece of explicit information such as an image or a word may cross the threshold of consciousness and become available to conscious report, but no matter how much bottom-up or top-down perceptual or emotional amplification of any cerebral areas takes place, no inherently implicit representation, such as any part of linguistic competence, will ever become available in the conscious neuronal workspace. The global neuronal workspace model applies only to conscious content. It specifies the neural conditions under which a given explicit representation is made potentially available to a wide variety of neural processes that give rise to a subjective feeling of conscious access. The “global broadcasting” of signals from neurons with long-distance connections occurs within the conscious neuronal workspace. A prerequisite for inclusion in the workspace is that the representation or process involved be inherently capable of reaching consciousness. Implicit linguistic competence is not of such a nature. The procedures of implicit linguistic competence (i.e., implicit representations) are outside the realm of application of this neural workspace, by virtue of not being types of representations that have the inherent potential to become conscious. Only in the conscious workspace do processes exchange information. Hence, information from metalinguistic knowledge cannot be exchanged with implicit competence because (1) implicit competence has no access to the workspace and (2) the types of information are not compatible with each other: metalinguistic-type information is irrelevant3 to the structure and processing of implicit linguistic competence. According to Luu, Kelley, and Levtin (2001), consciousness is a property of all brains with the necessary neural hardware. I would specify that the property is restricted to all cerebral representations endowed with the necessary neural software, namely those that are inherently susceptible of reaching consciousness.

3.  What we have in our mind as knowledge are only representations of rules we construct on the basis of observable regularities in our data, but these are not the rules on which linguistic ability relies (Bartsch, 2002: 35).



Chapter 2.  Consciousness in L2 appropriation 

Subjective conscious experience arises from the activation of cerebral processes associated with explicit representations. Any given representation is part of either an implicit or an explicit network; it never partakes of both. An explicit entity (motives and emotions, executive functions) may trigger the activation of an unconscious representation or routine; unconscious networks may constrain conscious events (Baars & Franklin, 2003), for instance by contributing to the summation of impulses that allow the conscious representation to reach its activation threshold. With respect to language, motivations, emotions, or intentions, may select, within the implicit language neurofunctional modules, the schemata relevant to the optimal encoding of the message. The schemata themselves remain non conscious.4 To procure awareness, the brain has to construct an explicit representation. Much of the neural computational activity needed to construct an explicit representation is unconscious; the representation itself is the result of these computations (Crick & Koch, 1998) and it alone reaches consciousness. Note that, throughout the process, unconscious mechanisms remain unconscious and only representations with the inherent potential to become conscious can in fact do so. The choice of the appropriate schemata is dictated by the (possibly conscious) “contexts” but under no circumstances do these conscious and unconscious contexts modify the structure of the implicit linguistic competence system (i.e., the nature and form of the underlying procedures). All these contexts serve as input to unconscious modules that implicitly modulate cells that underlie perception (i.e., feed their contribution into the selection of the appropriate perception). No such process occurs between metalinguistic knowledge and implicit linguistic competence. Metalinguistic knowledge cannot feed its contribution to implicit linguistic competence because, unlike the visual parietal maps that modulate visual features of color, its content is incongruent (fundamentally differs from) that of implicit linguistic competence; therefore it is irrelevant (the explicit rule is not isomorphic5 in any way with the implicit procedure; it cannot select, replace, or affect it by any means).

4.  The output of semantics, namely the retrieval of meaning, is explicit in a way that the output of syntax is not: One is conscious of the overall meaning of a sentence (which combines phonological and morphosyntactic processes) but not of the syntax or phonology per se. There is no overt output of syntax in isolation. What we perceive is a combination of the outcome of all the modules involved, from which the meaning is automatically abstracted. We are conscious of the message, not of the underlying computations that generated its form. 5.  “The biases and established states of neural network connections are by themselves not rules or principles, nor can they be understood as algorithmic programs” (Bartsch, 2002: 42).

 Declarative and procedural determinants of second languages

Baars’s (1988) theory distinguishes a vast array of unconscious specialized processors running in parallel, and a single limited-capacity serial workspace that allows them to exchange information. However, this information remains unconscious when either (1) the threshold of consciousness has not been reached (i.e., its activation level is not sufficient), or (2) the representation to be processed is not one that can potentially reach consciousness. Inherently implicit representations are outside both Baars’s global theater and Dehaene and Changeux’s conscious neuronal workspace. They are not part of the set of workspace neurons; they never occupy the workspace. Only the output of implicit modular systems can cross the threshold of consciousness. The only way to manipulate conscious items is to first recall them to consciousness (Baars, 2003). Implicit linguistic competence cannot be called to consciousness; it cannot be deliberately rehearsed, perceived, or reported. It is not temporarily but inherently unconscious. Moreover, metalinguistic knowledge and implicit linguistic competence, by their nature – their difference in kind – are intrinsically incapable of affecting each other’s content or structure. Only the output of implicit competence can interact with metalinguistic knowledge because it is observable and hence conscious. Unlike the unconscious functions referred to in the workspace literature, implicit linguistic competence is inherently unable to influence the activation or inhibition of other functions. Implicit linguistic competence, an automatic function, is rigidly constrained by invariable goals, is impervious to outside interference (however strongly broadcast throughout the brain), and has nothing to contribute to a conspiracy bent on activating a conscious item other than its own output.6 Dienes and Perner (2003) propose that there is a hierarchy of explicitness and argue that each step in the hierarchy is related to the unity of consciousness. However, representations that are intrinsically not amenable to becoming explicit (such as implicit linguistic competence) cannot constitute a step in such a hierarchy. Linguistic competence does not take part in the unity of consciousness by virtue of its implicitness (Dienes & Perner, 2003: 214). The notion of a hierarchy of explicitness is applied specifically to “what it is to have knowledge” (p. 217). All aspects of knowledge can be made explicit (p. 217). Thus, we could speak of implicit knowledge (i.e., temporarily implicit aspects capable of being made explicit) as opposed to implicit competence (all of whose aspects are inherently,

6.  Remember that the concepts (the input to) and the uttered sentences (the output from) the modular implicit linguistic competence system are explicit. The concepts to be verbalized do not modify the linguistic system, they only select the most suitable among the lexical and grammatical items available in the modular system (Paradis, 2004).



Chapter 2.  Consciousness in L2 appropriation 

and hence permanently, implicit and cannot in principle be made explicit under any circumstance). Implicit competence procedures do not form a mental state of which we are conscious; we have only inferential knowledge of them. They never enter the stream of consciousness. 1.1  Only a subset of explicit representations are active at any given time Awareness is the consequence of memory representations being the focus of attention (Frensch et al., 2003). “Assuming that a concept comprises all the knowledge that an individual possesses about a thing or event, it is never activated in its entirety at any given time. Only those aspects that are relevant to the particular situation in which it is evoked are activated (Damasio, 1989). Thus, the exact same portion of the relevant neural network is not activated every time that a given word is heard or uttered” (Paradis, 2004: 200). The neurons that subserve explicit representations constitute the set of workspace neurons. At any given moment, the state of activity of the workspace is characterized by the intense activation of a subset of workspace neurons (that represent the content of consciousness), while the rest of workspace neurons are actively inhibited (Dehaene & Changeux, 2004). Thus, at any given time, either we are, or we are not, conscious of a given item. 1.2  The threshold of consciousness Whatever goes on below the threshold of consciousness is by definition not conscious. There may be levels of analysis, but not degrees of consciousness. We are either aware of something or we are not. “Access to the workspace is all-or-none and exclusive of other representations” (Dehaene & Changeux, 2004: 1153). No matter how close a representation gets to being activated, if it does nor reach threshold, it does not become conscious – its activation level may contribute to the summing up of several sources of activation/inhibition, but until a particular representation reaches threshold, it does not attain consciousness. Even then, it does so only if it is of a kind that is susceptible of reaching consciousness. Indeed, a certain type of representation is either susceptible of being brought into awareness or it is not. As Dennett (1981) put it, “we do not have access to many things that various parts of our nervous systems are shown to have access to” (p. 151). Below the threshold of consciousness, fast propagation of sensory information can occur, but without causing a conscious experience (Dehaene & Changeux, 2004). Subliminal processing can be extensive but remains confined to specialized processes, whereas it is the entry of inputs into the global workspace that constitutes the neural basis of access to consciousness (p. 1147). Specialized (modular) processes, attuned to the processing of a particular type of information

 Declarative and procedural determinants of second languages

(e.g., syntax, phonology), all share characteristics of automaticity and fast feedforward processing.7 The activity within these modules does not enter consciousness. Inherently implicit representations are barred from entry into the workspace. Fast propagation of sensory information can occur below the threshold of consciousness,8 but it does not trigger “global ignition” – that is, cause a conscious experience. Subliminal activity9 from various areas of the brain may result in the activation of an explicit representation (its crossing of the threshold of consciousness), but no amount of such activity, however amplified by attentional, perceptual or any other activity, will cause an inherently unconscious representation (such as implicit linguistic procedures) to become conscious. No matter how much processing goes on, it remains implicit until it produces a conscious perception. Psycholinguistic experiments, such as priming, may demonstrate that the activation of a given representation is heightened, but cerebral activity that affects any given representation remains unconscious unless the activation reaches that representation’s consciousness threshold. Whatever is below the threshold is not available to consciousness and the subject remains unaware of it. (Entities that are not amenable to consciousness will of course never become conscious, as their activation will lead to the accomplishment of an implicit procedure.) What seems necessary if one is to have a conscious experience of a visual image, for instance, are widespread exchanges of information between areas processing different attributes of the visual scene that support perceptual grouping (Engel & Singer, 2001; Pascual-Leone & Walsh, 2001; Lamme, 2006a,b). Visual awareness (conscious experience) does not occur until the (unconscious) processes have been completed. Nothing is conscious until the various exchanges of information between the different areas have taken place (Haynes, 2005; Dehaene et al., 2006). But what applies to explicit representations does not hold for inherently implicit ones. Activation of unconscious processes may lead to the activation of an inherently explicit representation but not of an inherently implicit one – only of its explicit output. Degrees apply only to what can be perceived, not to our awareness of the perception. We are aware of the relative strength of a stimulus, namely whether it barely reaches the perceptual threshold or is loud and clear, but the instant that we become aware of a perception, however feeble, we are conscious of it (we are

7.  Feed-forward refers to a kind of system that reacts in a pre-determined way to a particular change in its environment: it responds to a variation without first having to be affected by it. 8.  Only what crosses the threshold of consciousness becomes available for conscious report (Dehaene & Changeux, 2004). 9.  Activity that is inadequate to produce mental awareness, functioning below the threshold of consciousness.



Chapter 2.  Consciousness in L2 appropriation 

conscious of a weak perception10). No matter how many of one’s cerebral areas have their activation threshold raised, until a given representation has reached its threshold, it is not activated and one is not aware of it. Awareness corresponds to the activation of mental representations that reach a required threshold level. But, as argued in the preceding section, not all cerebral processes are capable of being conscious. No level of cerebral activation, however high, will ever allow a representation to become conscious if it does not belong to the set of representations whose nature makes them amenable to reaching consciousness. Schemata underlying implicit linguistic processes are not members of this set. They are, and of necessity (i.e., by their very nature) remain, unconscious processes. In other words, while mental states that are sustained by declarative memory can become conscious, those that are sustained by procedural memory cannot; they are not accessible to awareness under any circumstances. This implies that in order for one to be conscious of something, (1) not only does one need to be awake, perceptive, and capable of thought, will or perception, but also (2) the corresponding mental representations must be of the kind that are processed in declarative memory. Consciousness depends not only on the vigor of sensory inputs, pre-existing weighted connections, degree of competition with other stimuli, and the release of chemicals such as peptides, hormones, and amines, of which the size of the synchronous assemblies of brain cells are an index (Greenfield & Collins, 2005) but on the inherent capacity of a particular representation to become conscious – a sine qua non. No matter how many millions of neurons synchronize in activity over any period of time, no type, number, or concentration of chemicals can force an inherent implicit representation to become conscious. Consciousness applies only to the range of what one can explicitly know. Some underlying implicit processes, such as the communication between neurons or sensory or motor cell assemblies, may contribute to the activation (the “firing”) of an inherently explicit representation but (1) they themselves remain unconscious and (2) they will never cause an inherently implicit representation to become conscious.

 onsciousness of input and output but not of implicit processes 1.3  C in between Dienes et al. (1995) report that participants trained on an artificial grammar lacked conscious awareness of the competence they had acquired:  Participants classified items substantially above chance even when they believed that they were

10.  The sounds of the person playing the piano next door may be extremely faint or very loud, but as long as I am able to distinguish between piano playing and silence, I am aware. In other words, awareness arises only when (but as soon as) the perceptual threshold is reached.

 Declarative and procedural determinants of second languages

literally guessing. Yet they had a large degree of strategic control over their competence: Participants trained in two artificial grammars could decide which grammar (of which they had no explicit knowledge) to apply in a test. In other words, Dienes et al.’s participants knew what task they had to perform but not what the applicable rule was. This prompted Dienes and Perner (1999) to ask why we can sometimes control, in limited ways, knowledge (here, competence) of which we are not aware. The situation is comparable to a young native bilingual child who is able to speak English or German according to the circumstance, without having any conscious knowledge of the grammar of either; or of an individual who uses the appropriate motor control procedures when skating or riding a bicycle (including proprioceptive control of balance) without knowing the implicit laws of gravity in either case. We are always aware of the intention to perform a certain task, not of the procedures that underlie the performance. For example, we know what we want to say (we are aware of the information we wish to encode) but not how we grammatically encode the message in its on-line verbalized form. Executive control processes (the intention to do X) set in motion the automatic implicit computational procedures but do not control their operations. The speaker is aware of the desired outcome, then of the actual outcome, but not of the automatic procedures that served to achieve that outcome. This lack of awareness is illustrated by the fact that two radically different types of inferred explanations can co-exist, such as generative grammar on the one hand, and connectionist networks on the other, with radically different types of rules or types of networks within each camp. The proposed neurolinguistic theory of bilingualism holds irrespective of which view turns out to be correct, if either of them does. The actual processes will remain unconscious and automatic, just as the processes of speech articulation do even after one has been made conscious of them by X-rays of the articulatory system (see note 1). Knowing how things are done does not necessarily give one the ability to implement these processes automatically, as illustrated by the anecdote reported in Paradis (2004) of a Brazilian instructor explaining to a group of native English students how they produced the sounds of their language, but doing so with his Portuguese accent. Note that afterwards, even the native speakers who had successfully learned how these processes worked continued to speak their language automatically, without awareness of the milliseconds of vibration of their vocal flaps or the exact position of their tongue relative to their palate, teeth, or cheeks – and he continued to speak with his foreign accent. All that one is aware of, that can be observed (and hence can be noticed), is the input (what explicit linguistic material the acquirer receives) and the output of the implicit linguistic competence system (in the form of utterances). We are also aware of what we wish to say (the conceptual content of the message to be



Chapter 2.  Consciousness in L2 appropriation 

expressed – i.e., the input to the implicit linguistic computational system) and of the conceptual content of any linguistically and pragmatically decoded message (i.e., the conceptual meaning of heard utterances). Awareness is constrained by the limits of what can be perceived, observed, noticed, or felt. Something must be noticeable to be noticed, observable to be observed, and perceptible to be perceived. For entities that are amenable to consciousness, information processing among layers of neurons results in the conscious experience of something other than the processing mechanisms themselves, which remain implicit. We are conscious of the result, not the structure of the process. The interaction between conscious and unconscious aspects of verbal communication is limited to the explicit conceptual input to, and explicit output from, the implicit linguistic competence system. Some researchers insist that, whereas consciousness is largely absent from the acquisition (“implicit learning”) process, it is not totally absent (Reber, 1997). Indeed, learners are alert and conscious, not asleep or comatose; they hear the utterances addressed to them – but they are not (to any degree, however minimal) conscious of the process of acquiring the underlying structural characteristics of the utterances they are hearing. 1.4  Consciousness and working memory Memory functioning is composed of distinct subsystems, only some of which are involved in consciousness; working memory is the short-term retention of items held in consciousness for immediate use (Young & Pigott, 1999; Stewart, 2002). Dehaene and Changeux (2004) have established a clear link between the content of working memory and of consciousness. Consciousness recruits specialized unconscious networks that carry out detailed working memory functions (Baars & Franklin, 2003) in the same way that a conscious intention to communicate a message recruits the implicit language modules to verbally encode that message. “In Global Workspace theory, conscious working memory elements recruit unconscious resources to carry out their jobs” (Baars, 2003: 11). These unconscious resources remain unconscious and do not enter the workspace or conscious working memory. They merely contribute to the sum of positive and negative impulses that results in the activation of an entity that is inherently capable of giving rise to a conscious experience. Working memory activities are always mediated by conscious elements (Baars, 2003: 19). All active components of cognitive working memory are declarative, hence conscious (Baars & Franklin, 2003): “all active working memory operations involve consciousness” (Baars, 2003: 12); thus, unconscious processes are out of consideration. Working memory is awareness (Osaka, 2003: 28); the content of working memory always enters awareness (Frensch, et al., 2003); working memory requires

 Declarative and procedural determinants of second languages

conscious processing (Baars, 2003). The processes of working memory are therefore closed to implicit linguistic competence representations:  There is no possibility that inherently implicit and explicit representations can interact in working memory – a fact essential to the point made in Chapter 3. The conscious components of working memory help mobilize and guide unconscious routines that carry out working memory functions (Baars & Franklin, 2003), just as the conscious components of verbal communication (the intention to communicate a conceptual message) mobilize unconscious procedures (the implicit schemata of the procedural system) that carry out verbal communication functions. Voluntary manifestations of working memory functions (Baars & Franklin, 2003) correspond to the voluntary manipulations of verbal communication functions. A “fringe conscious” experience (Baars & Franklin, 2003), such as the intention to recall and report working memory items, corresponds to a fully conscious intent, for which the conscious content has not yet been recovered. Note that only explicit content (i.e., capable of being made conscious) can be successfully recalled in working memory. It is therefore not surprising that working memory is mediated in L1 and L2 by a unitary (frontoparietal) neural system (Xue et al., 2004b). Verbal working memory is essential to most conscious cognitive processes, whether they involve categorization, problem solving, deduction, or any form of reasoning (Jonides, 1995).

2.  Perception, attention and noticing Attention is a prerequisite for consciousness (Dehaene & Naccache, 2001). Inasmuch as perception is defined as the apprehension or recognition of sensory stimuli, it involves awareness of what is perceived. In fact, perception is often defined as the conscious mental registration of an input sensory stimulus (e.g., Dorland, 2003) or even the condition of being aware (e.g., Parker, 2005), or becoming aware of something via the senses (Miller, 2007). Therefore, “perception without awareness” (Merikle, Smilek, & Eastwood, 2001) can only refer to neurological processes (such as the implicit tallying of the frequency of occurrence of an item proposed by Ellis, 2002, 2005, 2006) that leave an implicit memory trace. So, when it is said that unattended stimuli are perceived even when there is no awareness of perceiving (Mack & Rock, 1998), strictly speaking, these stimuli are not perceived but implicitly registered by the brain. To be perceived, a stimulus must be consciously detected by the subject. Perception refers to those subjective (hence conscious) experiences of objects and events that ordinarily result from stimulation of the receptor organs of the body (Parker, 2005). It is the mental organization and interpretation of sensory information (Columbia Electronic Encyclopedia, 6th edition, 2003) of which the subject is conscious.



Chapter 2.  Consciousness in L2 appropriation

Attention is a necessary condition for noticing and noticing is a necessary condition for learning, but neither attention nor noticing is a condition for acquisition: What is attended to and noticed is different from what is acquired. In other words, acquisition is truly incidental. 2.1  Attention in second language acquisition and learning The indiscriminate use of the terms acquisition and learning, language and competence (especially in contexts where authors acknowledge the difference) can only lead to confusion. Each time one of these words is used, the reader must wonder whether the author claims the statement to apply (only) to learning or (only) to acquisition, or indifferently to both. For example, in the statement “SLA is largely driven by what learners pay attention to and notice in target language input” (Schmidt, 2001: 3–4), the A in SLA stands for acquisition. Since we are in the context of second language instruction to adults, the statement is true of learning, as stated – though only indirectly, by providing means to consciously construct correct utterances whose frequent use will allow the implicit acquisition of the procedures corresponding to their underlying structure, as will be discussed in the next chapter. What is actually driven by what learners pay attention to and notice is explicit metalinguistic knowledge. On both the Universal Grammar account and the connectionist account, acquisition is unconscious (Schmidt, 2001: 4). Noticing is the necessary and sufficient condition for the conversion of input into intake for learning (Schmidt, 1993) – but only for learning, since what is acquired is something other than what is available to observation. Schmidt (2001) suggests that “learning lexicon and morphology require attention and awareness in ways that learning syntax does not” (p. 24). Indeed, vocabulary, insofar as the meaning-form relationship between words is consciously known, is subserved by declarative memory and requires attention to be learned.11 The morphosyntactic properties of these words, internalized as part of the lexicon, remain implicit in both the native and the second language until, possibly, they are taught as part of the grammar curriculum in school or in Linguistics 101. Tomlin and Villa (1993) argue that detection is necessary before other cognitive processing (storage in memory, hypothesis formation, etc.) can occur. Again, this is true of explicit learning, not of implicit acquisition. Following Paradis’ (2004) proposed neurolinguistic theory of bilingualism, detection or focal attention entails conscious registration of what is detected. If there is no conscious 11.  Even when vocabulary is learned from context, relying on reading and listening, once the words are learned, they become explicit items (both in L1 and L2).



 Declarative and procedural determinants of second languages

registration, then “detection” is not explicit and can only refer to the implicit tallying (Ellis, 2002) of the underlying form – or to any other implicit acquisition device. More noticing leads to more learning (Baars, 1988), but not to acquisition, as demonstrated by the tens of thousands of learners of a second language who have learned the syntactic rules, have noticed the inflected forms, know (and can describe) the mechanisms used in articulation and phonation, but have not internalized the corresponding grammar in the form of implicit linguistic competence and therefore continue to speak with a foreign accent, using faulty syntax and attaching the wrong inflections. Paying attention to input may help learners become aware of the mismatch between what they produce and what native speakers of the target language produce. As soon as one is no longer able to detect any discrepancy between a second language learner’s output and that of native speakers, learning can be said to be at an end (Klein, 1986) – but not necessarily acquisition. Negative evidence that is noticed may correct the learner’s explicit grammar and help produce correct forms but it does not directly lead to a realignment of the internal grammar. It is not unknown for learners to be told time and time again that what they produce is incorrect and yet to continue to automatically produce the incorrect form, even though they have explicit knowledge of the correct form and can say what it is. White (1992), for instance, concedes that accurate performance subsequent to negative evidence provided in the classroom does not necessarily indicate changes in L2 implicit linguistic competence because it may reflect explicit learning instead. Schmidt (2001) asks whether, in order to acquire language, it is sufficient for learners to focus their attention on meaning, while they pick up the message form without paying attention to it. Whereas this is the case when children acquire their first language, it may no longer be sufficient in any practical sense for adults attempting to acquire a second language. They may need to have their attention drawn to some features of the input so that they may consciously focus on the form and thus (explicitly) learn it. In that way, they will be able to consciously construct sentences containing that form, and, through much practice, eventually (and independently) internalize the form as part of their L2 competence (Ellis, 2005), as will be discussed in Chapter 3. Note that the acquisition process per se is the same: An item is acquired by dint of using it, though it is more difficult for adults to do so directly through interaction with speakers of the target language, as children do without the need for prior metalinguistic knowledge. (This speaks to the issue of an optimal period for language acquisition, which will be discussed in Chapter 4.) Schmidt contrasts this position with Gass’s (1997) argument that apperceived input that is processed only semantically and receives no syntactic processing will not lead to development of syntax. This view is not incompatible with attention’s



Chapter 2.  Consciousness in L2 appropriation 

not being focused on the form of input: While children focus on the meaning of the utterances addressed to them, the implicit intake consists in abstracting the underlying structure that leads to the development of what linguists describe as (implicit) syntax. It is not the (conscious) meaning that leads to the development of syntax but the (non-conscious) intake, as we are about to see.

3.  Explicit input is not implicit intake Becoming conscious of material is indeed “the sovereign remedy for learning anything” as Baars (1997) elegantly puts it (cited by N.C. Ellis, 2005: 312) – but not for acquisition: Incidental acquisition relies on attention to something other than what is internalized, and what is acquired is often not even there to be noticed (or attended to) in the first place. Becoming conscious of some material is a requisite for learning (Ellis, 2005: 312) but is outside the scope of implicit acquisition (Ellis, 2005: 309). What is attended to is not what is internalized: For example, the articulation of language sounds is acquired through proprioceptive feedback of which we are not aware. What we attend to are acoustic properties while what we acquire are implicit motor programs for articulation and phonation. Note that, in perception, one does not pay attention to the underlying acoustic component features of speech sounds either, but to their globally perceived surface sound, that is, without conscious analysis of the distinctive acoustic features of the waveform, such as particular formants, noise bursts, and so on. Thus, what is attended to is entirely different in nature from what is internalized. The input is explicit (it is perceived) but the intake (i.e., what the acquirer picks up from the input) is implicit (it is not observable and not noticed). As a possible analogy, in breathing, air is the input, while oxygen is the intake. The explanation in Physiology 101 of the process by which oxygen is abstracted from air does not help you breathe. We are no more conscious of abstracting the underlying structure of perceived sentences than of extracting oxygen from the air we breathe. The metalinguistic knowledge we learn does not serve as intake for establishing implicit linguistic competence either. According to Schmidt (2001), what learners notice are utterances or parts of utterances that are exemplars of higher-level categories and principles of the linguistic system, but not the principles or the system itself. So what is noticed (the input and/or its explicit analysis) is not the intake (what is internalized). Ellis (2005: 323) argues that if a cue in L2 is not noticed, it will not be tallied by implicit acquisition through use thereafter. To be tallied, something must be used. The underlying structure of what is used is what is internalized. So, if an element

 Declarative and procedural determinants of second languages

is not used, its underlying structure will not be internalized. Note that it is not the element itself (the input) that has an effect but the implicit tallying of its abstracted underlying pattern (the intake). What is tallied is the number of times the underlying structure is implicitly abstracted, not the surface form, because (1) the surface form of each of the various exemplars contains different lexical items,12 and (2) what is extracted is not the construction as it is explicitly formulated, but its underlying structure,13 on which attention is not focused when it is tallied. Moreover, the same surface structure may have different underlying structures. The difference is not apparent on the surface. For example, the surface structure the boy hit the man with an umbrella has a different underlying structure – and corresponding meaning: [The boy hit the man] [with an umbrella]; [The boy hit] [the man with an umbrella] – depending on whether the boy did the hitting with an umbrella or hit the man who was holding an umbrella; only the latter has the same underlying structure as the boy saw the man with an umbrella. The word order is the same, yet the utterances with the first interpretation and the second interpretation are tallied as distinct constructions.14 The nature of what is tallied differs from that of the perceived exemplars (i.e., each sentence with its different words) and from their surface grammatical forms (but instead, the corresponding implicit procedures). The fact that native speakers as young as four years of age automatically (and accurately) inflect non-words that they have never heard before (Berko-Gleason, 1958) demonstrates that their ability to handle the underlying rule does not depend on the frequency of occurrence of combinations of specific words, but on their abstracted underlying properties. These children have internalized the subtle phonemic contexts that induce voicing, devoicing, or adding an epenthetic

12.  The following sentences share the same surface structure but each contains different words: The horse jumps over the fence; The cat sneaks under the cover; The shutter bangs against the wall; The sewage empties into the river. 13.  “Our ‘know how’ capabilities are not processes governed by rules, though the regularities in their results may be described, in idealization, as rules … idealizations of the relationships between the results of the processes that in reality take place without being governed by rules” (Bartsch, 2002: 31). 14.  Besides, one does not pay attention to the form of utterances, but to their meaning: One does not remember sentences verbatim, only the message they convey. A bilingual may not even remember which language the sentence was produced in, in the absence of a compelling pragmatic context, which may help one to figure out which language it must have been; even then, that does not help recover the utterance itself (see Haritos, 2003). Adult native speakers are rarely able to repeat just-heard complex sentences that are more than 15 words long even though they were well understood (Tomioka, 2002).



Chapter 2.  Consciousness in L2 appropriation 

vowel to the voiced form (/z/, /s/, /6z/) when marking the plural of nouns; they have not learned that some specific previously encountered words take one form while other specific words take another form – rather, they have abstracted an implicit rule. In sum, the surface form, as it is consciously perceived, is not what serves as intake. The establishment of the strengths of the various connections of the neural network underlying the implicit competence procedures or schemata proposed by Ellis (2002, 2005) is not based on the surface structure but on the implicit tally of features that are not attended to and hence not noticed: We do not spend our time counting the units of language; mechanisms underlying unconscious counting do it for us (Ellis, 2006). Information about the underlying structure of a complex stimulus environment is implicitly acquired by a process that takes place without conscious operations (Ellis, 1994b, 2002). According to Ellis (2002, 2005), acquisition takes place by implicitly tallying the frequency of occurrence of each form, necessarily irrespective of the actual words that constitute that form on any occasion of its occurrence: What can be observed (the input) are actual strings of words, not the structure they actualize. Unlike what happens with the learning of words, which is sustained by declarative memory, the stimulus that is attended to is not the material that lays down the underlying grammatical structure; instead, this is a function of the frequency of its (abstract) underlying form. Grammar acquisition is not mere sequence learning (as would be the case if the observed surface form served as intake). Note that in English, for instance, the same form can be a noun, a verb or an adjective and the underlying structure of the sentence that includes it changes accordingly (e.g., some paupers live in houses; this shed houses paupers.) The objects of attention and noticing are elements of the surface structure of utterances in the input, rather than the abstract rules or principles of which such instances are exemplars (Schmidt, 2001). In other words, rules are not the target of noticing. Schmidt explicitly distinguishes between attention to the surface structure of utterances and understanding of the abstract rules and principles. The former corresponds to the input, the latter to the abstracted underlying structures that constitute the intake which results in implicit linguistic competence. If attention is the mechanism responsible for access to awareness (Baars, 1988, 1997), then it cannot be used to access intake (whose stimuli are not represented in conscious awareness), which results in implicit linguistic competence. “The spotlight or zoom lens” of the global workspace cannot focus on any portion of intake – whether the contents or the process. The individual pays attention to the input but not to the intake. The elements that serve as intake are not noticeable. The learner may be aware of the structural regularities of the surface structure of the language and formulate constituting

 Declarative and procedural determinants of second languages

metalinguistic reflections that do not contribute to the contents of intake either. Learners may only make explicit hypotheses and test them against further input, thereby increasing their knowledge without directly contributing to intake. Explicit components of input may become intake for learning, but only implicit components of input become intake for acquisition – not what is noticed, but what is implicitly abstracted. As will be discussed at length in the next chapter, what is noticed results in metalinguistic knowledge that allows the learner to construct correct sentences whose use then serves as a means to implicit intake. Implicit aspects of input are not noticed but are non-consciously registered. They serve as the basis for the development of implicit procedures that will allow the automatic comprehension and production of novel utterances corresponding to the surface structure of the particular input.15 3.1  The double implicitness of intake Learners are not conscious of either the underlying structures or their statistical frequency of occurrence: (1) Explicit constructions are not the material upon which some implicit mechanism performs some transformation, conversion, manipulation, adjustment, alteration, modification, or tuning. (2) The frequency with which the underlying structures of utterances are abstracted is what provides intake, and that too is implicit: Intake is the result of “an unconscious frequency analysis” (Ellis, 2006: 104). It is generally agreed that not all input serves as intake:  Only a portion of input (i.e., of all the language stimuli that a learner perceives) is used as intake (i.e., as what is internalized or integrated into the speaker’s linguistic competence). Intake is what goes in (Corder, 1967), the process of assimilating linguistic material (Gass, 1988), and thus what leads to the formation of implicit competence. It is not just a subset of input; rather the two terms refer to two fundamentally different phenomena (Gass, 1988). It is not merely that only a certain percentage of input is used as intake, but that what serves as intake is actually not perceivable in the input. The intake cannot be noticed because it is not observable; it is only implicitly abstracted. The explicit utterance that is perceived (its surface form, made up of a sequence of specific words) does not serve as intake; its implicit underlying structure does. Rarely are any two utterances composed of the same string of words. What is abstracted from these ever-changing surface forms is the

15.  For example, given the inputs: I am looking forward to hearing from you; he is looking forward to receiving your letter; she is looking forward to seeing you again; etc., a French learner of English may eventually automatically produce:  They are looking forward to returning home (as opposed to: *they are looking forward to return home).



Chapter 2.  Consciousness in L2 appropriation 

common underlying syntactic structure whose features (e.g., government and binding relationships) are not apparent. These implicit relationships are abstracted either as generative rules (as proposed by generative linguistics) or as statistically based weighted connections resulting from tallying not the frequency of occurrence of actual sentences (made up, as they are, of different words each time) but of abstracted structural commonalities. Exemplar-based connectionist models (Ellis, 2002, 2005) do not tally actual word strings but the underlying properties they share; (if they were to tally the actual identical surface forms, they would hardly ever get a count of greater than one and hence the weight of the connections would never build up. Only stereotypes occur repeatedly in the exact same words, and they are stored declaratively as large lexical items – they are outside of the generative grammar or the statistical connectionist network. An item-based schema (Ellis, 2005: 329) is not built on the repetition of word-for-word identical sentences, but on the frequency of occurrence of the shared underlying structure that is implicitly abstracted by connectionist mechanisms. In order to establish a schema or set of connections that permits the generation of novel sentences on the basis of tallying a large number of sentences, each of which contains different words, the acquisition device necessarily relies on an abstractive mechanism. Intake is thus not a subset of input. It cannot be regarded as “that subset of input that actually gets in” (Ellis, 2008: 238). None of what is perceived “gets in” as it is. Rather, the intake is abstracted from the perceived input;16 it is not a subset of what is perceived. The intake does not consist of any part of the input as such, but is the result of an implicit process of tallying the frequency of the input utterances’ implicit underlying patterns (and those patterns can only be inferred). The perceptible input, or a subset thereof, may contribute only to explicit knowledge. Noticing does not bear on what is internalized but on explicit forms from which the underlying grammatical representation is abstracted. (Most of the intake is not observable; it is not explicit and it is not noticed at any time.) Thus, the intake is not what is noticed, observed or focused upon in the input. The input consists of utterances, but the intake is not the perceived utterance. It is the tallied characteristics of the utterance’s implicit underlying structure. Note that both the tallying and the underlying structures that are tallied, that is, both the process and the representation, are implicit. Nothing explicit serves as intake; hence there is no direct connection between explicit knowledge or explicit processes and implicit competence or implicit processes, as will be argued in the next chapter.

16.  Acquirers statistically abstract information from masses of input data as generalizations from stored exemplars; this frequency-biased abstracting of schematic constructions is an unconscious process (Ellis, 2006).

 Declarative and procedural determinants of second languages

4.  Neurobiological and neurochemical bases of consciousness Ullman (2006b) reports what might appear at first glance to be two separate possibilities:  either there may be functional sub-specialization within the circuitry passing through posterior/dorsal regions of Broca’s area (BA 44), with distinct tracks subserving each function, or these distinct functions may represent different aspects of the same underlying function, namely procedural memory. As he emphatically remarks, “the existence of such procedural system does not imply that all its parts have the same functional roles” (p. 482). The apparent alternative (either distinct functions or one underlying function) finds a plausible account in the neurofunctional modularity of procedural memory and the subsystems within neurofunctional modules (Paradis, 2004), a solution implied in Ullman’s remark. One of the characteristics of procedural memory is that it is task-specific. Indeed, there is a high abstract level at which procedural memory constitutes an umbrella for a particular kind of function; its processes are non-conscious and automatic, as opposed to declarative memory whose contents are capable of being brought into consciousness and of being consciously controlled. However, each procedural function is neurofunctionally modular and impenetrable to interference from other functional domains and it is not implausible that it should also be microneuroanatomically modular. Ullman (2006b) in fact does not preclude the possibility that further topographic and functional sub-segregation may exist within the procedural circuitry:  different domains “may each be subserved by distinct parallel sub-circuits within procedural memory” (p. 483). The representation of knowledge and the representation of skills are different in nature, involving different cerebral structures and probably different neurotransmitters and other neurochemical agents (Ullman, 2001; Stewart, 2002). The brain creates the unity we perceive out of the diversity of perceptions that impinge on our senses. We are aware of each independent perception and of the (centrally) constructed whole, but not of the workings of the underlying mechanisms that result in the perceived unity. The hippocampus and related memory structures bind or integrate into a memory trace the neural elements that mediate a conscious experience. Memory mediated by non-hippocampal structures cannot be conscious; these structures can influence behavior but without giving rise to conscious recollection. Two fundamentally different forms of memory are mediated by hippocampal and nonhippocampal structures, respectively; explicit, but not implicit, memories are accompanied by consciousness (Moscovitch, 2004). Consciousness is bound to the information-containing elements of the explicit memory trace, whereas implicit memories are not mediated by hippocampal structures; they are processed and elaborated elsewhere (Young & Pigott, 1999). Unconscious processes can in fact



Chapter 2.  Consciousness in L2 appropriation 

be found in diverse areas throughout the brain (Hardcastle, 1995). In the next chapter, it will be claimed that these two systems, which respectively sustain explicit metalinguistic knowledge and implicit linguistic competence, do not – because they cannot – interface. Cognitive functions that give rise to conscious experience occur throughout the brain, depending on which features of perception reach their activation threshold. Consciousness depends upon the integration of affect and motivation with representations of sensory, motor, and symbolic information (Damasio, 1994; Luu et al., 2001). Neural correlates of conscious states must have access to explicitly encoded information and directly project into the cerebral areas involved in planning and explicit control, such as prefrontal cortex and anterior cingular cortex (Crick & Koch, 1998).

5.  Conclusion We have established that knowledge and competence operate independently of one another; that explicit representations are stored in an explicitly accessible form and that their activation gives rise to conscious experiences; that consciousness is restricted to this type of representation; and that explicit, but not implicit, memories may be accompanied by consciousness. There are thus two types of representations: inherently implicit and inherently capable of becoming explicit. Nothing outside of implicit linguistic competence can have an influence of any kind on the structure of the grammatical system. Outside entities (such as pragmatic considerations) may influence the selection of particular morpho-syntactic elements, but cannot alter their structure. By definition, nothing implicit (such as the underlying structure of utterances) can be observed, let alone noticed. Within N.C. Ellis’s framework, it is the implicit tally of the frequency of occurrence of these non-observable underlying structures that must serve as intake in establishing the implicit procedural system. Linguistic competence is assumed to emerge from the implicit tallying of the frequency of occurrence of the structures (intake) underlying perceived surface constructions (input) – not of the input itself, which is explicit. Acquisition thus takes place without awareness by an unconscious process of induction that results in implicit competence. We are aware of neither the content nor the process of intake. The creating of implicit hierarchical organizational structures is based on manipulating not the perceived input but the abstracted (non-observable) intake.

 Declarative and procedural determinants of second languages

What you heard may alter your (declarative) memory of what you saw (and vice versa), but nothing other than intake from (oft-repeated) linguistic input, perceived during meaningful communicative interaction, can alter the schemata of implicit linguistic competence. Neither the various steps toward learning by different methods nor the resulting explicit knowledge directly serves as intake to the acquisition mechanism, which is based not on knowledge, nor on noticing, nor on any conscious operation, but on statistical tallying of the use of constructions (whether or not the learner is aware of their explicit grammatical structure). One does not intake a portion of the explicit input. The intake corresponds to something that is not visible on the surface. Intake is doubly unconscious: both its contents (what is tallied) and its mechanism (the tallying itself) are implicit. Acquisition remains incidental:  there is no intake of explicit material from the noticeable input. Some components of a verbal communication act are conscious, others are not. One is aware of the desired result and of the outcome but not of how the outcome is brought about (one is unaware of the workings of the underlying mechanisms involved). In both novice behavior (for which controlled processing is required) and expert behavior (which can be carried out automatically, without attention), the decision to behave in a certain way may be conscious and deliberate and require attention, but the execution of the automatic behavior itself does not require attention.

chapter 3

The disintegration of the explicit/implicit interface debate (or interface newspeak?) In this chapter it will be argued, contra the claims of N.C. Ellis (2005) and Ellis and Larsen Freeman (2006), that (1) an interface between explicit metalinguistic knowledge and implicit linguistic competence cannot exist in consciousness (as claimed by these authors); (2) the so-called “dynamic interface” is no interface; and (3) an indirect influence, no matter how substantial, is not an interface. But first, let us look into what the term interface means – what it is used to refer to – so as to be able to continue meaningful, collaborative research, and so that all investigators involved will know what their colleagues are talking about when this term is used in a particular publication.

1.  The meaning of interface By and large, an interface is the locus of interaction between two systems, the overlap where two systems affect each other. The term interface is also used to refer to a link that connects two systems and allows for the communication of data between them. In either case, interface implies the conversion of signals from one form to another; a transmission of information from one system to another; some interaction between the two; or some form of interoperation.1 The interface

1.  The term interface began to be used in the sixties in the computer industry to designate the point of interaction between two systems. In generative linguistics, interface usually refers to the locus of interaction between the computational systems and the systems that govern the interpretation and use of the output of those systems; it refers to the way in which two or more systems interact (Unsworth, 2004). The interaction between the three grammatical modules (syntax, lexicon, semantics), as well as the interaction of these modules with the pragmatic system, is often referred to as an “interface” (Bos, Hollbrandse, & Sleeman 2004). In their discussion of the syntax-morphology interface, Baerman, Brown, & Corbett (2005) imply that there are interactions between the core components of morphology and syntax. As for the lexicon-syntax interface, van Hout, Hulk and Kuiken (2003) consider precisely how the lexicon and syntax are linked; the kind of information they exchange; the relationship between lexicon and syntax.

 Declarative and procedural determinants of second languages

enables systems to share information or exchange data. An interface connects systems that are not directly compatible; it converts signals from one system to another; it is the interaction between two elements. There are therefore basically two definitions of interface: (1) There may be a direct interface between two systems: A is said to interface with B (Figure 3.1). A

B

Figure 3.1.  Direct interface A demarcation between two systems where signals meet and interact. A shared boundary across which two systems communicate.

(2) An item (I) that connects two systems (A, B) is said to be an interface between the two systems (Figure  3.2). Either definition implies a direct relationship: in the first instance, between A and B; in the second instance, between I and both A and B. In Figure 3.1, System A interfaces with System B. In Figure 3.2, System A does not interface with System B, but both A and B interface with device I (called the interface). A

I

B

Figure 3.2.  Connection via an interface link Connections between A and B are possible only via the interface device I which interconnects systems that are not directly compatible; it converts signals from A format to B format.

We shall see that, in the context of an alleged interface between explicit conscious metalinguistic knowledge and implicit non-conscious linguistic competence of a second language, neither (1) nor (2) applies. Explicit knowledge and implicit competence do not share information; they do not exchange data; they do not interact. In keeping with the generally accepted characteristics of an interface, there are a number of ways by which one might wish to claim that there is an interface between explicit metalinguistic knowledge and implicit linguistic competence in the process of L2 acquisition. (1) The output of explicit knowledge serves as the input to implicit competence (e.g., as the term is used in the syntax-semantics interface); (2) explicit knowledge has a direct influence on implicit competence (e.g., as the term is used in chemistry); (3) there is a device that interfaces between the two systems, which are not directly compatible, and that transforms the signal from one to the other (e.g., as the term is used in computer science). In all types of



Chapter 3.  The explicit/implicit interface debate 

interface, there is direct contact, followed by an effect. An interface implies some form of direct interaction, interconnection, intercommunication, data exchange between systems, sharing of information, or signal conversion – none of which exists between explicit metalinguistic knowledge (or learning thereof) and implicit linguistic competence (or acquisition thereof). Claiming to demonstrate an interface between metalinguistic knowledge and implicit linguistic competence, N.C. Ellis (2005) in fact reviews numerous means by which one learns a second language by explicit means, but never shows how or when an interface occurs, that is, how any of the explicit ways of learning interfaces with the acquisition of implicit linguistic competence, or how any element of explicit metalinguistic knowledge interfaces with an underlying implicit procedure, or what such interface would consist of. At times, the text might inadvertently lead the reader to believe that a demonstration is given, but this would be an illusion. Therefore, let us consider the arguments one by one, so as to dissipate any possible ambiguity. According to Ellis, we acquire language by using language. Implicit acquisition occurs during fluent comprehension and production. Explicit learning of language occurs in our conscious efforts to negotiate meaning and construct communication. The bulk of language acquisition consists in implicit acquisition from use. Linguistic competence emerges in the spreading activation of the cognitive unconscious. Over thousands of hours, unconscious acquisition processes occur automatically during language use. “Explicit and implicit knowledge are distinct and dissociated; they involve different types of representation and are substantiated in separate parts of the brain … Explicit knowledge does not become implicit knowledge, nor can it be converted into it” (p. 307). If something cannot be converted into something else, by definition, there is no interface. Yet, what Ellis then proposes is that “Nevertheless, there is interaction. However unalike they are, these two types of knowledge interact” (p. 307). This seems to contradict what precedes and most of what follows. By carefully going through the arguments and demonstrations provided by Ellis, it will be shown that, by his own account, there cannot be a direct interface between metalinguistic knowledge and implicit competence. During the appropriation of an L2, the use of competence may replace the use of metalinguistic knowledge over time, as proficiency increases, or in rapid succession, when speakers revert to their use of metalinguistic knowledge to compensate for a gap in their incomplete implicit competence. This is not an interface but the substitution of the use of one mechanism for the use of another. One does not directly feed the other, transmit information to the other, impact upon the other, operate on the other or directly influence the other in any way. This is not to deny that explicit metalinguistic knowledge has an indirect influence on the acquisition of implicit linguistic competence, and it will be argued that this is exactly what Ellis demonstrates: Explicit instruction has a direct effect on

 Declarative and procedural determinants of second languages

explicit language learning and an indirect effect on implicit language acquisition. The claim here is simply that there is no direct interface between explicit metalinguistic knowledge and implicit linguistic competence; that the actual contribution of explicit knowledge to implicit competence is indirect – possibly extensive, but indirect. All of Ellis’s (2005) assertions point to exactly this: No piece of knowledge ever has a direct effect on any aspect of linguistic competence. Hence there is no interface in any of the possible meanings of the word. An indirect influence does not have any of the relevant characteristics of an interface. The claim that there is an interface between explicit and implicit aspects of L2 acquisition is not compatible with the premises or any of the data provided by Ellis (2005), as we are about to see. It will thus be argued that, based on the author’s own premises, none of what Ellis (2005) discusses can be interpreted as an interface between explicit metalinguistic knowledge and implicit linguistic competence in L2 acquisition.

 e premises: Learning and acquisition are distinct; explicit 1.1  Th knowledge is not transformed into implicit competence Ellis (2005) stipulates that explicit memories are used in the conscious building of novel utterances and that explicit knowledge and explicit educational activities contribute to the conscious creation of utterances that then partake in subsequent implicit learning and proceduralization (p. 308). Yet, no part of explicit knowledge is proceduralized: “explicit knowledge does not become implicit knowledge” (p. 307). The internalized procedures do not correspond to the explicit rules that allow the conscious building of sentences. The building of implicit procedures is driven, not by the explicit knowledge of how to construct sentences, but by implicitly tallying the frequency of occurrence of items and their collocation (Ellis, 2002, 2005). All of the evidence presented by Ellis shows that there are different memory systems that are independent of each other, each performing a specific type of operation – not that a particular system performs a certain task at the beginning and another task later, after practice, but that some neural structures act on new material and other neural structures work on material that has been repeatedly presented. No single structure performs both tasks. One continues to process new material and the other develops rapid retrieval. As Ellis reiterates, “It is worth emphasizing again that – of themselves – metalinguistic descriptions do not impact implicit language knowledge. They are of a different stuff, and they are stored in different parts of the brain” (p. 331). Likewise, when production is revised thanks to monitoring and correcting, it is not the revision itself that enters into the implicit acquisition process, but the implicit statistical tally of the subsequent use of the structures in question. Hence, there is no direct impact, no direct contact, no interface. Note that when



Chapter 3.  The explicit/implicit interface debate 

Ellis writes, immediately after the above quote, “However, metalinguistic descriptions and other explicit knowledge that come to consciousness at the appropriate moment to influence the processing of language form and its corresponding interpretation do impact language knowledge,” the word however might lead the reader to believe that “language knowledge” here refers to implicit language competence. But it does not:  As quoted above, explicit knowledge has no influence on the structures in implicit competence. You may use either implicit or explicit analysis, or both in parallel, but they do not interact, in that one does not influence the internal structure or modus operandi of the other: Each process provides data, independently, to a different functional system. Comprehensible output, like the various means of guided production, contributes to L2 appropriation. The gist of the story is that the more metalinguistic knowledge you possess, the more correct the utterances you produce will be. The frequency of use (both comprehension and production) of these utterances will help build up your implicit competence for the generation of constructions of that type. The influence of metalinguistic knowledge may be significant, but it is nevertheless indirect. Because there is no direct connection between the knowledge itself (or its means of appropriation) and the implicit linguistic competence system or its development, we cannot speak of an interface. The automatic use of implicit linguistic competence does not correspond to an accelerated use of metalinguistic rules (or anything explicitly known). The underlying neural computational procedures are not the implementation of explicitly known rules. A metalinguistic rule corresponds to a theoretical construct inferred from the observable output of implicit linguistic competence, not to the description of an actual procedure that is implemented in order to generate sentences of a given type. Nor is the generation of language sounds the super-accelerated use of elements of a particular phonological theory. 2.  The so-called “dynamic interface” is no interface In Ellis (2005), in addition to the title, the dynamic nature of the alleged interface is mentioned six times. First, in the abstract:  “The interface is dynamic: It happens transiently during conscious processing.” This claim is repeated on pages 307 and 308:  “Interface is a dynamic process; it happens in conscious processing.” Which component(s) of explicit knowledge transiently interact with implicit linguistic competence during the dynamic process and how they connect are nowhere specified. Third appearance: “At any one time, our state of mind reflects complex dynamic interactions of explicit and implicit knowledge” (p. 313). Unfortunately the

 Declarative and procedural determinants of second languages

nature, locus and manner of the interaction are not specified here either. Dynamic interactions, the nature of which is not identified, are alleged to be reflected in our state of mind. Then comes a quote from Scott Kelso that will be discussed in the section on inapplicable analogies and metaphors below, which does not address the problem of interface; on the contrary, she indicates elsewhere that implicit skills are independent of conscious processes and hence not in phase with them. Fourth appearance: “Metalinguistic information connects with implicit learning, and they meet and interact in processing. It is a dynamic interface” (p. 325). What information connects how with language acquisition is left to the imagination of the reader. Acquisition is a process; metalinguistic information is a representation, a piece of knowledge. Where and how which piece of knowledge interacts with which stage of the implicit acquisition process and the nature of this interaction are not disclosed. Where do they meet? Fifth appearance: “The interface is both dynamic and dialectic” (p. 332). The reader is again referred to various passages in Scott Kelso (2002) that explore the basic forms of biological coordination that arise due to self-organizing synergetic processes, and deal with human brain states and the way information is transmitted from cells to cells. No such coordination applies in this case (as will also be seen in the inapplicable analogies and metaphors section below). Nor is there a dialectic back-and-forth exchange of information between explicit knowledge and implicit linguistic competence. None is described, and none is logically possible given the premises presented on p. 307. Sixth appearance:  On page 339, the alleged interface is now presented as a given:  “Given … the interactions of explicit and implicit language learning in dynamic interface as analyzed in this review … .” However, Ellis’s review is an exhaustive analysis of measures to enhance explicit knowledge. No interactions between explicit knowledge and implicit competence have been described, nor could they be, given Ellis’s own premises. From the contexts in which it is used, let us try to figure out what the denotation of the alleged dynamic interface could be. It appears to refer to the fact that at time T1 an explicit formula is explicitly learned; it is subsequently used until T2, at which time it has been integrated into the implicit memory system as an implicit procedure that will allow the speaker to automatically understand and produce novel utterances of that form. The alleged interface thus appears to be the sum of the events that took place between T1 and T2. What relevant events did take place? Two sets of events have occurred in parallel (Figure 3.3), each subserved by a different memory system with its own nature and characteristic operating procedures: (1) The explicit formula has been noticed, analyzed, consciously decomposed and reconstructed, drilled, etc., in declarative memory; (2) the implicit acquisition system has tallied the frequency of occurrence of the patterns abstracted from the input utterances and has thereby



Chapter 3.  The explicit/implicit interface debate 

established an implicit procedure in a connectionist network whose activation will allow the speaker to automatically understand and produce novel utterances containing the relevant structure. explicit formula taught E-memory

I-memory

T1 | | | | | T1

formula used dynamic interface?

T2 | | | | |

→

→ T2 procedure used

By time T1, an explicit formula has been learned; it is subsequently used until T2, at which time it has been integrated into the implicit memory system as an implicit procedure. No evidence has been provided that contact occurred during that period between elements in explicit memory and elements in implicit memory. Figure 3.3.  Dynamic interface?

Learning is a dynamic process. So is acquisition. But no part of the dynamic process of learning has been shown to directly affect the dynamic process of acquisition.2 The dynamics of the acquisition process are of a different nature, using different types of intake and different kinds of operations. There is never a time in the acquisition process when a component of explicit processes or their outcome is passed on to, or transformed into, competence (as is the case with an interface in the original computer science sense – something in electronic language A is translated so it is readable by electronic language B). Nor does the output of the metalinguistic system serve as input to the implicit linguistic competence system (as is the case with the term as used in generative linguistics – System A interfaces with System B when the output of A serves as input to B). Based on all the evidence and arguments presented, it appears that transient dynamic interface refers to the indirect influence that the resulting knowledge of explicit operations exerts on the development of L2 implicit competence.3 Interface calls for a direct transmission from one system (here, explicit metalinguistic knowledge) to another (here, implicit linguistic competence). There is no direct transmission or conversion (which would contradict the author’s premises) to be found in the 37 pages of text of Ellis’s article. There are a number of explicit

2.  “The dynamics of language learning are inextricably linked to the dynamics of consciousness” (Ellis, 2008: 242) – but the dynamics of language acquisition only indirectly so, in that attention is focused on something other than what is acquired. 3.  A suspicion later confirmed when Ellis proposes that the degree of influence of metalinguistic knowledge on implicit acquisition justifies the claims of interface (pp. 325, 326).

 Declarative and procedural determinants of second languages

processes and outcomes thereof that may eventually have an indirect, albeit significant (“profound”) effect on the development of implicit competence. But, as the author acknowledges, the two are and remain of a different nature and explicit knowledge does not become implicit competence. Never do the twain actually meet. They are used in sequence or in parallel, but do not interface. This becomes an unsolvable conundrum when it is stipulated that the dynamic interface “happens in conscious processing” (p. 308). Explicit knowledge is conscious; implicit competence, by definition, is not. Only explicit phenomena can interface, interact, or connect, whether dynamically or otherwise, in consciousness – unless we have a mechanism such as the pineal gland, as Descartes envisaged it, that converts conscious knowledge into implicit competence (as I had suggested would be necessary); however this is ruled out (Ellis & Larsen-Freeman, 2006: 569). We are left with a self-contradiction. Conscious (declarative) knowledge can be brought into consciousness; implicit competence cannot. Unconscious mental events (inactivated declarative memories) may be activated and become part of current awareness (consciousness), but activated procedural competence does not become part of awareness. Only its output is observable and hence no part of implicit (procedural) competence can interface in consciousness with metalinguistic knowledge, as will be confirmed in the next section. At no point is there direct communication between the two systems or between one system and the other via an interface (i.e., some device that converts elements from one system into elements readable by the other). The two systems are first used sequentially, as metalinguistic knowledge is appropriated before implicit linguistic competence arises; then they continue to coexist in parallel, and the switches from using implicit linguistic competence to using metalinguistic knowledge decrease in number and duration over time, as implicit linguistic competence increases in volume (Figure 3.4). Learn MLK

MLK available for use Use available ILC

→ →

Metalinguistic knowledge (MLK) remains in place but is used less and less as implicit linguistic competence (ILC) increases and recourse to MLK becomes less frequent. Figure 3.4.  Shift in reliance

2.1  No interface but switching from one to the other There is no continuum between conscious control as a property and automaticity. The perceived continuum, namely the gradual increase in fluency in L2 over time,



Chapter 3.  The explicit/implicit interface debate 

reflects a gradual shift from total reliance on explicit metalinguistic knowledge to a progressively greater reliance on implicit competence (when such competence has been internalized through practice). Note that explicit metalinguistic knowledge remains available throughout, but because automatic processes are faster, more efficient, and require less attentional resources, the use of implicit competence is favored, whenever possible. But there is no continuum between metalinguistic knowledge and implicit competence, between controlled and automatic processes. A person progressively recovering from blindness (say, by the removal of congenital cataracts) may gradually replace Braille reading by learning to read visually. Braille reading does not become visual reading; the use of the first is replaced by the use of the second. The knowledge of Braille remains (it was not transformed into visual reading). Surely, you would not want to claim that there is an interface between touch and sight in this case. Ellis argues that learners can utilize their explicit metalinguistic knowledge “at points of dysfluency, when they do not yet have the implicit fluent knowledge” (p. 330). This is not an interface but the use of one system (explicit metalinguistic knowledge) when the other (implicit linguistic competence) is not available. Indeed, when L2 speakers’ linguistic competence does not suffice to communicate their message, they switch to using explicit knowledge to consciously encode the problematic portion of the message. They fill the gap in their implicit linguistic competence by resorting to consciously known grammar. This is not an interface, by any definition. As in the course of development, it is a switch from the use of one system to the use of another system, whenever the first (preferred, faster) system fails. Similarly, “recruiting consciousness to overcome the implicit routines that are non-optimal for L2” (Ellis & Larsen-Freeman, 2006: 571) is not an interface but the use of metalinguistic knowledge whenever implicit linguistic competence is not available; in other words, substituting the use of one system for that of the other. If you fry your eggs in butter when you run out of oil, you would not claim that oil and butter interface. Metalinguistic knowledge remains accessible, unless it is forgotten (Schmidt, 1994). According to Schmidt, as learners progress, more attention is devoted to what they want to say, while the process of grammatization becomes more and more automatic. In other words, they gradually switch from consciously groping for structures to express their intentions (using explicit knowledge) to using acquired implicit procedural competence. This does not mean that metalinguistic knowledge turns into linguistic competence. The declarative system does not feed into the procedural system. The process is one of (possible) gradual acquisition of competence, which, once acquired, is used instinctively – for two reasons:  (1) the use of competence is automatic and hence cannot be cut off and (2) it is faster and less effortful than the old

 Declarative and procedural determinants of second languages

process of consciously controlling production. The speaker switches from using the controlled performance to using the (now available) automatic performance. This switch is not a deliberate one: You cannot decide not to automatically understand what you understand; nor can you choose to use competence other than automatically. You can choose to control your speech if you want to be sure to avoid errors (as in a formal address, even in your native language). You can then check the output of your procedural system before voicing it and consciously change anything you consider to be less than optimum in the circumstances. The data cited by Ellis from Frackowiak et al. (2004; Chapters 23–24) are compatible with the notion that what is learned becomes automatic through practice (and the result of practice is subserved by a different neural mechanism):  The medial temporal lobes are very active early in training; and later, through practice, subcortical structures (the striatum and other basal ganglia) take over. In other words, the use of the procedural system (acquired incidentally) replaces that of the declarative. Neuroimaging studies have also shown activation in the hippocampal system (parahippocampal gyri and mesial temporal lobes) in early stages of L2 learning; then, when native-like proficiency has been achieved, the activation occurs in areas associated with procedural memory – the classical perisylvian language areas (Perani et al., 1998) that sustain implicit linguistic competence representations. Repeated memories result in activation elsewhere: Implicit memory reflects processes in separate neo-cortical regions (Ellis, 2005: 318). Experienceinduced changes are illustrative of the shift from reliance on one system (explicit, declarative) to another (implicit, procedural). There is nothing in all the descriptions on page 318 that is not compatible with the development of two independent systems, with greater reliance on explicit knowledge at the beginning and a switch to reliance on implicit, automatic processes after a good deal of practice. Whenever working memory is used in language production, controlled processes are substituted for the automatic use of implicit competence. This happens not only in the case of second language speakers who have not acquired complete implicit competence, but with native speakers in very formal situations, when they monitor the output of their implicit competence system before vocalizing their utterance. This has the side effect of slowing down their speech. Furthermore, and independently, an interface (albeit dynamic) cannot happen transiently “during conscious processing” because implicit processes are not available to conscious inspection. Linguistic competence is called implicit precisely because it is not observable. Linguists can only infer that there must be an underlying system that can be described in terms of rules, based on the regularities of the perceivable output. But the actual procedures that give rise to the output remain opaque to introspection. By definition, nothing implicit (such as the underlying structure of utterances) can be observed, let alone noticed. It is



Chapter 3.  The explicit/implicit interface debate 

the frequency of occurrence of these non-observable underlying structures that provides the intake to the establishment of the implicit procedural system. The implicit forging of serial associations; the synthesizing of collocations, larger formulas and composite constructions; the creating of implicit hierarchical organizational structures, are all based on manipulating, not the perceived input, but the abstracted (non-observable) intake. Linguistic competence emerges from the implicit tallying of the frequency of occurrence of the underlying structures of perceived surface constructions – not of the input itself, which is explicit.

3.  Consciousness cannot possibly be the interface Referring to Baars’s (1997) Global Workspace theory, Ellis (2005) concludes that “consciousness is the interface” (p. 312). Consciousness is said to be a facility for exchanging information (p. 312) – explicit information, no doubt. In order to avoid plays on words and inappropriate shifts from one level of discourse to another, let us remember that there are at least three types of information the brain is said to manipulate: (1) the information transmitted between cells or groups of cells; (2) that transmitted within the autonomic brain, that is, the part of the nervous system that regulates involuntary functions such as the control of glands and heartbeats (the same lack of conscious control characterizes processes that govern the implicit schemata that underlie implicit linguistic competence); and (3) conscious information in declarative memory. Each type of information is different in nature and applies to different types of entities. It would be a category mistake to apply the properties of one type to another. According to Baars’s theory, as quoted by Ellis, the entire stage of the theater corresponds to working memory. Now, working memory deals with conscious elements. It is a truism that one cannot focus consciousness “as a bright spot” (p. 312) on something that is invisible to consciousness in the theater of working memory, a construction from which implicit elements are absent. Only that which can in principle be conscious can enter the conscious workspace; no part of implicit linguistic competence falls into that category:  Consciousness “is not needed in highly skilled and routine actions” (Baars, 1988: 23). Information that is clearly represented in the nervous system but that cannot be retrieved and reported, such as syntax and other highly automatic tasks, is unconscious (Baars, 1988:  390). There is thus no logical possibility that it can interface with anything in the conscious workspace. The working theater is yet another metaphor that does not actually entail an interface between metalinguistic knowledge and implicit linguistic competence. There is no place in the theater for the metalinguistic knowledge to interact with

 Declarative and procedural determinants of second languages

(translate or convert into, affect the structure or process of) implicit linguistic competence. How conscious information “creates access to” non-conscious linguistic competence (Ellis & Larsen-Freeman, 2006:  571) remains to be explained. It goes against the description of the workspace as provided by Baars (1988, 2003, 2005) and Dehaene and Changeux (2004), cited by Ellis (2005) and Ellis and Larsen-Freeman (2006). Some (and only some) processors (of a particular type), such as sensory systems in the brain, distribute information to the system as a whole (Baars, 2005). Metalinguistic knowledge is not such a system. It is a subset of declarative knowledge whose information is not relevant to the contents of the implicit linguistic competence system, to which it has nothing to offer. Ellis’s implicit underlying weighted neural connections are not the application of an explicitly known rule; nor are they its conversion, translation, transformation, re-codification, or notational variant. An interface between an element of explicit knowledge and an element of implicit linguistic competence would require at least one of these contingencies to obtain. Metalinguistic knowledge is not the kind of information whose underlying neural population, when mobilized by top-down attentional amplification into a brain-scale state of activity, can make it available to inherently non-conscious autonomous representations such as implicit linguistic competence: Some cognitive and cerebral representations are permanently inaccessible to consciousness (Dehaene & Naccache, 2001) and implicit linguistic competence is one of them. Even if the conscious knowledge of a metalinguistic item could be broadcast to one or more of the implicit linguistic competence modules, it would be of no use (given that there is no correspondence between the explicit item and the implicit procedure). Even when contexts influence the selection of an item, they do not interfere with its contents. An element of metalinguistic knowledge cannot alter an element of implicit competence by virtue of the modules’ property of information encapsulation acknowledged by the proponents of neural workspace theory (Dehaene & Naccache, 2001: 12). Within this framework, specialized modules may contribute to the elaboration of a conscious representation in the global workspace – but the reverse is not the case: A conscious representation cannot contribute to the shaping of the internal structure of an encapsulated automatic functional module. Modular processes can be temporarily mobilized and made available to the global workspace, and therefore to consciousness – however, the modular processes themselves cannot reach consciousness if they are not inherently capable of doing so. There is therefore no way in which an item of explicit metalinguistic knowledge could meet an item of implicit competence in the global workspace (the latter has no access to it), let alone affect it there, even if it could have access (because of the latter’s encapsulation).



Chapter 3.  The explicit/implicit interface debate 

In addition, as discussed in Chapter 2, the spotlight of conscious processing does not reach non-conscious operations. Where is the interface between focused attention on conscious processing of form-meaning connections and the resulting implicit procedure? What is the mechanism or process that is the interface between explicit metalinguistic knowledge and implicit linguistic competence? How does attention focused on relevant form-meaning connections interact, connect, or interface with implicit phenomena of any kind, in the limelight of conscious processing? It does not. Neither the memorized construction itself (i.e., the representation of an explicit item) nor the conscious operations that led to its memorization directly influence the implicit system. Only the repeated use of the particular form in different sentences does. Frequency of use is not part of explicit metalinguistic knowledge. Ellis (p. 314) quotes Dennett (1978) as saying, “That of which I am conscious is that to which I have access.” This entails that what I have no access to is not conscious. We have access to declarative knowledge; we do not have access to procedural knowledge. What is more relevant to the issues at hand is what a few (unquoted) lines later Dennett calls “computational access” and defines as a link between subroutines, in that one routine has access to the output of another, namely, “the results of computation of one subroutine are available for further computation by another subroutine” (Dennett, 1978:  150). This process is consonant with one definition of an interface: the output of one system serves as the input of another – a criterion that is not fulfilled between explicit knowledge and implicit competence. The interface cannot be consciousness. Implicit linguistic competence procedures are opaque to conscious introspection. Implicit acquisition “is an unconscious process” (N. Ellis, 1994a: 39). There is thus no possible direct connection between consciousness and implicit linguistic competence, let alone any interaction. We are aware only of the intended message and of the output of implicit linguistic competence (not of the operations of the mechanism in between). Even “the fundamental role of context on the contents of consciousness” (p. 325), thrown in for good measure, is imputable to pragmatics, not implicit linguistic competence. Consciousness is perhaps the prototypical example of an emergent phenomenon and the units of consciousness might be identifiable as patterns of brain synchrony (Ellis & Larsen-Freeman, 2006:  570) but, as discussed in Chapter 2, implicit linguistic competence never enters into the category of phenomena that reach consciousness. It belongs to a different type of phenomena which, like those of the autonomic systems, never reach consciousness but result in actions that are observable. Implicit linguistic competence is by definition not within reach of consciousness. What does it mean to say that consciousness is the interface between explicit metalinguistic knowledge and implicit linguistic competence? If there were an

 Declarative and procedural determinants of second languages

interface, it would certainly not be in consciousness. Consciousness cannot be the connection because it is restricted to explicit knowledge (“conscious” knowledge, awareness) in contrast with implicit competence (unconscious, outside awareness). It would be a bridge projecting from consciousness with nowhere to land on the other side. Consciousness does not transfer, convert, or translate metalinguistic knowledge into implicit linguistic competence. It can only link two items or processes capable of entering awareness. Implicit, non-conscious procedures cannot enter awareness. The flow of consciousness does not reach implicit linguistic competence, hence introspection can never confirm or refute any information from implicit linguistic competence. It is not part of “the stuff of noticing” (Ellis & Larsen-Freeman, 2006: 570). It never partakes in our conscious experience of what allows us to build novel linguistic representations from usage. Only the output of implicit linguistic competence can be noticed, not the way novel representations are acquired or used, nor their underlying structure, nor the processes that generate the novel sentences used in automatically produced utterances. Ellis and Larsen-Freeman (2006) also proclaim that conscious events monopolize time in the limelight, surrounded by a fringe or penumbra of associated, vaguely conscious events. Such “vaguely” conscious events have the potential to become “fully” conscious when the “bright spot” is focused on them. However, implicit linguistic competence is not “vaguely conscious,” it is simply not susceptible of being brought into consciousness at all; in other words, it is never able to appear in the spotlight (and hence cannot interface with anything there). We are conscious of the desire to speak and of the subsequent output of the implicit linguistic system; we may consciously trigger the activation of the unconscious system and notice its output, but the unconscious modules themselves (i.e., their structure and operations) remain forever unconscious. Consciousness is a facility for accessing, disseminating, and exchanging explicit information and for exercising global coordination and control over explicit material. The unsubstantiated claim that this is the interface between the explicit and the implicit needs to be justified. It is indeed “the stuff of learning,” not the stuff of acquisition, whose different nature Ellis (2005) has explicitly acknowledged (pp. 307, 331). Consciousness has no control over the contents of implicit grammar. Ellis and Larsen-Freeman (2006) finally concede that “with hindsight, interface was an unfortunate appellation for this issue” (p. 569), but go on to address “the mechanisms of interface:  noticing, selective attending, noticing the gap, skillbuilding, coaching, processing” all of which are mental actions involved in explicit learning; thus they lose sight of implicit acquisition with which explicit learning is claimed to interface. As long as competence is defined as implicit, unconscious, and not susceptible of reaching consciousness, consciousness cannot be the



Chapter 3.  The explicit/implicit interface debate 

interface. Learning is a process. So is acquisition, but the former is conscious and knowledge (its outcome) is capable of reaching awareness, whereas the latter is not conscious and competence (its outcome) cannot reach awareness. So, in plain English, what does it mean that consciousness is the interface, given that one of the two members of the interface is excluded from consciousness, and the first member cannot be transformed into the second. How do they interact? When would they do so, however transiently? Where, exactly, in the 80 pages of the joint articles (almost exclusively devoted to the mechanisms of explicit learning) is the answer to be found? When and how do the two processes meet (in order to interface)? How does the interface between explicit language learning and implicit tallying occur “through consciousness itself ” (Ellis & Larsen-Freeman, 2006: 570)? How can consciousness, the capacity of being aware of something, be the locus of the interface between explicit (i.e., conscious) processes and implicit processes that are unconscious and incapable of reaching consciousness? Or should we change the meanings of these words too – war is peace, freedom is slavery, ignorance is strength … implicitness is consciousness? Our conscious experience is what allows us to consciously build novel explicit linguistic representations. The implicit tallying  4 is what allows us to build novel implicit linguistic representations from usage (Ellis, 2005). Indeed, as Ellis and Larsen-Freeman reiterate, “our conscious experience is what allows us to build novel linguistic representations from usage” (p. 570). What conscious experience does is to allow the use of explicit knowledge to construct sentences. But, as Ellis (2005) emphasizes throughout,5 it is the subsequent use of these sentences, not the knowledge, that will independently establish implicit linguistic competence. “Consciousness is experiencing” (Ellis, 2005:  314). Tallying (i.e., the acquisition mechanism) is implicit (pp. 321, 324); linguistic competence is implicit. What is implicit is not experienced, not consciously perceived. The neurobiology

4.  This is not necessarily an endorsement of Ellis’s account of acquisition through tallying (though it appears at least as plausible as many other accounts and more plausible than some). It simply shows that, within the framework of Ellis’s own premises, there is no interface between L2 explicit knowledge and implicit language acquisition. The nature of implicit competence (whether a generative grammar of any form or a connectionist-type neural network) remains a question to be eventually empirically verified, but whatever the actual form of its internal structure, implicit linguistic competence is independent of, and does not interface with, explicit metalinguistic knowledge. 5.  “Explicit learning results in explicit memories” (pp. 308, 317). “Mere usage in processing is enough for implicit tallying” (pp. 321, 324). Conscious activities are “out of the scope of implicit learning” (p. 309). “Metalinguistic descriptions do not impact implicit language knowledge” (p. 331).

 Declarative and procedural determinants of second languages

of implicit tallying is distinct from (and unconnected to) the neural correlates of consciousness (pp. 310–312). By Ellis’s (2005) own stipulation, explicit metalinguistic knowledge cannot be converted into implicit linguistic competence (p. 307). Nor can consciousness be an interface because consciousness has no access to the implicit workings of linguistic competence (implicit linguistic competence remains forever unavailable to conscious introspection) according to the workspace theorists cited by Ellis (2005) and Ellis and Larsen-Freeman (2006). Implicit linguistic competence is automatized and carries out “routines within a module” in the absence of direct conscious sensation or control (Ellis, 2008: 241). “In contrast, conscious processing … unifies otherwise disparate areas in a synchronized focus of activity” (p. 241, my emphasis). In other words, conscious processing has no access to the modules of implicit competence, which are excluded from the synchronized activity that affects metalinguistic knowledge. Hence, conscious processing is not where they could interface. According to Koch’s (2004) metaphor of democratic elections for neuronal competition (invoked by Ellis 2005, 2008), a conscious state is the activity associated with the winning coalition of cell assemblies. But all these coalitions contribute only to the election of elements that inherently can become conscious; inherently implicit automatic systems are not on the electoral list and are outside their zone of influence. No implicit linguistic procedures can be a winner; no neural substrate of metalinguistic knowledge can exercise control over them:  They are not represented within the same network – they belong to different states (pun intended). Yet, Ellis (2008: 241) reiterates the incongruous affirmation that “consciousness is the interface.” According to the evidence, an interface could not possibly be in consciousness.

4.  An indirect influence is not an interface The third rationale invoked in support of an interface is that the degree of influence of metalinguistic information is so profound that “claims to interface and interaction seem fully justified” (Ellis, 2005: 325). It would be a fallacy to pretend that an indirect influence constitutes an interface. The degree of influence is not at issue. There is no critical mass of indirect influences that would turn them into an interface. The effectiveness of instruction in the development of L2 (p. 325) is not denied. The question is whether that influence, however extensive, is direct or indirect. The degree of influence of metalinguistic information, by providing adequate material to be used in conversation, may be considerable. However, this process does not transform or integrate any part of metalinguistic knowledge into implicit



Chapter 3.  The explicit/implicit interface debate 

linguistic competence. No direct connection between metalinguistic knowledge and implicit linguistic competence has been demonstrated, or even hinted at.6 If you read in a book that a particular plant, when processed in a certain way, cures a particular disease that affects you, and you then procure, process and consume the plant in question, and consequently are healed, one cannot say that the book interfaces with your being cured. The cause of the cure, what had a direct effect on the disease, is the substance contained in the plant, not the book or your reading of it. No doubt, your reading the book has indirectly resulted in your getting rid of the disease, by giving you the knowledge that allowed you to look for, find, process and consume the plant that provided the substance that is directly responsible for eradicating the disease. But you remain unaware of the chain of chemical reactions that resulted in the cure. You are only aware of (1) having been sick, (2) eating the plant, and (3) being rid of the symptoms. Your acquaintance with the knowledge found in the book, in itself, has no direct effect on your illness. If you had not put that knowledge to practice, you would not have been cured. Was the book instrumental to your getting rid of the disease? Absolutely. Did it interface with the disease? The very question seems silly. What did interface with the disease, and interact with it, is the substance contained in the plant’s molecules, not the book or your knowledge of what is in it. Your explicit knowledge of the book’s content is different in nature from the content of the invisible molecule. Your knowledge does not interface with your health. Unlike chimpanzees in the wild who spontaneously chew that plant when they are affected by the disease, you would never have been cured had you not read the book. Unlike children who acquire a given language construction without explicit knowledge or negative evidence, adults very rarely acquire that construction without explicitly noticing it. But the explicit knowledge does not interface with the corresponding implicit procedure any more than your knowledge of the

6.  Ellis (2008) provides another description of the locus of the alleged interface: The tension between our implicitly controlled system and the evidence of over-generalization of which we have been made aware when we receive negative evidence. But the tension is in fact between our awareness of the outcome of our implicit system (not the system itself) and our awareness of the errors in our production. This awareness does not serve as “the interface allowing system change” (p. 240). There is no implicit-explicit system interaction. Negative evidence simply provides us with an opportunity, on the basis of awareness of our produced errors, to consciously construct a variety of utterances in the correct form, whose underlying structure will be implicitly abstracted and tallied. This is another way of stating that explicit knowledge may have an indirect influence on implicit competence (assuming that it is indeed automatized and not just speeded up).

 Declarative and procedural determinants of second languages

book interfaces with the healing process. Nor do adults acquire the implicit procedure if they do not practice (even though they have the knowledge). Monitoring, or any type of negative evidence, can result in an immediate explicit correction. The revision itself then enters into the implicit learning process (p. 331), not directly, as such, but only by subsequent use and implicit statistical tallying of the underlying structure of the corrected utterance. The correction and revision do not directly impact implicit linguistic competence. What is perceived are various utterances containing a particular structure (of which the hearer is not necessarily aware at the time of input), whereas what serves as intake to an implicit process (i.e., what is tallied) is the frequency of occurrence of an implicit underlying structure, a statistical computation that lays the foundation for implicit procedures. The steps by which any pedagogical grammar or linguistic theory explains the structuring of a given construction do not correspond to the way in which the implicit system operates to generate sentences containing such constructions in their output state. The knowledge of pedagogical rules cannot guide the internalization of the corresponding implicit procedure because there is no piece of knowledge or analytical step that could become part of the implicit procedure – there is no possible interface because of the lack of isomorphism between explicit and implicit elements at any level of structure or stage of development. Ellis declares that “The weight of the evidence to date is in favor of significant interface by means of attention being focused on relevant form-meaning connections in the limelight of conscious processing” (p. 326). The evidence referred to is the indirect influence of metalinguistic knowledge following instruction, on the eventual acquisition of the ability to produce those forms that have been explicitly taught. Interface thus comes to mean “influence,” whether direct or not. This amounts to a contradiction in terms. “Interface” cannot mean indirect influence. The “evidence” alleged here is the (indirect) influence of explicit knowledge on implicit acquisition – a position which the claim that there is an interface is supposed to contradict. If “interface” is not the appropriate concept, then what is? Whatever it is should be called by a suitable name, and so far, the candidate that best fits the descriptions of it found in Ellis (2005) and Ellis and Larsen-Freeman (2006) seems to be an “indirect influence.” Why then insist on calling it an “interface,” when that is acknowledged to be the wrong appellation for the concept, and continue to claim, in the rest of the paragraph, that consciousness is the interface? To top it all, the mechanisms of interface are listed as “noticing, selective attending, noticing the gap, skill-building, coaching, processing” (p. 569) – every one of them a conscious mental action. It would be an incongruity to call an indirect influence an interface. If words are to keep their basic meanings, even metaphorically, then an interface must at



Chapter 3.  The explicit/implicit interface debate 

the very least have a direct connection. Why do I make such a fuss over the use of a word? Because when you use interface, you appear to be saying the opposite of what the “non-interface” advocates claim (namely, that there is no direct influence such as a transformation of explicit knowledge into implicit competence, that the influence is only an indirect one), when in fact, you would be saying exactly the same thing. This amounts to a sophism:  There is no relation of opposition (as would deceptively appear to be the case) but one of similitude between Ellis’s and the non-interface positions. Ellis’s (2005) and Ellis and Larsen-Freeman’s (2006) position is identical to that proposed by Krashen over 25 years ago, against which the interface position has been pitted ever since. In fact Ellis and Larsen-Freeman provide evidence for what they refer to as “the extreme non-interface position” proposed by Krashen (1982; see also Lamendella, 1977). The only difference might be that Krashen downplayed the importance of this indirect influence whereas Ellis highlights it. The non-interface position states that there is no direct operation performed by metalinguistic knowledge upon implicit linguistic competence; no direct contact, no direct interaction, no direct effect. Its only influence is that it serves as a model for practice and a monitor to check the output (Krashen, 1981; Ellis, 2005). It is through practice alone (i.e., the use of language in meaningful communicative environments) that language is acquired (Krashen, 1981; Ellis, 2005). Ellis goes on to specify the actual device at work: the implicit tallying of the frequency of occurrence in language use of the various forms to be acquired.

 one of the proposed characterizations are compatible 5.  N with an interface Ellis declares that metalinguistic information and implicit acquisition “meet and interact in processing” (p. 325). Where is this demonstrated? Where and how do they interact? In reality, they do not interact: as I have argued, the use of one (implicit linguistic competence) is occasionally replaced by the use of the other (metalinguistic knowledge) whenever there is a gap in the former. One may decide to call the relationship between knowledge and acquisition a gavagoosh if one wishes, so long as it is defined as an indirect influence, but one may not use a word that already has such a radically different meaning in the language (both lay and scientific) that it denotes the opposite of what it generally means in this debate. Calling an indirect influence an interface violates the language conventions. If it is the author’s wish that the new, unconventional technical meaning with which he wants to endow the word interface signifies “an indirect influence,” then he must clearly state this at the beginning of his

 Declarative and procedural determinants of second languages

paper. If this were the case, the claim would cease to be controversial, since no one (including Krashen and Paradis, cited as non-interface proponents) denies that explicit learning may have some indirect influence on subsequent second language acquisition. The issue is then simply a discussion about the extent of this indirect influence, its nature, and the circumstances under which it operates (as Ellis for instance, following Krashen, describes the role of the monitor). But, barring newspeak, it would not be an interface, as it would be identical with the traditional non-interface position. Ellis’s interfacedicy,7 the defense of an interface between explicit and implicit language in the face of a self-contradiction, is no more successful than Leibniz’s theodicy. Does any of the means of gaining metalinguistic knowledge interface with the procedures of implicit linguistic competence? Does the result of instruction – metalinguistic knowledge itself – interface with the procedures of implicit linguistic competence? Do the consciously constructed utterances serve as input to the procedures of implicit linguistic competence? The answer to all these questions is no. What serves as input is the frequency with which particular constructions are encountered, irrespective of their surface form (i.e., with different words, possibly different tenses, different pronouns). As discussed in Chapter 2, the surface form, as it is consciously perceived, is not what serves as intake. The establishment of the strengths of the various connections of the neural network underlying the implicit competence procedures or schemata championed by Ellis (2002, 2005) is not based on the surface structure, but on the implicit tally of features that are not attended to and hence not noticed: We do not spend our time counting the units of language – mechanisms underlying unconscious counting do it for us. Information about the underlying structure of a complex stimulus environment is implicitly acquired by a process that takes place without conscious operations (Ellis, 1994a, 2002). The hypothesized implicit acquisition device does not establish the weight of associations on the collocation of specific words; it establishes the ability to automatically generate novel sentences whose structure is independent of the actual word collocations of the input from which the underlying structure common to all encountered exemplars (each exemplar having a different surface form) is abstracted (the intake) and tallied. The intake for the inferred associative probability matrix

7.  Coined after theodicy, the defense of God’s benevolence, omniscience and omnipotence in the face of the existence of evil. In theosophy, you invent a monad, or you say “it’s a mystery,” and that solves that. In scientific endeavor, you do not have this option. Leibnitz, who did not wish to take the fideist option, failed to solve the conundrum because it cannot be solved by logical reasoning (Paradis, 1969). In psycholinguistics, reason, logical argument, and empirical evidence should prevail.



Chapter 3.  The explicit/implicit interface debate 

must be something like: (agent) N + V + determiner + (object) N, not “Mary had a little lamb”, “Peter has a lousy haircut” or “Cindy saw a blue balloon.” According to Ellis and Larsen Freeman (2006), the interface is not physical. It is a process. Now, unless it is metaphysical or spiritual, a process, be it neural, takes place in the physical world, involving physical entities that are being processed (be they phase-locked cell assemblies or long-distance neuronal connections, whatever). The question, then, is:  Exactly what does metalinguistic knowledge contribute to implicit linguistic competence and at what point in the process does it do so? As is described in Ellis (2005), a grammatical form is explicitly learned, it is subsequently consciously applied for use, and then, after many uses, the underlying structure is abstracted and implicitly tallied, and this tallying, quite independently of the knowledge that allowed the conscious construction of each utterance containing the structure of the given type, is what configures the implicit procedures of automatic linguistic competence. The interface process is not to be found in Baars’s (1988) global workspace (the spotlight does not reach unconscious representations) or in Scott Kelso’s (2002) phase-locked rhythmic synchrony (automatized skills are explicitly excluded), or in Dehaene and Changeux’s (2004) conscious neuronal workspace (modular unconscious functions remain incommunicado). Avowedly, there is no process of conversion, transformation, transmission, or communication between the two entities in question. If the process is not an indirect influence, then what is it? If it is, why not clearly say so? Both Ellis (2005) and Ellis and Larsen-Freeman (2006) purport to reveal the nature of the metalinguistic knowledge/implicit linguistic competence interface and both follow the same strategy of describing learning processes in detail, discussing a large array of issues pertaining to the role of social context and the functioning of neural communication (at the cell and cell-assembly levels), socioculturally situated cognition, consciousness and development as social constructions, dialectics, context dependence, socialized consciousness, which are all orthogonal to the issue of interface between metalinguistic knowledge and implicit linguistic competence. They never describe how metalinguistic knowledge directly interacts, determines, or modifies implicit linguistic competence or is converted into it. They are content with stipulating that the alleged interface is dynamic (without showing how the mechanisms dynamically interact) and that consciousness is the interface (without showing how something avowedly non-conscious by nature is nevertheless part of what is conscious, that is, of what one is aware at any given time). In fact, Ellis and Larsen-Freeman (2006) unwittingly provide further evidence against the notion of an interface. We might consider the syntax-pragmatics interface to be the selection of the available form within the grammar in keeping with the pragmatic situation; the

 Declarative and procedural determinants of second languages

semantic-pragmatics interface as the selection of the available meaning within the semantic system that corresponds to the pragmatic situation; the phonologypragmatics interface, the selection of the available prosodic features within the phonological system corresponding to the pragmatic situation. But there is no comparable interface between declarative knowledge and procedural competence: Implicit linguistic computational procedures do not select explicit metalinguistic elements during the generation of a sentence or in the course of acquisition. Controlled metalinguistic processing does not – in fact cannot – select implicit (i.e., unconscious) computational procedures (of which the would-be controller, the speaker, is not aware). Only after a sentence is generated (produced orally or in covert speech) can one monitor, correct, change anything – consciously. In other words, again, there is no interface between metalinguistic knowledge and implicit linguistic competence. Ellis and Larsen-Freeman (2006) state that learning is a process; learning is dynamic. Both these facts are irrelevant to the issue of interface between metalinguistic knowledge and implicit linguistic competence. Both are true of learning, both are true of acquisition, but this says nothing about any actual (even transient, dynamic, however swift) interaction, a direct effect of one upon the other. There is not a millisecond during which metalinguistic knowledge directly affects implicit linguistic competence acquisition or representation. Explicit/implicit interface is not weak, it is not dynamic, it is not in consciousness, it is simply nonexistent, and neither Ellis (2005) nor Ellis and Larsen-Freeman (2006) have shown otherwise. No criterion of interfacing is met: There is no point of interaction between the two at any time, no transfer, transmission, exchange or conversion of data. The fact that they are different in nature entails that explicit knowledge does not “talk” to implicit competence and hence that there is no direct interface. The fact that explicit knowledge is not converted into implicit competence entails that there is no interface mediating between the two. The question of how consciousness or any explicit knowledge might directly influence implicit systems is left unanswered.

6.  Illusory and untenable would-be evidence In order to eliminate possible ambiguities (or at least attempt to reduce possible misinterpretations), avoid misunderstandings, dissipate possible smoke, and remove potential mirrors, let us examine some passages from Ellis (2005) that contain expressions that might suggest supporting evidence and must be interpreted with caution. This will show how important it is to avoid ambiguity (especially when much material not relevant to the task at hand is presented). It will therefore be examined here in detail. Metaphors can be stretched only so far. There is no room for poetic license in Scientific English.



Chapter 3.  The explicit/implicit interface debate 

6.1  From seeds to trees In a section cleverly called “Formulas: The concrete seeds of abstract trees,” Ellis suggests that “explicit memories seed what will later become schematic linguistic constructions” (p. 320). Note that, given the premise that “explicit knowledge does not become implicit knowledge, nor can it be converted to it” (p. 307), the concrete seed (the formula, explicit metalinguistic knowledge) cannot develop into an implicit tree (the implicit procedures). Explicit formula seeds cannot yield implicit schemata trees – only explicit abstract trees representing formulas (such as linguists create). The formula will later become a schematic construction via the standard indirect route (i.e., repeated conscious use of correct form provides input from which intake is implicitly abstracted and tallied). Ellis goes on to state that such a formula “is acquired early and thus serves as the concrete beginnings of what later will be a more general construction schema” (p. 320). The use of acquire might deceptively lead the reader to assume that the author refers to the implicit schemata that underlie constructions. This can only be an illusion. The resulting abstract trees are as explicit as the concrete seeds from which they derive. Eventually, acquisition may take place independently, but the concrete formulas do not become implicit patterns. The memorized formula, even when explicitly analyzed, could not be converted into an implicit schema. The initial step is described as the explicit learning of a formula; a second step involves the learning of related meaning, followed by a conscious explicit analysis of how the construction might work. Eventually, independently, implicit acquisition takes place by the unconscious “connectionist frequency-tuned abstraction of patterns” (p. 320). According to Ellis, in L2, the usual route of naturalistic acquisition is from formula learning and conscious analysis to creative construction. Creative construction can be conscious; new sentences may be constructed by explicit analogy. It is also the case that, eventually, one may implicitly internalize procedures that will permit the creative use of constructions of the given type. But let us not be misled into believing that the route is from explicit formula to implicit creativity, that a formula (the seed) eventually becomes an implicit procedure (the abstract tree). This route from conscious processing to implicit acquisition is indeed an accurate description of the onset and end point of the appropriation process, but let us not forget that along the way there are two parallel processes taking place:  the explicit learning and analysis of observable data and the implicit acquisition of structures on every subsequent occasion of use, which occurs not through the manipulation of any part of the explicit knowledge about the construction, but through the implicit tallying of the intake (“the frequency-tuned abstraction of patterns,” p. 320).

 Declarative and procedural determinants of second languages

6.2  Tuning The title of the section that begins on page 321, “Explicit focus on implicit tuning,” is also potentially misleading. It should not be taken to mean that L2 learners focus on implicit tuning of known constructions in the process of acquiring them. It can only refer to the author’s focus in this section on the acquisition of the underlying implicit procedures that subserve the ability to use constructions that were previously only explicitly known – acquisition which proceeds from every occasion of their use: “Mere usage in processing is enough for implicit tallying” (p. 321). Once constructions have been learned, “noticing is no longer necessary” (p. 321). We should recall the explicit input/implicit intake argument mentioned earlier. Noticing does not bear on what is internalized but on explicit forms from which the underlying grammatical representation is abstracted. (Most of the intake is not observable; it is not explicit and it is therefore not noticed at any time.) There is consequently no direct contact between the explicit constructions that are noticed and the acquired implicit competence that allows one to generate corresponding implicit constructions; they are, as Ellis submits, “of a different stuff ” (p. 331). The explicit constructions do not serve as the material upon which some implicit mechanism performs some transformation, conversion, manipulation, adjustment, alteration, modification, or tuning. The frequency of implicit abstraction of the underlying structure of utterances is what provides intake. Learners are not conscious of either the underlying structures or their statistical frequency of occurrence. Ellis states that “changing the cues that learners focus on in their language processing changes what their implicit learning systems tune” (p. 327). Indeed, what is focused on is henceforth used correctly, and repeatedly so, following the usual pattern. The implicit system is not tuned by explicit processing, but via implicit processing (p. 325). The implicit acquisition systems are tuned by the intake, not what learners focus on. For most complex forms, involving long-distance dependencies and agreement, the implicit underlying procedure derived from the tally is very unlikely to be of a form anywhere close to the learned rule for the explicit formation of such structures. The proverbial rule for past participle agreement in French is a good example. The past participle agrees in gender and number with the direct object when the object precedes the verb but not when it follows it, when the auxiliary verb is avoir (‘to have’), but it agrees with the subject when the auxiliary verb is être (‘to be’); with reflexive verbs, which also take être as auxiliary, the past participle agrees with the preceding direct object (the reflexive pronoun) but not when the direct object follows (as in the case of parts of the body). “Tuning” of this knowledge is surely not what allows a speaker to automatically produce constructions



Chapter 3.  The explicit/implicit interface debate 

such as la pelle qu’il a prise [priz]; elle s’est pris [pri] le pouls; elle a pris [pri] la pelle; elle s’est prise [priz] pour cible (‘the shovel that he took’; ‘she took her (own) pulse’; ‘she took the shovel’; ‘she took herself as target’). Whereas this rule has greater importance for writing than speaking, agreement is aurally perceived and orally produced with participles whose feminine form ends in a consonantal sound. 6.3  Proceduralization Ellis speaks of “how the construction steps are proceduralized” (p. 328). What is (implicitly) proceduralized cannot be the construction steps of explicit instruction or analysis for several reasons: (1) To be proceduralized, explicit knowledge would have to become implicit knowledge, a possibility denied on page 307. (2) The process of proceduralization (via the implicit acquisition mechanisms) is given as tallying the frequency statistics (p. 324), and mere use is enough for tallying (p. 321). And (3) the construction steps cannot become proceduralized because the implicit procedures do not correspond to the explicitly taught account of the construction: The procedures do not mimic, match, or copy pedagogical grammar descriptions. There is no one-to-one correspondence between the explicit construction steps of pedagogical grammar statements and the implicit procedures that generate constructions of the given type. It is, as Ellis reminds the reader (p. 328), the utterance of the construction or formula – not the construction or our explicit knowledge of its analysis – that is used to build implicit procedures. What Ellis describes as examples of “proceduralization” are mechanisms of explicit learning. Indeed, “by various means, the learners can use explicit knowledge to consciously construct an utterance in working memory” (p. 332). According to Sharwood Smith (1978, 1981), apparently cited approvingly, the utterances themselves provide input to implicit knowledge (p. 332) in the following indirect way: (1) sentences are consciously constructed in working memory; (2) sufficient repetition of these preplanned utterances results in fluency (i.e., their production is speeded up). But, as Ellis explains, “executing different conscious procedures a little faster as a result of practice is not an automatization of explicit knowledge” (p. 333). Finally, (3) “after enough practice, they can eventually do it automatically” (p. 332) – thanks, no doubt, to the abstraction of patterns by the implicit acquisition systems. In other words, through practice of utterances of the same structural type, the collocations have been implicitly tallied and a generative grammar (in the form of a matrix of statistical probabilities) has developed and is used automatically. But this is a process that relates only very indirectly to the learner’s previous conscious construction of utterances. The output of implicit competence is similar to that of consciously planned utterances but the process, not directly derived from metalinguistic knowledge but arrived at independently, is different.

 Declarative and procedural determinants of second languages

Ellis refers to Anderson’s work (p. 333). The first two of the three stages in the move from declarative to procedural knowledge in Anderson’s (1983, 1992, 1996) ACT model, namely the cognitive and the associated stages, are conscious; the third, the autonomous stage, constitutes the acquisition of implicit competence that is automatic, hence without conscious control (i.e., autonomic). There is no interface here. These are sequential steps: learners switch from learning and using metalinguistic knowledge to acquiring and using implicit competence. The passing from (1) a stage where rules are explicit, through (2) an associative phase where rules are applied repeatedly, to (3) an autonomous stage where the rules are no longer explicit and are executed automatically, entails a change in the nature of the rules involved. In stages (1) and (2), explicit metalinguistic rules are learned, but, in parallel with the repeated use of these rules during stage (2), a set of implicit procedures is acquired, leading to (3), an autonomous stage in which the rules are not of the same type as in (1) and (2) but are implicit procedures. Only their output corresponds to the output of the explicit rules – namely sentences of the same grammatical type – but not their structure nor mode of operation. In fact, Anderson’s model of skill learning is not readily applicable to language acquisition because, as noted above, unlike most motor skills, acquisition is not the automatization of the same entities that were previously practiced. In a motor skill, the same sequence of movements that was relatively slowly and consciously performed is eventually performed automatically. In language acquisition, however, it is not the explicit rules that learners have inductively or deductively constructed for themselves (or have formally learned) that are automatized, but implicit computational procedures that were never consciously practiced. Implicit linguistic competence is not compatible with the general framework of Anderson’s ACT production system. It is noteworthy that Anderson’s memory system in which proceduralization is supposed to take place is labeled “production memory”8 (1983:  19). It refers to a way of processing declarative language information that is more efficient because the information no longer has to be processed through working memory. Production is speeded up, but the product is still derived from declarative memory. The implicit computational procedures of linguistic competence are not made up of the same schemata as the known declarative rules. The steps used in sentence construction based on declarative knowledge are not the same as those that generate sentences in implicit competence, namely the computational procedures. It can therefore not result from the proceduralization of previously rehearsed metalinguistic knowledge.

8.  “An ACT production system consists of three memories:  working, declarative, and production” (Anderson, 1983: 20).



Chapter 3.  The explicit/implicit interface debate 

In Anderson’s model, after the declarative knowledge has been applied interpretively a number of times, a set of productions that can apply the knowledge directly is compiled. In other words, the same knowledge that is first applied interpretively is later applied directly, faster and more efficiently (i.e., with fewer errors). Conscious experience may be converted into automatic routines provided that what is automatized is the same thing that has been consciously experienced, which is not the case in language acquisition. Implicit competence is not a production system that operates in the way precise motor sequences are automatically produced after having been consciously rehearsed. It is a generative system that does not implement explicit grammatical information. The ACT may apply to learning, but not to acquisition. The term proceduralization cannot refer to the automatization of metalinguistic knowledge that would result in implicit linguistic competence procedures. Proceduralization takes place only by acquiring implicit procedures. The elementary components of routines and procedures in Anderson’s model are declarative data. The conversion from declarative to what he calls procedural (productions) is not the conversion of a set of one type of knowledge (declarative) into a set of a different type (implicit competence) but to the production of the same type of knowledge, but prepackaged, faster and more efficient. Proceduralization can refer either to the substitution of the use of implicit competence for the use of metalinguistic knowledge or to the speeding up of the use of metalinguistic knowledge. In either case, there is no conversion of metalinguistic knowledge into implicit competence, hence no interface. McLaughlin (1987), also referred to by Ellis, describes the observed transition from slow and halting production by means of attentive control of a construction in working memory to fluent automatic processing. Again, we have two sequential phenomena: The use of one replaces the use of the other over time. No interface is specified here. 6.4  Ambiguities The paragraph on “Differentiation and restructuring” reproduced below is a good example of the ambiguity caused by the unsystematic (and misleading) use of the terms learning and acquisition. On page 334, Ellis writes: Another scope for explicit learning involves explicit memories as object of scrutiny for later analysis [presumably, explicit analysis]. Formulas are initially learned as whole constructions. Only later, when the learner has acquired [presumably, explicitly learned] some of the relevant component building blocks, might the learner realize [explicitly] that these wholes might indeed be dissected into their component parts. Language acquisition does not just rely on integration; it is repeated cycles of integration and differentiation leading to a restructuring of knowledge. Much of this analysis and restructuring comes from the tallying and distributional analysis of implicit learning. (All italics and square brackets are mine.)

 Declarative and procedural determinants of second languages

There seems to be a surreptitious shift here from explicit learning to implicit acquisition, since we are talking about tallying, but where and how did this change take place? There is no becoming or conversion (p. 307) so it can only happen in the usual sequence of explicit learning and analysis followed by practice of what has been learned and, in parallel, developing implicit competence through implicit tallying of the underlying structure of the relevant utterances. For these statements to apply to implicit phenomena, they would have to correspond to the configuration of internalized procedures, which is ruled out by the premise that the internalized procedures are not the same as either the whole construction of memorized formulas or the explicit dissection into their component parts. The explicit constructs that are analyzed are not isomorphic with the design of their underlying implicit procedures; they represent different algorithms. 6.5  Inapplicable analogies and metaphors The following references found in Ellis (2005) would deserve a place in the hall of fame of the Belles Infidèles9 (Paradis, 2006). In support of the claim that “at any one time, our state of mind reflects complex dynamic interactions in implicit and explicit knowledge,” Ellis (p. 313) cites a passage from Scott Kelso (2002) about coordination dynamics, phase synchrony and the stability of information over time, whose relevance is not immediately apparent. So let us examine what Scott Kelso says in that cited article. The activities of cells scattered all over the visual system, which respond to various features of an object (form, color, motion, etc.), are integrated to create a complete image of the object. Distant neurons responding to features of an object pool information together to create a coherent percept. No such firing of neurons in unison occurs to bind, blend, coordinate or join together explicit and implicit procedures or representations. There is no explicit/implicit amalgam, analogous to the way there is a whole percept combining various parts. Explicit and implicit representations and processes do not combine to form an entity (such as a visual image). There is no participation of explicit knowledge in the automatic generation of sentences or the establishment of underlying procedures that would allow them to do so. There is no need for such integration. Each system is sufficient, at different levels of efficiency, to produce utterances. Children produce complex sentences by the age of 6 without prior explicit metalinguistic knowledge and some L2 speakers produce sentences without implicit competence. There is no correspondence between the alleged dynamic interface and the biophysics of the different frequency bands of oscillatory processes in the human brain, the subject

9.  In the “irrelevant citations” and/or “selective citations” sections.



Chapter 3.  The explicit/implicit interface debate 

of Scott Kelso’s paper. Our state of mind reflects complex dynamic interactions of implicit competence (on the one hand) and interactions of explicit knowledge (on the other hand); these different interactions occur independently, and the quoted paragraph from Scott Kelso does not suggest otherwise. Any analogy with the visual (or any other sensory) system is flawed from the start because it describes how various pieces of information relevant to the eventual awareness of a visual image are transmitted from peripheral sensors to the grouping together of the various components. These components are all relevant parts of the visual system. By contrast, metalinguistic knowledge is not a component of implicit competence. Unlike spatial frequencies, depth, texture, etc., which are necessary for the building of the conscious visual image, metalinguistic rules and any other form of explicit information are not elements constitutive of implicit linguistic procedures. Moreover, as Scott Kelso (2002: 367) remarks, rhythmic synchrony (the medium of coordination) appears to be crucial for conscious awareness or a decision to focus attention – but there is no such phase-locked synchrony between explicit knowledge and implicit linguistic competence or its acquisition. Ellis’s proposed “dynamic interface” cannot be understood in terms of Scott Kelso’s notion of either absolute or partial coordination, that is, a kind of functional order or pattern in space and time that brings various activations together. Rather, the “coordination” between explicit knowledge and implicit competence corresponds to the third type (which Ellis does not mention), namely what Scott Kelso refers to as the functional order in which the component parts behave quite independently, with no coordination at all,10 “as occurs after persistent, long term practice in playing the piano or the violin” (p. 365) … or using language in interactive communication. Metalinguistic knowledge and implicit knowledge are not “coupled” at any time; they are never “in phase” let alone “in phase-locked states,” however far we wish to stretch the metaphor. Nothing corresponds to “ghosts of coordinated states” since they have never been in a phase-locked state. There is no “bistability,” not even transient (all the quoted expressions in this paragraph are from Scott Kelso, 2002:  368). Each system goes its merry way, independently of the other, before, during, and after the acquisition of implicit linguistic competence: Second language learners may have much metalinguistic baggage long before they start to develop implicit competence, and if they manage to reach near-complete

10.  Dehaene, Kerszberg, & Changeux (1998:  14533) also hypothesize that automatic processing activates specialized processors “without requiring coordination by global workspace neurons” – a hypothesis for which the authors find support in “recent images of brain activity during unconscious processing.”

 Declarative and procedural determinants of second languages

competence years later thanks to high integrative motivation and considerable practice, they either retain their metalinguistic knowledge in declarative memory (i.e., independently) or they may by then have forgotten most of the metalinguistic knowledge they learned in school. When a metalinguistic representation is created, it is not implicit linguistic knowledge that is made explicit, but rather, knowledge that describes and/or analyzes the noticeable input (namely, the output of one’s own or others’ implicit linguistic competence11). Metalinguistic knowledge does not arise directly from implicit competence. Inherently implicit representations do not subserve explicit representations. Because “explicit knowledge about language may contribute indirectly to the development of implicit knowledge of language,” (Roehr, 2008a: 83), and only indirectly, explicit and implicit competence do not interact, and they certainly do not interface during conscious processing. Following N. Ellis’s (2005) misrepresentation of Scott Kelso’s account of synchronized firing of a number of neurons in different brain regions,12 Roehr (2008a) assumes that the line between stimulus detection and noticing can be regarded as “the point of interface between implicit and explicit processes and representations” (p. 70). The line between detection and noticing may be, as she assumes, the threshold of conscious awareness, but given that implicit stimuli (i.e., the underlying structures that serve as intake) are neither detected nor noticed, there is no interface between implicit linguistic and explicit metalinguistic processes and representations. The implicit process of detection can only apply to stimuli that

11.  Only the content of attended discourse can access the global workspace (Dehaene et al., 1998:14530; my emphasis), namely the output of implicit linguistic competence available for comprehension. The structure of implicit competence, not amenable to awareness, is not. Only aspects of high-level speech production endowed with “the property of reportability, namely the fact that it can be described or commented upon using words” have any active representation in the workspace (p. 14530). 12.  Roehr’s reference to Engel (2003) is equally inapplicable to the implicit/explicit interface, for the same reasons. Contrary to various cognitive functions that are amenable to becoming conscious, metalinguistic knowledge and implicit linguistic competence do not share processes. No information related to objects or events external to its module can influence the structure of implicit linguistic competence schemata during processing. No neurons or network nodes underlying metalinguistic knowledge participate in modular implicit linguistic processes; the latter never give rise to sensory awareness and no binding with metalinguistic neural networks can take place. Specific binding phenomena that allow the generation of syntactic structure occur only within the implicit module. No element of metalinguistic knowledge is constitutive of implicit linguistic competence. Thus there is no reason for binding, hence no interface. Engel is concerned with the kind of binding mechanism that may be critical “for the establishment of conscious mental states” (p. 133, my emphasis).



Chapter 3.  The explicit/implicit interface debate 

are capable of being noticed, that is, of becoming conscious. Implicit linguistic competence items do not fall into this category. The distinction between detection and noticing can relate only to the level of focused attention required for data capable of reaching awareness and therefore of entering explicit knowledge. In the context of the posited interface between implicit linguistic competence and metalinguistic knowledge, the only relevant aspect of the cited dynamic binding by transient synchronization of neural discharges is its dissociative function, namely keeping apart information related to different objects or events that need to be kept separate. Such binding is a mechanism not only to activate disparate relevant items scattered throughout the brain, but also to distinguish neuronal activity pertaining to a particular type of object from that pertaining to unrelated information (such as metalinguistic knowledge and implicit linguistic competence). Pace Heraclitus, whose characterization of creativity (“All things come into being through opposition, and all are in flux like a river”) is quoted by Ellis on page 332 (and used again by Ellis, 2006 and Ellis & Larsen-Freeman, 2006), even though explicit knowledge and implicit competence are both in a constant flux, their itineraries remain parallel. In spite of Heraclitus’ cosmology and that of his successor Anaxagoras, who taught that everything is in everything (and presumably the reverse), the explicit stream does not flow into the implicit river. The metaphor works for either explicit knowledge or implicit competence, but the flow of consciousness does not run into the unconscious river. The stream of consciousness indeed refers to what enters into awareness at any moment, ever bringing different (explicit) items to the fore. But it can never bring any parcel of implicit processes into awareness; only their result achieves consciousness, not the operations that brought it about, nor the underlying structure of the procedures that yield this observable result. Because Ellis has ruled out the possibility of explicit knowledge’s becoming implicit competence, this whole incursion into Ancient Greece and biophysics leads us back to square one: The conditions for an interface are not met. The information flow between cells and cell assemblies, the contributions of the right or left hemisphere to the summation of an implicit process never make it into consciousness. We are not conscious of the struggle for supremacy between the possibly contradictory (electrical) signals sent from the right and left hemispheres whose sum results in an observable outcome. Such contradictory signals are illustrated in the observed behavior of split-brain individuals whose left hand fights the attempted movements of the right hand (Sperry, 1974). Implicit linguistic competence is as remote from the content of consciousness as the competition between neural spiking discharges. The stream of consciousness contains only what reaches consciousness, namely those items, processes and phenomena of which we are aware at any given moment. What enters consciousness is indeed in a constant flux, but only that

 Declarative and procedural determinants of second languages

which can enter consciousness (i.e., whose nature is such that it can be consciously apprehended) does enter consciousness. What is of a nature such that it cannot become conscious (i.e., cannot be contemplated “with the mind’s eye”) is not at any moment, even if only for milliseconds, “a snapshot of consciousness” (Ellis & Larsen-Freeman, 2006). Because of its non-conscious nature, there is no way that implicit linguistic competence could enter (the flow of) consciousness, any more than other nonconscious cerebral autonomic activities such as the kidney or liver functions. To claim that the connection between metalinguistic knowledge and implicit linguistic competence is consciousness is not just an oxymoron (i.e., an apparent contradiction) but an actual logical contradiction in terms. The entire discussion of the firing activity of neurons, the synchronization of their spiking discharge, their competition, the re-entrant signaling feedback, the involvement of thalamo-cortical loops and its synchronized rhythmic action do not speak to the issue of a metalinguistic knowledge/implicit linguistic competence interface, and certainly do not establish its possibility, let alone its existence. Whereas motivation, for instance, emerges from the interaction of many agents, no internal or external agent (including metalinguistic knowledge) can influence the procedure that sustains the adjective-noun agreement in gender and number in French or case assignment in German. The procedures of implicit linguistic competence are not the result of the summation of outside factors of different sorts, only of the tallying of the underlying structure of collocations heard and produced. The knowledge of the rule of past-participle agreement will not help a learner to generate it on each appropriate occasion; only the statisticalprobability-governed weight of the associations will. The communication between neurons is obviously not conscious. Some of this activity results in conscious thoughts or observable behavior. But only those phenomena whose nature is amenable to becoming conscious do become conscious (when a certain activation threshold is reached). However, the brain’s electrochemical messages within the autonomic system (including implicit linguistic competence) are not within the set of phenomena that can reach awareness, and hence they do not. There is no locus where, or time when, an explicit piece of information is capable of manipulating information of an autonomic nature. Metalinguistic knowledge is sustained by certain neural circuits with their own ways of operating (including the transmission of electrochemical messages from cells to cells within the specific circuits devoted to declarative material) and implicit linguistic competence is sustained by other neural circuits with their specific functioning mode; the conscious explicit information carried by the first does not directly modify the contents and operations of the second. They are of a different nature (Ellis, 2005) and the first does not in any way become the second (Ellis, 2005).



Chapter 3.  The explicit/implicit interface debate 

“The dynamic mutual influence of inter-related types of information as they activate and inhibit each other over time” (Ellis & Larsen-Freeman, 2006:  571) does not apply to the implicit/explicit interface because no external factor can affect the structure of encapsulated representations (Dehaene & Naccache, 2001). Metalinguistic knowledge and implicit linguistic competence have not been shown to inter-relate. This is precisely what should be demonstrated. It is not demonstrated for the simple reason that there is no possible interaction between explicit metalinguistic knowledge and implicit linguistic competence. The implicit procedures are not just a different notation of the explicit rules. They are not the same rules. At no point in time do they interact. No one has ever suggested how an explicit rule could conceivably morph into the implicit computational procedure that allows a speaker to automatically comprehend and produce sentences that can be described by linguists as implementing that rule. Again, the “shifting harmony of sub-patterns” argument is irrelevant to the issue of an interface between metalinguistic knowledge and implicit linguistic competence. It has been shown (Ellis, 2005; Ellis & Larsen-Freeman, 2006) that learning is dynamic, acquisition is dynamic, not that they interact, not that one determines or modifies the structure of the other.13 The signals we receive from cosmic background microwave electromagnetic radiation14 are very weak, but they are received. No signals of any kind from metalinguistic knowledge are picked up by implicit linguistic competence, however weak.

 escription of explicit phenomena contributing 7.  D to metalinguistic knowledge Flawed output can prompt focused feedback that provides data for explicit analysis. Memorized expressions can provide material that can be used to explicitly construct well-formed sentences that will then serve as material for the repetition of correct structures, and allow the implicit tallying of their frequency, thus indirectly contributing to the development of the implicit grammatical system. L2 learning contributes to L2 acquisition only in providing means to produce a larger number of exemplars of correct utterances from which, indirectly, the

13.  Ellis (2008: 240) further refers to the cognitive neuroscientific investigation of the neural correlates of consciousness, the global workspace theory, and the massively parallel constituency of the conscious mind, without showing how conscious representations could be capable of influencing modularized automatic systems, an eventuality which, as we have seen, the cited authors either ignore or explicitly exclude from their models. 14.  Moving from a Heraclitian to a more contemporary vision of the cosmos …

 Declarative and procedural determinants of second languages

acquisition mechanism will implicitly tally the relevant unnoticed underlying characteristics. Explicit knowledge is several steps removed from having any direct influence on implicit competence: (1) rules are used to construct utterances (and to monitor output); (2) utterances serve as input; (3) intake is abstracted from input; (4) the frequency of abstracted patterns is tallied to establish the weighted connections that constitute the implicit procedures required to generate novel sentences. In the four different effects of practice outlined by Ellis (p. 333), the first three are explicit, whereas the fourth, which takes place subsequently and independently, is implicit: (1) improved access to explicit memories; (2) slightly faster execution of conscious procedures (explicitly stated not to be evidence of implicit competence); (3) consolidation of memorized whole utterances (hence in declarative memory); and (4) automatization of production as a result of implicit tallying. The effects on explicit material are independent of those on implicit competence. Production shifts from the controlled use of metalinguistic knowledge to the automatic use of incidentally acquired implicit competence. Drilled patterns, conjugations and declensions, mnemonics, feedback, negative evidence, recasts, use of analogy, dissection of formulas into their component parts, rehearsal of the phonological loop, declarative statements of pedagogical grammar and monitoring of one’s output can all contribute to the conscious construction of a desired utterance (Ellis, 2005:  327–332). But they are “out of the scope of implicit learning” (p. 309). “Explicit memories and declarative knowledge can partake in [conscious] utterance building and monitoring” (p. 328). How does this feed into implicit acquisition? Recasts, for example, present learners with data ready for acquisition because they highlight the relevant element of the form (p. 332). The form is explicit, the desired meaning to be expressed is conscious. Thus, recast data are indeed ready for learning, but there is still a long convoluted road to acquisition – namely, the usual route of internalizing the relevant procedure through implicit tallying. The recast presented to the learner is one exemplar whose underlying structure can be tallied. It will take many more presentations of utterances containing the particular construction before the implicit procedures needed to generate it are acquired – again, without any direct relation or contact between the perceived form and the acquired procedure. “Inspection and monitoring of prearticulatory output” (p. 331) is a distinguishing characteristic of controlled processing, involving the anterior cingulate cortex – as opposed to automatic processing of implicit competence (which does not require the anterior cingulate cortex, but portions of the right cerebellum and left basal ganglia). Self-repair, in L1 or L2, is also a conscious process bearing on the output of implicit competence (and/or of conscious utterance building in the case of L2). The output of implicit linguistic competence is consciously perceivable but its internal structure and functioning are not.



Chapter 3.  The explicit/implicit interface debate 

Ellis devotes Section 5 to “Working memory in language acquisition” (p. 337–338; my italics). Working memory is characterized as the home of explicit deduction, hypothesis formation, analogical reasoning, prioritization, control, and decision making – in other words, it is the site of conscious operations. Differences in working memory capacity as measured by sentence span relate to language learning aptitude, which refers to success on school tests, on which individuals generally perform better in the written than the oral sections (VanPatten, 2002). They relate to second language knowledge, not to second language competence. For the same reason, the fact that individual differences in measures of attentional processing predict individual differences in SLA supports the role of explicit learning and conscious processing, not in second language acquisition, but in second language learning (possibly as a first step toward eventual language acquisition). The attentional system, the phonological loop, the visuospatial scratch pad, and the cross-modal store (cross-modality being a characteristic of declarative memory that is absent from procedural memory) – in other words, all the constituents of the British working memory model – work on conscious elements. It is not surprising that they should play a role in the learning of vocabulary and formulaic utterances (Ellis, 2005: 338) given that they are the declarative elements of language, whether first or second (as discussed in Chapter 1). The explicit learning that takes place in the consolidation of explicit linguistic constructions is thoroughly investigated. This, however, says nothing about how this knowledge and/or the means to learn it interface with implicit acquisition. In the section on explicit memories of linguistic constructions (pp. 317–320), Ellis describes how neural systems in the hippocampal and limbic structures allow the consolidation of explicit memories. In adults, explicit memories are learned rapidly. In contrast, implicit memories are acquired slowly because this requires a very large number of encounters with each particular form. The learning rate of anterograde amnesiacs is known to be “grossly slow in comparison with normal subjects” (p. 319). Normal adult subjects are shown to be able to learn explicit associations better than anterograde amnesiacs, which should not be surprising given that this is the characteristic symptom of amnesia. It has not been shown that amnesiacs are slower than normal adults at acquiring rules implicitly. Both are slow at that. Van der Linden et al. (1994) have shown that patients with anterograde amnesia are able to acquire new words incidentally (and only when assessed by tests of implicit memory: Williams, 2005), but not explicitly (Cohen & Eichenbaum, 1993). Studies that show an influence of explicit metalinguistic knowledge on implicit acquisition are alleged to “provide good reasons to consider an interface whereby explicit knowledge affects implicit learning during output” (Ellis, 2005: 337). Indeed, in their output, learners provide exemplars of utterances containing a construction

 Declarative and procedural determinants of second languages

to be tallied, but the utterances or their conscious processing do not directly affect acquisition. No part of the observable utterance15 serves as intake. Only its underlying structure does. None of the cited evidence of influence is shown to be direct. In no case does any of these genuinely good reasons to provide metalinguistic knowledge (pp. 335–336) establish the existence of an interface between such knowledge and implicit linguistic competence. Interface suggests a direct connection between systems. An indirect influence does not qualify. If the point of this demonstration is that teachers should provide negative feedback to accelerate the appropriation of a non-obvious contrast (without an interface), that is fine, but one should say so.

8.  Why adults should need explicit metalinguistic knowledge We may assume that most children are not taught grammar before they go to school. They acquire an implicit grammar without instruction. By that time, they have acquired the structures of their native language, including long-distance discontinuous dependencies, such as questions, binding, negative polarity across clauses, and so on. Hence, explicit knowledge is not a prerequisite for language acquisition. However, because of the interference of the already acquired L1 grammar, which serves as the default in interpreting whatever L2 features have not been internalized, adult learners may incorporate deviant items in their L2 language subsystem, often the corresponding L1 parameter. In such cases, noticing the non-salient L2 feature through instruction or any other type of explicit evidence is necessary to provide learners with the opportunity to produce exemplars of such features in their utterances and to perceive them when they are heard. Whereas perception of a feature is a requisite for providing a target, it is the repetition of this target, not its knowledge, that indirectly establishes implicit competence. An explicit construction (the target) does not contact, connect, communicate or interface with implicit competence or its acquisition process: The repeated use of a construction leads to the tallying of its abstract underlying structure, to which consciousness does not have access. This strengthens the argument that, at least for some grammatical features, instruction (drawing attention to a particular form) is necessary for adult learners of an L2 in order to allow them to practice sentences that contain the given 15.  For example, take the sentences considered earlier: (a) The horse jumps over the fence; (b) The cat sneaks under the couch; (c) The shutter bangs against the wall; (d) The sewage empties into the river. Sentences (a)–(d) have the same underlying structure though the words that constitute them are different. The tallying cannot be of the collocation of the words (as heard) but of the abstracted underlying pattern (not immediately apparent).



Chapter 3.  The explicit/implicit interface debate 

construction and eventually, through practice (i.e., frequent use), to implicitly incorporate the underlying structure of that construction in their L2 implicit linguistic competence. Metalinguistic descriptions do (indirectly) impact the acquisition of linguistic competence but, as discussed above, their indirect effect does not constitute an interface. It appears that adults, unlike children, are not able (or find it very difficult) to acquire complex long-distance discontinuous dependencies in a second language. This means that they do not build implicit competence procedures for the most complex aspects of syntax incidentally, the way children do. Ellis (2005) hence proposes that explicit knowledge is a prerequisite for acquiring such dependencies in L2. Adults need negative evidence (e.g., corrections) so as to incorporate these syntactic aspects into their metalinguistic knowledge and possibly, eventually, acquire them through repeated use. (Of course, before one can speak of competence, it must be demonstrated that the adult L2 speaker’s observed success reflects automatization rather than speeded-up processing. This speaks to the critical period hypothesis and will be considered in Chapter 4.)

9.  I ndirect influence of metalinguistic knowledge on acquisition not denied Notwithstanding the lack of interface, metalinguistic knowledge is not only necessary to consciously learn a language; it may also help to incidentally acquire it, if only indirectly, by focusing attention on items that need to be practiced – even though what is focused upon is not what is internalized. Metalinguistic knowledge also allows learners to monitor the output of linguistic competence and thus increase their production of correct forms, the frequency of which may eventually (though indirectly) establish the implicit procedures that will sustain their automatic use. The need for focus on form after the age of 7 has been most obvious in early immersion programs. The findings of a comparative analysis of close to 1200 Canadian immersion students ranging in age from 7 to 14 showed that learners benefit from instruction that provides opportunities for noticing, language awareness, and controlled practice with feedback, in order to move beyond the use of interlanguage forms (Lyster, 2004). As Lyster points out, instruction allows for more target-like declarative representations and communicatively rich practice activities then enable the learners to proceduralize their knowledge (i.e., to incidentally acquire the corresponding implicit procedures). Inasmuch as some acquisition does take place in a formal, instructional setting, metalinguistic knowledge has an undeniable influence on language acquisition. It is equally undeniable that this influence is indirect and that therefore we

 Declarative and procedural determinants of second languages

cannot speak of an interface or direct interaction between the two. And if there were one, it could not be in the limelight of consciousness. Explicit instruction thus influences implicit language acquisition indirectly, by pointing to negative evidence and furnishing adequate material for practice (through numerous pedagogical and mnemonic exercises); explicit knowledge then continues to help by serving as a monitor to correct one’s output. But, during practice (i.e., comprehending and producing utterances), nothing explicit enters into contact with (let alone affects) any part of the implicit system at any time. As mentioned in Paradis (2004), foreign language instruction has an unquestionable, though multiply indirect, influence on second language acquisition, whenever acquisition eventually occurs. But (1) what is acquired is not what is explicitly known, and (2) what is known is not transformed or converted in any way into what is acquired (the way potatoes become French fries or fertilizer becomes part of the potato). There is no direct connection between metalinguistic knowledge and implicit competence, whether static, dynamic, in consciousness or in afterlife. Metalinguistic knowledge does not become implicit competence: It does not communicate or transmit information to the implicit system. They are of a different nature and do not share a common code or a common content. As Ellis (2005) puts it, “Formulas, slot-and-frame patterns, drills, and declarative grammar rules contribute to the conscious creation of utterances whose subsequent usage promotes implicit learning and proceduralization” (p. 305). In other words, explicit learning operations do not connect with implicit acquisition, that is, the development of implicit procedures. They only serve to consciously construct utterances (utterances of which we are also conscious). Subsequently, the use of the utterances promotes implicit acquisition – not by converting utterances into procedures, in fact not by using the utterances themselves in any way, but, according to Ellis, by implicitly abstracting statistical probabilities of the frequency of occurrence of underlying structures, thereby building weighted connections in a neural network system, the result of which will correspond to implicit procedures (unobserved schemata) that underlie the generation of sentences, allowing for the automatic comprehension and production of novel sentences of a particular form. As a consequence of the premise that metalinguistic knowledge does not become implicit competence and is not converted into it, at no stage during the appropriation of an L2 does any part of metalinguistic knowledge directly affect, modify, act upon, make changes to, or interfere with any implicit linguistic procedure or its acquisition. What seems to happen, in accordance with Ellis’s proposals, is that L2 is acquired in three stages: (1) obtain metalinguistic knowledge, (2) use that knowledge to construct utterances, and (3) through the use of these utterances, gradually acquire implicit competence incidentally, in interactive communicative environments. At that stage, the two systems are used alternatingly



Chapter 3.  The explicit/implicit interface debate 

as the speaker relies more and more on the faster, more efficient acquired competence, intermittently switching to the controlled use of metalinguistic knowledge whenever competence does not automatically generate the desired form (Figure 3.4). Some individuals may develop very little implicit competence and will continue to use explicit knowledge to construct utterances in a speeded-up albeit still controlled manner when they communicate in their second language. The bottom line is thus as follows:  Explicit learning yields explicit knowledge. Incidental acquisition yields implicit competence. Explicit learning does not yield implicit competence but knowledge against which the output of implicit competence can be compared. It also provides specimens that contain certain constructions, the repetition of which leads to the build-up of implicit procedures whose configuration differs from the declarative enunciation of the corresponding rule but allows the generation of constructions that can be described in terms of that explicit rule. There is, however, no direct link between the rules (as explicitly learned), the explicit processing of sentences, or the utterances themselves (as perceived) and the implicit statistical tallying that establishes linguistic competence, or the resulting implicit procedures that underlie the generation of sentences containing constructions corresponding to those of their surface counterparts. No indirect influence, however extensive, constitutes an interface. Are we witnessing the disintegration of the interface debate, in that interface proponents are now saying the same thing as their opponents, or do words now have a 1984-type reading, in that a term that until now referred to a direct contact resulting in direct influence is now used to mean an indirect one? In either case, the interface debate as we knew it has dissolved into self-contradiction.

10.  How explicit knowledge benefits implicit acquisition – indirectly “Interaction in which participants’ attention is focused on resolving a communication problem” (Ellis & Larsen-Freeman, 2006: 572) contributes to increasing metalinguistic knowledge which will serve as a model and a monitor, and thus provide opportunities for using well-formed structures of interest whose increased frequency of use will implicitly (i.e., without awareness and through an altogether different system) lead to the acquisition of the relevant underlying procedures (i.e., set the proper weight between connections or lay down the appropriate procedures to generate the required form). Here again, metalinguistic knowledge does not interact with the required structures or processes that underlie implicit linguistic competence. Ellis’s (2005) “Creative construction in working memory” section (pp. 328–329) refers to the conscious building of novel utterances on the basis of known explicit operations, a process that differs from the creativity of implicit competence. As Ellis

 Declarative and procedural determinants of second languages

demonstrates, speakers have the explicit ability to construct a novel utterance in Welsh by analogy, on the basis of a number of Welsh phrases. Independently, the ability to generate novel utterances can be implicitly acquired by connectionist mechanisms (i.e., implicit tallying of frequency of use). The more abstract the schema, the more the second language learner must have an explicit understanding of the functional interrelationships between the structures in order to successfully learn the proper forms that then serve as input to the implicit mechanisms for tallying. Examining explicit memories of linguistic constructions, Ellis notes that working memory is required in both the initial encoding and the ultimate recall of explicit knowledge and emphasizes that explicit learning results in explicit memories (N.C. Ellis, 2005: 317; Ellis 2006: 112) – and only explicit memories. Exaggerated stimuli and other teaching innovations draw conscious attention to a phonological contrast so as to make it easier for learners to perceive (p. 327). Learners focus their attention on the differences and thus are eventually able to (consciously) perceive them. But note that what ends up being automatized is of an entirely different nature and is in fact not even noticeable. L2 learners acquire the ability to produce the sounds by implicitly controlling their articulatory and phonatory organs, the mechanisms of which they remain unaware. Speakers do not consciously know whether their tongue touches their palate, and if so, where; how many milliseconds their vocal flaps vibrate; how far apart their lips are; or what intensity of breath is applied. Their attention was drawn to one thing (acoustic properties) and they have acquired another (the ability to produce the sound, without any awareness of how they are doing it). This ability has been acquired independently, by trial and error, and through repeated use once the produced sound matched the acoustic properties of the target. Nothing that was actually noticed has been automatized. The required control over motricity was developed through use, without any part of the knowledge about the sound serving as input to the proprioceptive system. The learners’ perception of the sounds, whether produced by them or by others, remains conscious. This is a further example of the fact that metalinguistic knowledge remains as such (i.e., explicit knowledge), while something else (implicit competence) is acquired. Articulatory competence and its automatic use remain non-conscious and independent (some individuals are able to perceive some acoustic contrasts but unable to produce them – it is this discrepancy between perception and ability to reproduce that guides them in their attempts to reach the target during the first stages of phoneme acquisition). Perception is a requisite in providing the target, but it is the repetition of the imitation of the target, not the knowledge, that produces competence. Consciousness does not act upon automatization. Negative evidence draws attention to the correct form but it is the repeated use of utterances containing it (in communicational contexts) that directly contributes to its acquisition (more precisely, to the acquisition of the relevant



Chapter 3.  The explicit/implicit interface debate 

underlying implicit procedures). Metalinguistic knowledge allows the monitoring and correction of utterances whose repeated use will contribute to the establishment of implicit linguistic competence. Neither the knowledge of the rule, nor the use of the rule when consciously constructing sentences, directly contributes to acquisition – only the repeated use of the resulting utterances serves as the input from which linguistic competence is implicitly abstracted (as discussed in Chapter 2). Negative evidence and the resulting metalinguistic knowledge thus do not interface with either the acquisition process or the ensuing implicit linguistic competence. In late L2 appropriation, metalinguistic knowledge precedes implicit linguistic competence. In the best of circumstances, implicit linguistic competence is acquired independently and its use gradually replaces that of metalinguistic knowledge. The two may be used in alternation or in parallel; they complement each other but do not interact. Some aspects (e.g., morphology or syntax) may be automatized while others (e.g., phonology) continue to be controlled – one more illustration of the neurofunctional modularity of the components of linguistic competence.

11.  The contexts of learning and acquisition Ellis and Larsen-Freeman (2006) rightly point out that language takes place in a social context, involving action, reaction, collaborative interaction, intersubjectivity, and mutually assisted performance; that individual learning is an emergent, holistic property of a dynamic system comprising many social, individual, and contextual influences. At the same time, acquisition is always independent of knowledge and proceeds from frequency tallying (Ellis, 2005). Learning and acquisition proceed independently, each using different aspects of the social, individual, dialectic and cultural influences, and develop in parallel, in the process of setting up two distinct systems (one explicit, the other implicit). The social context determines the input (language, socio-topolect, Paradis, 1998a), and possibly the output (opportunities, cultural constraints). To this extent, sociocultural variables will indirectly affect the form of the implicit procedures, depending on the type of input. A child raised in Japan acquires Japanese and a child raised in Portugal acquires Portuguese. Adult speakers of Japanese who live in Brazil will learn Brazilian Portuguese, helped by negative feedback and all other means of improving their knowledge, and thus will notice and produce an increasing number of correct sentences which may eventually lead to the acquisition of the relevant forms through the implicit tallying of their exemplars. The linguistic evidence received from feedback is metalinguistic evidence (which has no direct impact on the underlying structure or implicit procedures,

 Declarative and procedural determinants of second languages

of which native interlocutors are not even aware); together with pragmatic feedback, it allows the socially scaffolded development of metalinguistic awareness. Note that in this process, again, there is no direct contact between metalinguistic knowledge and implicit linguistic competence. In Ellis’s theory (Ellis, 2002, 2005), what directly brings implicit linguistic competence into being is not any kind of declarative knowledge or declarative process. Metalinguistic descriptions indirectly impact the acquisition of linguistic competence but their indirect effect does not constitute an interface. The various neural entities (representations, processes, states) that contribute to the activation of a conscious perception (or the various conscious perceptions and feelings that contribute to the activation of an unconscious representation) are germane to the representation that ends up being activated – either a component of the whole (as in vision) or a relevant factor in the equation of the desirability of triggering a particular response. No such relationship exists between elements of metalinguistic knowledge and implicit linguistic competence:  Metalinguistic knowledge is not a part of implicit linguistic competence nor a factor in selecting an appropriate implicit procedure. During the performance of effortful tasks individuals can temporarily inhibit the automatic activation of some processors and enter into a “controlled” mode of processing (Dehaene & Naccache, 2001). This is what native speakers do when they encounter difficulties: They switch from using implicit competence to using metalinguistic knowledge. This does not constitute an interaction between competence and knowledge; it is the temporary reliance on one or the other. Second language speakers do that in the course of conversation every time they are confronted with a gap in their developing L2 implicit grammar. On the other hand, second language teaching and language rehabilitation methods should not systematically implement any practice suggested by hypothetical psycholinguistic or neurolinguistic constructs before their application has been independently validated in the classroom and in the clinic, respectively. Even though these hypotheses may prove to be well founded, they are not necessarily directly applicable to the classroom. Alternatively, the classroom techniques that take these hypothetical constructs as rationale may prove serendipitously successful for unsuspected reasons. In this case, no harm is done. The story goes that people who lived around a small lake in a remote valley were plagued by a devastating endemic disease. They eventually surmised that it was caused by the putrid fumes that often emanated from the water. Therefore, they decided to move up the mountain where they could no longer smell the fumes, and indeed they no longer contracted the disease. In fact, it was transmitted by mosquitoes that thrived in the lake region where they found the stagnant water they needed to reproduce. Never mind why, the people are happy to be rid



Chapter 3.  The explicit/implicit interface debate 

of their disease. Their move up the mountain was a good decision. The reason why is of importance only for scientists – though for them, it is important. It is also important for future planning: Dwellings must be constructed away from swamps or the swamps must be drained. Similarly, once it has been ascertained that adults need to be made aware of forms before they can hope to acquire them, forms may usefully be taught. Moreover, if the aim of appropriation of a second language is to be able to communicate, and if one manages to do so with minimal use of automatic competence but with very efficient and speedy controlled metalinguistic knowledge, the end justifies the means. The distinction between automatic and speeded-up is important as long as it is a theoretical question, but for practical purposes, successful L2 speakers do not mind16 how and by what means they are able to communicate, as long as they do so efficiently. The distinction may, however, have practical uses, such as determining, after brain damage, which function is damaged or lost and how it can best be remedied by using intact functions to compensate (e.g., use metalinguistic knowledge when implicit competence is not available). If formal instruction makes a difference in achieving native-like competence in L2 grammar (Marinova-Todd, 2003), for practical purposes, it does not really matter whether you use implicit memory or explicit knowledge. Who cares how you manage to pass for a native speaker of L2? As long as you are able to successfully communicate in the second language, and the more accurately and fluently the better, the question is moot. However, from a scientific point of view, if you are interested in how things work, then it is crucial.

12.  Conclusion Ellis (2005) acknowledges that explicit and implicit knowledge are distinct and dissociated, that they involve different types of representations, and that explicit knowledge does not become implicit knowledge, but then holds that, nevertheless, the two types of knowledge interface. In this chapter, it has been argued that this is a contradiction. Ellis’s alleged interface is said to possess three characteristics: (1) It is dynamic (a misnomer: The process is not an interface but a substitution of the use of one system by another); (2) it happens in conscious processing (a logical contradiction: It is incompatible with the premises); (3) it is justified by the degree of indirect influence (a contradiction in terms: The alleged interface is opposed to the

16.  Any more than most car drivers worry about how the engine functions.

 Declarative and procedural determinants of second languages

non-interface position that makes the same claim, namely no direct influence). Ellis and Larsen-Freeman (2006) concede that interface is an unfortunate appellation for this issue but nevertheless continue to maintain that the interface is dynamic and that consciousness is the interface (so does Ellis, 2008). How conscious metalinguistic knowledge interfaces with unconscious linguistic competence in consciousness remains to be explained. Ellis’s account of an implicit acquisition mechanism is speculative (e.g., neurolinguists of the generative linguistics school radically disagree with the associative account of language acquisition) and awaits empirical verification. Nevertheless, even assuming that the theoretical construct were eventually validated to the last detail, Ellis’s account of an interface would still remain internally inconsistent. The studies cited in support are not relevant to the interface issue: They state that their constructs apply to conscious phenomena only, explicitly excluding non-conscious automatic functions from (the cited aspects of) their model (Baars, 1988; Scott Kelso, 2002; Dehaene & Changeux, 2004). Metalinguistic knowledge does not directly act on, hence interface with, implicit linguistic competence because the elements of metalinguistic knowledge (rules, paradigms, forms) are not the same as the procedures of implicit linguistic competence, especially if (1) the procedures are believed to be weighted connections governed by a statistical probability matrix (Ellis, 2000, 2002); and (2) the acquisition device consists of the implicit tallying of underlying structures rather than actual utterances – since the acquired structure is perceived in the form of many utterances, each with different words, so that the very same utterance is rarely encountered twice, as in Heraclitus’s river metaphor (Ellis, 2005; Ellis & Larsen-Freeman, 2006) in which the concept that is abstracted (the river) is different from all the perceived exemplars of flowing drops of water. To have an interface, one would have to assume that the implicit computational procedures are identical with the corresponding explicitly formulated rules (the way different drops of water are alike). Ellis (2005) has stated (but not shown) that the interface is dynamic and dialectic. In fact, he has not shown evidence of any type of interface. A transient dynamic interface entails that at some point during acquisition metalinguistic knowledge directly affects implicit linguistic competence. Not only has this not been demonstrated, but its very possibility has been explicitly denied:  Metalinguistic knowledge cannot be converted to implicit linguistic competence (Ellis, 2005:  307). Not here, not now, not ever. Conversely, implicit linguistic competence does not feed into metalinguistic knowledge. The latter is derived only indirectly, from the former’s explicit output; it is based on observation and rooted in declarative memory. Implicit competence does not affect metalinguistic knowledge directly. Metalinguistic knowledge is not based on the introspection of



Chapter 3.  The explicit/implicit interface debate 

implicit competence, but on a conscious recollection of observed language data. There is no one-way interface in either direction (let alone a two-way interface). An interface requires direct contact, either immediately between two entities, or between two systems via an interface, which itself directly transmits or converts information from one system to the other. None of Ellis’s accounts of memory processes meet any of these requirements. There is no direct contact between explicit knowledge and implicit competence. Acquisition remains incidental: there is no intake of explicit material from the noticeable input: The input is what is perceived, the intake is the frequency-tuned abstraction of patterns from this input. In sum, both Ellis (2005) and Ellis and Larsen-Freeman (2006) affirm that the interface is dynamic and that consciousness is the interface, but provide no justification, explanation, or demonstration. They fail to extricate themselves from the contradiction between the premises set up in Ellis (2005) and the claim that “nevertheless” explicit metalinguistic knowledge and implicit linguistic competence interface. The same strategy is used in both papers: Learning is discussed in detail, a number of genuine facts from various domains (albeit without relevance to the issue) are presented, but the interface as such is nowhere identified. The authors do not show how metalinguistic knowledge and implicit linguistic competence are mutually dependent. The two are complementary in that metalinguistic knowledge is used whenever implicit linguistic competence is not available; but metalinguistic knowledge as such does not interfere with the development of the underlying structure of the language that is acquired. When it is said that controlled processing becomes automatized, in fact a new processing mode has developed and the speaker switches from using controlled processing to relying on the (different) automatized system. Metalinguistic knowledge has not been automatized. Acquisition is dynamic and takes place during conscious processing, but it takes place independently whilst the acquirer is consciously processing something else. Notwithstanding the fact that neither the knowledge of the rule nor the use of it in order to consciously construct sentences contributes directly to (i.e., interfaces with) the development of implicit linguistic competence, there is a profound beneficial influence of metalinguistic knowledge on L2 acquisition. But this undeniable influence is not direct. No part of explicit knowledge makes direct contact with any part of implicit linguistic competence or its acquisition. Neither the various steps toward learning via various methods nor their resulting explicit knowledge directly serves as intake to the acquisition mechanism, which is based, not on knowledge, nor on noticing, nor on any conscious operation, but on the implicit tallying of patterns abstracted from the repeated use of various constructions (whether or not one is aware of their explicit grammatical structure).

 Declarative and procedural determinants of second languages

Nick Ellis is successful in showing how explicit metalinguistic knowledge indirectly facilitates the subsequent acquisition of implicit linguistic competence in L2, as it provides the means to correct errors and focuses attention on what needs to be practiced, but he fails to identify an interface between metalinguistic knowledge and linguistic competence. No matter how you slice it, the influence of metalinguistic knowledge, while extensive, remains indirect. Consequently, there is no interface in any conventional definition of the word. Whatever is observed, if noticed, is recorded in declarative memory. Metalinguistic knowledge remains in declarative memory unless it is forgotten. It can be used to infer an explicit grammar that can be consciously applied to the construction of sentences. There is no means by which it can directly influence (interfere with, contribute to, in any way modify, shape or tune) the procedures that sustain implicit linguistic competence. However, adult second language learners do need negative feedback to allow them to notice and use features of grammar that they would otherwise not perceive. These will then be incorporated in explicitly constructed utterances. The repeated use of different sentences containing such features will increase the input to the hypothesized implicit acquisition mechanisms. These mechanisms will abstract patterns from that input and establish frequency-tuned associative network connections. It is in this indirect fashion that metalinguistic knowledge has an influence on the development of implicit linguistic competence. Despite the fact that metalinguistic knowledge is not transformed into metalinguistic competence, there is no denying that language acquisition can be speeded up by explicit instruction, and that focused L2 instruction can result in substantial target-oriented gains. These gains generally take the form of speededup controlled use of metalinguistic knowledge but may possibly correspond to eventual improvements in competence. Ellis (2005) is correct in stating that a particular form will not be acquired unless it is used, and that the form is not used by an adult second language learner until it has been noticed. However, it is not by noticing that the form is acquired. The learning itself does not participate in the acquisition of the underlying structure and the knowledge itself does not contribute to the shaping of the implicit mechanisms responsible for the generation of the relevant procedures. It is, according to Ellis (2002, 2005), not the learnervs awareness of the surface structure (or of its explicit analysis) but its frequent use in context that will eventually result in the acquisition of its underlying structure. The acquisition of the structure is doubly independent of its learning:  The mechanism of its acquisition differs from that of its learning, and what is acquired (either a generative rule or a pattern of weighted connections – both implicit and



Chapter 3.  The explicit/implicit interface debate 

non-conscious) is different from what is learned (the surface structure and/or its analysis – both explicit and conscious). What this demonstrates is that adults do not acquire the grammar of a second language from mere verbal interaction with speakers of the language, the way children acquire their native language, but need to explicitly learn (at least some aspects of) the language, so as to be able to use a sufficient number of utterances incorporating a certain feature before it is acquired (in the long run, by the same mechanism that children use, possibly less efficiently). This is what we will examine in the next chapter.

chapter 4

Ultimate attainment in L2 proficiency It is no longer possible to speak of language without taking into consideration the facts that (1) the language system is one component of verbal communication, (2) the language system contains dissociable modules that have their own intrinsic properties; and (3) implicit components of language are different in nature and subject to different types of control than explicit components (in particular vocabulary) – they develop independently of each other according to their own genetically programmed timetable, and are susceptible to different external factors. These considerations are essential in the investigation of a critical period for language acquisition. There are basically two schools of thought on the critical period hypothesis: neuroscientists assume that there is some form of critical period that is too obvious to warrant discussion, let alone a controversial debate; language teachers and social psychologists are adamant that there is no such thing as a critical period and refuse to even consider neurological data. The term “critical period” is so controversial that it would be counterproductive to use it here. Suffice it to mention the differences between ultimate attainment in L2 as compared to L1 and investigate the neurophysiological factors that account for the readily observable and widely acknowledged differences. In many animals, it has been shown (and uncontroversially accepted) that neural circuits are shaped by experience during restricted periods in early life (Yazaki-Sugiyama, et al., 2007): These include the cat (Hubel & Wiesel, 1962), mouse (Iwai & Lester, 2006), and some songbirds (Hensch, 2004). It would be surprising if the human brain were exempt from such restrictions (though less surprising that their existence is not easily accepted – the wish for free will is overwhelming). If there is a critical period, it is certainly different from that observed in birdsong or cats’ ability to acquire the perception of vertical lines. For one thing, the human brain takes much longer to mature than any other animal’s, including other primates. Therefore, the notion of an optimal period for acquiring languages will be proposed after discussion of the available data. The least controversial observation is that every individual without severe mental defects has acquired a native language. Some have even acquired two or more. Not everyone who has acquired an L1 manages to acquire an L2. Some individuals find it excruciatingly difficult and some never get beyond the most basic rudiments. Why is this so? The non-neurophysiological factors that are proposed by researchers are in fact, as we are about to see, direct consequences of a variety of neural underpinnings.

 Declarative and procedural determinants of second languages

Not just the manner of appropriation, but the nature of what is appropriated (competence vs. knowledge) is affected by age of L2 appropriation. Children exposed to L2 interaction starting any time before the age of 4 or 5 (and the younger the better) acquire the second language implicitly, like the first, using procedural memory. For example, early German-French bilinguals were found to have no problem with acquiring French gender before age 3, but to have problems when French is acquired after the age of 3 (Möhring, 2001). After age 6 or 7, second language appropriation relies more and more on conscious learning, thus involving declarative memory.

1.  Ultimate attainment in L1 and L2 Birdsong (2006) remarks that there is a widespread belief that native-like attainment by late L2 learners will be confined to one or a few tasks and that an individual will not display native-likeness across the full range of linguistic behaviors or experimental performances. His overview of empirical findings does not show otherwise. Interestingly, Birdsong (2007) reports that most tasks on which late L2 learners are able to approach or achieve native-likeness are off-line tasks. In any case, the learners often do better on off-line tasks. Also, there seems to be no limit to the ability to learn L2 vocabulary (though the exact semantic boundaries of words, their various connotations, and the constraints on their uses in proper contexts, which depend on experience (including incidental experience1) rather than explicit instruction, remain incomplete). This would suggest that there is something peculiar to such tasks. One aspect worth considering is that off-line tasks are known to rely on explicit knowledge, hence on declarative memory rather than implicit competence. This might imply that, perhaps, the capacity to acquire implicit linguistic competence is susceptible to decline with increasing age of onset of L2 appropriation. Whether there is a critical or sensitive period or a gradual decline up to a certain age, there appears to be a rather early age after which a second language does not reach native-likeness in all aspects of use, as shown in Birdsong’s (2006) overview and (2007) presentation. This is not to say that, after many years of total immersion and practice, L2 never becomes very close to native-like in many aspects – sufficiently so for everyday practical purposes.

1.  New words can be acquired incidentally when listening to, and reading, a story while focusing on comprehension. The meaning of words is learned even if one does not remember having encountered them in the text (Horst, Cobb, & Meara, 1998).



Chapter 4.  Ultimate attainment in L2 proficiency 

Birdsong proposes that there may not be a single linguistic task that some individual cannot fully master (in relative isolation) in L2. This means that, in principle, any given linguistic task can be mastered by someone. But there definitely seems to be a dearth of individuals able to master all components. Native-like proficiency in the second language is almost never acquired and second-language processing is slower (Toppelberg, 1997). Why should this be? Different individuals are able to deal with any one or two components of second language processing, but not all components at the same time. This suggests that these individuals are not able to consciously control so many components simultaneously, and hence at least some components are not automatized. Let us remember that divided attention (i.e., having to pay attention to more than one task at the same time) interferes with performance on explicit tasks, not on automatized processes: A native speaker has no problem processing in parallel the various phonological, morphological, and syntactic components during lexical retrieval. At the very least, this is an indication that a different mechanism is at work, at least partially, in L2 processing – and a good candidate is the use of declarative memory to compensate for the gaps in L2 implicit linguistic competence. According to Roehr (2008a), implicit linguistic competence is stored in and retrieved from an associative network during parallel distributed processing, whereas explicit knowledge is processed sequentially with the help of rule-based algorithms. The difference in kind between these two processes results in phonology, morphology, syntax, and lexical retrieval being processed in parallel (hence simultaneously) by linguistic competence, while metalinguistic knowledge is processed only one item at a time; metalinguistic knowledge requires attention, whereas linguistic competence does not. Studies that specifically examine the ability of L2 users to pass for native speakers indicate that passing for a native speaker is a temporary, context-, audience-, and medium-dependent performance (Piller, 2002; Marinova-Todd, 2003). This reinforces the notion that even expert L2 users’ performance is controlled to some extent (i.e., not as automatic as native speakers’), as previously suspected, given that a late-learned L2 is more vulnerable to noise, fatigue, stress, and declarativememory impairments (amnesia, Alzheimer’s disease, even normal aging). It is also positively related, among other things, to amount of formal L2 study and level of formal education (Marinova-Todd, 2003), factors that do not affect the acquisition and concurrent use of phonology, morphology, syntax and pragmatics of L1. Many studies do show that adults are able to learn one or another aspect of language to a native-like level, independent of the other components of grammar. In L2, different components of the implicit language system are appropriated independently of each other at different rates and to different extents. This contrasts with the way very young children simultaneously acquire phonology, morphology

 Declarative and procedural determinants of second languages

and syntax without paying attention to these components, while focusing their attention on the semantic and pragmatic aspects of verbal communication. Typically, when one skill component has been mastered in a second language, it is after much concentration on that particular component, independently of all others. Effort, hence conscious control, is exerted in developing a particular language component in a way that is not the natural way of developing implicit linguistic competence. The latter is achieved by developing the various components of language structure in parallel, incidentally, without paying attention, and without focusing one’s efforts on any specific subcomponent (phonology, morphology, syntax, semantics and the grammatization of the lexicon). They are all internalized simultaneously (possibly at different levels of development in each, but concurrently nevertheless). Thus, they are integrated into implicit linguistic competence and can be used in unison when performing the normal task of comprehending and producing the complex construct that an utterance represents, with each part contributing to the whole (prosody, segmental phonology, morphology, syntax, lexicon). Moreover, each component is selected automatically in accordance with the pragmatics of the situation, including the intention to communicate a particular message, modulated by the specific situational context, the knowledge of the interlocutor’s beliefs, and placing emphasis on the appropriate concept through all means afforded by the grammar – i.e., prosody, word order, inflectional morphology, loudness, speed of delivery, etc. In Marinova-Todd’s (2003) study, some L2 learners failed to achieve nativelike levels of proficiency in grammar knowledge, but scored within the native range on pronunciation measures, whereas other L2 learners achieved nativelike scores on grammar measures and failed to achieve native accent in the L2. Thus, some highly proficient L2 speakers tend to be stronger in some areas of L2 knowledge and weaker in others, and score within the native range in only some domains, which means that they have not internalized the L2 as a whole. When much effort is exerted on one particular aspect, it may reach native levels, but it is a skill that is not integrated into the general implicit linguistic competence system for that language. In this study, out of a group of 30 participants selected from among very highly proficient L2 speakers, only 3 participants consistently achieved scores within the native range. And even then, scores within the native range do not necessarily imply that these participants used the same means to achieve similar results. The reasoning that, because some individuals are able to attain native-likeness in some aspects of L2 performance, it can be assumed that it is possible for some individuals to attain native-likeness in all language tasks, is fundamentally flawed. Normal language performance incorporates all components and puts them into action simultaneously. This is made possible because the various integrated



Chapter 4.  Ultimate attainment in L2 proficiency 

functions are automatic, without conscious control, in which case, there is no dispersion of effort. As Hyltenstam and Abrahamsson (2003) point out, the subtle differences that seem to exist between native and native-like proficiency “are probably highly insignificant in all aspects of the second language speaker’s life and endeavors, although very significant for a theory of human capacity for language learning” (p. 580).

2.  The optimal period The notion of optimal period retains the general characteristics of the traditional critical period (it applies to skills, it is time-sensitive (if not time-locked) and depends on properties of the brain), but is more flexible in that it is not categorical (all-or-nothing) and admits of variability among individuals with respect to maturational deadlines and length of developmental stages (themselves determined by complex interactions between genetic and experiential factors). The fact that there are (rare) exceptions does not mean that there is not a general principle at work. (Some sheep are born with five legs;2 this does not prevent encyclopedias and veterinary handbooks from describing sheep as four-legged.) In the context of interest, the optimal period hypothesis thus applies to implicit linguistic competence, which depends in large part on the expression of the gene FOXP2. The gradual decline in procedural memory for language forces late second-language learners to rely on explicit learning, which results in the use of a different cognitive system from that which supports their native language. It is the acquisition of implicit competence that is affected by age, both biologically (gradual loss of plasticity of the procedural memory for language after about age 5) and cognitively (greater reliance on conscious declarative memory for learning in general and, consequently, for learning a language from about age 7). If we assume that normal language acquisition and use refer to the incidental internalization and automatic use of implicit linguistic competence, then an optimal period affects the acquisition of language. It is a gradual process within a window between the ages of 2 and 5 years, give or take a few months in view of the considerable interindividual variability in the rate of maturation in general and of development of the language areas in particular. In fact, even birdsong critical periods are not chronologically invariant and their duration can be regulated by the amount of tutor song exposure, vocal practice, and the brain’s steroidal milieu (Mooney, Prather, & Roberts, 2008).

2.  One specimen is exhibited at the Musée Cantonal de Zoologie, Lausanne, Switzerland.

 Declarative and procedural determinants of second languages

Human FOXP2 is a gene whose integrity is necessary (but not sufficient) for the acquisition of implicit linguistic competence. It determines the expression of various genes at specific times during brain development and at diverse time-points during the lifetime of an organism (Marcus & Fisher, 2003), including that of areas of the cerebellum and basal ganglia that subserve the acquisition and subsequent processing of implicit linguistic competence. Its mutation (or a lack of exposure to language input at the time of its programmed triggering of the relevant genes) disrupts language acquisition. Individuals must then have recourse to compensatory mechanisms in order to appropriate language through learning. The optimal period thus refers to the period during which individuals must be exposed to language interaction if they are to acquire linguistic competence. This period has an upper limit that varies with respect to which component of the implicit language system is acquired, namely, in chronological order, prosody, phonology, morphology, and syntax (including syntactic features of the lexicon). But the vocabulary, that is, the sound-meaning pairing of words, is conscious and hence subserved by declarative memory; consequently, it is not susceptible to the optimal periods that apply to the various components of implicit competence. Systematic performance in real-time3 language processing is the litmus test of implicit linguistic competence. “Age of exposure during language acquisition seems to have a dramatic impact on the subsequent real-time processing of sentences” (Friederici, Steinhauer, & Pfeifer, 2002:  529). The optimal period that applies to implicit linguistic competence can be masked to some extent by reliance on compensatory mechanisms whose control can be considerably speeded-up. To the extent that proficient L2 is subserved by declarative memory, it is not susceptible to the optimal period. Not only does L2 performance differ from L1, but it differs along the implicit/ explicit dimension. Given that vocabulary learning is sustained by declarative memory (in both L1 and L2), there is no optimal period for learning new words or explicit grammatical rules, except for the gradual decline of declarative memory function with advanced aging, culminating in senility (and accelerated in Alzheimer’s disease).

3.  Optimal window of opportunity To ascertain the learner’s potential in post-adolescent L2 acquisition is a legitimate goal and a commendable enterprise. The very fact that the question is posed

3.  As opposed to off-line, when individuals have the opportunity to consciously control what they are doing (or saying), as in written tasks in general and grammaticality judgment tasks in particular. Tasks performed in real time (i.e., on-line tasks) are assumed to be performed automatically.



Chapter 4.  Ultimate attainment in L2 proficiency 

highlights the fact that there is a difference between implicit linguistic competence attainment in L1 and L2. One would not propose to study the potential for native language acquisition in normal (i.e., not brain-damaged or profoundly mentally impaired) individuals without FOXP2 genetic anomaly. Note that although richness of vocabulary varies between native speakers, the ability to fully acquire the basic phonology, morphology, and syntax of the individual’s topo-/sociolect does not. There are indeed differences in what individuals can do with their native language, in how colorfully they are able to express their ideas, but all have mastered the components of the implicit grammar of their language and are able to use them simultaneously to understand and produce utterances automatically. That means that their output is consistent, that is, without variability: they do not vibrate their vocal flaps for the right number of milliseconds or place the adverb in the correct position only 75% of the time (or “above chance,” as is often reported with obvious satisfaction in L2 studies, incorrectly interpreted as evidence of incorporation of the tested element into the subjects’ implicit competence). If one day you started violating subject-verb agreement 25% of the time, your close friends and relatives would no doubt be alarmed. There seems to be a period from birth to age 4 or 5 after which nativelikeness becomes progressively rarer and attainment less successful. In other words, between the ages of 2 and 5, children acquire the basic grammar of their native language(s). When individuals are exposed to a second language after that age, native-likeness is rarely, if ever, achieved on all language tasks even though behavioral measures may improve especially after years of total immersion in an L2 environment. But as Birdsong (2006) rightly points out, native-likeness at the L2 acquisition end state does not imply access to Universal Grammar (or implicit linguistic competence) – especially in the light of better results on off-line than on-line tasks and poorer results on those aspects that are more difficult to control consciously (e.g., phonology) than on those like syntax, where surface word order and other features are observable and can be volitionally controlled (and, with practice, speeded up). Based on their study of Nicaraguan sign language (and on studies by Kegl, Senghas, & Coppola, 1999; Newport, Bavelier, & Neville, 2001; Senghas & Coppola, 2001; and Mayberry, Lock, & Kazmi, 2002), Morgan and Kegl (2006) estimate the window of time for language acquisition to be less than 6 years for native-like acquisition, and less than 10 to gain some acquisition benefits. According to Mayberry et al. (2002), language learning ability is determined by the onset of language experience during early brain development, independent of the modality of experience (spoken or ASL). The ability to acquire language arises from a synergy between early brain development and language experience. It is seriously compromised when language is not experienced during early life. The timing of the initial language experience during human development strongly

 Declarative and procedural determinants of second languages

influences the capacity to learn language throughout life. Newman et al. (2002) have demonstrated that the right hemisphere angular gyrus is active during ASL processing only in native signers. Right-hemisphere damage in native signers leads to impairments in the processing of syntactic constructions and classifiers that rely on spatial relationships. This region is less susceptible to modification by experience after puberty. Adolescent ASL first-language learners cannot process the language as efficiently as native signers due to their lack of grammatical competence and related problems in processing (Morford, 2003). Similarly, Grimshaw et al. (1994, 1998) describe the case of a young man who, profoundly deaf since birth, was fitted with auditory aids at the age of 15. His subsequent language development has demonstrated growth of vocabulary and semantically related syntax, but he has considerable difficulty with syntactic structures that cannot be semantically mediated. The authors conclude that his development is consistent with the hypothesis that there is a critical period for first language acquisition, especially with respect to syntax. They point out that individuals who were not exposed to language until after the optimal period present a failure to comprehend some syntactic structures (pro-forms, movement rules, verb tense) and a large disparity between comprehension and production (thanks no doubt to the availability of context and pragmatic cues). Mayberry (1993) investigated whether the long-range outcome of L1 appropriated after early childhood is similar to that of L2 learning in deaf individuals. Participants born with normal hearing subsequently lost in late childhood, who had then learned ASL, outperformed those who appropriated ASL as a first language at the same age. Moreover, the performance of the latter declined with increasing age of appropriation. Similarly, children who had otitis at age 1 have identifiable language deficits at age 9 (Hyltenstam & Abrahamson, 2003). The authors consider that these data support the notion of an optimal period beyond which a natural language can no longer be normally acquired. In the study by Rönnberg et al. (2004), only the subgroup of subjects who started sign language at birth (as opposed to those who started at primary school) evinced a clear left-hemisphere dominance for a working memory task performed in sign language, in line with the findings for working memory in spoken language. Those who started learning sign language at school seemed to apply explicit treatment to the visuospatial processing involved in generating the virtual spatial array needed to complete working memory tasks in sign language, rather than handling it implicitly as the early bilinguals did. The authors suggest that the difference in their results for the two groups might reflect an age-of-acquisition effect. As delay in exposure to a first language increases, accuracy of grammaticality judgments decreases, independent of ASL syntactic structure. “The onset of first language acquisition affects the ultimate outcome of syntactic knowledge for all



Chapter 4.  Ultimate attainment in L2 proficiency 

subsequent language acquisition” (Boudreault & Mayberry, 2006: 608). Adults have largely lost the ability to learn a language without reflecting on its structure and have to use alternative mechanisms to learn a second language (DeKeyser, 2000). In Harley and Hart’s (1997) study of students enrolled in L2 early (starting in grade 1) and late (starting in grade 7) immersion programs, analytical language ability (relying on conscious memory) was the only significant predictor of L2 proficiency (tested in grade 11) in the case of late but not early immersion students. Ross and Bever’s (2004) study comparing the sensitive period for language acquisition in two populations of deaf individuals (with familial right-  or lefthandedness) found that when both populations are exposed to language in early childhood, comparable levels of proficiency are attained. However, individuals with familial left-handedness show evidence of a shorter sensitive period. Age of acquisition rather than years of experience determined sign language proficiency. This suggests that genetic factors are involved that apply to both handedness and language, and that their expression is sensitive to time of first language exposure. Sundara and Polka (2008) have shown that advanced early L2 learners (i.e., L2 exposure onset by 5 or 6 years of age) discriminated /d/-initial syllables in Canadian French (dental /d/) and Canadian English (alveolar /d/) in a way consistent with a merged category, whereas simultaneous bilinguals were at least as good at discriminating between them as unilingual speakers. This suggests that, at least at the phonological level, simultaneous bilinguals acquire each language as unilinguals do, whereas early L2 learners do not. Even by 6 months of age, well before word meanings are acquired, infants’ phonetic perception has been altered by exposure to a specific language, which results in language-specific prototypes that assist infants in organizing speech sounds into categories (Kuhl et al., 1992). The claim is not that adults cannot master foreign languages, but that their achievement is mainly the result of conscious learning and conscious control of their output. After many years of total immersion in an exclusively L2-speaking environment, without contact with speakers of L1, some, possibly most, of the components of implicit linguistic competence may eventually be automatized. As Hyltenstam and Abrahamson (2001) point out, even late learners can achieve native-like behavior for individual tasks, structures, or domains. Nevertheless, published studies have still not identified a single adult learner who is indistinguishable from a native speaker in all relevant aspects of the L2.

4.  The optimal period is restricted to implicit linguistic competence The optimal period applies to the normal acquisition of language, which results in implicit linguistic competence. But language can also be learned, using cerebral

 Declarative and procedural determinants of second languages

mechanisms other than those used to acquire implicit linguistic competence, and resulting in conscious knowledge about form, namely explicit metalinguistic knowledge that can be mastered to a high degree of proficiency. Its controlled use can be sufficiently speeded up to be perceived as native-like – as is the case with the L1 of intelligent genetic dysphasic individuals (Paradis & Gopnik, 1997). The use of declarative memory to compensate for gaps in L2 implicit competence is reflected in the considerable inter-individual variability in attainment between late L2 learners – compared to considerable inter-individual homogeneity in the acquisition of the native language(s) – the greater reliance on working memory, the role of education, the success in semantics relative to syntax and phonology, the success in off-line relative to on-line tasks, the decline with age, and in general, the ease with L1 vs. the difficulty with L2. 4.1   Inter-individual variability in attainment Acquisition via procedural memory is available to everyone up to about 5 years of age, after which the use of procedural memory to acquire language rapidly declines and individuals rely on declarative memory. Note that the decline in the use of procedural memory when appropriating a second language is not necessarily due to a deficiency in procedural memory for language per se (though it may be at least partially so), but possibly also to a number of psychological factors such as the propensity to use general learning ability (as applied to the many other things learned from that age on), the presence of the L1 system and the general difficulty of acquiring new habits of the same general kind as existing ones (e.g., for a tennis champion to acquire badminton skills), which drives the speaker to continue to rely on L1 procedures when generating L2 sentences (at each level of language structure) and to apply L1 meanings to quasi-equivalent L2 lexical items. Some implicit linguistic competence in L2 can probably be acquired in certain aspects of linguistic structure (syntax, morphology, phonology, in that order of probability) though not completely at any level. This is one reason why there is great variability in individual success at learning a second language. By contrast, any one without severe mental retardation and with an intact FOXP2 gene acquires a first language fully and easily, with hardly any inter-individual variation. Learning a second language is dependent on general intellectual capacity – and is positively correlated with the individual’s IQ (Mayberry, Taylor, & Obrien-Malone, 1995), another source of variance. Extensive practice over long periods of time may help with the acquisition of some components of the grammar and speed up the controlled use of the rest. Some rare L2 speakers may achieve native-like proficiency (i.e., mastery of phonology, morphology, syntax and the lexicon) but by other means (cf. Rieber & Vetter, 1995).



Chapter 4.  Ultimate attainment in L2 proficiency 

As Hyltenstam and Abrahamsson (2003) point out, adult learners do not acquire a second language from mere exposure but learn it indirectly. Young learners perform more similarly to each other whereas older learners show greater variation in their rate of appropriation and their ultimate attainment in their L2 (Marinova-Todd, Marshall, & Snow, 2000, 2001). Speakers process a late-learned second language differently than their native language and the resulting performance is rarely (if ever) the same. Even if their second-language production and comprehension were observably identical to those of L1 speakers, the fact that they use speeded-up control rather than automatic processing would be evidence that, after a certain age, one has to resort to an altogether different processing mechanism because the acquisition of implicit competence is no longer possible (or extremely time-consuming and inefficient). Whereas procedural memories are more resistant to loss over time than declarative memories (which are especially vulnerable in aging), procedural memory for language acquisition becomes less efficient and takes longer with increasing age – for a number of reasons, including L1 entrenchment as discussed below. Even when the second language is acquired at a very early age, differences between the processing and/or representation of L1 and L2 have been reported. Perani et al. (2003), for instance, found that bilingual speakers who had been exposed to the second language from the age of 3 (and who had used both languages in daily life ever since, with comparable levels of proficiency in the comprehension of both) showed less extensive cerebral activation during lexical search and retrieval in the language acquired first, suggesting that additional resources were recruited within a dedicated network when generating words in L2. (See Mack, 1984, 1986, for experimental evidence of differences in relatively early bilinguals.) Even individuals with a very young onset of L2 experience diverge at the level of fine linguistic detail from native speakers (Singleton, 2001). With respect to some measures of phonetic performance, “extremely early exposure” is required to perform like native unilinguals (Mack, 2003). There are at least two possible basic reasons for deviance from the native norm in early L2 acquirers: (1) the quality of the L2 spoken in the child’s environment (parents, relatives, sometimes a whole immigrant community), which becomes the norm for the acquirer (just as students in Montreal immersion classes in the sixties picked up the pidgin of their peers4); (2) generalization across the two language

4.  The usefulness of form-focused instruction, as discussed in Chapter 3, is exemplified in the success of the more recent introduction of explicit grammar in immersion classes where, before, high levels of fluency were accompanied by notoriously poor accuracy (Lyster, 1990, 2004; Spada, 1997; Day & Shapson, 2001; Jean, 2005).

 Declarative and procedural determinants of second languages

subsystems and influence of the items first acquired and most often activated on the corresponding items in the other language. The undeniable influence of social and educational variables on L2 appropriation at a later age stems from a fundamental neurobiological phenomenon, namely the apparent gradual (or not so gradual) loss of the ability to acquire language incidentally, to use procedural memory so that it would become available for automatic use. This inability is compensated for by relying on conscious learning, using declarative memory. The numerous causes of inter-individual differences in attainment are a direct result of a number of internal and external factors5 to which the acquisition of a native language is impervious. 4.2   The impact of working memory and level of education In tasks that tap working memory and episodic memory (i.e., that rely on declarative memory), there is an observed performance decline with age, whereas on tasks involving procedural memory, age-related effects, when observed, are comparatively mild (Birdsong, 2006). Yet, it is on tasks involving procedural memory par excellence (e.g., pronunciation, but also automatic (hence systematic) use of all aspects of the grammar), that late L2 learners do worst. This again suggests, consonant with the reported “relatively low degree of automaticity in L2” (Birdsong, 2006: 29), that some of the language processes that are sustained by procedural memory in L1 are dependent (at least in part) on declarative memory in L2 – especially when one considers that “the entorhinal cortex and hippocampus appear to incur greater annual shrinkage than other areas of the brain” (Birdsong, 2006:  31). “This decrease is linked to age-related cognitive deficits across domains such as working memory and executive function” (p. 33), both involved in explicit tasks. Declines in the anterior cingulate cortex (p. 33) are related to problems with conscious control. It is noteworthy that “executive control processes associated with prefrontal and cingulate cortices can operate only on consciously perceived stimuli” (Dehaene & Changeux, 2004: 1152). These data would suggest that adults have recourse to declarative memory to learn (rather 5.  Internal: Cognitive style, motivation, attitude, aptitude, IQ, level of education. External:  age at exposure, degree of exposure relative to L1, degree of exposure to L1 during the appropriation of L2, high-/low-prestige status of L2 in the community, ethnic and political factors associated with L2, structural distance between the languages, quality of the L2 spoken in the individual’s environment. Internal and external factors interact in that, for instance, the sociolinguistic and political status of L2 will affect an individual’s attitude and motivation. The level of education, imposed by external circumstances, nevertheless has an impact on the individual’s ability to learn, in that it develops reasoning capacity.



Chapter 4.  Ultimate attainment in L2 proficiency 

than acquire) a second language, and that this task becomes more difficult with age as the underlying cerebral structures that sustain declarative functions wane. In a pilot experimental investigation, Fehringer and Fry (2007) found that highly proficient German speakers of English produced a significantly higher overall rate of hesitation phenomena in their second language than in their first (p = .000). The difference was most noticeable in the types of phenomena (repetition, corrections, expansions) that would indicate extra planning demands, as shown by the increased necessity for reformulation in L2. This is taken to indicate that an additional cognitive load was imposed by working memory in L2. The tasks that showed significant differences between languages are the ones that demand most attention. In L1, greater production of optional complementizer phrases (whose embedding is considered a particularly demanding task) is significantly correlated (p = .033) with fewer hesitation phenomena (suggesting automatic processing) whereas in L2 the correlation is not significant. Working memory scores were significantly higher in L1 than in L2 (p = .005). To account for the interference from L1, which plays an important role in constraining the native-like performance of L2 speakers, the authors surmise that speakers with poorer working memory resources for L2 are likely to find it difficult to control their language subsystems: the native language that is supposed to be suppressed might interfere with the second language that is selected for use. Factors other than working memory that may have influenced the results, such as depletion of energy and anxiety are also indicative of extra reliance on consciously controlled processes (Dewaele, 2007). Fehringer and Fry observe that their subjects’ L2 is not quite the same as their native language in spite of their extremely high level of ability in L2 grammar. They conclude that, in fact, L2 users rarely reach a level of fluency approaching that of native speakers. Speakers are said to need to “work harder” in order to display ease and fluency in their second language; L2 working memory may lack sufficient attentional resources. Interestingly, effort and attention are associated with conscious, non-automatic processes. Level of education is often cited as a significant predictor of high proficiency achievement in L2. The advantage gained by the study of L2 as a foreign language prior to immersion in the L2 environment is noticeable even after decades of exposure (Urponen, 2004). Instruction has a direct influence on learning, not on acquisition (cf. Harley & Hart, 1997). Instruction benefits proficiency, but with a focus on explicit learning (Bialystok, 1997). 4.3   The success in semantics relative to syntax and phonology In an ERP study, Sanders and Neville (2003) show that native speakers and late bilinguals process words similarly, whereas syntactic processing is strongly

 Declarative and procedural determinants of second languages

impacted by age of acquisition. They conclude that these findings support the proposal that subsystems within language display varying degrees of plasticity. Indeed, what they show is that implicit systems subserved by procedural memory (here, syntax) are affected by age of acquisition, whereas explicit semantic systems subserved by declarative memory (here, words) are not. Hahne (2001) also reports an ERP experiment, in which native Russian speakers who had learned German as a second language differed significantly from native listeners in various aspects. For semantically correct sentences, the N400 negativity was more pronounced, extended to frontal electrode sites, and was delayed by about 100 ms in the L2 group, as compared to the native German controls. Moreover, the difference between correct and incorrect sentences was much smaller in the L2 group. Thus, with regard to semantic aspects, the ERP differences were only quantitative. However, with regard to syntactic aspects, the differences were qualitatively different: Phrase structure violations elicited an early negativity in comparison to correct sentences in the native listeners, an effect interpreted as reflecting automaticity. There was no such modulation of the anterior negativity in the L2 group, suggesting a deficiency in automaticity. As in previous studies with Japanese and French speakers (Hahne & Friederici, 2001), language learners did not process syntactic categories in the same way as native listeners did. In a lexical decision task in which target words were primed by adjectives that were correctly or incorrectly inflected for gender (the morphosyntactic condition) or by adjectives that were semantically associated or not associated with the target word (the semantic condition), Scherag et al. (2004) found that native German speakers gained from both morpho-syntactically and semantically congruent primes. In contrast, long-term English immigrants to Germany did not benefit from morphosyntactic primes, whereas their semantic priming effects were similar to those of the native German speakers. Also worthy of note is the fact that, in addition, the L2 participants’ overall processing time was longer, another indication of reliance on non-automatic processes. The authors interpret their data as suggesting that the full acquisition of at least some syntactic functions may be restricted to limited periods in life, whereas the elaboration of semantic functions is based on associative learning mechanisms that permit learning throughout life. 4.4  The decline in L2 performance with increasing age According to Birdsong (2006), a review of the literature reveals that for late L2 learners there is either (1) a random array of scores or (2) a persistent decline in performance with increasing age of appropriation. The first is consistent with declarative learning in general: in contrast with native language acquisition, individuals differ considerably in various domains of cognitive ability (including explicit language learning). The second corresponds to the observed growing



Chapter 4.  Ultimate attainment in L2 proficiency 

difficulty of using declarative memory as age increases (associated with the reported waning of hippocampal structures and the anterior cingulate cortex). Because declarative memory abilities decline more with age than procedural memory functions (Birdsong, 2005, and citations therein), to the extent that L2 is subserved by declarative memory, L2 should decline more than L1 in advanced aging, as control becomes less effective. We may expect elderly speakers to show a decline in fluency, accuracy and phonology in the production of L2, of which they may be aware as they hear their own output – since production is more sensitive to decline than comprehension (and perception, in the case of a foreign accent, as errors of lexical stress application or phoneme production are noticed after faulty production). 4.5  The ease of appropriation and use of L1 vs. L2 In the face of (1) the considerable difficulty in acquiring a second language in adulthood and (2) the fact that two or more languages can be acquired as effortlessly as one when the child is exposed to them at a very early age (the earlier, the better, i.e., from the crib), it is not unreasonable to consider that the time constraints imposed on the acquisition of native implicit linguistic competence – as demonstrated by studies of L1 acquisition delay (Lebrun, 1978; Mayberry, 1993; Mayberry et al., 2002; Boudreault & Mayberry, 2006), – must also apply to the acquisition of implicit linguistic competence in a second and third language. The fact that young children are slow at appropriating a second language and need a longer period to achieve levels that adolescents and adults can achieve faster, even though they tend to surpass adults in the long run6 (Nikolov & Mihaljevic Djigunovic, 2006), suggests that young children acquire the language (incidental acquisition takes time) whereas adults, to a great extent, learn it (they reach a certain level of accuracy by means of explicit learning, which is faster, but this knowledge is limited and is not converted into competence; nor, most of the time, is much competence acquired in parallel, as learners continue to rely on declarative memory). Children acquire their native language(s) long before they have any explicit knowledge of language. They are not more efficient L2 learners than adults, but they are more efficient L2 acquirers. Even though they are slower at acquiring a second language – implicit acquisition “is slow because it needs a large sample” (N.C. Ellis, 2005:  315) – they eventually internalize it better than adults. Unlike young children, adults find it very difficult to incidentally acquire the competence

6.  “There is enough evidence to show that child second language acquirers are indeed superior [to adult learners] in terms of ultimate ability” (Patkowski, 1990: 73).

 Declarative and procedural determinants of second languages

that allows them to use language constructions automatically. “Native-like proficiency is almost never acquired, and second-language processing is slower” (Toppelberg, 1997: 1328). Some adult L2 learners “are impervious to years of input that evidences tens of thousands of exemplars of high-frequency form-function patterns” (N.C. Ellis, 2005: 322). Most adults’ faster appropriation is the result of the use of speeded-up metalinguistic knowledge. As Ellis (2005) reminds us, accuracy and fluency are not necessarily an indication of implicit linguistic competence. In Montrul et al.’s (2006) study on the use of clitics, early and late bilinguals performed alike (yet late bilinguals were more inaccurate than early bilinguals at rejecting sentences in some conditions, and there was a clear advantage for early bilinguals with clitic left dissociations) on an off-line task (grammaticality judgment). On an on-line task, with all sentence types containing clitics and objects in sentence-initial position, the reaction times of the late bilinguals were slower than those of the early bilinguals (whose RTs did not differ from the unilingual control group’s). The slower responses of late bilinguals are consonant with the use of controlled rather than automatic processing, hence a lack of implicit competence for those items. Montrul and collaborators note that early bilinguals appear to have more native-like knowledge of clitics than late bilinguals, even when they have low-to-intermediate proficiency in the language. According to the authors, this may be due to the fact that the clitic system was acquired before age 4, as in unilingual children. By contrast, late bilinguals use more metalinguistic knowledge. “The early bilinguals are more accurate and faster on clitic-left dissociation, as if their linguistic knowledge were automatic” (p. 227). Indeed, the hallmark of automaticity is systematic accuracy conjoined with speedy processing. In a study by Flege et al. (2006), immigrant children scored lower than natives but higher than adult immigrants, though they still had a detectable accent after 4 years of English-medium schools. Very few of the 57 adult Hungarian-speaking immigrants in DeKeyser’s (2000) study scored within the range of child immigrants on a grammaticality judgment task, and the few who did had high levels of analytical skills (suggesting that they probably used their metalinguistic knowledge). 4.6  You don’t learn L2 the way you acquired L1, do you? How come? Whatever one’s opinion about the existence and the nature of an optimal period for language acquisition, one thing is clear and does not seem controversial: Adult L2 learners do not appropriate their L2 in the same way as they acquired their L1. Everybody admits that it is hard work. In addition to transferring structures from L1, learners of a second language have a very hard time automatizing their L2. Surely they would if they could. Automatized language is so much more efficient.



Chapter 4.  Ultimate attainment in L2 proficiency 

Automatic processing always takes over when it is available: it is faster, effortless, allows the speaker to focus attention elsewhere and tolerates a good deal of noise. So, if the acquisition system were still at learners’ disposal, no doubt they would avail themselves of it. Anyone who immerses himself or herself in a second language environment for months and years manages to learn the language, and possibly, after long practice with controlled speech, manages to acquire a good portion of the language, but does not acquire it directly, from scratch, the way children before four or five years of age do, and rarely all components of language structure. (A tall, blond, blue-eyed colleague, who specializes in child language, used to say how frustrated she was when, after years of Dutch immersion in the Netherlands, and in spite of her high motivation to pass for a native, salespeople in Amsterdam would invariably answer her Dutch queries in English.) As suggested by Seidenberg and Zevin (2006), “computational and biological accounts play complementary roles in understanding at least some major cognitive phenomena” (p. 608), but with respect to first and second language appropriation, the biological account of the roles of procedural and declarative memory cannot be dodged. The computational explanation might clarify how L1 competence interferes with L2 acquisition, but must also account for why a first language cannot be fully acquired after age 6, as shown not only by the few cases of hearing children deprived of language input but also by the numerous deaf children not exposed to sign language early enough (Mayberry, 2006). The onset of language acquisition in early human development dramatically alters the capacity to learn language throughout life (Mayberry & Lock, 2003). I am not speaking of the form of the utterance (foreign accent, interference from L1 and other deviances in morphosyntax and lexical semantics) but of the system used to perform both comprehension and production. Speaking with a foreign accent is not a sign of a lack of automaticity. A deviant phonological and articulatory system could be automatized. But in addition to the contents of the grammar, what makes the appropriation and use of L2 different from L1 is the lack of automaticity and consequent reliance on conscious (albeit possibly considerably speeded up) control of one or more of the components of grammar. The greater the number of components that necessitate control, the slower and less systematic the performance. The most perceptible difference between the grammar of a speaker of L2 and that of a native speaker is the deviance in contents (accent, grammatical errors, inappropriate semantic boundaries of lexical items). Less observable, in very proficient late second language speakers, is the greater reliance on metalinguistic knowledge and control, which results in reduced speed and increased variability (not readily perceived in conversational situations).

 Declarative and procedural determinants of second languages

The argument that the bilingual’s performance should not be compared to that of unilingual native speakers (Grosjean, 1989, 2008; Cook, 1992; Piller, 2002) is applicable to the contents of language representations (i.e., uni- or bidirectional interference in what is stored as implicit competence), not to the means by which language is represented (i.e., procedurally or declaratively; automatic or controlled).

5.  Optimal period and the right hemisphere The optimal period is hypothesized to apply to implicit linguistic competence (Paradis, 2004). Implicit linguistic competence is subserved by cortical and subcortical structures of the left hemisphere and areas of the right cerebellum. Unless the various components of verbal communication are distinguished (implicit linguistic competence, metalinguistic knowledge and pragmatics), claims about “language” will necessarily be muddled, as their truth or falsity depends on which component they refer to. Discussions of the critical period hypothesis and the role of the right hemisphere are no exception. Lenneberg (1967) associated the critical period with maturation, as reflected in language lateralization.7 His proposed laterality shift from the right to the left hemisphere was soon shown to be incorrect (Krashen, 1973), and it applies to none of the three main components of verbal communication. Barring early cerebral injury, implicit linguistic competence is sustained by procedural memory in the left hemisphere from the start (i.e., between 1;6 and 2 years of age, when the first two-word constructions appear, before which there was no grammar), irrespective of modality (signed or spoken). At the earliest stages of verbal communication, pragmatics becomes associated with speech sounds; and as linguistic pragmatics develops, it continues to be sustained by right-hemisphere structures. Language awareness, sustained by declarative memory, does not undergo progressive lateralization either. Two conditions will result in the absence of development of implicit linguistic competence: (1) a deviant FOXP2 gene8 or (2) the absence of language

7.  “The limiting factors postulated are cerebral immaturity at the one end and termination of a state of organizational plasticity linked with laterality of [language] function at the other end of the critical period” (Lenneberg, 1967: 176). 8.  Leading (among other things) to genetic dysphasia in which many morphological and phonological aspects of implicit linguistic competence are compromised and are made up for by the use of explicitly learned metalinguistic knowledge.



Chapter 4.  Ultimate attainment in L2 proficiency 

interaction at the age during which implicit linguistic competence normally develops.9 As we saw in Chapter 3, two factors may conspire to make acquiring implicit competence in a second language in adulthood difficult: (1) age, leading to the use of declarative memory when learning anything new, and (2) the fact that implicit competence has already been established along the parameters of L1. The evidence (both the end-age of language impairment subsequent to righthemisphere lesion and the notion of gradual language lateralization itself) that was originally the rationale for setting the end point of a critical period for language acquisition at 12 years of age has been shown to be invalid. It is clear that there is no critical period that ends at puberty. If there is a critical period, it terminates much earlier (and it has nothing to do with lateralization). Yet, many authors continue to use Lenneberg’s (1967) definition, especially when arguing against the existence of a critical period (Marinova-Todd et al., 2000; Komarova & Nowak, 2001; Flege et al., 2006); and hence, they include individuals aged from 6 to 12 years (sometimes even up to age 14, e.g., Flege et al., 2006) in the “early L2 acquisition onset group” in contrast to individuals older than 12 in the “late L2 onset group.” Both groups contain individuals who have passed the incidental acquisition age.10 As a result, children who were 6 when they arrived in their new country and still had detectable accents after 3 or 5 years of residence in an English environment are considered to provide evidence inconsistent with the critical period hypothesis (Flege et al., 2006). Two reasons may jointly account for their accents: (1) they were exposed to the accented L2 of their parents, relatives and friends, and (2) they were exposed to the L2 after the age of 6 years. The second reason may be in effect even in the absence of the first: Munro and Mann (2005) report that a foreign accent is perceived in speakers who started immersion in L2 from about age 5 on, after which the degree of perceived accent increases with age at onset of L2 exposure. In a study by Flege, Yeni-Komshian, and Liu (1999), the foreign accents of the participants grew stronger as age of immersion in an L2 environment (e.g., immigration) increased. While grammatical scores also decreased steadily, unlike 9.  A normal FOXP2 gene and reliance on procedural memory are necessary but, in order to acquire a language (first or second), sufficient interaction opportunities are also required. 10.  In a study conducted by Palij (1990), early bilinguals (L2 acquired before 6) did not differ from native speakers on any of the measures, but both differed consistently from groups who had appropriated L2 after the age of 6. The patterns of differences among the groups are clear and striking. Native and early bilinguals differ significantly from late learners on all tests ( p < .0001). Palij concludes that these differences are important and should be considered when selecting subjects for language experiments.

 Declarative and procedural determinants of second languages

accent, they were influenced by other variables, such as extent of education received in the L2 environment. This supports the notion that pronunciation of a second language is not only subject to an earlier onset deadline but also more difficult to control than other aspects of language structure, such as syntax. However, grammatical scores too show increased reliance on declarative memory (as evidenced by the influence of education) with increasing age of first exposure.

6.  Evidence adduced against a critical period Some adult second language learners are reported to be able to attain native proficiency in some aspects of language or on some language tasks. Some learners whose exposure to L2 occurs after age 12 are still able to acquire an L2 accent that is perceived as native by native speakers. Neufeld’s work is often cited in support of native-like attainment in the pronunciation of a foreign language. Yet, as the author acknowledges, not only is it the outcome of a “highly artificial learning situation” (Neufeld, 1977: 48), but the learners’ performance does not correspond to what can be considered as speaking a second language without a detectable foreign accent. Rather, it is an exercise in psittacism, a task some species of parrots and myna birds are able to perform. The participants were trained to repeat one-to-eight-syllable stock phrases. They were specifically told not to expect to learn their meaning or grammatical rules. All this notwithstanding, out of 20 rated participants, the production of only 3 for Japanese and 1 for Chinese was judged to be native-like (in spite of the fact that the judges might have expected a majority of native samples, having been told beforehand that, “although improbable,” many samples they were to hear, and conceivably all, might be non-native (p. 53)). The adults in Neufeld’s (1980) experiments did not acquire phonology (as claimed in the paper’s title). All they acquired was the ability to reproduce specific strings of sounds (corresponding to Japanese sentences to the extent that native speakers thought they were spoken by Japanese). An acquired phonology would entail the ability to use phonological rules (not just to imitate sounds) in extemporaneous sentence production, a task the subjects were absolutely unable to do, since they could not speak Japanese. They would not have been able to use in novel contexts the phonological rules involved in the passages they could imitate (albeit perfectly). Pronunciation is a skill that most L2 learners find difficult to integrate with the simultaneous selection of morphosyntactic rules and lexical items (Lamendella, 1979). The fact that the amount of phonological training has a significant positive effect on the pronunciation of a group of university students learning an L2 shows



Chapter 4.  Ultimate attainment in L2 proficiency 

that adults are able to learn to control their production of L2 speech sounds. This, however, needs to be integrated with the other components of grammar into implicit linguistic competence and automatized. Note that the conditions of Neufeld’s (1979) study did not replicate the learning situation of young children (as claimed by Marinova-Todd et al., 2000). The students were not involved in communication with native speakers but received specific phonological training. “Subjects’ performance [in the 1977–1979 studies] involved imitation only, with no creative use of language” (Neufeld, 1979: 234). Neufeld admits that the studies do not provide “sufficient evidence to definitely reject the ‘strong version’ of the critical period hypothesis” (p. 235). Some studies claim that with short, intensive training, it is possible to acquire a second language’s phonological contrasts. But as Sebastián-Gallés and Bosch (2001) point out, in spite of a 10-  to 20-percent increase in performance on identification or discrimination tasks, performance does not reach the native speakers’ level. The improvement likely reflects controlled performance, in which case it would be additional evidence in favor of a biologically based critical (i.e., optimal) period. The case studies reviewed by Nikolov and Mihaljevic Djigunovic (2006) document that all the post-puberty learners who were frequently mistaken for native speakers definitely strove for unaccented proficiency and worked actively to master their new language (Bongaerts et al., 1997; Ioup et al., 1994). Individuals who have been found to successfully attain ultimate native-like proficiency are reported to have been highly motivated to pass for L2 native speakers (Moyer, 1999) and to have worked on their language development consciously (Nikolov, 2000; Moyer, 2004). The many pieces of evidence from a wide array of different domains pointing to increased explicit processing in the second language are too numerous to be ignored. They can no longer be swept under the carpet. Any theory of second language appropriation must be able to account for them.

 actors invoked in lieu of a neurobiological critical period 7.  F to account for poor performance in L2 are actually the consequences of an optimal period The various phenomena proposed to explain the differences between first and second language ultimate attainment play a role only because, for a number of genetically programmed cerebral events, procedural memory, which allows language to be acquired, becomes far less available after an optimal period. The consequent reliance on declarative memory renders the appropriation of language

 Declarative and procedural determinants of second languages

contingent upon the various factors suggested by authors as being responsible (instead of a neural-based reason) for arduous and eventually poor attainment. 7.1  Effects due to age are a consequence of brain processes The observed age-related phenomena probably result from the interaction of multiple causes (Singleton, 1989). “Factors other than a biologically determined C[ritical] P[eriod] play a role in the variability of the ultimate attainment of older learners” (Marinova-Todd et al., 2001: 174). These alternatives to a neurological account – motivation, explicit language instruction, the very knowledge of another language (Singleton, 2001) – become relevant only because of the biologically determined advent of declarative memory on which late learners must rely to compensate for their lost ability to incidentally acquire L2 as young children do. Variability in learning is caused by factors intrinsic to declarative memory (working memory functions, IQ, focused attention, executive control, etc.). Adults may eventually achieve near-native (or even native-like) proficiency, though not necessarily full implicit linguistic competence the way early bilinguals do. The difference is not only, or necessarily, one of content (deviant items incorporated in the grammar at any level) but of lack of automatic use of all aspects of language. L2 attainment continues to negatively correlate with age of learning (Birdsong & Molis, 2001). At the end of early childhood, learners no longer rely almost exclusively on procedural memory for incidental language acquisition and start learning a second language explicitly, relying on declarative memory. If the age of onset of learning is further postponed until middle-age, declarative memory, on which learning relies, gradually declines. Hence, as one gets older, not only has reliance on incidental acquisition long ceased, but explicit L2 learning becomes progressively more difficult (as does learning in any domain). The aging process is thus doubly responsible for lack of success (or the increased difficulty of attaining it), (1) because of the end of the period when language can be acquired easily, and (2) because of the decline with age of the means by which learners can compensate (i.e., the decay of the hippocampal-system-dependent declarative memory). Both processes are determined by brain maturation, physiology, and genetically built-in obsolescence. As with any genetically programmed process, variability owing to differing experiential conditions is possible, within limits. Marinova-Todd et al. (2000) rightly point out that “myriad factors are involved in successful learning,” but then add: “many of which may be correlated with age but have nothing to do with changes in the brain” (p. 24). First, let us remember that these numerous factors are not involved in successful L1 and early L2 acquisition (except for the opportunity to interact with speakers of the language). The



Chapter 4.  Ultimate attainment in L2 proficiency 

reason why they become relevant is because, at a certain age (and to this extent, age is a factor, and age effects have to do with changes in the brain), declarative memory becomes available (and this represents a change in the physiological properties of the brain) and individuals tend to rely increasingly on conscious learning. At the same time, incidental learning ceases to be efficient. To that extent, one can say that there is a period (from about age 2 to about age 5) during which acquisition relies on one cerebral entity (procedural memory); after that, acquisition becomes less efficient, at successive periods for the various components of implicit linguistic competence, until about adolescence. Meanwhile, the individual compensates by consciously learning and controlling the use of those aspects of L2 that are no longer acquired incidentally. From then on, the factors involved in learning come into play and those involved in general cognitive capacities (working memory, IQ, etc.) become relevant, resulting in considerable variability in rate and degree of success. Among the factors that typically lead to native-like proficiency in L2, aptitude, meaning the ability to learn explicitly, becomes one of the major variables. The fact that cognitive aptitude strongly correlates with success of L2 learning (Ehrman & Oxford, 1995) again suggests that high attainment in L2 is the result of learning rather than acquisition. All these factors are associated with learning performance in any knowledge domain subserved by declarative memory. The brain is responsible for the aging process and its consequences, the availability and decline of procedural memory for the acquisition of implicit linguistic competence, and the availability and decline of declarative memory (modulated by its inherent constraints on learning: working memory capacity, aptitude, attitude, motivation, etc., which vary across individuals) for learning foreign languages. We might therefore agree with Marinova-Todd et al. (2000) that, literally, “age does not influence language learning” (p. 28), at least until declarative memory declines with advanced age, but it does considerably influence language acquisition. As in every domain involving genetic makeup, brain maturation and concomitant cognitive development, and experience, we cannot make absolute claims about age onset for a specific phenomenon. Many factors do interact, but within limits, and outcomes fall within a certain range. One may then consider any statement about chronological age as referring to a norm (i.e., the vast majority, with some standard deviation in either direction). Such limits apply to the availability of procedural memory for acquiring a language, whether first or second. It is true that a number of age-related factors are at work (Singleton, 2001). These age effects that are assumed not to rely on neurolinguistic arguments are in fact caused by neurophysiological phenomena such as the diverging development of procedural and declarative memory that sustain language functions:  (1) The availability of procedural memory for language acquisition gradually declines

 Declarative and procedural determinants of second languages

from around age 5; (2) from then on, factors associated with declarative memory gradually enter into play; (3) through adulthood and old age, the gradual deterioration of declarative memory negatively affects second language learning and use. The uncontroversial difference between appropriating a native language (or two) and further languages after the age of about 5 years is determined by the timeframe of the availability of procedural memory for implicit linguistic competence and the advent of declarative memory. It may be modulated by a number of factors,11 such as motivation (Schumann, 1998), attention, effort, aptitude,12 education level (Urponen, 2004), working memory capacity (McDonald, 2006), verbal analytical ability (Harley & Hart, 1997; DeKeyser, 2000), and environmental dynamics (Flege et al., 1999). Ironically, these factors become relevant to the appropriation13 of L2 (though not the acquisition of L1) precisely because of reliance on different memory systems when acquiring L1 (procedural memory for implicit linguistic competence) and learning a subsequent language (declarative memory for all aspects of language, including metalinguistic knowledge and some features of pragmatics). The deadline for the incidental acquisition of implicit linguistic competence varies with each component module. There is not one optimal period14 but several, respectively for prosody, phonology, inflectional morphology, and syntax (Weber-Fox & Neville, 1996, 2001), in this order of early termination. This only partly coincides with the order of acquisition, some aspects of word order being generally acquired before inflectional morphology, but more complex aspects of syntax after some features of agreement (Tavano, Fabritiis, & Fabbro, 2005). McDonald (2006) proposes “inadequate processing speed” (p. 381) as an explanation for the poor grammaticality judgments of late second language learners, as opposed to their being beyond the critical period for language acquisition. Slow 11.  These factors, which are generally invoked as being responsible for the difference between acquiring L1 and L2 in arguments against a critical period, are not relevant when acquiring one or more languages before the age of 5 years. They only come into play when L2 has to be learned. 12.  All children without severe mental defects have the aptitude to acquire, and do acquire, the languages to which they are exposed and in which they interact. 13.  The word “appropriation” is used here throughout because (1) whatever is said is valid for both acquisition and learning, and (2) it is difficult to ascertain by mere behavioral criteria whether what has been appropriated has been acquired or learned, and if acquired, whether it was acquired incidentally from the start or subsequent to a long period of explicit processing from which competence was gradually abstracted – as discussed in Chapter 3. 14.  Meaning that the period during which Z can be acquired does not start before X months and ends at Y years of age.



Chapter 4.  Ultimate attainment in L2 proficiency 

processing suggests that they use controlled metalinguistic knowledge instead of automatic implicit competence. The use of automatic processing is preferred whenever it is available. If late second language learners use controlled processing, it means that automatic competence is not available. The inadequate processing speed is a result of a period after which procedural memory for language acquisition is no longer efficient. Thus again, the explanation proposed in lieu of a critical period to account for late learners’ failure to appropriate a second language the way children do happens to be the necessary use of compensatory strategies as a result of an optimal period for language acquisition. 7.2  Native language entrenchment MacWhinney (2005) emphasizes the extent to which a second language speaker’s repeated use of L1 leads to its ongoing entrenchment. As he notes, this entrenchment operates differentially across linguistic areas, with the strongest effect occurring in output phonology (i.e., the most implicit of language processes) and the least in the area of vocabulary (i.e., the explicit component of language), for which new learning continues to occur lifelong. Seidenberg and Zevin (2006) note that “acquiring language early in life seems patently easier than learning it later” (p. 595). They propose that interference due to increasing entrenchment of L1 provides a basis for the decline in plasticity associated with the closing of the critical period for language acquisition. According to these authors, the loss of plasticity associated with this phenomenon seems to be specifically related to the capacity to generalize. However, no correlation with any specific cognitive handicap has been found in individuals with genetic dysphasia. Although these individuals are unable to acquire the simplest regular rules of inflectional morphology, such as marking the plural on regular nouns and forming the past tense of regular verbs in English (Ullman & Gopnik, 1994), forming regular compounds in Greek (Dalalakis, 1999), or rendaku in Japanese (Fukuda & Fukuda, 1999), they do not necessarily show such deficits in cognitive domains other than language. It thus appears that generalizations in language are independent of a non-domain-specific capacity to generalize (consonant with the taskspecificity of procedural memory). Seidenberg and Zevin (2006) remark that “PDP networks have not, as yet, incorporated facts about neurological development” (p. 597). One piece of evidence that the rapid acquisition of language with gradual loss of capacity to acquire other languages is “tied up to biological development on a maturational timetable” (p. 606) comes precisely from the coincidental emergence of declarative memory. The concept of entrenchment, which was the result of considering the phenomena computationally rather than biologically, may account for what

 Declarative and procedural determinants of second languages

and how Parallel Distributed Processing computers learn, but when it comes to humans learning language, biological facts cannot be excluded from the equation. The involvement of cerebellum and basal ganglia versus hippocampal and mesial temporal lobe structures in subserving, respectively, procedural and declarative systems cannot be ignored. A small system of artificial grammar rules may be syntactically instantiated by the adult speaker in a way that strongly resembles native-like sentence processing (Friederici et al., 2002), but L2 processing requires the automatic (i.e., native-like) processing of several independent systems, including a large set of complex rules. Vocabulary and basic word order may be acquired whereas relatively more complex structures may not be so readily acquired (see Yokoyama et al.’s (2006) and Suh et al.’s (2007) studies, discussed in the next chapter). If Seidenberg and Zevin’s computational account turned out to be the sole explanation of why the acquisition of a second language is so difficult, it would be a justification for why the use of declarative memory becomes necessary and would thus support the procedural/declarative sequence, showing why a compensatory means (i.e., conscious declarative learning) is needed (i.e., when implicit linguistic memory is entrenched). Irrespective of the applicability of their model, the following effects on language appropriation and loss (p. 602) obtain: Maintenance of L1 interferes with appropriating L2; the continued experience with L1 keeps the language entrenched; proactive interference from L1 affects appropriation of L2; and retroactive interference from L2 causes attrition of L1 if L1 ceases to be used. These effects obtain for both incidental acquisition and conscious learning. In the case of acquisition, it is difficult to modify existing automatized procedures; in the case of learning, one consciously applies the explicit rules inferred from one’s own and other L1 speakers’ output. In the case of L1 entrenchment and attrition, these effects are compatible with the activation threshold hypothesis (Paradis, 2007a). The result is that L2 is learned rather than acquired (at least at first, and its use remains controlled for a long time) before components of the grammar (with or without inappropriate transfer of features from L1) are eventually internalized.

8.  Conclusion The optimal period refers to the time during which a second language can be acquired incidentally as implicit linguistic competence that will be used automatically. After that window of opportunity, learners rely on declarative memory, which leads to the various findings listed by Harley and Wang (1997) as needing



Chapter 4.  Ultimate attainment in L2 proficiency 

an adequate account: (1) the graduated age of onset differences between children and older learners; (2) the differences between adult onset ages; (3) the contrasting findings concerning initial rate advantages for older learners; (4) the ability of some adult learners, but not others, to achieve native or near-native levels of success; and (5) the variance found among learners in a bilingual setting. The numerous factors generally invoked to account for differences in attainment between the native and a later-learned second language arise from the L2 learners’ switching from reliance on procedural memory – as young children do – to reliance on declarative memory functions. All children without a serious mental deficiency who are exposed to language acquire a native language. Not everyone who has acquired an L1 manages to acquire an L2. The immediate reasons are numerous and varied but all stem from genetically programmed neurobiological events and none plays a significant role in the acquisition of a language during the optimal period. The factors that affect achievements in L2 appropriation are common to all declarative tasks and irrelevant to automatic achievements. One major consequence of relying on declarative memory is a considerable degree of interindividual variability brought about by differences in working memory capacity, education, attitude, and several other internal and external factors. The proactive negative influence of L1 (entrenchment) may be one of the reasons why the appropriation of a second language is difficult and depends on recourse to declarative memory. This would speak to cerebral plasticity (i.e., capacity and resource management). The advent of, and gradual increased reliance on, declarative memory, with a concomitant decreased reliance on procedural memory, would explain why only some types of knowledge (e.g., syntax versus words) show optimal-period-like loss of plasticity. Skills in general (and implicit linguistic competence processing in particular) acquired during their optimal period are more resistant to attrition through disuse than learned material, but the acquisition of skills after their optimal period becomes more difficult with increasing age. Neither puberty nor language lateralization marks the deadline for an optimal period for second language acquisition. Yet, much evidence adduced against an optimal period is predicated on Lenneberg’s (1967) premises. On the one hand, participants in most experiments are late versus later learners rather than early versus late, and tasks tap only a single aspect of language or even a non-linguistic type of performance. In addition, irrespective of the subject population, whether the participants’ performance is speeded-up or automatic is never ascertained. Some authors emphasize overall deficiencies in ultimate attainment, others focus on cases of high achievement on several tasks – but whether one considers

 Declarative and procedural determinants of second languages

the late second language implicit competence to be half-full or half-empty, the implication is that it is not full. This is not to deny the possibility of reaching (quasi) native-like execution via speeded-up controlled processing. As was mentioned in the previous chapter, this may not be of any practical import, but it is nevertheless essential from a neuroscience perspective. The ultimate observable output may be the same, even though it is achieved by different neurofunctional means.

chapter 5

The pervasive relevance of the distinction between implicit competence and explicit knowledge We have considered the impact of the declarative/procedural model on consciousness research in Chapter 2, on second language teaching in Chapter 3, and on ultimate attainment in a second language in Chapter 4. In this chapter, the crucial and ubiquitous contribution of declarative and procedural memory to the acquisition, processing, and loss of languages will be examined. It will be shown that one important methodological consequence for laterality, neuroimaging and any types of experimental studies that purport to determine the representation and processing of language is that single words cannot be used as stimuli. It will also be argued that a failure to specify whether a theoretical construct is assumed to apply to implicit or explicit processes is a major source of confusion when it comes to ascertaining what claims are actually being made and, consequently, which cerebral structures should be involved.

1.  I mplications of the declarative/procedural distinction for laterality studies The reasons why the results of experimental bilingual laterality studies and their meta-analyses cannot be considered valid are multiple, plain and simple: (1) the results are contradictory; consequently, (2) successive meta-analyses yield different findings, depending on which studies they choose to include; (3) the results contradict those of the Wada test, electrical stimulation of the brain and aphasia data; and most importantly, for the present discussion, (4) no rationale has been presented that would allow one to generalize results from single words to the language system. Let us quickly recap each point. (1) Laterality differences have been reported for very specific subgroups of bilinguals and/or under very specific conditions, such as only for early or for late bilinguals, or for early or late bilingual women but only late bilingual men. Decreased asymmetry has been claimed to hold (exclusively) for just about every

 Declarative and procedural determinants of second languages

possible subgroup of bilinguals and its opposite: for instance, differences only in proficient late bilinguals (as opposed to beginners) who have learned their second language in a formal setting; or conversely, no difference in such a group (nor in beginners) but only in those at the beginning of the informal acquisition of a second language. In addition to the contradictory results among studies that found some difference between bilinguals and unilinguals, about an equal number of studies found no difference. Equal involvement in each language, more or, alternatively, less involvement of the right hemisphere have all been claimed for L1 only, for L2 only, or only for early or late stages of informal acquisition or of formal learning. Evidence contradicting the stage hypothesis abounds (see Vaid, 1983). There is also evidence inconsistent with the manner of acquisition, the stage of acquisition and the age hypotheses (see Paradis, 1990). (2) Previous suggestions from literature reviews and meta-analyses abound: (a) Vaid and Genesee (1980) conclude that right hemisphere involvement (i.e., less asymmetry) will be more likely the later the second language is learned relative to the first; left hemisphere involvement (i.e., more asymmetry) is more likely the earlier the second language is learned. (b) Vaid’s (1983) investigation indicates that right hemisphere participation is more likely the more informal the exposure to the second language is; left hemisphere involvement in second language processing is more likely the more advanced the stage of second language acquisition. (c) Vaid and Hall (1991) find no significant support for greater right hemisphere involvement in the second language, no increased left hemisphere involvement with increased proficiency, and no reliable difference in informal L2 acquisition, but greater right hemisphere participation in early bilinguals. Hull and Vaid (2006) deduce that early bilinguals are bilaterally organized whereas late bilinguals, like unilinguals, are left-hemisphere dominant for language. (3) The methodology that consists in measuring degrees of laterality of cerebral representation runs counter to the rationale and arguments provided in the very studies that are touted as allegedly supporting the paradigm, namely the Wada test. Hull and Vaid (2007), for instance claim that evidence for the reliability and validity of laterality measures can be marshaled and they cite Segalowitz (1986); Segalowitz and Bryden (1983); and Kosaka et al. (1993) in support. These studies clearly indicate that, for a number of reasons,1 these experimental paradigms can only be used to determine left- or right-dominance, not degrees of laterality among individuals with left dominance for speech.

1.  Note that unless the paradigm has been shown to be valid, all methodological considerations (such as subject selection, language type, word length, frequency, etc.) usually adduced to be responsible for the reported multiple contradictions are moot.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

Kosaka et al. (1993) make it clear that the dual task paradigm cannot be used to distinguish degrees of dominance among left-dominant patients. Segalowitz (1986) assesses only measures of gross dominance (i.e., whether individuals are left- or right-dominant) and therefore his assessment does not concern differences in degree of lateralization. In fact, he considers that diagnostic use of non-invasive laterality measures to determine hemispheric specialization (let alone degree of lateralization) has always been a chancy affair and that any measure of lateralization from such tasks reflects more than just brain organization. Segalowitz and Bryden (1983) propose a rationale that is the opposite of what Hull and Vaid (2007) imply: Not only do they not support the validity of a measure of degree of hemispheric dominance, but precisely because of the considerable variance associated with experimental laterality measures, they argue that one ought to consider relative measures of laterality only as indicative of left-dominance for speech (i.e., ignoring differences in degree of dominance), and they conclude “Only lesion studies can examine absolute representation of functions” (p. 365). For the record, studies using Wada tests on bilingual individuals report left-hemisphere representation for both languages (Rapport, Tan, & Whitaker, 1983; Black & Ronner, 1987; Berthier et al., 1990; Gomez-Tortosa et al., 1995; Kho et al., 2007), even when one of the languages is American Sign Language (Damasio et al.,1986). Studies involving electrical stimulation of the brain likewise consistently report both languages of bilinguals to be lateralized in the left hemisphere (Ojemann & Whitaker, 1978; Rapport et al., 1983; Ojemann, Ojemann, & Lettich, 2002; Roux & Trémoulet, 2002; Lucas et al., 2004; Lubrano, Roux, & Démonet, 2004; Walker et al., 2004). April and Han (1980), who report a case of crossed aphasia in a right-handed bilingual Chinese man, nevertheless stipulate that a preliminary sampling in two separate Chinese populations has failed to demonstrate a greater incidence of crossed aphasia. Ijalba, Obler, and Chengappa (2004) conclude that no convincing, methodologically sound study has supported an increased incidence of crossed aphasia in bilinguals. (See also Karanth & Rangamani, 1988; Rangamani, 1989; Solin, 1989.) (4) Not only can results obtained with single-word stimuli not be generalized to the representation and processing of language (as one generally cannot generalize from a part to the whole), but experiments that use such stimuli address a component that differs radically from the language system. What makes the native language special is that it is acquired incidentally, stored implicitly, and processed automatically in the course of using it. It is subserved by procedural memory circuits dedicated to phonology, morphology, morphosyntax, and the semantic and grammatical constraints on the lexicon. Single-word stimuli (i.e., words in isolation) are explicitly known form-meaning associations and as such are subserved by declarative memory. These are two distinct memory systems involving distinct

 Declarative and procedural determinants of second languages

neurofunctional mechanisms, sustained by distinct neuroanatomical structures (i.e., the perisylvian cortex, portions of the right cerebellum, left neostriatum and other basal ganglia for the normal use of language, as opposed to the hippocampal system, medial temporal and anterior cingulate cortex for words in isolation). In addition, the normal use of language involves cortical areas of the right hemisphere to process the pragmatic aspects of utterances, a process which becomes irrelevant in processing single words, as they are deprived of a context. Thus, even if the laterality paradigms were shown to be valid, that is, if it were demonstrated that they actually measured degrees of laterality of cerebral representation and processing of stimuli (which they do not), studies using single words as stimuli could only speak to the representation of words (i.e., to their explicitly known form-meaning association subserved by declarative memory), not to the representation of language (whose cerebral representation is subserved by a type of memory different from that subserving single words, namely procedural memory). In their response to Paradis’ (2008d) comments, Hull and Vaid (2008) do not address this issue. They discuss the relative merits of meta-analyses over individual studies but (1) they do not show that the paradigm has been validated, and (2) they do not explain how a meta-analysis of studies that used a non-validated instrument could be considered valid – hence how a measured degree of difference obtained by a non-validated instrument with single-word stimuli corresponds to a degree of language lateralization. They state that what their study does is to identify the sources of variance – but the sources of variance are moot, as are all well-controlled methodological factors, if the measuring instrument has not been shown to actually measure what it purports to measure. Hull and Vaid (2008) conclude that the differences in degree of lateralization based on lateralization measures “may well be” valid indicators of differences in the degree of underlying hemispheric asymmetry. On the other hand, they may very well be indicators of something else, as suggested in the Segalowitz (1986) paper cited by Hull and Vaid (2007) in support of the reliability and validity of laterality measures. There is a high error of measurement in most lateralization tasks; the tests measure aspects of brain organization that reflect processes beyond the use of the hemispheric specialization of interest (p. 203) – and yet, the strength of the conclusion concerning brain lateralization depends on the validity of the measure (p. 192).

2.  I mplications of the declarative/procedural distinction for imaging studies The same reasoning that applies to laterality studies applies to neuroimaging studies. Yet again, the trend reported in Paradis (2004) continues: well over two-thirds of



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

the studies published since (18/27)2 use single words as stimuli (though a very recent trend seems at last to favor sentences). Single-word studies may be easier to control experimentally, but they cannot address the issue of the representation of languages in the bilingual brain. Fortunately, some authors (though far from all) are beginning to stipulate that their findings speak to the “single word level” of organization of languages (Klein, 2003:  424) or the “level of word meaning” (Crinion et al., 2006) when they use single words as stimuli. Similarly, Hernandez and Meschyan (2006) and Meschyan and Hernandez (2006) appropriately specify that their study investigates how language proficiency and orthographic transparency modulate neural activity during bilingual single-word reading and do not extrapolate their results to reading in general. Halsband (2006) notes that in his study (purported to address the representation of language in the brain of bilinguals) only “the retrieval of word pairs without syntax was investigated” (p. 365). Klein et al. (2006b), though their abstract purports to examine first and second language processing, conclude that their results support the idea that, at the lexical level, the neural substrates for L1 and L2 are shared. But we should not forget that the processing of single words relies on an organization that differs from that of the rest of language. Roux and Lubrano (2006) call attention to the limits of neuroimaging studies that use single words as stimuli: “The use of single word tasks should not find differences between unilinguals and bilinguals (or between bilinguals), irrespective of fluency in both languages” (p. 118). 2.1  Words of caution about the interpretation of neuroimaging studies A long list of reasons for caution when using neuroimaging data as evidence about language representation and processing was presented in Paradis (2004). In a recent special issue of Brain and Language devoted to neuroimaging (see editorial by Sidtis, 2007a), a number of additional problems are considered: Sidtis (2007b) argues that the contrast approach is flawed. Drake and Iadecola (2007) and Baslow and Guilfoyle (2007) provide a reminder that blood flow and other derived hemodynamic responses are not direct measures of neuronal activity. Hansen (2007) demonstrates that imaging results are model-dependent. Chen and Small (2007)

2.  Single words:  De Bleser et al. (2003); Briellmann et al. (2004); Pillai et al. (2004); Rönnberg et al, (2004); Xue et al. (2004a,b); Emmorey et al. (2005); Rodriguez-Fornells et al. (2005); Tatsuno & Sakai (2005); Tham et al. (2005); Crinion et al. (2006); Halsband (2006); Hernandez & Meschyan (2006) Klein et al. (2006a,b); Meschyan & Hernandez (2006); Marian et al. (2007); Kovelman et al. (2008b). Sentences:  Rüschemeyer et al. (2005); Golestani et al. (2006); Scherer et al. (2006); Yokoyama et al. (2006); Gandour et al. (2007); Jeong et al. (2007a,b); Suh et al. (2007); Kovelman, Baker, & Petitto (2008a).

 Declarative and procedural determinants of second languages

point out that the magnitude of the typical BOLD response is not very different from the baseline noise level of an fMRI time series and report not only differences in reproducibility between groups but also differences in group reproducibility effects under different task conditions. When a task is accomplished covertly (e.g., mentally name a picture; generate a word, describe past events – without oral production), it is difficult to ascertain whether the subjects actually performed the task as prescribed. Hasegawa, Carpenter, and Just (2002) had remarked that the analysis of cortical processing of languages is limited by the spatial resolution of the technology: When the location of the activation is similar in both languages, it is possible that the networks that are reflected in the activation are spatially adjacent within a 52-mm3 voxel in which fMRI data are acquired. The situation is no better with PET. The results of ERP studies are also rather inconsistent (Stowe & Sabourin, 2005). When “regions of interest” (ROI) are selected for examination in a neuroimaging experiment, one runs the risk of not detecting the areas outside the ROI (i.e., outside the regions that subserve known automatized language functions) that are recruited in order to overcome gaps in implicit competence, such as the anterior cingulate cortex (executive control), the hippocampal system (declarative knowledge) and areas of the contralateral hemisphere (pragmatic inference). Yet, these are precisely the areas that may reflect the involvement of other cognitive functions (e.g., attention, memory) associated with the language task that would indicate the extent to which speakers use mechanisms other than implicit linguistic competence to communicate verbally. This would allow one to identify subgroups such as early versus late appropriation onset bilinguals. The artificiality of some tasks precludes the possibility of deriving information about normal language processing. In an fMRI study using word generation (given a letter of the alphabet) and sentence generation (given two prepositions or conjunctions), Mahendra et al. (2003) report that early and late bilingual subjects with equivalent task performance nevertheless differ in the overall amount of activation associated with language performance. This might possibly reflect automatic versus speeded-up control: an overall difference was found for the extent of activation by early and late bilinguals. However, the generation of sentences given two prepositions would involve cognitive functions other than those required in the natural use of language. Many neuroimaging studies are based on written material, yet the manipulation of writing mobilizes aspects of metalinguistic knowledge that are not mobilized spontaneously by speech (Gombert, 1993). A major problem is that none of these studies have been replicated. They cannot be compared because they use different tasks, different techniques, different paradigms to determine what counts as an activation, different thresholds, etc. It would be easy to neutralize this criticism by replicating a study, using the very



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

same methodology, task, stimuli, type of subjects (along the relevant parameters of age of L2 onset, proficiency, and exposure) and all the other variables adduced as justifications for finding incongruent results. Until we do, we will never find out whether a given technique is reliable and valid. So far, it would appear that many reported activations are task- and technique-dependent rather than indicative of language representation and processing per se. However, while we should remain fully aware of the limitations of such studies, a consideration of the role of implicit and explicit memory may help us interpret the disparate and sometimes contradictory results and discover some patterns. For example, in a recent review of the literature, Abutalebi (2008) reports that the available evidence indicates that a second language seems to be acquired through the same neural structures responsible for native language acquisition. Indeed, both languages rely on the same mechanisms of verbal communication, namely implicit competence (though, at the beginning, to a much lesser extent), metalinguistic knowledge and pragmatics (Paradis, 2004). Abutalebi also reports that neural differences for a second language may be observed, in terms of more extended activity, and that these differences may disappear once a more native-like proficiency is established, reflecting a change in language processing mechanisms:  from controlled processing for a less proficient L2 system to eventually more automatic processing. He concludes that the reviewed neuroimaging data support the notion that language control is a crucial aspect specific to the processing of a weak second language.

 onsequences of not distinguishing word studies 2.2  C from sentence studies In a review of the cerebral processing of a second language, Stowe and Sabourin (2005) draw their conclusions from seven of the over 60 bilingual imaging studies available at the time. This restricted sample contains four studies on single-word processing (Chee et al., 2000; Ding et al., 2003; Perani et al., 2003; Xue et al., 2004b), one on phonological contrasts (Callan et al., 2004), and only two on sentence processing (Dehaene et al, 1997; Hasegawa et al., 2002) – the only studies in their sample that concurrently tap the use of implicit linguistic competence, metalinguistic knowledge, and pragmatic processing. This sample is far from reflecting the overall picture, which is one of multiple contradictions. As reported in Paradis (2004), some studies found a difference between early and late bilinguals, others did not; those that did find differences found different ones. Similarly, some neuroimaging studies report less activation in the classical language areas for the weaker language (Perani et al., 1996; Dehaene et al., 1997; Halsband et al., 2002), while others report greater activation (Yetkin et al., 1996; Nakai et al., 1999; Chee et al., 2001; Hasegawa et al., 2002).

 Declarative and procedural determinants of second languages

Whether they use words, sentences or short narratives as stimuli, most studies find that early bilinguals activate common cortical areas when processing L1 and L2 (e.g., Kim et al., 1997; Chee et al., 1999b, 2000; Hernandez, Martinez, & Kohnert, 2000; Urbanik et al., 2001), even when L1 is not the dominant language (Hernandez et al., 2001), and irrespective of the differences in orthography, phonology and syntax between the two languages (Chee et al., 1999a). Conversely, late bilinguals with moderate fluency in L2 are reported to use different areas (Perani et al., 1996; Dehaene et al., 1997; Kim et al., 1997; Nakai et al., 1999; Urbanik et al., 2001) or at least do not show complete overlap (Simos et al., 2001). However, some have found that late fluent bilinguals activate the same cortical areas for L1 and L2 (Klein et al., 1994, 1995a, b, 1999; Perani et al., 1998; Illes et al., 1999; Pu et al., 2001; Tan et al., 2001, 2003), even if they have only moderate fluency (Hasegawa et al., 2002). Note that most studies that found no difference in late bilinguals used single-word processing tasks. Whether they activate the same or different areas, some studies report higher activation for L1 words or sentences (Perani et al., 1996; Dehaene et al., 1997; Chee et al., 2001), while other studies find higher activation for L2 words or sentences (Yetkin et al., 1996; Nakai et al., 1999; Hernandez et al., 2001; Urbanik et al., 2001; Hasegawa et al., 2002; Luke et al., 2002; Scherer et al. 2006). Some report that activated language areas increase with proficiency (Perani et al., 1998), others that they decrease with proficiency (Yetkin et al., 1996; Hasegawa et al., 2002). Kovelman et al. (2008a,b) report higher activation in both languages of early proficient bilinguals (as compared to unilinguals of each language). It is interesting to note that, whereas the studies by both Chee et al. (2000) and Xue et al. (2004a) (cited by Stowe and Sabourin, 2005) found activations that were identical for Chinese and English, the activated areas differed across studies: left frontal, and temporal areas, as well as fusiform gyrus in the former, left inferior frontal gyrus and two areas in the left parietal lobe in the latter. Do we conclude that the individuals in each study had both their languages in the same location, but that the locations differed across groups? If the difference were simply due to task differences, as the authors surmise, should we not expect some overlap if both tasks reveal the activation of language? Or is each study tapping something other than language? If not, what is it that each is tapping within the language system? At the very least, one cannot generalize about the representation and processing of language on the basis of such tasks with these techniques. Stowe and Sabourin consider Ding et al’s (2003) findings to be similar to those of Xue et al. (2004a), in that for both languages the most significant activation was in the left inferior temporal lobe and fusiform gyri. First, only the fusiform gyrus overlaps with Xue’s findings. Second, differences in the processing of the L1 and L2 tasks are apparent: additional right inferior frontal lobe activation together



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

with less left middle temporal gyrus activation for L2. Yet these three studies (Chee et al., 2000; Xue et al., 2004a, and Ding et al. (2003) are considered to “use relatively similar materials and tasks” (Stowe & Sabourin, 2005: 334). Then, according to the rationale, one would expect relatively similar areas of activation in all three. Yet, this is not the case, and the inconsistency gets worse when one looks beyond the four studies selected for this overview (see Paradis, 2004). Even in the selected studies presented by Stowe and Sabourin, the majority of them (4/7) report distinct/different areas of activation for L2. Stowe and Sabourin (2005) consider additional areas of activation in L2 to be interesting (i.e., relevant) only if L1 does not activate the area at all. Given that the major difference between L1 and L2 processing seems to be a greater reliance (if not complete reliance in the beginning) on explicit processes and pragmatics in L2 when implicit competence is not available, these areas will necessarily also be activated in the processing of L1 since it too relies (albeit to a much lesser extent) on these functional systems. Not a single utterance can be processed without involving pragmatics and speakers generally monitor their output to some extent. The authors acknowledge that studies restricted to word-level meaning examine an aspect of language processing that “provides us with limited information as word level meaning is something that we keep learning across the lifespan, even for an L1” (p. 334). This is precisely because words (as meaning-form pairings) are conscious and as such are sustained by declarative memory in both L1 and L2. It is therefore not surprising that similar processing areas should be activated in both languages. The differences obtained in two of the four reported studies using single-word tasks may reflect the differences in word meaning (at the lexical and, consequently, conceptual level) between the two languages, and/or may reflect the questionable validity of the tasks and imaging techniques employed to determine whether the same language areas are used to process a second language as the first. Note that there is a difference between the same language areas and the same neuronal connections (within a same area). For instance, the proposed neurolinguistic theory of bilingualism (Paradis, 2004) adopted here hypothesizes that, for items in linguistic competence, each language is subserved by micro-anatomically distinct substrates within the same gross anatomical area. However, when L2 elements are not part of implicit competence, they are represented in, and processed by, different mechanisms altogether, and hence activate different cerebral areas.3 Thus, for relatively late L2

3.  Namely, the hippocampal system, parahippocampal gyri, mesial temporal and anterior cingulate cortex, as opposed to the cerebellum, neostriatum and other basal ganglia, and perisylvian cortical areas.

 Declarative and procedural determinants of second languages

learners, the same gross anatomical areas will be used for sentence comprehension and phonological processing in L1 and L2 only to the extent that elements have been incorporated into implicit linguistic competence. Moreover, some pathways involved in processing implicit and explicit materials respectively are so close to each other, though functionally and micro-anatomically separate (Ullman, 2006b), that current neuroimaging techniques fail to distinguish between them (Hasegawa et al.,  2002). Klein et al. (2006b) also purport to examine first-  and second-language processing (Abstract, p. 366) in bilingual individuals who acquired their second language between 4 and 12 years of age and whose proficiency ranges from high to low. The actual task was within- and across-language adaptation (i.e., reduction of brain activity with repeated presentation of the same stimulus) for spoken words (i.e., listening to six identical words in succession or five identical words followed by a different word; in the bilingual condition, the final word would be in the other language), using functional magnetic resonance adaptation. In fact, the goal of the study was “to determine whether overlap exists in the brain regions responsible for processing heard words in L1 and L2” (p. 367). Their findings are that “at the lexical level,” the neural substrates for L1 and L2 in bilinguals are shared, but with some populations of neurons within these shared regions showing language-specific responses – consonant with the findings of recent electrocortical studies (Roux & Trémoulet, 2002; Lucas et al., 2004; Roux et al., 2004; Walker et al., 2004; Bello et al., 2006; Giussani et al., 2007). Halsband (2006) does not exclude the possibility that, within the same cortical areas, distinct neural circuits independently subserve different language sets. As will be discussed in more detail below, the subsystems hypothesis (Paradis, 2004) assumes that this is the case for the implicit linguistic competence subsystems. What has not been internalized, however, is not part of the subsystem and is subserved by extra-linguistic mechanisms on which speakers of L2 rely to a much greater extent than for L1. The common observation of different centers of activation within a single cortical area associated with processing language material in L1 and L2 may reflect the micro-anatomical representation of different circuitry for each language subsystem within the same functional language macro-anatomical area (e.g., Dehaene et al., 1997; Halsband, 2006; Marian et al., 2007).  Halsband (2006) republishes the data from an experiment described in Halsband et al. (2002) in a paper that purports to address the question of the representation of language in the brains of bilingual and multilingual subjects. The reported experiment is, in fact, a typical memory experiment involving the learning and retrieval of word pairs that are not semantically related. The encoding phase produced activation in prefrontal (associated with working memory) and hippocampal (associated with declarative memory) areas. During retrieval, the



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

precuneus (known to subserve episodic memory retrieval – Cavanna & Trimble, 2006) showed activation for both languages, but differential activation for each language in Broca’s area, the cerebellum, and the angular and supramarginal gyri. Interestingly, retrieval of words in the native (but not the foreign) language activated both Broca’s area and the right cerebellum (involved in the procedural memory for language), whereas the left cingulate gyrus was activated for the foreign (but not the native) language (indicative of non-automatic control). These data suggest that, while bilinguals rely on controlled processes in a memory task for both languages, they were able to rely on implicit linguistic competence to a greater extent for their L1 than their L2. Even though Perani et al. (2003) used a single-word task with proficient early bilinguals (L2 onset at 3 years of age, with different degrees of exposure to L2) they found a difference between L1 and L2 processing: more left frontal and right temporal activation for L2 than L1; more left temporal lobe activation for L1 than L2. The combination of some shared and some unique areas of activation reflects an overlap of implicit linguistic competence in a common gross anatomical area and other areas used to alleviate the lack of competence for the various components of the implicit L2 grammar, causing considerable variation among individuals in the sites and extent of areas activated by L2 processing. The two neuroimaging studies involving sentence processing examined by Stowe and Sabourin (2005) reveal both quantitative (Hasegawa et al., 2002) and qualitative (Dehaene et al., 1997) differences. Less automatized processing4 does not simply activate more cortical surface but also different cortical loci corresponding to explicit control (anterior cingulate cortex) or explicit knowledge (hippocampal system). “More native-like performance goes hand in hand with a more nativelike use of the neural underpinnings of language” (Stowe & Sabourin, 2005: 239), namely from controlled to automatic (barring speeded-up control, which accounts for a good proportion of L2 performance). Different component modules of the second language grammar (e.g., phonology, morphology, syntax), may be incorporated to different extents into implicit linguistic competence, giving rise to inter-individual variability in terms of what appears to overlap and what appears to be separate (inasmuch as these modules are sustained in different cerebral areas). Although studies report right-hemisphere activation in both languages, processing in a less proficient L2 tends to elicit more activation than processing in

4.  In other words, fewer elements are part of the automatized competence. A function is said to be less automatized to the extent that some components that are automatized in L1 are not automatized but explicitly controlled in L2.

 Declarative and procedural determinants of second languages

L1 (Dehaene et al., 1997; Urbanik et al., 2001; Hasegawa et al., 2002; Luke et al., 2002). These data are consistent with the hypothesis that both languages rely on pragmatic features to process sentences, but that L2 relies on them more. De Bleser et al. (2003) seek to explore the organization of the bilingual lexicon (in fact, vocabulary) by means of a covert picture-naming task in an fMRI study. The authors claim (1) that their results are not compatible with “entirely distinct neural substrates for the different languages of a bilingual individual” (p. 439, my italics); and (2) that activation was not less left-lateralized for the weaker L2, contradicting other studies such as Perani et al. (1998). However, as mentioned in the previous section on language laterality, unlike studies using sentences or short narratives as stimuli, there is no reason why naming pictures should activate right-hemispheres areas, given that there is no context, and hence no pragmatic inference, involved. Studies using single words as stimuli cannot be compared to studies using sentences because they do not tap the representation and processing of the language system (implicit linguistic competence); they tap only the vocabulary. In an fMRI study of 14 Russian non-native speakers of German who had lived for 5 years in Germany, Rüschemeyer et al. (2005) report a more similar pattern of increased activation in response to semantic than to syntactic violations. The non-native speakers showed a different pattern of activation from native speakers in all experimental conditions, namely greater involvement of several cortical and subcortical areas. Contrary to native speakers, non-native participants showed no reliable differences in the processing of syntactically anomalous versus correct sentences. The authors postulate that this was a reflection of increased syntactic processing costs associated with parsing a second language. Native speakers use “highly automatized and efficient processes” (p. 282); non-native speakers are forced to employ additional resources in the inferior frontal gyrus. Increased levels of activation were also observed in the head of the caudate nuclei in both hemispheres. The caudate nucleus works together with the prefrontal cortex to support cognitive function. The pattern of distribution in the right hemisphere was much more variable (and quite dispersed) in the L2. Moreover, again unlike native speakers, non-native participants showed no areas of differential activation between syntactically anomalous and correct sentences. In contrast, semantically anomalous sentences did trigger higher levels of activation than correct sentences, as in native speakers. In other words, non-native speakers show the same pattern of activation for semantic (more declarative, explicit) processing but not for syntactic processing (more automatic, implicit). Yokoyama et al. (2006), also in an fMRI study, used active and passive sentences to investigate whether late bilinguals process structurally complex sentences in L1 and L2 in the same or in different cortical networks. They found activation for both L1 and L2 in the left-hemisphere language-related regions, but different



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

activation patterns between L1 and L2 in the processing of passive sentences. The authors hypothesize that it is task difficulty that elicits greater activation during sentence comprehension in L2 than in L1 and they conclude that cortical representation reflects task difficulty. In the second language, task difficulty represents the degree to which L2 structures have not been internalized, which is a function of syntactic complexity. When a language is internalized, namely when the speaker has acquired implicit linguistic competence in that language, comprehension is automatic, and thus effortless. If speakers’ cortical representation reflects greater task difficulty in sentence comprehension in L2, as Yokoyama et al. suggest, then it implies that parts of L2 grammar have not been internalized and declarative knowledge, which is sensitive to degrees of complexity, must come to the rescue. The second language speakers may have acquired basic word order but not functional categories. Investigating the neural mechanisms underlying written sentence comprehension processing in an fMRI study, Suh et al. (2007) also found that the amount of activation was greater for more complex (embedded) sentences than for simpler (conjoined) sentences in L1, while no difference was found in L2. Whereas a large proportion of certain areas was commonly activated for the processing of L1 and L2 (in particular areas involved in the reading process per se in the occipital lobe), some areas were activated for L1 only (left medial/superior frontal gyrus, left lingual gyrus, and right fusiform gyrus); other areas were activated for L2 only (left cingulate and left fusiform gyrus). (Note that the cingulate gyrus is associated with conscious control, and that the fusiform gyrus, implicated in visual word recognition, is demarcated by the anterior part of the parahippocampal gyrus, known to be associated with declarative memory functions.) The authors suggest that the difference between the first and the second language could be due to the fact that certain syntactic processes are automatic in the first but not in the second. The processing of sentences in the second language is not likely to be automatized regardless of the difficulty level, which is why no difference due to complexity between the two types of structures was observed. The fact that subjects responded faster in the first than in the second language lends support to such an interpretation. The second language is sometimes reported to produce greater activation in the left cerebellum than the first (Pillai et al, 2004; Yokoyama, 2006). Whereas the right cerebellum is known to be involved in language processing,5 the left is not. The activation of the left cerebellum in L2 is likely to reflect the use of some

5.  Silveri, Leggio, and Molinari (1994); Zettin et al. (1997); Mariën et al. (2001); Docking, Murdoch, and Ward (2003); Jansen et al. (2005); Booth et al. (2007).

 Declarative and procedural determinants of second languages

compensatory strategy possibly associated with the right cerebral hemisphere (e.g., pragmatics or some form of attention: Tabor-Connor et al., 2006, observed learning-related attention in the left postero-lateral cerebellum consistent with learning-related effects in the right frontal cortex). In some cases, the right cerebellum is activated for L1 (automatic language processing) while there is no cerebellar activation for L2 (Halsband, 2006). In an fMRI study, Frenck-Mestre et al. (2005) found no significant differences as a function of language, in early or late bilinguals, in the activation of an established pattern of areas of the cortex, basal ganglia and cerebellum, for overt articulation of words and sentences. They concluded that the motor loop involved in articulation for the two languages of early and late bilinguals is essentially identical. There is always, of course, the question of whether the selected statistical threshold was adequate6 to evince a difference if there were one. On the other hand, even if the source of the impulses to the effectors were different7 (e.g., from automatic procedural competence for one language and from controlled performance for the other, at the speech motor planning phase), it would not be unexpected that the areas associated with the motor control of articulation (speech motor execution) should be activated for both. Finding that bilinguals use “the same areas” (Stowe & Sabourin, 2005: 350) for tasks using words but different areas for grammatical tasks is not surprising. It is predicted by the proposed neurolinguistic theory (Paradis, 2004), based on the evidence that “there is a shared semantic system for L1 and L2 at a single word level” (Xue et al., 2004b: 796). Stowe and Sabourin’s conclusion is that the same brain areas are used for processing both first and second language, though late learners are more likely to draw on additional resources during L2 processing. In other words, whatever has been automatized is represented and processed in the same brain areas; whatever has not been automatized uses additional (that is, different) resources (namely, controlled declarative-memory-based metalinguistic knowledge) during L2 processing. Proficiency and fluency (ease of access) will be modulated by the activation threshold (frequency and recency of L2 use). The consequences of later acquisition are more obvious at the level of automatic production skills (Briellmann et al., 2004). The overlap corresponds to automatized competence, the additional cortical resources correspond to whatever is substituted for what has not been automatized, namely controlled

6.  A change of the statistical threshold setting can result in an increase in activated areas and even cause a shift in the asymmetry index (Chee et al., 1999a). 7.  The difference would be upstream (afferent) from the motor effectors’ (actual speech muscle-movement control) activation.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

metalinguistic knowledge or pragmatic inference. The attention and working memory neural systems required to compensate for difficulty in processing L2 (Wang et al., 2003; Callan et al., 2004; Wang et al., 2003; Xue et al., 2004a; Gandour et al., 2007) are precisely those required for the explicit handling of aspects other than implicit competence.

 e nature of the additional cortical resources reported 2.3  Th to be recruited for L2 The majority of researchers who report overlapping activations for L1 and L2 are beginning to concede that additional cortical resources in which activation occurs in L2 only are recruited to meet the increased computational demands caused by lower L2 proficiency (Gandour et al., 2007). Later-learned or less proficient languages also depend on regions beyond the common activated zone such as the anterior cingulate gyrus and the contralateral hemisphere (Briellmann et al., 2004). In an fMRI study, Vingerhoets et al. (2003) report that cerebral activation during the use of the foreign language showed a tendency toward more extensive recruitment of areas activated in the native language and also the activation of a greater number of regions. Golestani et al. (2006) consider that their findings suggest differential involvement of the means modulating syntactic processing in the L2 versus L1 of moderately fluent bilinguals. They further suggest that attention and/or executive aspects of sentence production are more demanding in the L2 than the L1. Indeed, since automatic processing is neither open to introspection nor amenable to external control, the greater demands imposed on L2 processing cannot constitute a slower rate or other type of manipulation of automatic processing; therefore, they must involve mechanisms other than implicit linguistic competence (and not a different way of processing implicit competence). The often reported increased volume of activation or extended spatial areas cannot refer to a different way of dealing with implicit competence but must reflect computations other than those of the automatic procedural system that sustains implicit competence. The assumption that the differences between the processing of L1 and L2 stem only from the additional effort required for L2 implies that the nature of the representation of L1 and L2 is the same. If it were only a question of the need for more neural impulses (a higher activation threshold) to activate L2, it would presuppose that the L1 and L2 language systems process items of the same kind in the same way, with simply more difficulty in L2. This would assume that metalinguistic knowledge is of the same nature as implicit linguistic competence, that it is processed in the same way as implicit linguistic competence, or that it is

 Declarative and procedural determinants of second languages

converted into implicit linguistic competence. None of these assumptions are viable (Paradis, 2004). Metalinguistic knowledge cannot be sustained by the system that supports implicit linguistic competence since they are distinct and dissociated, involve different types of representation and “are substantiated in separate parts of the brain” (N.C. Ellis, 2005: 307). As long as L2 has not been internalized (automatized), it cannot be processed by the same neural mechanisms as L1; only implicit linguistic competence for L2 can. Additional activation (effort, cognitive computations) can only refer to activation of mechanisms outside the classical language areas proper that process the grammar, such as pragmatics and metalinguistic knowledge. No doubt these areas are also used to supplement the grammar in native verbal communication, but to a much lesser extent. In L1 they serve only to adapt the output (or input) to the appropriate circumstances of the discourse, not (as in L2) to overcome the lack of grammatical means to express the selected propositions (or comprehend the utterances of others). While it is true that, in a sense, “L2 processing is carried out through the same brain computational devices as those in L1 processing” (Perani & Abutalebi, 2005:  203), some precisions are in order. This would be true only of early bilinguals who developed two implicit subsystems from the start. It can only be the case when at least some implicit linguistic competence has been acquired for L2. The additional resources required for L2 grammatical processing are not part of the same neurofunctional system as L1 implicit linguistic competence. Rather, they take the form of other components of verbal communication, which use brain devices that differ from those that subserve implicit linguistic competence, at least until the L2 grammar has been internalized. Thus, it cannot be said that L2 grammatical processing is carried out “through the same computational brain devices underlying L1 grammatical processing” (Perani & Abutalebi, 2005: 204), because some parts of the grammar that are processed automatically by implicit computational procedures in L1 are processed explicitly by declarative memory procedures in L2. The second language may be said to use the same neural systems that sustain verbal communication in L1 (to different extents), even though grammatical processing, until internalized, is not carried out by the devices that underlie L1 grammatical processing (i.e., procedural memory for language, implicit linguistic competence). This is indeed the way in which “proficiency, age of acquisition, and amount of exposure can affect the cerebral representation of each language” (Perani & Abutalebi, 2005:  205), by using each of the components of verbal communication to different extents. Probably the most common finding of recent neuroimaging studies is that the cortical representations of different languages in bilinguals partially overlap. The part that overlaps is likely to represent the L2 competence acquired by the speaker and the areas that do not overlap represent the various compensatory strategies –



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

which accounts for the considerable variability across subjects found in the representation of L2, contrasted with the great similarity in the areas subserving L1. The extent of overlap corresponds to the extent of automatized L2 implicit linguistic competence. It is difficult to infer what an increase in the spatial extent of activation in a cortical area implies. Some studies report that an increase in activation co-varies with better mastery (Wang et al., 2003), whereas others report that such an increase correlates with lower mastery (De Bleser et al., 2003; Briellmann et al., 2004; Gandour et al., 2007). Moreover, given the known predominance of inhibitory neural processing and inhibitory cognitive mechanisms, the interpretation of increase/ decrease cortical blood flow seems problematic (Chertkow & Murtha, 1997). Wang et al. (2003) report that their English subjects’ improved performance at learning lexical tone in Mandarin involved not only the expansion of preexisting language-related areas but also the recruitment of additional cortical regions. This could be interpreted as evidence of learning rather than acquisition, with the additional regions being devoted to focused attention, explicit control, and pragmatics. In an optical imaging experiment using semantic (correct/incorrect attribution of meaning) and syntactic (subject-verb agreement) tasks, Scherer et al. (2006) report a wider network sustaining L2 overall and, particularly for the semantic condition, more extensive activation in the right hemisphere for L2 than for L1. This result suggests the greater use of metalinguistic knowledge for the syntactic task and of pragmatics for the semantic task as compensatory strategies. Neuroimaging studies consistently report a greater volume of activation during the processing of L2 than during the processing of L1. Two types of explanation can be envisaged. Either (1) speakers need to recruit other mechanisms to compensate for the gaps in their second language implicit linguistic competence, or (2) more impulses are required to activate second language representations whose activation threshold is higher than in the corresponding native language representations. If the latter were the case, the underlying grammatical procedures involved would have to exist before they could be activated. In late L2 learners, this would imply that they could use the same procedures that are used automatically by native speakers but “with additional effort”8 (presumably involving attention

8.  For example, “attentional control for selection” (Callan et al., 2004: 1190); “increased computational command caused by low L2 proficiency” (Xue et al., 2004b:  8); “working memory/sequencing demands” (Golestani et al., 2006: 1036); “increased efforts necessary for syntactic processing” . . . “increase in cognitive demands” (Jeong et al., 2007b: 181, 184).

 Declarative and procedural determinants of second languages

and declarative memory, as is generally assumed when speaking of effort). But a task performed by using automatic functions cannot be executed in a way other than automatic. Since speakers who are not fully proficient have not internalized the required procedures, they cannot manipulate them one way or another. They certainly cannot control them, since these procedures are not explicitly known. Voluntary control requires explicit representations of one’s knowledge, whereas automatic action can be sustained by procedural know-how (Dienes & Perner, 1999: 741). The moment a performance is not automatic, a different system (one relying on explicit processing) takes over. The use of additional effort entails the use of mechanisms that require conscious effort (concentrated attention, control) and that involves declarative memory. Given that the use of implicit linguistic competence is fast and effortless, it is used whenever it is available. Only when it cannot be used to express a particular message (or understand a message expressed by others) does one have recourse to non-automatic means. To the extent that a second language has not yet been fully acquired (i.e., internalized in the form of implicit competence that is available for automatic use), the same type of processes cannot be used in L1 and L2. The additional effort must necessarily involve processes additional to the ones used effortlessly for the native language. The observed extra neural activity is more likely to reflect processing demands other than those related specifically to L2 language processing per se (Callan et al., 2004), namely metalinguistic knowledge and pragmatics. The differences usually observed do not indicate differential intra- or interhemispheric language9 localization but a difference in the extent of reliance on cognitive functions other than implicit competence to compensate for lacunae in any of the components10 of implicit grammar. Additional right-hemisphere involvement may not only be indicative of increased use of pragmatic elements, but also of extra cognitive task difficulty. Rajah and McIntosh (2006) found right dorsolateral prefrontal cortex involvement in domain-general executive control tasks that impose strenuous cognitive demands, reflecting general strategic organizational or monitoring processes. In a more recent study, increased task difficulty across domains caused greater right lateral premotor cortex activity (Rajah, Ames, & D’Esposito, 2008). Given the functional heterogeneity of the frontal cortex, including various cognitive control processes, these data would suggest that L2 speakers exhibiting greater activation in

9.  Qua language functional system, implicit linguistic competence. 10.  Phonology, morphology, syntax (e.g., degree of fronting, palatalization, voice onset time, etc., within phonology; passive construction, relative clause, phrase structure, etc., within syntax – each of which may be acquired at a different time).



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

these areas do not process their second language (or, at least, parts of it) as implicit linguistic competence but rely on declarative-memory-based cognitive processes. The extra effort and additional resources that neuroimaging studies report as allocated to the processing of a second language cannot be devoted to the activation of the same neurofunctional system that sustains L1 implicit competence, for at least two reasons, one logical, the other empirical. (1) At the very least, in order to activate a system more, the system must be there to be activated. A system simply cannot be activated before it has been acquired. Therefore, only to the extent that L2 implicit competence has been acquired could it require more effort to activate (in the form of more impulses, a lowering of its activation threshold). In order to communicate verbally, metalinguistic knowledge and pragmatics need to be substituted for whatever implicit linguistic competence has not yet been acquired, resulting in clearly separate areas of activation. (2) Empirically, there is a large body of evidence from various sources (neuroimaging, electrical cortical stimulation) that strongly suggests that, even once the L2 system has been acquired, the two languages are each subserved by distinct language-specific circuits within the shared neurofunctional language areas.

3.  Procedural and declarative language switching and mixing There are different types of language switching. Some switches are deliberate and controlled, others are automatic. It is therefore not unlikely that there should be a different cerebral mechanism involved in the implementation of each type of switching. Controlled switching would thus rely on anterior cingulate and dorsal prefrontal cortex (Badgaiyan, 2000; Elsinger, Harrington, & Rao, 2006; Faw, 2003) whereas automatic switching would involve the same mechanisms that sustain automatic language functions (comprehension and production tasks in conversational settings). 3.1  Types of switches and consequences As with other functions, switching will be either (1) conscious, deliberate, and hence subserved by cerebral structures that sustain declarative memory processes; or (2) implicit, like the use of grammar, in the course of normal language use, and hence subserved by cerebral structures that sustain procedural memory processes. It is also likely to receive pragmatic input related to the context of use, with consequent right-hemisphere involvement. Intra-sentential code switches are rule-governed and systematic, displaying dependency relations that reflect the operation of underlying syntactic principles

 Declarative and procedural determinants of second languages

and can be imputed to unconscious, implicit linguistic competence (Toribio, 2001). Mixing (including nonce11 borrowing) is thus most likely generally implicit, as part of the normal functioning of grammar within the language system, with rare exceptions, such as when one deliberately chooses to borrow a word or phrase for special effects (jokes, emphasis, particular reference, political statement, etc.). Both switching and mixing have received substantial attention lately, in studies using both natural and contrived tasks. A message and the intention to communicate it are conscious. Thus, the decision to speak either French or Dutch, to alternate between the two, or to deliberately introduce a Dutch word or phrase into a French sentence for a particular purpose, is of the same type as the decision to speak in the first place or to remain silent. Such a decision to switch or mix and its implementation rely on an executive function like any other conscious decision to change one’s behavior. Switching and mixing can thus be subserved by either a declarative or a procedural memory system, depending on the circumstances, as with other functions, such as the slow, deliberate performance of a sequence of movements for the purpose of demonstration versus the automatic performance of the same sequence of movements in the normal course of action on the soccer field. In addition, in the case of unreflected, automatic switching and mixing in the normal course of conversation, switching is governed by the pragmatics of the situation (e.g., a unilingual friend joining an up-to-then bilingual conversation). The switches involved in normal mixing are governed by the implicit grammar of the language system, which automatically inserts the Dutch element in the French sentence in accordance with the grammatical constraints of the two language subsystems. This is especially likely to be the case for individuals who live in a community where mixing is frequent. Switching from one language to another when a unilingual friend joins in can be automatic, the way one lifts one’s foot off the gas pedal when the rear lights of the car in front light up. Mixing on-line is as implicit/procedural as any other grammatical process. Switches induced by word-finding difficulty or by an item with a much lower activation threshold (due to more frequent and/or more recent use) are automatic. Conversely, switching on request (whether words or sentences) is a consciously controlled task. Legitimate (i.e., natural) intra-sentential switches are subject to the constraints of the grammar of both language subsystems. Automatic nonce borrowing, in an

11.  A nonce borrowing refers to a word borrowed on the fly, that is not an official part of the language into which it is inserted and hence is not used by native speakers of that language, the way long integrated loan words are.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

English-Spanish bilingual speaker for example, depends on a significant imbalance between the English item’s activation threshold and that of the corresponding Spanish item. When a speaker is in a unilingual speech mode (i.e., a mind-set to speak only one language), the high activation threshold of an item in the currently spoken language will result in what happens with unilingual speakers: hesitation pauses for hundreds of milliseconds or circumlocutions. When the speaker is in a bilingual mode (i.e., mind-set to use whatever language is best available in the circumstance), the high-threshold item of a currently spoken language will automatically be replaced by the corresponding low-threshold item from the other language; this may either result in a nonce borrowing or trigger a switch to the other language (Clyne, 1980), causing a language switch in mid-sentence (e.g., the speaker starts a sentence in Polish and finishes it in Yiddish). After L2 learners have achieved the change from controlled to automatic L2 processing, that is, when they have internalized L2, implicit linguistic competence for each of L1 and L2 constitutes a subsystem within the language neurofunctional system (Paradis, 2004). Items may then be automatically borrowed from the L2 subsystem when the individual is in a bilingual mode of communication and the L1 item is temporarily unavailable (or the reverse). Whenever attention is involved, the performance is conscious. The control exercised by bilingual speakers on language switching can be automatic or conscious, depending on whether it involves an automatic selection or a conscious decision. Under experimental conditions of switching on cue, switching costs are reported to increase, reflecting the difference between deliberate, conscious switching, with attention focused on the (language external) cue, and automatic switching (with no more cost than implicit linguistic competence processing in a single language), in individuals accustomed to frequent switching. The controlled switch produces the same output (albeit with increased latencies) as automatic switching in a natural switching environment. This is homologous to speaking an L2 in either a controlled or an automatic way, with the same consequences: controlled switching will involve general mechanisms of explicit task-switching that rely on declarative memory in the same way that controlled speaking does; automatic switching will involve mechanisms of implicit linguistic competence (with automatic control of inhibition/disinhibition) in the same way that normal unilingual speaking does. 3.2  Switching data from neuroimaging studies The findings of neuroimaging studies depend crucially on the type of task: Contrived tasks should activate areas consistent with declarative memory processes; natural tasks should activate areas consistent with procedural memory systems. The use of any task other than the natural use of language (including natural

 Declarative and procedural determinants of second languages

switching and mixing) has the same consequences as using single words:  The task does not tap the normal automatic processes that sustain the natural use of language, including the contribution of pragmatics and its neural underpinnings. When the switch occurs on-line, without conscious control, in the normal course of events, the control mechanism will be the same as that which selects an active, passive or cleft sentence, namely the automatic inhibition of the nonselected item, in this case the entire language subsystem. In this situation, the process need not involve executive control. The neuroanatomical bases of cerebral language control are being intensively explored in an effort to understand not only mixing and switching but also the diversity of bilingual aphasia recovery patterns (Green, 1986, 1998, 2002, 2005). It will be argued that Green’s inhibition control model is applicable to both automatic and controlled situations, involving two distinct cerebral mechanisms, depending on whether the switching and/or mixing depends on conscious decision or automatic selection. In most neuroimaging studies of language switching published since the year 2000 (Hernandez et al., 2000, 2001; Costa & Santesteban, 2004;  Costa, Santesteban, & Ivanova, 2006; Abutalebi et al., 2007a; Christoffels, Firk, & Schiller, 2007; Khateb et al., 2007; Wang et al., 2007), the cards are stacked against natural language switching: (1) the task involves the use of a single word to name a picture; and (2) even more importantly, the subjects are required to switch on command – two contexts that favor the use of declarative memory and conscious executive control. For example, Abutalebi et al. (2007a) investigated the neural correlates of language selection processes in German-French bilingual individuals during picture naming in different unilingual and bilingual selection contexts. In the unilingual condition, participants were shown pictures and, on the basis of a cue word that immediately followed, had to either name the picture (in L1) or produce a related verb (in L1). In the bilingual condition, they had to name pictures either in L1 or L2 on the basis of a cue word (Deutsch or Français) that immediately followed. The authors report that naming in the first language in the bilingual experimental context increased activation in the left caudate and anterior cingulate cortex, relative to the unilingual context. These areas were even more extended when participants were naming in L2. Now, brain activity differs between naming in a natural situation and naming on cue. Nevertheless, the naming task in the native condition is closer to natural language use than either (1) generating a verb from a noun on request or (2) switching to one or the other language on cue. In this bilingual context, the very randomization of the L1 and L2 cue increases the need for attention. The inter-language selection context is explicitly guided (by the cue) whereas the intralanguage selection is different in nature, based on natural implicit word retrieval processes, which take less time, as confirmed by the reported statistically shorter



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

reaction times. A “selection cost” reflects the conscious decision-making process. In the L1 naming condition, the selection process (i.e., finding the right word for the picture) is implicit; in the bilingual condition, the selection of the language is explicit before the word is implicitly retrieved. The same applies to L2 picture naming, with additional involvement of declarative memory in the (effortful) search within the (less-known) set of L2 words. The authors note that the fact that the production of L1 names in the bilingual versus the unilingual context, as well as in the direct comparison between L1 naming in the bilingual context and the name/verb generating condition all induced activation in the left anterior cingulate cortex and caudate nucleus, strongly suggests that the cognitive processes underlying lexical retrieval might differ between the two selection contexts (i.e., simple naming versus generating a verb or switching to L2). One possible interpretation is that producing something on cue (a selection between two actions: naming a picture or generating a verb in one case; naming in L1 or L2 in the other) requires conscious attentional control, whereas simply naming a picture in L1 in an L1 context is a more natural, implicit task.12 The selection of a noun in L1 in an L1 naming context is a more natural task than naming in one or the other language on the basis of a cue word (Deutsch or français) that follows each picture for 300 ms. In fact, the naming in the bilingual condition is a conscious, deliberate choice to abide by the command signaled by means of the cue. It is therefore not surprising that naming pictures in L1 in the bilingual (controlled) context activated a larger neural network than in the unilingual (more natural) context, in particular in the anterior cingulate cortex, a sign of executive control, and the caudate nucleus, which many functional neuroimaging studies associate with executive functions (Seger & Cincotta, 2005), as well as even greater participation of the anterior cingulate cortex and caudate nucleus for naming in L2. These results do not speak to the situation when bilinguals are placed in a normal mixed context, namely in the normal use of both languages in a bilingual conversation (natural implicit switching, with automatic control). The absence of an L2 naming condition precludes ascertaining whether the areas activated in

12.  Though it is still different from using the word on-line in the course of producing an utterance in the absence of any experimental constraints, representing a difference between internally initiated and externally specified (i.e., internally generated and externally guided) verbal responses (Lau et al., 2004; Elsinger, Harrington, & Rao, 2006; Tremblay & Gracco, 2006). In addition to the fact that the task (1) concerned single words and (2) required an externally guided response, the left caudate is known to be activated when the language processing system cannot rely entirely on automatic mechanisms and has to recruit controlled processes as well (Friederici, 2006).

 Declarative and procedural determinants of second languages

the bilingual experimental context would have also been revealed in a unilingual context. Given that the subjects’ L2 was definitely less proficient than their L1 (they obtained only 66% correct on proficiency tests), chances are that these areas would be activated, possibly to a lesser extent, but still to a larger extent than for L1 naming in an L1 context. As expected, the results of neuroimaging cued switching studies consistently show increased activation of brain regions related to executive functions, namely the left dorsolateral prefrontal cortex (Hernandez et al., 2000, 2001) and left dorsolateral prefrontal cortex, in addition to the anterior cingulate cortex (Abutalebi et al., 2007a; Khateb et al., 2007; Wang et al., 2007). Not only do cued switching studies not represent language switching in natural contexts but, from the declarative/procedural viewpoint, they correspond to opposite kinds of processes. Cued switching is deliberate, calls for focused attention, and is consciously controlled; natural switching occurs offhand, heedlessly, and automatically. 3.3  Switching data from clinical studies Any skilled behavior acquired without formal learning or focused observation is subserved by procedural memory devoted to that function. If a particular function is implicit (e.g., the implementation of implicit rules of code-switching), it is subserved by procedural memory and will be vulnerable to impairment when the structures that subserve procedural memory are damaged. It is preserved when the damage is elsewhere (e.g., affecting structures that subserve declarative memory). The reverse is true of behavior that is consciously learned and/or explicitly controlled. We may therefore expect that automatic language switches will be subserved by implicit activation and inhibition mechanisms, whereas deliberate switches will be subserved by cerebral mechanisms implicated in conscious control. Three types of pathological language-switching problems are encountered:  (1) Frequent violations of grammaticality constraints, that is, switching at junctures where non-brain-damaged individuals never switch, for instance between a clitic pronoun and a verb (je don’t understand), or borrowing function words in isolation (je parle anglais to mes enfants) (Paradis & Lecours, 1979). Blending may occur at all levels of linguistic structure. At the phonological level, patients may speak one of their languages with phonological interference from (one of) the other(s); in other words, they speak with a foreign accent that was not present premorbidly (Kho et al., 2007). At the word level, they may produce blended forms, such as twörpo from German Zwerg and Hungarian törpe (‘dwarf ’) (Gloning & Gloning, 1965) or use a root from one language with an inflection from another such as se fallió from English fall and Spanish -ió simple past inflection (se cayó: he fell), or speakear from English speak and Spanish -ear, infinitive marker (Ansaldo & Marcotte, 2007). At the syntactic level, patients may use structures from one



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

language while speaking another (e.g., he lives in Toronto since ten years), a type of error they never made before. These are deficits of the implicit control mechanisms of linguistic competence. (2) Another kind of problem involves mixing two languages when speaking to non-bilinguals, failing to take into account the pragmatics of the situation. This is again a problem of implicit control, especially in cases when the patient is aware of the problem but mixes languages uncontrollably (e.g., Ansaldo & Marcotte, 2007). (3) Patients may also simply switch more than they used to, either as a conscious way of coping with word-finding difficulty, or without awareness (as is often the case in non-brain-damaged individuals); the former involves explicit executive control, the latter, implicit mechanisms. There has been an attempt in the recent literature to distinguish pathological mixing (intermingling different languages within a single utterance) from pathological language switching (alternating between languages at utterance boundaries even when the patient’s interlocutors do not understand one of the languages). It has been suggested that these symptoms follow lesions in different brain areas: left temporoparietal cortex for mixing; left anterior cingulate and frontal lobe for switching (Fabbro, Skrap, & Aglioti, 2000). Both pathological switching and mixing in a single patient without multiple lesion sites have also been described (Mariën et al., 2005). These authors infer from the parallelism between the evolu­ tion  of the language symptoms and their SPECT findings that subcortical left frontal lobe circuitry may be crucially involved in both language switching and mixing. In fact there are four types of processes:  (1) implicit language switching; (2) explicit language switching; (3) implicit language mixing; and (4) explicit language mixing. All four processes occur in normal bilingual conversation. Explicit switching and mixing may be deliberate on the part of the speakers, guided by their particular communicative intentions and requiring conscious executive control. Spontaneous, on-line implicit mixing reflects the nonce borrowing of words from another language whose activation threshold is lower. Spontaneous switching may be triggered, mid-sentence, by a borrowed word, or, at a sentence boundary, by implicit pragmatic cues such as when a familiar unilingual speaker joins the group. These are affected by two types of deficits: impairment of the inhibitory mechanisms of implicit control (responsible for raising or lowering the activation threshold) and/or pragmatic deficits (in implicit switching and/or mixing); impairment of voluntary control (in explicit switching and/or mixing). When IQ and education level are a factor in the prognosis for recovery from aphasia (Green, 2005), it is an indication that metalinguistic knowledge, subserved by the declarative memory system, is at work (since these factors have been shown to have an influence on the capacity to learn, but not to acquire a language). Let us now examine how these phenomena are presented in recent cases of pathological switching (see also Paradis, 2008b).

 Declarative and procedural determinants of second languages

Riccardi, Fabbro, and Obler (2004) describe a patient who still languageswitched frequently five years post-stroke but who nevertheless seemed to be sensitive to the pragmatics of the switching situation as he was able to switch in appropriate circumstances. Rossi, Denes, and Bastiaanse (2003) report the case of a patient who did not exhibit any phenomena resembling pathological mixing during standardized testing but who nevertheless constantly mixed her three languages during unilingual conversations, despite being persistently reminded that she had to use the language of the examiner. The authors conclude that a conversational setting may be more revealing of pathological mixing than a formal test (on which patients may exercise control). This provides support for the suspicion that, in any testing situation, including neuroimaging sessions (see discussion of Abutalebi et al.’s 2007b results below), attention is mobilized. Abutalebi, Miozzo, and Cappa (2000) describe a trilingual woman who would inadvertently switch from one language to another, even in conversation with unilingual speakers, despite being fully aware of her impairment. A circumscribed infarct was located in the periventricular white matter surrounding the left caudate nucleus. The authors therefore speculate that damage to the dominant basal ganglia may disrupt the (presumably automatic) inhibition/disinhibition process required to switch between languages, an area suspected of being involved in access to and control of lexical representations. Ansaldo and Marcotte (2007) describe a patient who was equally at ease in Spanish and English but did not customarily mix or switch before insult. Subsequent to a left subcortical perisylvian embolic lesion, the patient frequently mixed within a single utterance, even when speaking to a unilingual conversational partner. The patient was aware of his inappropriate mixing but could not help it, even when asked to use only one language. His inability to deliberately avoid mixing would suggest that the mixing process is automatic, reflecting the use of impaired implicit linguistic competence, where borrowing and switching are governed by the activation threshold levels across languages. Interestingly, the authors report that attempts to control mixing voluntarily resulted in an exacerbation of word-finding difficulty and anomia-related signs (e.g., more gaps, silences and groping behaviors), thus confirming that mixing is controlled automatically and is impervious to conscious control. The role of pragmatics in the appropriate selection of which language to use in unilingual and bilingual discourse is generally recognized (Rossi et al., 2003; Riccardi et al., 2004; Kohnert, 2004; Costa et al., 2006; Ansaldo & Marcotte, 2007; Centeno et al., 2007). As with other aspects of language use, in language switching and mixing, the relevant pragmatic information may be conscious and deliberately applied, or it may be implicit and automatically brought to bear. From the various data on switching language based on a designated cue cited by Green (2005: 521), “it follows that damage to frontal structures should impair



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

the ability either to maintain a given language or to avoid switching between languages” on request. Such an entailment need not exist for natural switching or maintenance of a language currently spoken. From Fabbro et al.’s (2000) patient’s inability to speak only Friulian or only Italian, Green infers that the deficit was the consequence of a lesion in the anterior cingulate. However, the patient had a rather extensive lesion which affected the white matter underlying the left inferior, middle, and superior frontal gyri, the anterior callosum structures, the anterior arm of the internal capsule, the striatum, the premotor area, the frontal eye field, and the cingulate gyrus; the latter was also slightly involved in the right hemisphere. Clearly, many areas subserving implicit linguistic competence were involved (e.g., the striatum); in addition, it is not clear whether the anterior cingulate was specifically affected (it seems to have been only inferred, on theoretical grounds). If it were the case that the problem was specifically caused by the anterior cingulate, one might have expected the patient to have had deficits in task switching and decision making outside the specific domain of language. The patient did not present with attentional disorders and made no errors in tests assessing attention, nor did he make translation errors in either direction. The problem may have been an imbalance of activation thresholds or rapidly depleting resources attributed to each subsystem, as proposed by Green (1986). 3.4  Conscious and automatic control mechanisms in language switching Language control mechanisms come in two flavors: conscious and unconscious; the two have distinct neural underpinnings. Natural switching belongs to the latter, switching on request to the former. One cannot use a task involving controlled switching to identify the brain mechanisms that perform natural on-line processes (or the reverse: use natural on-line switching data to identify the conscious executive control mechanisms). In automatic selection, in natural uses of language, the activation/inhibition processes are generated within the neurofunctional language system, implicating the perisylvian cortex, basal ganglia and cerebellum; in deliberate selection, the processes are subject to conscious control, involving attention and general executive functions, thus engaging the prefrontal cortex and anterior cingulate. Language control is part of a system for the control of action in general (Green, 2005). However, a distinction must be made between deliberate versus automatic selection. Control of language and control of action have much in common: Both are either conscious or automatic. Memory schemas are controlled either automatically, as in the generation of sentences in implicit linguistic competence, or consciously, as in the construction of sentences from metalinguistic knowledge. A mechanism that regulates the outputs from the lexico-semantic system by altering the activation levels of representations within that system and by inhibiting outputs from the system

 Declarative and procedural determinants of second languages

(e.g., schemas of language tasks; Green, 1998: 69) is an automatic, implicit system; it can nevertheless be set in motion deliberately, as when one follows instructions to name a picture in one or the other language or when one chooses to switch languages or to translate. The same is true when, after a halt, one decides to start pedaling a bike, setting in motion the automatic sequences of rapid movements to keep balance. When the control is automatic, as in the comprehension and production of sentences in one’s native language or mixing and switching by early bilingual speakers in a natural setting, then the control of implicit schemata, sustained by procedural memory, is language-system-specific and impervious to conscious general executive control (as it is to influence/interference from any system external to its neurofunctional module). The inhibition/disinhibition processes within the specialized control mechanism may yield the same results as those of the deliberate control mechanism, but they function independently and are controlled automatically. It so happens that specific exemplars of action in general can also become implicit through frequent use (e.g., motor skills) and, like linguistic competence, including natural language switching, be controlled automatically. One may therefore say that language control, like control of action in general, can be either implicit or explicit. When it is explicit, then it is indeed part of a system for the conscious control of action in general; when it is implicit, by virtue of the task-specificity of procedural memory, its control is restricted to the particular functional system, as is the case with highly skilled actions in general. In keeping with the direct access hypothesis (Paradis, 2004: 203), according to which there is no a priori reason to believe that the mechanism for accessing a word and its translation equivalent should differ from that for accessing a word and its synonym within the same language, nonce borrowing of words from a language other than the one currently in use should be subject to the same basic principles of selection of the acceptable item with the lowest activation threshold (i.e., elements across both subsystems that are more or less suitable to fill the expressive need, that is, compatible with the intended message, while fitting into the current grammatical and semantic contexts). There is no need to involve attentional mechanisms in natural language switching. The control of the lexicalization process (i.e., the manipulation of activation thresholds) is automatic during natural language mixing and switching, as it is for lexical selection in unilingual speech. In the normal use of language, the selection, among appropriate items, of the one with the lowest activation threshold is an automatic process, not a deliberate, conscious choice. Contrary to deliberate borrowings or switches made to stress a point or upon request (e.g., an explicitly cued response), this kind of borrowing from, or the switch to, the other subsystem should therefore activate the cerebral structures responsible for implicit linguistic competence, not those involved in explicit executive decisions.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

The process that alternately inhibits one or the other language subsystem during speech follows the general laws of excitation and inhibition between neural phenomena of any type (Paradis, 1977:  90); it is automatically driven by the activation threshold levels. These thresholds are modulated by the mental set induced by the particular situation (unilingual or bilingual mode), as a blanket increase in activation threshold up to a given degree can be imposed on an entire language subsystem. Given the previous threshold levels within this now inhibited subsystem, the threshold of some items will nevertheless remain higher than that of the corresponding items in the non-inhibited (i.e., currently used) subsystem, making them targets for nonce borrowing when the speaker is in a bilingual mode. Bilingual speakers possess two subsets of neural connections, one for each language (and each can be activated or inhibited independently because of the strong associations between its elements), while at the same time they possess one larger set (that comprises both subsets) from which they are able to draw elements of either language at any time (Paradis, 1981). The selection of elements is automatically driven by their activation threshold levels. The selection of a word from Language B while speaking Language A in a bilingual speech mode follows the same nonconscious, automatic process as the selection of a close synonym by a unilingual speaker (or a bilingual speaker in a unilingual mode). But “the decision to speak in English or Russian is surely of the same order as the decision to speak at all or to remain silent, or the decision to wiggle one’s little finger or to keep it still” (Paradis, 1977: 91, italics added). Insofar as switching is deliberate, “bilingual control is a special case of the control of action in general” (De Groot & Christoffels, 2006: 189). On the other hand, non-deliberate switching on-line is hypothesized to proceed from Luria’s principle that all possible alternatives must be inhibited in order to activate the selected item, a process which is highly automatized (Luria, 1973b; Posner & Rothbart, 1992). This principle applies within a single language in unilingual speakers and across subsystems within the language system in multilingual speakers (to the extent that the non-native languages have been internalized). To the degree that switching is spontaneous (hence automatic), it is a special case of nonconscious control over implicit computational procedures, such as the selection of a lexical item or syntactic construction within a language, as with skilled motor action (e.g., bicycle riding). In terms of the proposed neurolinguistic theory of bilingualism (Paradis, 2004), Green’s (1998, 2005) language tag construct can easily be adapted to fit the subsystems hypothesis in which languages are represented in distinct neurofunctional language subsystems. A subsystem can be selectively inhibited as a whole (its activation threshold raised) when the other subsystem is in use. Alternatively, elements with a lower activation threshold may be selected from the other

 Declarative and procedural determinants of second languages

subsystem, by virtue of both subsystems being subserved by different circuits that are still part of the larger neurofunctional language system, when in the bilingual communicative mode. In other words, an entire subsystem or parts of it may be selectively deactivated. Lexical representations thus do not need an implicit language-specific tag, as they would if they were represented as part of a single extended language system. Lexical items are identified by virtue of being part of a distinctive subsystem network. If, as De Groot and Christoffels (2006) propose, the connections that constitute language subsystems are implicit and the language tags are explicit,13 then the notion of language tag concurs with that of Paradis (2004), namely that the language tag, in the sense of awareness of language identity, is a product of metalinguistic knowledge and is not used in on-line processing. The inhibition/disinhibition process of David Green’s control model (Green, 1986, 1998, 2002, 2005, 2008) applies to natural nonce borrowing (inasmuch as it concerns single words) and switching when the activation threshold of a language as a whole needs to be raised so as to overcome competition. These processes are an extension of the automatic inhibitory mechanisms at work in the use of unilingual implicit linguistic competence:  Competing non-selected items are inhibited so that the selected item can reach threshold while the others are suppressed (Luria, 1973b). In the bilingual situation, not only must all the competing items within the selected language be inhibited, but also those from the other language. This inhibitory control is as automatic as the selection of a passive over an active or cleft construction in the speech of a unilingual speaker. Conversely, when the switch is deliberate, another mechanism takes over, distinct from the cerebral substrate that sustains implicit linguistic competence, as it relies on explicit executive control. When bilinguals are said to exert control over their two language subsystems by differentially activating and/or inhibiting their languages, the actual control operations are implicit; their triggering can be either implicit or explicit, depending on the circumstances (i.e., whether the process is explicitly willed or automatically set in motion during online processing). One study (Abutalebi et al., 2007b) is particularly interesting in this regard. It is the first neuroimaging study not to use a task requiring switching on cue. The participants were required to listen to four passages containing sudden and unpredictable changes of language. The somewhat surprising finding is that the

13.  Referring to the fact that Green (1998) and De Bot (2000) assume the existence of both language subsystems and language tags, they write: “such a set up specifies language membership twice for each word, once implicitly, in terms of connections … and a second time explicitly, by the language tag” (p. 192).



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

comprehension task seemed to activate the anterior cingular cortex. It would appear that the mere fact of finding oneself in an experimental context may be sufficient to increase vigilance and attention to the task at hand (in a way analogous to the bilingual vs. unilingual mode, namely a mind-set that biases the activation thresholds). When the same task is performed in a natural context, the variations in activation thresholds are set automatically, in the same way that items are selected in the normal course of generating utterances (i.e., without conscious control). These thresholds are a function of frequency and recency of use; their raising or lowering in order to effect the selection of one item over another (within or across languages) are influenced by the bilingual versus unilingual mind-set that allows or inhibits switches. Correspondingly, in comprehension, one will expect to encounter switches in a bilingual environment where switches are socially accepted, though not when the individual is in a unilingual speech mode. In other words, the control over mixing and switching in natural situations need not differ from that over the choice of different constructions in the course of conversation. In an experimental setting, however, it is likely that the participants’ attention is focused on the task, even when it consists of listening to random legitimate (i.e., grammatical) or ungrammatical sentences, in a way somewhat analogous to the extra attention paid to monitoring one’s speech in a particularly formal occasion; in both cases, explicit control is engaged. Cognitive executive control, whether unitary14 (i.e., general, all-purpose) or discrete15 (i.e., function-specific, e.g., language-specific) is known to rely on the prefrontal cortex. For one thing, the prefrontal cortex implements attentional control (Nieuwenhuis & Yeung, 2005). It also implements working memory (Baddeley, 1986). The dorsolateral prefrontal cortex is related to general executive functions (Christoffels et al., 2007). Voluntary shifts of attention are mediated by supramodal processes in the parietal lobe; task switching is classically affected by prefrontal damage (Green, Teder-Sälejärvi, & McDonald, 2005). The prefrontal cortex is critically involved in the control of behavior (Cohen, Botvinick, & Carter, 2000). Prefrontal and cingulate cortices are implicated in executive control of conscious action (Badgaiyan, 2000):  The dorsal prefrontal cortex is associated with attention to the selection of action (Lau et al., 2004); the dorsolateral prefrontal cortex controls willed action and working memory (Faw, 2003) – two conscious functions. These systems are involved in decision-making over voluntarily controlled elements of the nervous system; they control tasks performed upon instruction (Baars, 2004) or willed actions (Hyder et al., 1997). The dorsolateral and prefrontal 14.  See, for example, Duncan (2001); Miller and Cohen (2001); Kane and Engle (2002). 15.  See, for example, Baddeley (1996); Miyake et al. (2000).

 Declarative and procedural determinants of second languages

cortices (the “coordinator” system) controls willed action and working memory (Faw, 2003). Cognitive control is engaged by selective attention. Selective attention is a conscious function: One is aware of what one is paying attention to. The supervisory attentional system, by virtue of engaging attention, engages awareness. This is a cardinal function of the prefrontal cortex; non-automatic processes rely on selective attention, automatic ones do not (Cohen & Servan-Schreiber, 1992). Willed control organizes, coordinates and monitors (with attention and awareness) the schemata required to implement responses to experimental tasks, including cued switches: Attention is paid to the language selection in willed, cued switching, and such attention to the selection of a required response is associated with dorsal prefrontal and anterior cingulate cortices (the “attender” system, Faw, 2003), though it does not play a unique role in the generation of on-line internally initiated switches (Lau et al., 2004). Experimental language switching paradigms involve attention to the cue (prefrontal cortex), conscious monitoring (anterior cingulate cortex) and working memory (prefrontal cortex). Attention and decision rely on the prefrontal cortex. A deliberate switch is conscious before the fact (as opposed to conscious after the fact, when one becomes aware of one’s output, which itself was automatically generated). In sum, control in language switching is implicit (hence automatic) in the natural use of languages in bilingual settings when speakers switch between their languages frequently in their daily lives; it is explicit (hence controlled) when switching is performed on request. Externally specified actions, such as switching on cue, involve different areas (premotor areas) than internally specified actions, such as natural automatic switches (medial supplementary motor cortex and lateral prefrontal cortex) (Tremblay & Gracco, 2006); internally generated on-line control of actions is mediated by the basal ganglia, particularly the striatum (Elsinger et al., 2006). The striatum is known to be important for implicitly coordinating sequences and implementing behavioral routines, whereas the prefrontal cortex is thought to be involved in making conscious choices (Ravel & Richmond, 2005). In other words, the prefrontal cortex is involved in triggering the programmed schemata that are carried out automatically by the basal ganglia (the caudate in particular). A deliberate decision to switch or borrow an item activates prefrontal cortex areas; the selection itself is automatic. In the natural use of language, both the trigger and the implementation are automatic (and nonconscious). When similar stimuli produce similar responses under both conscious and nonconscious conditions, since the conscious condition involves working memory and requires executive control, the nonconscious condition cannot, under penalty of self-contradiction, be regulated by the same process. We cannot consciously control things of which we are not aware. The supervisory attentional



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

system is required when a complex response is consciously processed. If the response is automatic, it cannot be subject to the control of deliberate attention. Actions associated with implicit (albeit complex) processes cannot be consciously executed. These automatic responses are assumed to operate by lateral inhibition of unselected alternatives (including those in the other subsystem). Even though neuroimaging techniques show that some implicit tasks activate the same prefrontal cortex regions involved in central executive processes, it cannot be concluded that they are controlled by the same processes that regulate conscious actions. It is more likely that the mechanisms of automatic control rely on distinct frontal/basal-ganglia circuitry that runs parallel, as increased activity has been reported in the dorsolateral prefrontal area (BA 9/46) during implicit acquisition (Rauch et al., 1997; Badgaiyan, Schacter, & Alpert, 1999; Schacter, Badgaiyan, & Alpert, 1999). This is also suggested by the existence of a procedural-memory-related frontal/basal-ganglia circuit, projecting from the basal ganglia to BA 44, and another projecting to BA 45/47, which sustains the implicit retrieval/selection of conscious lexical knowledge (Ullman, 2006b), as mentioned above in the context of implicit regular and explicit irregular forms. Green and Abutalebi (2008) in fact agrees that natural switching can be implicit: Speakers can automatically switch into their other language in the event of a failure of lexical access in the current language and may also use such retrieval deliberately, in a communicative setting where both parties speak the same languages. In its simplest form, nonce borrowing may have to do with the relative accessibility of a lexical form (given the current speech rate); but if a speaker becomes aware of a word-finding problem, an intentional process can be invoked. In his comment on the patient reported in Mariën et al. (2005), Green and Abutalebi write: “when speaking to monolingual speakers of either language, E.M. showed pathological language switching and mixing. L2 intrusions were common in a task requiring the use of L1 only. … Remission of the symptoms of language mixing and switching was associated with increased perfusion … In this later phase, where switching occurred it was under conscious control …,” (p. 572) thus assuming that mixing was not under conscious control before this phase.

4.  Data from clinical studies 4.1   Data from bilingual neuropsychiatric disorders Age and type of appropriation are often cited as crucial variables in bilingual psycholinguistic and neurolinguistic studies, but it has not always been clear why they should be. In the psychiatric literature of the seventies and eighties – and in more recent publications that still uncritically rely on those previous studies

 Declarative and procedural determinants of second languages

(e.g., Santiago-Rivera & Altarriba, 2002) – on the basis of invalid (and therefore, not surprisingly, contradictory) experimental data, and of preposterous claims (see Paradis, 2006) of a higher incidence of crossed aphasia in bilingual aphasic patients, it was believed that the difference between L1 and L2 was caused by the differential lateralization of the late-learned language system (Bruce, 1895; Hughes, 1981; De Zulueta, 1984; Heinemann & Assion, 1996; De Zulueta, Gene-Cos, & Grachev, 2001). To be sure, as stated by Santiago-Rivera and Altarriba (2002), various structures in the brain play a differential role in the processing of a late bilingual’s two languages, but not, as assumed, because “age of acquisition, dominance, proficiency and so forth seem to play a role in the degree of hemispheric lateralization that occurs with a first and second language” (p. 31). Within the framework of a neurolinguistic theory of bilingualism (Paradis, 2004), age and type of acquisition do indeed bear on the neuropsychological organization of verbal communication, but not for any reason related to laterality. Implicit linguistic competence in each language is subserved by the same neural substrates as those of unilingual speakers of the respective languages. For both languages, metalinguistic knowledge is subserved by a common neural substrate, except that in late-learned languages, at least in the first stages, only metalinguistic knowledge is appropriated, and speakers will continue to rely to a greater extent on this knowledge for verbal communication in L2 until they have acquired implicit linguistic competence for that language. However, L2 implicit competence will only rarely, if ever, reach the same level as in L1.16 Granted, at first, the quantitative difference is considerable, and tasks that are automatically accomplished by implicit linguistic competence (e.g., the production of a sentence) are deliberately conducted under executive control. Such a task involves the processing of prosody, phonology, morphology and syntax, each of which may eventually be replaced by automatized procedures (over time, in the reverse order), and some of which may never be automatized. The differences observed in psychotic conditions and dementias are caused by this increased reliance on declarative-memory-based (and hence consciously controlled) explicit metalinguistic knowledge (Paradis, 2008a). As has been judiciously observed by psychiatrists familiar with bilingual patients, the second language requires more attention and conscious processing than the native one. This would account for the

16.  Note that the difference between L1 and L2 is only quantitative. Even at the beginning of L2 learning, when implicit linguistic competence is nonexistent, L2 learners do not use any mechanism that is not also used by native speakers. One of the two mechanisms, namely metalinguistic knowledge, may be used exclusively until some implicit competence develops, but at no point does a cerebral mechanism specific to speaking a second language emerge.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

general observation of reduced symptoms in bilingual schizophrenics when they use their second language as compared to their first (Paradis, 2008a). It would be compatible with the explanations provided by psychiatrists to the effect that, in the second language, the speaker may be forced to pay more attention to the language process per se, and thus may be disengaged from the affective focus of the message. The observation that patients are less psychotic in L2 when they have inadequate knowledge of their L2 or when it has been learned later (De Zulueta, 1984), while symptoms tend to be equal in both languages when they were acquired concurrently, lends further support to such an interpretation. So does the fact that dominant bilingual patients often appear emotionally detached when speaking their second language (Marcos & Urcuyo, 1979), as they invest more affect – and extra attentional control – in how they say things and less in what they say, showing constant concern with wording, grammatical constructions, and pronunciation. In fact, most psychiatrists implicitly or explicitly recognize the need for greater effort and concentration in processing a second language. Based on their patients’ behavior, they infer that the second language is more effortful and requires more attention than the first. For example, Laski and Taleporos (1977) ascribe their patient’s exclusive use of L1 during the state of toxic reaction to a possible lower level of vigilance – thus implying that greater vigilance is required to process L2, a function indicative of the controlled use of metalinguistic knowledge. Marcos (1976) considers that communicating in the second language requires more elaborate mental work than communicating in the first language: Extra cognitive and attentional demands are placed on the low-proficiency second language speaker. The second language encoding process (involving consciously monitoring syntax and phonology, in addition to vocabulary) is considered to be more intricate. Peck (1973) investigates the relationship of disease and other stress factors with second language: L1 seems less vulnerable to such factors. The significant increase in stress-associated hand movements produced by patients during L2 interviews is interpreted as reflecting second-language encoding efforts (Grand et al., 1977; Pitta, Marcos, & Alpert, 1978). It is therefore reasonable to assume that the language dissociations described in psychiatry, like those described in aphasia, have their neural underpinnings in the differential participation of cerebral systems for languages acquired (from birth) and languages learned (later in life). 4.2  Data from bilingual aphasia Aphasia is the result of damage to the cognitive neurofunctional system that sustains implicit linguistic competence. Hence, a language will be vulnerable to aphasia to the extent that it was acquired incidentally and that it has been automatized. Lesions in the left basal ganglia, portions of the right cerebellum

 Declarative and procedural determinants of second languages

and the perisylvian cortical regions that constitute “the classical language areas” (broadly, Broca’s and Wernicke’s areas) will result in language-specific impairments (i.e., impairments in the grammar, namely phonology, morphology, syntax, and/or the lexicon). In contrast, amnesia is the result of damage to the declarative memory system that sustains all conscious memories, including metalinguistic knowledge (i.e., knowledge of pedagogical grammar rules and other facts about language). Hence, metalinguistic knowledge is vulnerable to amnesia. Lesions in the hippocampal system, including the parahippocampal gyri and the mesial temporal lobes, will result in the loss of previously known metalinguistic material (retrograde amnesia) or the inability to gain new metalinguistic knowledge (anterograde amnesia). Implicit linguistic competence remains available in amnesia, while metalinguistic knowledge may remain available in aphasia. Recently, the differential roles of declarative and procedural memory have become increasingly clear in multilingual language disorders. This difference is relevant to multilingual aphasia recovery patterns, and concerns about bilingual aphasia assessment and treatment. Looking back at cases published over the end of the 19th and early 20th centuries, one finds a number of cases in which the native language is significantly less well recovered than a language that was learned formally later, in which individuals are likely to have considerable metalinguistic knowledge to fall back on (Bychowski, 1919; Goldblum, 1928; Kauders, 1929; Minkowski, 1933; Weisenburg & McBride, 1935; Chlenov, 1948). Some patients who recovered their second language better were literate only in that language or were schooled in their L2, which would guarantee a greater amount of metalinguistic knowledge in that language than in their spoken native dialect (Pitres, 1895; Eskridge, 1896; Minkowski, 1927, 1928; Riese, 1928; Veyrac, 1931; Paradis & Goldblum, 1989). Obviously, reliance on metalinguistic knowledge is not the only possible cause of better recovery of one of a patient’s languages. It is to be expected only when conditions would induce the patient to rely on available declarative knowledge. This will be more likely in cases when (1) the second language was learned formally or (2) the second language was the language of instruction and the patient was not schooled in his or her native language, especially when the native language is a spoken dialect. Conversely, sometimes the native language is the one in which a speaker is likely to have more extensive metalinguistic knowledge because it is the language of instruction, and possibly the only language in which the patient is literate (the second language being a spoken non-standard dialect or having been acquired informally); and may thus be expected to appear to recover better. The outcome will also depend on the degree of internalization of L2, because whatever portion of the language has been internalized will be implicitly represented as a subsystem of implicit linguistic competence that will be vulnerable to aphasia in the same way as the native subsystem.



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

The most convincing clinical evidence so far of procedural acquisition of the native language and explicit learning of a later-learned language, and of their respective cerebral substrates, comes from a case published by Moretti et al. (2001). The patient was a 46-year-old female Croatian-Italian bilingual who exhibited symptoms that, at first sight, looked like antagonistic recovery. The patient, tested one month post-onset, had developed a severe impairment of her native language, while her second language was relatively well preserved (subsequent to an infarct limited to the left caudate nucleus – a section of the striatum, a subcortical structure implicated in implicit memory for language). Four months later, a dramatic improvement in the native language was accompanied by deterioration of the second language (following the extension of the ischemic lesion to the frontal and temporal cortices – associated with declarative memory functions). The authors interpret these phenomena as providing evidence that the native language may be acquired and used through procedures sustained by implicit memory whereas a later-learned second language depends on explicit learning and use of metalinguistic knowledge, which requires the integrity of the declarative memory system. They also suggest that this case demonstrates the different roles of the subcortical and cortical structures that regulate implicit and explicit language functions. Leeman et al. (2007) describe a patient living in the French-speaking part of Switzerland who recovered German, a language he had learned at school, which he had barely mastered and never used actively, sooner and better than French, his native language and the language of his parents, which he spoke at work. After a period of global aphasia, for about two months, before he started recovering some French, most of the patient’s speech was in German. Although the staff on the ward always spoke French to him, he responded in German, even to his wife, who did not speak the language. Even automatic speech was produced in German. Four months later, he was still using German sentences during spontaneous speech and picture naming. The authors conclude that the apparent better recovery of the second language may be explained by his use of his explicit metalinguistic knowledge of the language that he had formally learned. Within the framework of the implicit/explicit perspective, all late-learned languages (L2, L3, Ln) are probably sustained to a large extent by declarative memory. As such, they are more likely to manifest dynamic interference from one another than from the native language(s). Dynamic interference (i.e., during item selection) is made possible by the fact that items in one language subsystem are of the same nature as those in any other subsystem (i.e., parameters of the same principles). Items from other cognitive neurofunctional systems (music, arithmetic, but also metalinguistic items in declarative memory) cannot interfere with them. Goral et al. (2006) offer supportive evidence in their description of the case

 Declarative and procedural determinants of second languages

of a trilingual woman with aphasia who exhibited more interference between L2 and L3 than between either of those languages and her native language, which suggests that cerebral representation and processing are more similar among later-learned languages than between any of them and the native language. This may well be because non-native languages are sustained much more by declarative memory. Implicit linguistic competence is impervious to external interference (though not interference from another language subsystem), whereas declarative memory is known to be susceptible to multiple interference, even across modalities. Additional evidence can be found in a review of the literature on language transfer by Murphy (2003) that revealed that third-language acquisition is characterized in particular by the unintentional incorporation of second-language items during speech production, primarily function words and bound morphology. Aphasic patients typically perform better on grammaticality judgment tasks than on sentence comprehension tasks (Wilson & Saygin, 2004). This might reflect the fact that grammaticality judgments do not necessarily tap implicit linguistic competence (which is compromised in aphasic patients) but metalinguistic knowledge (which is independent). This seems all the more likely given that posterior temporal areas were more reliably implicated in poor grammaticality judgment performance than anterior areas in their experiment, which involved 22 aphasic patients and 26 controls. While, after focused therapy, some aphasic patients are capable of utilizing metalinguistic skills to achieve normal accuracy scores even many years poststroke, there is little prospect of reducing the time taken to return within the normal range (Crerar, 2004). The suggestion is that implicit competence is not regained but the use of metalinguistic knowledge is a good substitute. 4.3  Data from other cerebral accidents/conditions In addition to the converging evidence reported from bilingual psychoses, dementias, aphasias (Paradis, 2008a), and Kasper Hauser syndrome (Lebrun, 2002), many stutterers do not stammer when they use a foreign language with which they are not fully conversant (Lebrun, 1997). Bipolar patients examined while neither elated nor depressed performed worse than control subjects on a measure of declarative memory but did not differ from control subjects on either of the two procedural learning tasks administered, which suggests that the function of temporal lobe structures but not basal ganglia is disturbed (van Gorp et al., 1999). Further indications of the declarative/procedural distinction are found in the following circumstances. Foreign accent syndrome. Avila et al. (2004) report dramatic alteration of the articulation of the native language (Spanish), but no alteration in any of the other



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

three languages (French, English, Catalan) learned after the age of 12, in a patient with foreign accent syndrome. The authors interpret the difference as supporting a differential participation of procedural and declarative memory systems depending on age of acquisition. As the authors suggest, the data serve to demonstrate a different type of motor control for the native language than for other languages learned later. This proposal is further supported by the case of differential speech apraxia described by Alajouanine, Pichot, and Durand (1949). The symptoms of speech dyspraxia were much more severe in French (the patient’s native language) than in English (his second language), even for phonemes and consonant clusters that are virtually identical in the two languages. When the kinesthetic commands came from the automatized French system, the patient’s articulation was extremely deficient, whereas it was perfectly realized when they came from his later-learned English. In these cases, although the articulatory motor effectors are intact, one of the efferent alternative systems at the speech planning phase, the L1 procedural phonological competence (i.e., the source of the impulses to the effectors coming from L1) is impaired, whereas the other, the L2 declarative lead-in efferent process, is intact. This situation differs from cases in which pronunciation (the use of the phonological module) is automatic in the L1, but as long as it remains consciously controlled in L2, it must rely on neural mechanisms that differ from those used in L1. The final step in production, the implementation of the articulatory movements, may be the same (e.g., Frenck-Mestre et al., 2005), but the afferent planning (the lead-in cognitive process17) is different (implicit competence for L1, controlled declarative knowledge for L2) and hence involves different neural structures. Progressive dementias. In Alzheimer’s, Parkinson’s and Huntington’s diseases, the selective or greater impairment is caused by damage to either declarative memory, which affects later-learned languages more (Alzheimer’s), or procedural memory, which affects predominantly the native language(s) (Parkinson’s, Huntington’s). In Alzheimer’s dementia, where declarative memory is known to be more impaired, the second language is consequently more vulnerable. Meguro et al. (2003) report that their Japanese-Portuguese bilingual patients with Alzheimer’s disease had great difficulty with kanji in Japanese and with irregularly spelled words in Portuguese, whereas they were able to read Japanese kana normally and regularly spelled Portuguese words nearly normally. Unlike kana script and regular Portuguese spelling, kanji and irregular Portuguese words have no

17.  To use a phrase coined by Indefrey and Levelt (2004) to refer to a process that takes place before another, here, the implicit or explicit cognitive process that sends activation signals to implementing devices (articulatory and phonatory movements), the source of the impulses for the execution of a task.

 Declarative and procedural determinants of second languages

one-to-one correspondence between grapheme and sound and must be memorized. In other words, for these patients, the items that rely on declarative memory were selectively or considerably more impaired than the items that are derivable by internalized rules. L1 aphasia subsequent to anesthesia. Ward and Marshall (1999) report the case of a patient who exhibited a paradoxical fixation on a non-native language following anesthesia. It has been reported that anesthesia with isoflurane/oxygen spares implicit memory (Kihlstrom et al., 1990) whereas anesthesia with sufentanil/nitrous oxide does not (Cork, Kihlstrom & Hameroff, 1992; Cork, Kihlstrom & Schacter, 1992). Here, oxygen, nitrous oxide and isoflurane were used. It could be that the combination of nitrous oxide and isoflurane not only does not spare implicit memory for the duration of the anesthesia but actually continues to inhibit implicit memory for some time after the end of the anesthesia. The result is a transient aphasia for the native language (English) but spared declarative memory, which allowed the retrieval of some metalinguistic knowledge of the patient’s second language learned in school 40 years earlier. The phenomenon could also have been caused by the patient’s hypoglycemia diagnosed at the time, since he recovered the use of English as soon as his blood sugar level was brought back to normal. Any biochemical imbalance may perturb one or the other system differentially (Ward & Marshall, 1999). Specific language impairment (SLI). In SLI, the implicit linguistic competence of both languages is necessarily affected since the condition is the result of an altered gene that shapes the development of the language system. In fact, extensive assessment of Arabic-Swedish bilingual children has revealed that both languages are affected in individuals with SLI (Håkansson, Salameh, & Nettelbladt, 2003; Salameh, 2006). Each language of the bilingual SLI individual develops in the same way it develops in unilingual SLI individuals: Each language is treated in the same way a second language is treated by normal speakers, that is, involving declarative memory to a greater extent than for the native language(s). When it is said that individuals with genetic dysphasia use declarative memory to learn their first language the way second languages are usually learned, the similarity stops there: Both groups learn the language consciously. This does not mean that individuals with genetic dysphasia will follow the same developmental patterns in learning English as a native language as do learners of English as a second language. They obviously will not, for L2 learners of English already speak another language, some aspects of which they are conscious of and will apply (often incorrectly) to their incipient (declarative) L2 grammar. Native learners with SLI do not have that source of possible interference. In children with genetic dysphasia (SLI), because the developmental disorder is caused by a defective gene affecting the language function as a whole, both



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

languages are expected to be impaired to the same extent, with deficit manifestations modulated by the structure of each one. Hirschman (2000) reports that children with SLI who had received about 55 half-hour sessions of metalinguistic training spread over 12 months significantly improved, at both the written and oral levels, compared to SLI control groups. Those children who had the poorest complex sentence usage tended to benefit the most from metalinguistic training. The author interprets these results as supporting the hypothesis that metalinguistic training helps to overcome the presumed neurological deficit of the language disordered child by making linguistic rules conscious. It is further evidence that the deficit in SLI consists in impaired implicit linguistic competence, which children overcome by learning their L1 grammar consciously (as hypothesized by Paradis & Gopnik, 1997). Fukuda et al. (2007) report that Japanese SLI individuals treat scrambled sentences as if they were in the canonical word order, assigning nominative case (ga) to the first noun and object case to the second noun, reflecting a conscious strategy. Object case assignment also seems to follow a non-grammatical principle (i.e., animacy/inanimacy). In Japanese, there is a tendency for an inanimate object to appear with the dative case (-ni) and for an animate object to appear with the accusative (-o). Individuals with SLI assign the dative case (-ni) to animate objects, irrespective of the type of verb, including those that would require the accusative (-o). Similarly, Marini, Tavano, & Fabbro (2008) report that, in an experiment involving 62 Italian children diagnosed with SLI, compared to those provided by a group of 195 children with typical language development, produced an amount of words comparable to that produced by the control group, but with numerous omissions and/or substitutions of bound and free morphemes and impaired syntax. These recent data confirm previous findings from various languages (Kehayia, 1997; Dalalakis, 1999) indicating that individuals with SLI do not have access to implicit linguistic competence but adopt explicit strategies.

5.  The declarative/procedural distinction and the subsystems hypothesis Each automatized language is represented as a subsystem of the neurofunctional language system. In a late-learned second language, only those portions of the language that have been acquired will constitute a subsystem; what has not yet been internalized will remain part of metalinguistic knowledge and will hence be subserved by declarative memory. The subsystems hypothesis thus accounts for the organization of implicit linguistic competence, hence of language in early bilinguals. Of the four conceivable hypotheses about the representation of languages in the bilingual brain, only one, the subsystems hypothesis, was deemed acceptable

 Declarative and procedural determinants of second languages

because it is the only one compatible with all known aphasia recovery patterns and other bilingual phenomena. The other three were not considered valid. It is therefore not the case that, in accordance with this theory, some individuals would have their languages organized as an extended system, others as a dual system, and others still as a tripartite system, as Obler et al. (2007) seem to suggest. Each of the discarded hypotheses is compatible with some recovery pattern, but not all.18 This, in fact, is why they were not further considered, and the subsystems hypothesis was, and still is, considered our best working hypothesis. Obler et al. (2007) write: “Clearly, Paradis’ categories do not yield mutually exclusive theories of which only one can hold. Rather, it would appear that different patterns of recovery from aphasia … reflect differences in the way the two or more languages were acquired and maintained, resulting in slightly differing brain systems underlying their pre-morbid processing” (p. 23). First, from the beginning, three of the categories were considered to be at odds with the reported data, and thus only the subsystems hypothesis was maintained. As it is compatible with all the reported data, the other three were unnecessary as well as being consonant with only some of the data. In addition, it is most unlikely that the implicit linguistic competence neurofunctional system should be represented in radically different ways in different individuals.19 In the proposed theory, differences in the way the two or more languages were acquired and maintained would not affect the way their implicit competence neurofunctional systems are represented. In early bilinguals (and possibly some late bilinguals, to the extent that they have internalized their second language as implicit linguistic competence), languages are represented as implicit subsystems. Late bilinguals, to the extent that they do not possess implicit linguistic competence in their L2, compensate by using explicit metalinguistic knowledge and pragmatics, which rely on neural substrates other than the ones that subserve implicit competence. To this extent, indeed, they differ from early bilinguals, if only quantitatively. The differences in how the two or more languages were acquired and maintained determine the extent of reliance on implicit linguistic competence, explicit

18.  Note that all patterns were considered at the same time (Paradis, 1981) and not two (extended and dual) in 1987, and another two (tripartite and subsystems) in 2004; as Obler et al. (2007: 21–22) seem to imply. 19.  The once-considered differential representation of the languages of bilingual speakers depending on their context of acquisition (Lambert & Fillenbaum, 1959) has not been confirmed and was not supported by the authors’ own data (published in Paradis, 1983: 626–635), as shown in Paradis (1977: 100–101 and 111).



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

metalinguistic knowledge and pragmatics, but not how implicit competence is organized. The proposed neurolinguistic theory of bilingualism does not conceive of individuals organizing their implicit competence in ways other than as a subsystem, for reasons given from the start (Paradis, 1981) and subsequently (Paradis, 1987, 2004), irrespective of the contexts of acquisition and use. Different contexts may lead to differences in the contents of representations, but not in the way that these implicit representations are stored and processed (Paradis, 2004, 2007b). Contexts of acquisition and use, in particular age of acquisition (and its consequences), will determine whether a language is incidentally acquired or explicitly learned. Only to the extent that it is learned, and hence sustained by declarative memory, will the representation and processing (quantitatively) differ from those of early bilinguals, by involving hippocampal and anterior cingulate structures to a greater degree. Nor does the subsystems hypothesis postulate “1. a larger set of connections with features of both languages; and 2. two subsets of connections, one for each language” (Walters, 2005: 75). This would refer to the tripartite hypothesis, which was rejected on the same grounds as the extended system, namely because, if whatever features the two languages have in common were represented only once, in a system common to both, the model would be incompatible with nonparallel recovery patterns. One would have to explain why the features from the common system are available when speaking one language, but not the other. There is indeed one large system, the language neurofunctional system, of which each language is a subsystem. The language system is divided into as many subsystems as there are languages spoken by the individual. But the language system is made up of these subsystems, and does not constitute a third, independent system. In other words, the language system is but the sum of its subsystems. The two sets of connections constitute the language system. In the three-store hypothesis, the third store, common to all languages, is not part of the language system but of the conceptual system.20 Given that each neurofunctional language subsystem is

20.  Words, represented in the vocabulary, are part of declarative memory and lexical items, represented independently and to some extent redundantly, are part of the lexicon, which is incorporated in the implicit linguistic competence of each language subsystem. According to the Three-Store Hypothesis (Paradis, 2004), lexical items are implicitly connected to conceptual representations that correspond to their semantic constraints. Conceptual representations result from the implicit grouping together of conceptual features. The conceptual representations resulting from this implicit process are conscious in the way that an utterance resulting from implicit procedural computations is consciously observable. Words from the declarative vocabulary independently access their corresponding conceptual representations.

 Declarative and procedural determinants of second languages

independent of the other, as each can be selectively inhibited, there is no portion common to two or more subsystems at the level of any of the constituent modules (phonology, morphology, or syntax). Items whose structure is identical in two languages are incorporated separately into each subsystem. They are integrated as an intrinsic part of each subsystem and are therefore available in language A when language B, which contains its clone, is inhibited, and vice versa.

6.  Declarative and procedural translation strategies Speakers have two translation strategies at their disposal (Paradis, 1994b). In Strategy 1, the naive strategy, probably adopted by occasional interpreters who are used to speaking one language to one group of speakers and the other to another group, translation is accomplished via the conceptual system, according to the normal process of implicit linguistic decoding (comprehension) of the source language material followed by encoding (production) of the target language material (Figure 5.1). Professional simultaneous interpreters, in the course of their training, gain extensive metalinguistic knowledge in the form of learned translation equivalents and can thus use Strategy 2. In this scenario, translation is carried out directly via (learned) association links between the lexicons (or any level of linguistic structure, including syntactic constructions) without going through the encoding/decoding route. It relies on declarative memory, being the result of consciously learned associations between L1 forms and L2 forms, L1 surface structures and L2 surface structures (e.g., use “have been V–ing” to translate French “present tense + depuis” (literally: since) and vice versa). Ruiz, Paredes, Macizo, and Bajo (2008), for example, deduce from the results of their experiments that the horizontal view of translation includes code-to-code links between the source language and the target language involving at least the lexical and syntactic level of processing. De Groot and Christoffels (2006) investigate bilingual control in translation and simultaneous interpreting. With good reason, they stress the importance of good comprehension of the source language. They point out that, in simultaneous interpreting, comprehension of the source language and production of the target language take place simultaneously; hence, both languages must be activated at the same time, though not to the same extent. When using Strategy 1, because comprehension can tolerate a higher activation threshold (greater inhibition) than production, the across-the-board raising of the source language’s activation threshold (increasing inhibition) prevents the production of source language elements but does not compromise good comprehension. Differential activation of the two languages thus does not lead to sub-optimal comprehension



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

(of the source language), while it provides optimal production (of the target language). Under these circumstances, translation of discourse from the stronger into the weaker language may yield the best outcome:  Comprehension is guaranteed and the target text may be adapted to the translator’s proficiency by using circumlocutions and other devices, when necessary. The meaning (of the source language), correctly understood, may be accurately rendered in the weaker language, in spite of its possibly somewhat compromised morphosyntactic form. Conversely, word translation may be harder from the stronger language to the weaker language, because there are no alternatives to the translation equivalent (which in discourse can be dodged). (a-linguistic) cognitive system ↓ ↑ SL lexical semantics TL lexical semantics ↓ ↑ SL syntax TL syntax ↓ ↑ SL morphology TL morphology ↓ ↑ SL phonology TL phonology ↓ ↑ SL utterance TL utterance Strategy 1 (automatic) Linguistic decoding of SL until the message is understood (awareness of the meaning), followed by the linguistic encoding of the message.

(a-linguistic) cognitive system SL lexical semantics → TL lexical semantics SL syntax → TL syntax SL morphology → TL morphology SL phonology → TL phonology ↓ ↑ SL utterance TL utterance Strategy 2 (conscious) Direct transcoding by automatic application of rules, from one linguistic element in SL to its structural equivalent in TL (e.g.,) morphological level: English -ical → French -ique; syntactic level: without + gerund → sans + INF Lexical level: décevoir → disappoint.

Figure 5.1.  Two translation strategies (after Paradis, 1994b: 62)

In simultaneous interpretation, (1) the source language utterance must be decoded (using implicit linguistic competence, hence procedural memory) in order to (2) derive the message (conscious information, hence in declarative memory); (3) store the message in short-term memory and hold it there until it has been acceptably translated (because the interpreter must ascertain that what was translated into the target language corresponds to the facts of the source message – a conscious operation, which relies on declarative memory); (4) convey in the target language the message decoded from the source language, namely encode (using linguistic competence, if one has automatized the target language; using declarative memory for those aspects that have not been automatized, if one is not

 Declarative and procedural determinants of second languages

a native speaker of both languages); (5) check that what was rendered in the target language matches the message of the source language (retrieving the information from short-term memory). Note that while the interpreter processes tasks (2) to (5), the source language speaker continues to produce utterances that must be added to what is already stored in short-term memory. This multitasking taxes working memory but also the declarative knowledge of the target language itself if it is not the interpreter’s native language. Speaking (and understanding) one’s native language should require minimal effort. Most cognitive effort is devoted to keeping information in short-term memory. On the other hand, many professional interpreters do not use the implicit route from decoding to message to recoding (Strategy 1), but use a declarative strategy throughout (Strategy 2), not relying on the meaning as much as on the form (see de Groot & Christoffels, 2006). Whereas declarative memory does not interfere with procedural memory when the two are used concurrently, if the interpreter is a native bilingual, then both language subsystems may to some extent interfere with each other – a problem to be avoided, and thus the interpreter must carefully (consciously) monitor her speech (i.e., mentally check the output of her automatic language processing before voicing it). Marshall et al. (2005) reason that only Strategy 2 can allow for dissociations between naming, translation, and comprehension, because in Strategy 1, translation abilities should not exceed naming. (The authors limit their discussion to the lexicon, referring to Gollan and Kroll’s, 2001, Model 1 and Model 2, but the reasoning is the same as with the extension of the model to the entire language system.) Price, Green, and Von Studnitz (1999) found that translation correlated with activation in cerebral structures associated with control mechanisms. This is interpreted by Marshall et al. (2005) as evidence that links that bypass semantic processing can be used when translating between two languages. This corresponds to Strategy 2, a controlled route. Translation may be available when word retrieval is not (Paradis, Goldblum, & Abidi, 1982; Lalor & Kirsner, 2001; Detry, Pillon, & de Partz, 2005), suggesting that words can be accessed from the vocabulary and, in some cases, sentences can be accessed via declarative Route 2 by aphasic patients.

7.  Further indications of declarative/procedural relevance Somnopolygraphical experiments have shown dissociated specific effects of early (Slow Wave Sleep) and late (Rapid Eye Movement) sleep on declarative and procedural memory in humans. Recall of paired-associate lists improved more during early sleep, and recall of mirror-tracing skills improved more during late



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

sleep (Plihal & Born, 1997). Loss of sleep affects implicit learning but not verbal report; degree of mental fatigue affects declarative memory but not implicit learning (Heuer et al., 1998). Acquisition has been shown to yield not only activation in the caudate nucleus – a structure known to be involved in procedural memory function – but also deactivation in the medial temporal lobe – a structure known to be involved in declarative memory function (Poldrack et al., 1999, 2001; Ullman et al., 2005). The interaction generally reported between procedural and declarative memory is one of mutual inhibition/disinhibition – one system inhibits the other when stimulated:  “learning in one system depresses functionality of the other” and “a dysfunction of one system leads to enhanced learning in the other” (Ullman, 2004: 243). In other words, the interaction is one of switching from using one memory system to using the other, not one of interference of one system in the internal structure of the other. 7.1  Variability in appropriation in L2 vs. systematicity in L1 Implicit cognitive functions show less individual variation than explicit cognitive functions and are less concordant with IQ measures (Mayberry et al., 1995; McGeorge, Crawford, & Kelly, 1997; Reber, Walkenfeld, & Hernstadt, 1991; Reber, 1997; Robinson, 2002) and are dramatically less disrupted under pressure (Masters, 1992; Reber, 1976). When, in an experiment, some students in a classroom appropriate a particular rule without instruction (e.g., the placement of manner and frequency adverbs in English) and others do not, it would seem that some individuals are better at it than others, which is never the case in L1 acquisition. It would thus appear that some students are able to notice that specific adverbs must be placed before the verb while others may be placed after it. Given that there is considerable variability in the ability to learn a second language (and that IQ is a factor in learning a language but not in acquiring it), it is not at all unlikely that some students will notice the difference in placement, while others will not. This might eventually help them acquire the rule by providing them with opportunities for more frequent use of the correct form. The fact that students do not all appropriate the same constructions at the same time increases the variability in individuals’ L2 performance.

 2 accent changes faster than L1 accent when speakers relocate 7.2  L to an area where a different variety is spoken An L2 accent will be modified by exposure to a new environment where a different accent is used in a way that an L1 accent will not. The accent of English as an L2 learned in England by a French speaker may change to an American accent over

 Declarative and procedural determinants of second languages

the years in ways that this speaker’s native French will not, after emigration to Quebec, Belgium or Switzerland. If the L1 accent changes, it will take much more time (i.e., after very long exposure, some new pronunciations may be acquired). This suggests that pronunciation of the native language is acquired (and automatized) whereas a second language with adult onset is learned (and controlled), in accordance with the general principle that what is implicit and automatized is more resistant to change than what is declarative and controlled.

 dditional evidence for L1 implicit procedural memory and L2 7.3  A explicit declarative memory Implicit memory is vulnerable to aphasia and Parkinson’s disease. Individuals with Parkinson’s disease exhibit impairment of syntactic integration processes (Friederici et al., 2003). Bilingual Parkinsonian patients have been shown to exhibit greater syntactic impairments in their native language (Zanini et al., 2004). Non-demented people with Parkinson’s disease experience difficulties using the morphosyntactic aspects of language. In contrast to controls, a Parkinson’s disease group failed to experience priming effects for constructions that crossed phrasal boundaries (Arnott et al., 2005). In an analysis of the responses to a web questionnaire on emotion and bilingualism (Dewaele & Pavlenko, 2001–2003), Dewaele (2007) reports stable, significantly higher levels of anxiety among the participants in the use of their second language than of the first. From the tenor of the comments in the open-ended questions, he infers that the feelings of tension and apprehension specifically associated with second language speaking suggest that production is not proceduralized, in that L2 speakers rely on explicit, declarative knowledge rather than implicit linguistic competence. For more experimental evidence of dissociable implicit and explicit verbal memory systems, see Ellis’s (1994a) review. 8.  Conclusion Single words fundamentally differ from the rest of language. They constitute the (conscious) vocabulary, which contains only the words’ form-meaning relationships. The vocabulary is to be distinguished from the lexicon, which includes intrinsic morphosyntactic properties and is integrated within the neural network of each specific language subsystem. The vocabulary is sustained by explicit memory, the lexicon by implicit memory. This has a considerable impact on the interpretation of language laterality and neuroimaging studies. Results from studies using single words as stimuli can at best only speak to the representation and processing of vocabulary, not of language as implicit linguistic competence. Such



Chapter 5.  The pervasive relevance of the implicit/explicit distinction 

studies address a component of language that differs radically from the neurofunctional language system. Even if there were highly significant differences between particular subgroups of bilinguals in laterality measures using dichotic, dihaptic, or visual half-field procedures, one would have to demonstrate that the differences reflect language lateralization and not something else (e.g., differences in attention allocation, response strategies, etc.; cf. Segalowitz, 1986; Segalowitz & Bryden, 1983). But even then, the results would at best only speak to the representation of the vocabulary, not the lexicon, let alone the language system. Neuroimaging studies using single words as stimuli (irrespective of the task) find common areas of activation for both languages (though not necessarily the same areas in different studies). Those that use sentences or short stories find, at the gross anatomical level, areas of overlap (corresponding to implicit linguistic competence in both languages, to the extent that it has been acquired in L2) and separate areas (corresponding to the increased use of metalinguistic knowledge and pragmatics to compensate for lacunae in implicit linguistic competence). Both languages rely on the same types of processes:  Whatever has been acquired incidentally is processed by implicit computational devices (that include parts of the right cerebellum, left basal ganglia, and circumscribed left perisylvian cortical areas) and what is learned and controlled is represented in declarative memory (involving the hippocampal system and large cortical zones bilaterally) and controlled by conscious executive functions (implicating the anterior cingulate and prefrontal cortices). The differences between the processing of L1 and L2 are therefore only quantitative. Both languages rely on the same set of cerebral mechanisms, though to different degrees on each. However, even though the implicit linguistic competence of both languages is subserved by neural substrates in the same gross anatomical areas, each language is sustained by discrete neural substrates at the micro-anatomical level. The discrete neural circuitry of the two languages may be intertwined but each remains distinct, hence isolable and capable of being selectively inhibited. To the extent that additional effort is required to process the grammar system of L2, that grammar cannot have the same nature as the L1 grammar. Native implicit competence is processed swiftly and without conscious control, hence without effort. Conscious effort cannot be applied to an automatic process. Therefore, if effort must be devoted to the processing of L2 grammar, that grammar cannot be represented as implicit competence. Any additional cognitive resources will have to apply to the processing of what the speaker uses to compensate for the lack of automatic competence, namely explicit metalinguistic knowledge and pragmatics. The extent of L2 cortical representation that overlaps with L1 can only correspond to the portion of L2 grammar that has been automatized.

 Declarative and procedural determinants of second languages

There are two types of language switching and mixing: deliberate and automatic. Switching on cue is a deliberate task. Most neuroimaging studies of language switching use some form of explicit cuing, leading to conscious control of the participant’s performance, and hence to the activation of cerebral mechanisms involved in conscious control of action. Contrastingly, natural mixing and switching in bilingual environments is automatic and sustained by implicit linguistic competence. In bilingual aphasic patients, switching may be caused by word-finding difficulty, in which case items from the other language are either automatically introduced in the discourse or deliberately selected as a conscious compensatory strategy. Pathological mixing is a dysfunction of the implicit language system which causes patients to mix and/or switch compulsively. When pragmatic ability is compromised, patients may switch inappropriately and speak a language not understood by their interlocutor. There are also two types of simultaneous translation strategies: (1) the use of implicit linguistic competence to decode the source utterance and apprehend its message – the implicit way, and (2) the application of consciously learned morphosyntactic equivalence rules for translating source language chunks directly into target language chunks, bypassing the overall meaning decoding phase – the explicit way. Most psychiatrists recognize the need for greater effort and concentration in processing a second language. The differences observed in psychotic conditions are caused by increased reliance on declarative-memory-based (and hence consciously controlled) explicit metalinguistic knowledge. Patients are generally more detached from their second than their native language. Because they must pay more attention to the morphosyntactic and phonological form at the expense of the meaning and its associated affect, psychiatrists can take advantage of the different systems and switch from one to the other depending on whether they wish to increase or decrease anxiety. Other indications of declarative/procedural relevance are found in various activities: dissociations of specific effects of early (Slow Wave) and late (Rapid Eye Movement) sleep; variability in L2 contrasting with systematicity in L1 appropriation; faster accent changes in L2 than in L1 when relocating to a different area; and declarative and procedural translation strategies, to name a few. It is now clear that the differential impact of declarative and procedural memory on the acquisition and processing of languages cannot be ignored in any type of experimental study of bilingualism.

Summary of key proposals Acquisition and learning (as defined) and their respective outcomes (competence and knowledge) are qualitatively different. They are sustained by different types of memory systems (procedural and declarative), which are themselves subserved by different neural substrates (cerebellum, striatum, other basal ganglia, and perisylvian cortical areas versus hippocampal system, parahippocampal gyri, mesial temporal cortex, and anterior cingulate). The intake is not the input – not just in the sense that only a portion of input serves as intake, but that the intake is not any part of the perceived input. It is the implicit abstraction of non-perceived (because non-observable) underlying structures. If acquisition is the establishment of a system of weighted connections that are governed by a statistical probability matrix induced by tallying the frequency of occurrence of any combination of items, then the intake is doubly implicit: Both what is tallied (the intake) and the tallying process itself (the acquisition device) are implicit. There are two sets of representations, those that are capable of reaching consciousness and those that are not; implicit grammar is inherently not capable of reaching awareness. Most representations are latent (temporarily nonconscious). They enter consciousness when selected (in focused attention). But only representations that are inherently capable of reaching consciousness can ever enter awareness. Therefore, implicit linguistic competence and metalinguistic competence cannot interface “in consciousness,” which is outside the reach of implicit competence. Consciousness cannot be the interface. In fact, there is no possible interface because (1) implicit linguistic competence and metalinguistic knowledge are different in nature (the formulation of an explicit rule is not what is computed by an implicit procedure); (2) one does not serve as input to the other; and (3) one is not translated (converted or transformed) into the other. Some components of a verbal communication act are conscious, whereas others are not. None is partially conscious. Representations may have their activation raised, but as long as their threshold is not reached, they remain nonconscious. According to the works cited in support of an interface by N.C. Ellis (2005, 2006, 2008), Ellis & Larsen-Freeman (2006) and Roehr (2008a), none of the neural messages transmitted from cell to cell (whether or not in synchrony), whose nonconscious contributions could result in the activation of conscious representations, could possibly influence the structure of encapsulated skills such as implicit linguistic competence. Automatic systems such as implicit linguistic competence

 Declarative and procedural determinants of second languages

are explicitly excluded from Baars’s global workspace, Dehaene and colleagues’ neuronal workspace and Scott Kelso’s synchronized coordination. Executive control processes (the intention to do X) set the automatic implicit computational procedures in motion but do not control their internal operations. The speaker is aware of the desired outcome, then of the actual outcome, but not of the automatic procedures that served to achieve that outcome. The implicit grammar cannot be modified (either facilitated or interfered with) by anything implicit or explicit external to it. In contrast, any conscious perception (irrespective of modality) may facilitate or interfere with any other conscious task. Research on the representation and processing of single words is a perfectly legitimate scientific endeavor, precisely because words differ from the rest of language (i.e., from implicit linguistic competence): They are sustained by the declarative memory system. For that very reason, the results of studies bearing on single words cannot be compared to those that use natural language tasks such as comprehension and/or production of sentences (preferably in context, so as to tap the pragmatic dimension as well); nor can they be generalized to the representation and processing of language. They speak to words in isolation, and to words only. The lexicon, on the other hand, with its implicit grammatical features, is part of the linguistic competence that constitutes each language subsystem of a bilingual speaker. One important neurobiological constraint that Parallel Distributed Processing networks have not, as yet, incorporated, and that should not be ignored, is the distinct roles of the hippocampal system and the cerebellum-basal ganglia circuit in subserving the declarative and procedural systems, respectively. Neurobiologically, implicit linguistic competence and metalinguistic knowledge involve different types of representation and are substantiated in separate parts of the brain. This means that we cannot assume that, because connectionist computer networks can produce two phenomena with only one system, this is the way the human brain actually does it. There are two types of language switching, automatic (in on-line interaction with other bilinguals) and controlled (when the speaker consciously decides to switch or is required to do so on cue in an experiment). The mechanism and cerebral processes that sustain them are different in nature. The language system (as implicit linguistic competence) is not involved when speakers name pictures or numerals in one or the other language in accordance with a provided cue (be it a colored background or the name of the language). The brain mechanisms allowing the on-line selection of a target language in the course of a normal conversation among bilinguals differ from those involved in general executive functions that control the switching between any two conscious choices, including the conscious selection of a particular



Summary of key proposals 

language. Natural switching is an implicit process governed by the neurofunctional system that processes the grammar; switching on demand is an explicit task and depends on conscious executive function (implicating the prefrontal and cingulate cortices). The expression of human foxp2 is likely to be the cause of an optimal period for incidental language acquisition without focused attention on form. Either a mutated foxp2 or late exposure to language interaction interferes with the ability to acquire a language and requires individuals to learn the language, which results in variability in ultimate achievement (caused by factors not relevant for language acquisition). The explanations proposed in lieu of a critical period to account for late learners’ failure to appropriate a second language in the same way that children do are actually based on processes necessitated by the expiry of the optimal period for language acquisition. Researchers implicitly or explicitly recognize that speakers need to make more of an effort and concentrate more to process a second language. The greater demands imposed on the processing of L2 grammar (inferred from the reported increased volume of activation in neuroimaging studies) involve cerebral mechanisms other than those that sustain implicit linguistic competence. Difficulty, effort, variability, and increased activation, all of which are associated with the appropriation and use of a second language, are characteristic of the employment of declarative memory. Every child without a severe mental deficiency acquires a native language, given the opportunity of verbal interaction in a speech community. Not everyone manages to acquire a second language. Much evidence points to an optimal period in early childhood during which acquisition takes place incidentally, and after which acquisition becomes more difficult with increasing age. The turning point coincides with the cognitive developmental stage when individuals begin to use conscious learning strategies. They start paying attention to the form of language input and, consequently, consciously learn various aspects of the second language. The optimal period is restricted to the acquisition of implicit linguistic competence, and does not affect language learning. The use of declarative memory to learn a second language leads to interindividual variability in ultimate attainment, resulting from differences in working memory capacity, level of education, IQ, motivation, and other factors that do not affect first language acquisition. Neither speed nor accuracy are necessarily indicative of automatized implicit linguistic competence: The hallmark of automaticity is systematic output. Some individuals are able to reach high levels of proficiency through the efficient use of declarative knowledge, resulting in speeded-up control of their

 Declarative and procedural determinants of second languages

output. Once a language has been learned, if it is extensively used to interact in a community of L2 speakers (preferably with few opportunities for using L1), portions of the language may eventually be automatized (possibly at different levels in various components of language structure: pronunciation and morpho-syntax, including the grammatical features of the lexicon). This does not correspond to the automatization of metalinguistic knowledge but to the acquisition of an independent system, namely implicit linguistic competence. There are thus two ways to fluency: Speeded-up processing and automatization.

References Abutalebi, J. 2008. Neural aspects of second language representation and language control. Acta Psychologica, 128: 466–478. Abutalebi, J., Annoni, J.M., Zimine, I., Pegna, A.J., Seghier, M.I., Lee-Jahnke, H., Lazeyras, F., Cappa, S.F., & Khateb, A. 2007a. Language control and lexical competition in bilinguals: An event-related fMRI study. Cerebral Cortex (in press). Abutalebi, J., Brambati, S.M., Annoni, J.M., Moro, A., Cappa, S.F., & Perani, D. 2007b. The neural cost of the auditory perception of language switches: An event-related fMRI study in bilinguals. Journal of Neuroscience, 27: 13762–13769. Abutalebi, J., Miozzo, A., & Cappa, S.F. 2000. Do subcortical structures control “language selection” in polyglots? Evidence from pathological language mixing. Neurocase, 6: 51–56. Aladdin, Y., Snyder, T., & Nizam, A.S. 2008. Selective postictal aphasia–Cerebral language organization in bilingual patients. Neurology, 71 (7): e14–e17. Alajouanine, T., Pichot, P., & Durand, M. 1949. Dissociation des altérations phonétiques avec conservation relative de la langue la plus ancienne dans un cas d’anarthrie pure chez un sujet français bilingue. L’Encéphale, 28: 245–265. Anderson, J.M. 1973. Structural aspects of language change. London: Longman. Anderson, J.R. 1983. The architecture of cognition. Cambridge, MA: Harvard University Press. Anderson, J.R. 1992. Automaticity and the ACT-super (*) theory. American Journal of Psychology, 105: 165–180. Anderson, J.R. 1996. ACT:  A simple theory of complex cognition. American Psychologist, 51: 355–365. Ansaldo, A.I., & Marcotte, K. 2007. Language switching in the context of Spanish-English bilingual aphasia. In J.G. Centeno, R.T. Anderson, & L. Obler (Eds), Communication disorders in Spanish speakers (pp. 214–230). Clevedon, UK: Multilingual Matters. April, R., & Han, M. 1980. Crossed aphasia in a right-handed bilingual Chinese man. Archives of Neurology, 37: 342–346. Arabski, J. 2006. Language transfer in language learning and language contact. In J.  Arabski (Ed.), Cross-linguistic influences in the second language lexicon (pp. 12–21) Clevedon, UK: Multilingual Matters. Arabski, J. 2007. General trends in language transfer studies. In J. Arabski (Ed.), Challenging tasks for psycholinguistics in the new century. Katowice, Poland: University of Silesia Press. Arnott, W.L., Chenery, H.J., Murdoch, B.E., & Silburn, P.A. 2005. Morphosyntactic and syntactic priming: An investigation of underlying processing mechanisms and the effects of Parkinson’s disease. Journal of Neurolinguistics, 18: 1–28. Avila, C., González, J., Parcet, M.A., & Belloch, V. 2004. Selective alteration of native, but not second language articulation in a patient with foreign accent syndrome. NeuroReport, 15: 2267–2270. Baars, B.J. 1988. A cognitive theory of consciousness. Cambridge: Cambridge University Press. Baars, B.J. 1997. In the theatre of consciousness:  The workplace of the mind. Oxford:  Oxford University Press.

 Declarative and procedural determinants of second languages Baars, B.J. 2003. Working memory requires conscious processes, not vice versa. A Global Workspace account. In N.  Osaka (Ed.), Neural basis of consciousness (pp. 11–26). Amsterdam: John Benjamins. Baars, B.J. 2004. The evidence is overwhelming for an observing self in the brain. Response to Thomas Clark, “Is there an observing self?” Science and Consciousness Review, 15 February. http://sci-con.org/2004/02/the-evidence-is-overwhelming-for-an-observingself-in-the-brain/ Baars, B.J. 2005. Global Workspace theory of consciousness: Toward a cognitive neuroscience of human experience. In S. Laureys (Ed.), The boundaries of consciousness: Neurobiology and neuropathology (pp. 45–54). Amsterdam: Elsevier. Baars, B.J., & Franklin, S. 2003. How conscious experience and working memory interact. Trends in Cognitive Sciences, 7: 166–172. Baddeley, A.D. 1986. Working memory. Oxford: Oxford University Press. Baddeley, A.D. 1996. Exploring the central executive. Quarterly Journal of Experimental Psychology, A, 49: 5–28. Badgaiyan, R.D. 2000. Executive control, willed actions, and nonconscious processing. Human Brain Mapping, 9: 38–41. Badgaiyan, R.D., Schacter, D.L., & Alpert, N.M. 1999. Auditory priming within and across modalities: Evidence from positron emission tomography. Journal of Cognitive Neuroscience, 11: 337–348. Baerman, M., Brown, D., & Corbett, G.G. 2005. The syntax-morphology interface. A study of syncretism. Cambridge: Cambridge University Press. Bartsch, R. 2002. Consciousness emerging: The dynamics of perception, imagination, action, memory, thought, and language. Amsterdam: John Benjamins. Baslow, M.H., & Guilfoyle, D.N. 2007. Using proton magnetic resonance imaging and spectroscopy to understand brain “activation.” Brain and Language, 102: 153–164. Bello, L., Acerbi, F., Giussani, C., Baratta, P.  Taccone, P., Songa, V., Fava, M., Stocchetti, N., Papagno, C., & Gaini, S.M. 2006. Intraoperative language localization in multilingual patients with gliomas. Neurosurgery, 59: 115–123. Beretta, A. 2006. A neurolinguistic theory of bilingualism. Studies in Second Language Acquisition, 28: 525–527. Berko-Gleason, J. 1958. The child’s learning of English morphology, Word, 14: 150–177. Berthier, M., Starkstein, S., Lylyk, P. & Leiguarda, R. 1990. Differential recovery of languages in a bilingual patient:  A case study using selective amytal test. Brain and Language, 38: 449–453. Bialystok, E. 1997. The structure of age: In search of barriers to second language acquisition. Second Language Research, 13: 116–137. Birdsong, D. 2005. Interpreting age effects in second language acquisition. In J.F. Kroll & A.M.B. de Groot (Eds.), Handbook of bilingualism:  Psycholinguistic perspectives (pp. 109–127). Oxford: Oxford University Press. Birdsong, D. 2006. Age and second language acquisition and processing: A selective overview. In M. Gullberg & P. Indefrey (Eds), The cognitive neuroscience of second language acquisition (pp. 9–49). Oxford: Blackwell. Birdsong, D. 2007. Age and L2 acquisition and processing: Behavior, brain, and biology. Center for Research in Language, Mind, and Brain, McGill University, 30 March. Birdsong, D., & Molis, M. 2001. On the evidence for maturational constraints in secondlanguage acquisition. Journal of Memory and Language, 44: 235–249.



References 

Black, P.M., & Ronner, S.F. 1987. Cortical mapping for defining the limits of tumor resection. Neurosurgery, 20: 914–919. Bongaerts, T., van Summeren, C., Planken, B., & Schils, E. 1997. Age and ultimate attainment in the pronunciation of a foreign language. Studies in Second Language Acquisition, 19: 447–465. Bookheimer, S.Y., Zeffiro, T.A., Blaxton, T.A., Gaillard, W., & Theodore, W.H. 2000. Activation of language cortex with automatic speech tasks. Neurology 55: 1151–1157. Booth, J.R, Wood, L., Lu, D., Houk, J.C., & Bitan, T. 2007. The role of the basal ganglia and cerebellum in language processing. Brain Research, 1133: 136–144. Bos, P., Hollbrandse, B., & Sleeman, P. 2004. The pragmatics-syntax and the semantics-syntax interface in acquisition. IRAL, 42: 101–110. Boudreault, P., & Mayberry, R. 2006. Grammatical processing in American Sign Language: Age of first-language acquisition effects in relation to syntactic structure. Language and Cognitive Processes, 21: 608–635. Bouissac, P. 2007. Sign worship. Semiotix, 9 May. http://www.semioticon.com/semiotix/semiotix9/ index.html. Briellmann, R.S., Saling, M.M., Connell, A.B., Waites, A.B., Abbott, D.F., & Jackson, G.D. 2004. A high-field functional MRI study of quadrilingual subjects. Brain and Language, 89: 531–542. Bruce, L.C. 1895. Notes of a case of dual brain action. Brain, 18: 54–65. Bunge, M. 1980. The mind-body problem: A psychobiological approach. Oxford: Pergamon Press. Bunge, M. 2007. Blushing and the philosophy of mind. Journal of Physiology – Paris, 101: 247–256. Bychowski, Z. 1919. Über die Restitution der nach einem Schädelschuss verlorenen Umgangssprache bei einem Polyglotten. Monatsschrift für Psychiatrie und Neurologie, 45: 183–201. Translated in Paradis (ed.) (1983), 130–144. Callan, D., Jones, J.A., Callan, A.M., & Akahane-Yamada, R. 2004. Phonetic perceptual identification by native-  and second-language speakers differentially activates brain regions involved with acoustic phonetic processing and those involved with articulatory-auditory/ orosensory internal models. NeuroImage, 22: 1182–1194. Cavanna, A.E., & Trimble, M.R. 2006. The precuneus: A review of its functional anatomy and behavioural correlates. Brain, 129: 564–583. Centeno, J.G., Anderson, R.T., Restrepo, M.A., Jacobson, P.F., Guendouzi, J., Miller, N., Ansaldo, A.I., & Marcotte, K. 2007. Ethnographic and sociolinguistic aspects of communication: Research-praxis relationships. The ASHA Leader, 12 (9): 12–15. Chee, M.W, Caplan, D., Soon, C.S., Sriram, N., Tan, E.W.L., Thiel, T., & Weekes, B. 1999a. Processing of visually presented sentences in Mandarin and English studied with fMRI. Neuron, 23: 127–137. Chee, M.W, Hon, N., Lee, H.L., & Soon, C.S. 2001. Relative language proficiency modulates BOLD signal change when bilinguals perform semantic judgments. NeuroImage, 13: 1155–1163. Chee, M.W., Tan, E.W., & Thiel, T. 1999b. Mandarin and English single word processing studied with functional magnetic resonance imaging. The Journal of Neuroscience, 19: 3050–3056. Chee, M.W, Weekes, B., Lee, K.M., Soon, C.S., Schreiber, A., Hoon, J.J., & Chee, M. 2000. Overlap and dissociation of semantic processing of Chinese characters, English words, and pictures: Evidence from fMRI. NeuroImage, 12: 392–403. Chen, E.E., & Small, S.L. 2007. Test–retest reliability in fMRI of language: Group and task effects. Brain and Language, 102: 176–185. Chertkow, H., & Murtha, S. 1997. PET activation and language. Clinical Neuroscience, 4: 78–86. Chlenov, L.G. 1948. Ob afazii u poliglotov. Izvestiia Akademii Pedagogicheskikh NAUK RSFSR, 15: 783–790. Translated in Paradis (ed.) (1983), 446–454.

 Declarative and procedural determinants of second languages Christoffels, I.K., Firk, C., & Schiller, N.O. 2007. Bilingual language control: An event-related brain potential study. Brain Research, 1147: 192–208. Clahsen, H., & Felser, C. 2006a. Grammatical processing in language learners. Applied Psycholinguistics, 27: 3–42. Clahsen H, & Felser, C. 2006b. Continuity and shallow structures in language processing. Applied Psycholinguistics, 27: 107–126. Clyne, M.G. 1980. Triggering and language processing. Canadian Journal of Psychology, 34: 400–406. Cohen, J.D., Botvinick, M., & Carter, C.S. 2000. Anterior cingulate and prefrontal cortex: Who’s in control? Nature Neuroscience, 3: 421–423. Cohen, J.D., & Servan-Schreiber, D. 1992. Context, cortex, and dopamine:  A connectionist approach to behavior and biology in schizophrenia. Psychological Review, 99: 45–77. Cohen, N.J. 1991. Memory, amnesia and the hippocampal system. Paper presented at the Cognitive and Neuroscience Colloquium, McGill University, 6 November. Cohen, N.J., & Eichenbaum, H. 1993. Memory, amnesia, and the hippocampal system. Cambridge, MA: MIT Press. Cohen, N.J., & Eichenbaum, H. 1997. Memory for items and memory for relations in the procedural/declarative memory framework. Memory, 5: 131–178. Cook, V. 1992. Evidence for multi-competence. Language Learning, 42: 557–591. Corder, S. 1967. The significance of learners’ errors. IRAL, 5: 161–170. Cork, R.C., Kihlstrom, J.F., & Hameroff, S.R. 1992. Explicit and implicit memory dissociated by anesthetic technique. Paper presented at the 22nd Annual Meeting of the Society for Neuroscience, Anaheim, California, 26 October. Cork, R.C., Kihlstrom, J.F., & Schacter, D.L. 1992. Absence of explicit or implicit memory in patients anesthetized with sufentamil/nitrous oxide. Anesthesiology, 76: 892–898. Costa, A., & Santesteban, M. 2004. Lexical access in bilingual speech production: Evidence from language switching in highly proficient bilinguals and L2 learners. Journal of Memory and Language, 50: 491–511. Costa, A., Santesteban, M., & Ivanova, I. 2006. How do highly proficient bilinguals control their lexicalization process? Inhibitory and language-specific selection mechanisms are both functional. Journal of Experimental Psychology:  Learning, Memory and Cognition, 32: 1057–1074. Cowan, N. 1999. An embedded-process model of working memory. In A. Miyake & P. Shah (Eds), Models of working memory: Mechanisms of active maintenance and executive control (pp. 62–101). New York: Cambridge University Press. Crerar, A. 2004. Aphasia rehabilitation and the strange neglect of speed. Neuropsychological Rehabilitation, 14: 173–206. Crick, F., & Koch, C. 1998. Consciousness and neuroscience. Cerebral Cortex, 8: 97–107. Crinion, J.T., Lambon-Ralph, M.A., Warburton, E.A., Howard, D., & Wise, R.J.S. 2003. Temporal lobe regions engaged during normal speech comprehension. Brain, 126: 1193–1201. Crinion, J., Turner, R., Grogan, A., Hanakawa, T., Noppeney, U., Devlin, J.T., Aso, T., Urayama, S., Fukuyama, H., Stockton, K., Usui, K., Green, D.W., & Price, C.J. 2006. Language control in the bilingual brain. Science, 312: 1537–1540. Dalalakis, J.E. 1999. Morphological representation in specific language impairment: Evidence from Greek word formation. Folia Phoniatrica et Logopaedica, 51: 20–35. Damasio, A. 1989. Concepts in the brain. Mind and Language, 4: 24–28.



References 

Damasio, A. 1994. Descartes’ error:  Emotion, reason, and the human brain. New York:  G.P. Putnam’s Sons. Damasio, A., Bellugi, U., Damasio, H., Pozner, H., & van Gilcer, J. 1986. Sign language aphasia during left-hemisphere amytal injection. Nature, 322: 363–365. Day, E.M., & Shapson, S.M. 2001. Integrating formal and functional approaches to language teaching in French immersion:  An experimental study. In R.  Ellis (Ed.), Form-focused instruction and second language learning (pp. 47–80). Malden, MA: Blackwell Publishers. De Bleser, R., Dupont, P., Postler, J., Bormans, G., Speelman, D., Mortelmans, L., & Debrock, M. 2003. The organization of the bilingual lexicon: A PET study. Journal of Neurolinguistics, 16: 439–456. de Bot, K. 2000. Simultaneous interpreting as language production. In B.E. Dimitrova & K.  Hyltenstam (Eds.), Language processing and simultaneous interpreting (pp. 65–88). Amsterdam: Benjamins. de Bot, K. 2007. Dynamic systems theory, lifespan development and language attrition (pp. 53–68). In B. Köpke, M.S. Schmid, M. Keijzer, & S. Dostert (Eds), Language attrition: Theoretical perspectives. Amsterdam: John Benjamins. de Bot, K., Lowie, W., & Verspoor, M. 2007. A Dynamic Systems Theory approach to second language acquisition. Bilingualism: Language and Cognition, 10: 7–21. De Groot, A.M.B., & Christoffels, I. 2006. Language control in bilinguals: Mono-lingual tasks and simultaneous interpreting. Bilingualism: Language and Cognition. 9: 189–201. Dehaene, S., & Changeux, J.-P. 2004. Neural mechanisms for access to consciousness. In M. Gazzaniga (Ed.), The cognitive neurosciences III (pp. 1145–1157). Cambridge, MA: MIT Press. Dehaene, S., Changeux, J.-P., Naccache, L., Sackur, J., & Sergent, C. 2006. Conscious, preconscious, and subliminal processing: A testable taxonomy. Trends in Cognitive Sciences, 10: 204–211. Dehaene, S., Dupoux, E., Mehler, J., Cohen, L., Paulesu, E., Perani, D., van Moortele, P.-F., Lehéricy, S. & Le Bihan, D. 1997. Anatomical variability in the cortical representation of first and second language. NeuroReport, 8: 3809–3815. Dehaene, S., Kerszberg, M., & Changeux, J.-P. 1998. A neuronal model of a global workspace in effortful cognitive tasks. Proceedings of the National Academy of Sciences of the United States of America, 95: 14529–14534. Dehaene, S., & Naccache, L. 2001. Towards a cognitive neuroscience of consciousness:  Basic evidence and a workspace framework. Cognition, 79: 1–37. DeKeyser, R.M. 2000. The robustness of critical period effects in second language acquisition. Studies in Second Language Acquisition, 22: 499–533. Dennett, D.C. 1978. Brainstorms: Philosophical essays on mind and psychology. Cambridge, MA: MIT Press. Dennett, D.C. 1981. Brainstorms: Philosophical essays on mind and psychology. Cambridge, MA: MIT Press. Detry, C., Pillon, A., & de Partz, M.-P. 2005. A direct processing route to translate words from the first to the second language:  Evidence from a case of a bilingual aphasic. Brain and Language, 95: 40–41. Dewaele, J.-M. 2007. The effect of multilingualism, sociobiological, and situational factors on communicative anxiety and foreign language anxiety of mature language learners. International Journal of Bilingualism, 11: 391–409. Dewaele, J.-M., & Pavlenko A. 2001–2003. Web questionnaire on bilingualism and emotion. University of London.

 Declarative and procedural determinants of second languages De Zulueta, F.I.S. 1984. The implication of bilingualism in the study and treatment of psychiatric disorders: A review. Psychological Medicine, 14: 541–557. De Zulueta, F.I.S., Gene-Cos, N., & Grachev, S. 2001. Differential psychotic symptomatology in polyglot patients: Case reports and their implications. British Journal of Medical Psychology, 74: 277–292. Dienes, Z., Altmann, G.T.M., Kwan, L., & Goode, A. 1995. Unconscious knowledge of artificial grammars is applied strategically. Journal of Experimental Psychology—Learning, Memory and Cognition, 21: 1322–1338. Dienes, Z., & Perner, J. 1999. A theory of implicit and explicit knowledge. Behavioral and Brain Sciences, 22: 735–808. Dienes, Z., & Perner, J. 2003. Unifying consciousness with explicit knowledge. In A. Cleeremans (Ed.), The unity of consciousness (pp. 214–232). Oxford: Oxford University Press. Ding, G., Perry, C., Peng, D., Ma, L., Li, D., Xu, S., Luo, Q., Xu, D., & Yang, J. 2003. Neural mechanisms underlying semantic and orthographic processing in Chinese-English bilinguals. NeuroReport, 14: 1557–1562. Docking K., Murdoch, B., & Ward, E. 2003. Cerebellar language and cognitive functions in childhood: A comparative review of the clinical research. Aphasiology, 17: 1153–1161. Dorland, W.A. 2003. Medical dictionary for health care consumers. Philadelphia, PA: Saunders. Drake, T., & Iadecola, C. 2007. The role of neuronal signaling in controlling cerebral blood flow. Brain and Language, 102: 142–152. Duncan, J. 2001. An adaptive coding model of neural function in prefrontal cortex. Nature Reviews Neuroscience, 2: 820–829. Ehrman, M., & Oxford, R. 1995. Cognition plus:  Correlates of language learning success. Modern Language Journal, 79: 67–89. Ellis, N.C. 1994a. Consciousness in second language learning: Psychological perspectives on the role of conscious processes in vocabulary acquisition. AILA Review, 11: 37–56. Ellis, N.C. 1994b. Implicit and explicit processes in language acquisition:  An introduction. In N.C. Ellis (Ed.), Implicit and explicit learning of languages (pp. 1–32). San Diego, CA: Academic Press. Ellis, N.C. 2000. Frequency effects in language acquisition. Paper presented at the American Association for Applied Linguistics annual conference, Vancouver, 12 March. Ellis, N.C. 2002. Frequency effects in language processing. Studies in Second Language Acquisition, 24: 143–188. Ellis, N.C. 2005. At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition, 27: 305–352. Ellis, N.C. 2006. Cognitive perspectives on SLA: The associative-cognitive CREED. AILA Review, 19: 100–121. Ellis, N.C. 2008. The dynamics of second language emergence: Cycles of language use, language change, and language acquisition. The Modern Language Journal, 92: 232–249. Ellis, N.C., & Larsen-Freeman, D. 2006. Language emergence: Implications for applied linguistics— Introduction to the special issue. Applied Linguistics, 27: 558–589. Ellis, R. 2005. Principles of instructed language learning. System, 33: 209–224. Elsinger, C.L., Harrington, D.L., & Rao, S.M. 2006. From preparation to online control: Reappraisal of neural circuitry mediating internally generated and externally guided actions. NeuroImage, 31: 1177–1187. Emmorey, K., Grabowski, T., McCullough, S., Ponto, L.B., Hichwa, R.D., & Damasio, H. 2005. The neural correlates of spatial language in English and American Sign Language: A PET study with hearing bilinguals. NeuroImage, 24: 832–840.



References 

Engel, A.K. 2003. Temporal binding and the neural correlates of consciousness. In A. Cleeremans (Ed.), The unity of consciousness:  Binding, integration, and dissociation (pp. 132–152). Oxford: Oxford University Press. Engel, A.K., & Singer, W. 2001. Temporal binding and the neural correlates of sensory awareness. Trends in Cognitive Sciences, 5: 16–25. Eskridge, J.T. 1896. Mind and word deafness after depressed fracture of the skull with subcortical hemorrhage—Operation: Complete recovery. Medical News, 68: 698–702. Fabbro, F. 2001. The bilingual brain: Cerebral representation of languages. Brain and Language, 79: 211–222. Fabbro, F., Skrap, M., & Aglioti, S. 2000. Pathological switching between languages after frontal lesions in a bilingual patient. Journal of Neurology, Neurosurgery and Psychiatry, 68: 650–652. Faw, B. 2003. Pre-frontal executive committee for perception, working memory, attention, longterm memory, motor control, and thinking: A tutorial review. Consciousness and Cognition, 12: 83–139. Fehringer, C., & Fry, C. 2007. Hesitation phenomena in the language production of bilingual speakers: The role of working memory. Folia Linguistica, 41: 37–72. Fisher, S.E., Vargha-Khadem, F., Watkins, K.E., Monaco, A.P., & Pembrey, M.E. 1998. Localisation of a gene implicated in a severe speech and language disorder. Nature Genetics, 18: 168–170. Flege, J.E., Birdsong, D., Bialystok, E., Mack, M., Sung, H., & Tsukada, K. 2006. Degree of foreign accent in English sentences produced by Korean children and adults. Journal of Phonetics, 34: 153–175. Flege, J.E., Yeni-Komshian, G.H., & Liu, S. 1999. Age constraints on second-language acquisition. Journal of Memory and Language, 1: 78–104. Frackowiak, R.S.J., Friston, K.J., Frith, C.D., Dolan, R.J., Price, C.J., Zeki, S., et al. (Eds), 2004. Human brain function (2nd edition). New York: Elsevier. Frenck-Mestre, C., Anton, J.L., Roth, M., Vaid, J., & Viallet, F. 2005. Articulation in early and late bilinguals’ two languages:  Evidence from functional magnetic resonance imaging. NeuroReport, 16: 761–765. Frensch, P.A., Haider, H., Rünger, D., Neugebauer, U., Voigt, S., & Werg, J. 2003. The route from implicit learning to verbal expression of what has been learned: Verbal report of incidentally experienced environmental regularity. In L.  Jiménez, (Ed.), Attention and implicit learning (pp. 335–366). Amsterdam: John Benjamins. Friederici, A.D. 2006. What’s in control of language? Nature Neuroscience, 9: 991–992. Friederici, A.D., Steinhauer, K., & Pfeifer, E. 2002. Brain signatures of artificial language processing:  Evidence challenging the critical period hypothesis. Proceedings of the National Academy of Sciences of the United States of America, 99: 529–534. Friederici, A.D., Kotz, S., Werheid, K., Hein, G., & von Cramon, Y. 2003. Syntactic comprehension in Parkinson’s disease: Investigating early automatic and late integrational processes using event-related brain potentials. Neuropsychology, 17: 133–142. Fukuda, S.E., & Fukuda, S. 1999. The operation of rendaku in the Japanese specifically languageimpaired: A preliminary investigation. Folia Phoniatrica et Logopaedica, 51: 36–54. Fukuda, S., Fukuda, S.E., Ito, T., & Yamaguchi, Y. 2007. Grammatical impairment of case assignment in Japanese children with specific language impairment. Japan Journal of Logopedics and Phoniatrics, 48: 95–104. Gandour, J., Tong, Y.X., Talavage, T., Wong, D., Dzemidzic, M., Xu, Y.S., Li, X.J, & Lowe, M. 2007. Neural basis of first and second language processing of sentence-level linguistic prosody. Human Brain Mapping, 28: 94–108.

 Declarative and procedural determinants of second languages Gass, S. 1988. Integrating research areas:  A framework for second language studies. Applied Linguistics, 9: 198–217. Gass, S. 1997. Input, interaction, and the second language learner. Mahwah, NJ:  Lawrence Erlbaum Associates. Genesee, F., & Nicoladis, E. 2006. Bilingual acquisition. In E. Hoff & M. Shatz (Eds), Handbook of language development, Oxford: Blackwell. Giussani, C., Roux, F.-E., Lubrano, V., Gaini, S.M., & Bello, L. 2007. Review of language organisation in bilingual patients: what can we learn from direct brain mapping? Acta Neurochirurgica, 149: 1109–1116. Gloning, I., & Gloning, K. 1965. Aphasien bei Polyglotten. Beitrag zur Dynamik des Sprachabbaus sowie zur Lokalisationsfrage dieser Störungen. Wiener Zeitschrift für Nervenheilkunde, 22: 362–397. Goldblum, Z. 1928. Nach Trepanation aufgetretene motorische Aphasie (Hypolalie) mit Restitution bei progressiv wachsendem Endotheliom im linken Zentrofrontallappen. Schweizer Archiv für Neurologie, 22: 227–268. Translated in Paradis (ed.) (1983), 261–273. Golestani, N., Alario, F.X., Meriaux, S., Le Bihan, D., Dehaene, S., & Pallier, C. 2006. Syntax production in bilinguals. Neuropsychologia, 44: 1029–1040. Gollan, T.H., & Kroll, J.F. 2001. Bilingual lexical access. In B. Rapp (Ed.), The handbook of cognitive neuropsychology (pp. 321–345). Philadelphia: Psychology Press. Gombert, J.E. 1993. Metacognition, metalanguage and metapragmatics. International Journal of Psychology, 28: 571–580. Gomez-Tortosa, E., Martin, E.M., Gaviria, M., Charbel, F., & Ausman, J.I. 1995. Selective deficit in one language in a bilingual patient following surgery in the left perisylvian area. Brain and Language, 48: 320–325. Goral, M., Levy, E.S., Obler, L.K., & Cohen, E. 2006. Cross-language lexical connections in the mental lexicon:  Evidence from a case of trilingual aphasia. Brain and Language, 98: 235–247. Grand, S., Marcos, L.R., Freedman, N., & Narroso, F. 1977. Relation of psycho-pathology and bilingualism to kinesic aspects of interview behavior in schizophrenia. Journal of Abnormal Psychology, 86: 492–500. Green, D.W. 1986. Control, activation, and resource: A framework and a model for the control of speech in bilinguals. Brain and Language, 27: 210–223. Green, D.W. 1998. Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1: 77–82. Green, D.W. 2002. Representation and control: Exploring recovery patterns in bilingual aphasics. In F.  Fabbro (Ed.), Advances in the neurolinguistics of bilingualism (pp. 239–260). Udine: Udine University Press. Green, D.W. 2005. The neurocognition of recovery patterns in bilingual aphasics. In J.F. Kroll & A.M.B. de Groot (Eds), Handbook of bilingualism: Psycholinguistic perspectives (pp. 516–530). Oxford: Oxford University Press. Green, D.W. 2008. Bilingual aphasia:  Adapted language networks and their control. Annual Review of Applied Linguistics, 28: 25–48. Green, D.W., & Abutalebi, J. 2008. Understanding the link between bilingual aphasia and language control. Journal of Neurolinguistics, 21: 558–576. Green, J.J. , Teder-Sälejärvi, W.A., & McDonald, J.J. 2005. Control mechanisms mediating shifts of attention in auditory and visual space: A spatio-temporal ERP analysis. Experimental Brain Research, 166: 358–369.



References 

Greenfield, S.A., & Collins, T.F.T. 2005. A neuroscientific approach to consciousness. In S.  Laureys (Ed.), The boundaries of consciousness:  Neurobiology and neuropathology (pp. 11–23). Amsterdam: Elsevier. Grimshaw, G.M., Adelstein, A., Bryden, M.P., & McKinnon, E. 1994. First-language acquisition in adolescence—a test of the critical period hypothesis. Poster presented at the TENNET meeting, Montreal, May 29; short abstract available in Brain and Cognition (1995), 28: 190–191. Grimshaw, G.M., Adelstein, A., Bryden, M.P., & McKinnon, E. 1998. First-language acquisition in adolescence: Evidence for a critical period for verbal language development. Brain and Language, 63: 237–255. Grosjean, F. 1989. Neurolinguists, beware! The bilingual is not two monolinguals in one person. Brain and Language, 36: 1–15. Grosjean, F. 2008. Studying bilinguals. Oxford: Oxford University Press. Hahne, A., 2001. What’s different in second-language processing? Evidence from event-relayed brain potentials. Journal of Psycholinguistic Research, 30: 251–266. Hahne, A., & Friederici, A.D. 2001. Processing a second language: Late learners’ comprehension mechanisms as revealed by event-related brain potentials. Bilingualism: Language and Cognition, 4: 123–141. Håkansson, G., Salameh, E.-K. & Nettelbladt, U. 2003. Measuring language development in bilingual children: Swedish-Arabic children with and without language impairment. Linguistics, 41: 255–288. Halsband, U. 2006. Bilingual and multilingual language processing. Journal of Physiology – Paris, 99: 355–369. Halsband, U., Krause, B.J., Sipilä, H., Teräs, M., & Laihinen, A. 2002. PET studies on the memory processing of word pairs in bilingual Finnish-English subjects. Behavioural Brain Research, 132: 47–57. Hamberger, M.J. 2007. Cortical language mapping in epilepsy: A critical review. Neuropsychology Review, 17: 477–489. Hansen, L.K. 2007. Multivariate strategies in functional magnetic resonance imaging. Brain and Language, 102: 186–191. Hardcastle, V.G. 1995. Locating consciousness. Amsterdam: John Benjamins. Haritos, C. 2003. Listening, remembering, and speaking in two languages: How did you do that? Bilingual Research Journal, 27: 73–99. Harley, B., & Hart, D., 1997. Language aptitude and second language proficiency in classroom learners of different starting ages. Studies in Second Language Acquisition, 19: 379–400. Harley, B., & Wang, W. 1997. The critical period hypothesis: Where are we now? In A.M.B. de Groot & J.F. Kroll (Eds), Tutorials in bilingualism. Mahwah, NJ: Lawrence Erlbaum Associates. Hasegawa, M., Carpenter, P.A., & Just, M.A. 2002. An fMRI study of bilingual sentence comprehension and workload. NeuroImage, 15: 647–660. Haynes, J.D. 2005. Visibility reflects dynamic changes of effective connectivity between V1 and fusiform cortex. Neuron, 46: 811–821. Heinemann, F., & Assion, H.G. 1996. Linguistic regression to use of the native language only in bilingual persons during acute psychotic episodes. Nervenarzt, 67: 599–601. Hensch, T.K. 2004. Critical period regulation. Annual Review of Neuroscience, 27: 549–579. Hernandez, A.E, Dapretto, M., Mazziotta, J., & Bookheimer, S. 2001b. Language switching and language representation in Spanish-English bilinguals: An fMRI study. NeuroImage, 14: 510–520.

 Declarative and procedural determinants of second languages Hernandez A.E., Martinez, A., & Kohnert, K. 2000. In search of the language switch: An fMRI study of picture naming in Spanish-English bilinguals. Brain and Language, 73: 421–431. Hemandez, A.E., & Meschyan, G. 2006. Executive function is necessary to enhance lexical processing in a less proficient L2: Evidence from fMR1 during picture naming. Bilingualism: Language and Cognition, 9: 177–188. Heuer, H., Spijkers, W., Kiesswatter, E., & Schmidtke, V. 1998. Effects of sleep loss, time of day, and extended mental work on implicit and explicit learning of sequences. Journal of Experimental Psychology: Applied, 4: 139–162. Hirschman, M. 2000. Language repair via metalinguistic means. International Journal of Language and Communication Disorders, 35: 251–268. Horst, M., Cobb, T., & Meara, P. 1998. Beyond A Clockwork Orange: Acquiring second language vocabulary through reading. Reading in a Foreign Language, 11: 207–223. Hubel, D.H., & Wiesel, T.N. 1962. Receptive fields, binocular interaction and functional architecture in cats’ visual cortex. Journal of Physiology, 160: 106–154. Hughes, G.W. 1981. Neuropsychiatric aspects of bilingualism: A brief review. British Journal of Psychiatry, 139: 25–28. Hull, R., & Vaid, J. 2006. Laterality and language experience. Laterality, 11, 436–464. Hull, R., & Vaid, J. 2007. Bilingual language lateralization: A meta-analytic tale of two hemispheres. Neuropsychologia, 45: 1987–2008. Hull, R., & Vaid, 2008. Bilingual laterality and the matter of degree. A response to Paradis (2008). Neuropsychologia, 46: 1591–1593. Hulstijn, J.H. 2003. Incidental and intentional learning. In C.J. Doughty & M.H. Long (Eds), The handbook of second language acquisition and research (pp. 349–381). Oxford: Blackwell. Hulstijn, J.H. 2007. Psycholinguistic perspectives on language acquisition. In J.  Cummins & C. Davison (Eds), The international handbook on English language teaching (pp. 701–713). Norwell, MA: Springer. Hyder, F., Phelps, E.A., Wiggins, C.J., Labar, K.S., Blamire, A.M., & Shulman, R.G. 1997. “Willed action”:  A functional MRI study of the human prefrontal cortex during a sensorimotor task. Proceedings of the National Academy of Sciences of the United States of America, 94: 6989–6994. Hyltenstam, K., & Abrahamsson, N. 2001. Age and L2 learning: The hazards of matching practical “implications” with theoretical “facts.” TESOL Quarterly, 35: 151–170. Hyltenstam, K., & Abrahamsson, N. 2003. Maturational constraints in SLA. In C.  Doughty & M. Long (Eds), The handbook of second language acquisition (pp. 539–588). Oxford: Blackwell. Ijalba, E., Obler, L.K., & Chengappa, S. 2004. Bilingual aphasia. In T.K. Bhatia & W.C. Ritchie (Eds), The handbook of bilingualism (pp. 71–89). Oxford: Blackwell Publishing. Illes, J., Francis, W., Desmond, J., Gabrieli, J., Glover, G., et al. 1999. Convergent cortical representation of semantic processing in bilinguals. Brain and Language, 70: 347–363. Indefrey, P., & Levelt, W.J.M. 2004. The neural correlates of language production. Cognition, 92: 101–144. Ioup, G., Boustagui, E., Tigi, M.E., & Moselle, M. 1994. Reexamining the critical period hypothesis: A case study of successful adult SLA in a naturalistic environment. SSLA, 16: 73–98. Iwai, Y., & Lester, H. 2006. GABA uptake determines critical period onset in mouse visual cortex. The Journal of Physiological Sciences, 56: S34. Jackson, J.H. 1878. On affections of speech from disease of the brain. Brain, 1: 304–330. Jansen, A., Flöel, A., Van Randenborgh, J., Konrad, C., Rotte, M., Förster, A.F, Deppe, M., & Knecht, S. 2005. Crossed cerebro-cerebellar language dominance. Human Brain Mapping, 24: 165–172.



References 

Jean, G. 2005. Intégration de la grammaire dans l’enseignement des langues secondes: le cas des exercices grammaticaux. Revue canadienne des langues vivantes, 61: 519–542. Jeong, H., Sugiura, M., Sassa, Y., Haji, T., Usui, N., Taira, M., Horie, K., Sato, S., & Kawashima, R. 2007a. Effect of syntactic similarity on cortical activation during second language processing: A comparison of English and Japanese among native Korean trilinguals. Human Brain Mapping, 28: 194–204. Jeong, H., Sugiura, M., Sassa, Y., Yokoyama, S., Horie, K., Sato, S., Taira, M., & Kawashima, R. 2007b. Cross-linguistic influence on brain activation during second language processing: An fMRI study. Bilingualism: Language and Cognition, 10: 175–187. Jonides, J. 1995. Working memory and thinking. In E.E Smith & D.N. Osherson (Eds), Invitation to cognitive science (vol. 3: Thinking, pp. 215–265). Cambridge, MA: MIT Press. Kane, M.J., & Engle, R.W. 2002. The role of prefrontal cortex in working-memory capacity, executive attention, and general fluid intelligence: An individual-differences perspective. Psychonomic Bulletin and Review, 9: 637–671. Karanth, P., & Rangamani, G.N. 1988. Crossed aphasia in multilinguals. Brain and Language, 34: 169–180. Kauders O. 1929. Über polyglotte Reaktionen bei einer sensorischen Aphasie. Zeitschrift für die gesamte Neurologie und Psychiatrie, 122: 651–666, Translated in Paradis (ed.) (1983), 286–300. Kegl, J., Senghas, A., & Coppola, M. 1999. Creation through contact: Sign language emergence and sign language change in Nicaragua. In M. DeGraff (Ed.), Language creation and language change (pp. 179–237). Cambridge, MA: MIT Press. Kehayia, E. 1997. Lexical access and representation in individuals with developmental language impairment: A cross-linguistic study. Journal of Neurolinguistics, 10: 139–149. Khateb, A.S., Abutalebi, J., Michel, C.M., Pegna, A.J., Lee-Janke, H., & Annoni, J.M. 2007. Language selection in bilinguals:  A spatio-temporal analysis of electric brain activity. International Journal of Psychophysiology, 65: 201–213. Kho, K.H., Duffau, H., Gatignol, P., Leijten, F.S.S., Ramsey, N.F., van Rijen, P.C., & Rutten, G.J.M. 2007. Involuntary language switching in two bilingual patients during the Wada test and intraoperative electrocortical stimulation. Brain and Language, 101: 31–37. Kihlstrom, J.F., Schacter, D.L., Cork, R.C., Hurt, C.A., & Behr, S.E. 1990. Implicit and explicit memory following surgical anesthesia. Psychological Sciences, 1: 303–306. Kim, K.H., Relkin, N.R., Lee, K.-M., & Hirsch, J. 1997. Distinct cortical areas associated with native and second languages. Nature, 388: 171–174. Klein, D. 2003. A positron emission tomography study of presurgical language mapping in a bilingual patient with a left posterior temporal cavernous angioma. Journal of Neurolinguistics, 16: 417–427. Klein, D., Milner, B., Zatorre, R., Zhao, V., & Nikelski, J. 1999. Cerebral organization in bilinguals: A PET study of Chinese-English verb generation. NeuroReport, 10: 2841–2846. Klein, D., Watkins, K., Zatorre, R.J., & Milner, B. 2006a. Word and nonword repetition in bilingual subjects: A PET study. Journal of Human Brain Mapping, 27, 153–161. Klein, D., Zatorre, R.J., Chen, J.K., Milner, B., Crane, J., Belin, P., & Bouffard, M. 2006b. Bilingual brain organization:  A functional magnetic resonance adaptation study. NeuroImage, 31: 366–375. Klein, D., Zatorre, R.J., Milner, B., Meyer, E., & Evans, A.C. 1994. Left putaminal activation when speaking a second language: Evidence from PET. NeuroReport, 5: 2295–2297. Klein, D., Zatorre, R.J., Milner, B., Meyer, E., & Evans, A.C. 1995a. The neural substrates of bilingual language processing:  Evidence from positron emission tomography. In M. Paradis (Ed.), Aspects of bilingual aphasia (pp. 23–36). Oxford: Pergamon Press.

 Declarative and procedural determinants of second languages Klein, D., Zatorre, R.J., Milner, B., Meyer, E., & Evans, A.C. 1995b. The neural substrates underlying word generation: A bilingual functional-imaging study. Proceedings of the National Academy of Science USA, 92: 2899–2903. Klein, W. 1986. Second language acquisition. Cambridge: Cambridge University Press. Koch, C. 2004. The quest for consciousness: A neurobiological approach. Englewood, CO: Roberts and Co. Kohnert, K. 2004. Cognitive and cognate-based treatments for bilingual aphasia: A case study. Brain and Language, 91: 294–302. Kohnert, K. 2008. Language disorders in bilingual children and adults. Abingdon, UK: Plural Publishing. Komarova, N.L., & Nowak, M.A. 2001. Natural selection of the critical period for language acquisition. Proceedings of the Royal Society, London, 268: 1189–1196. Kosaka, B., Hiscock, M., Straus, E., Wada, J.A., & Purves, S. 1993. Dual task performance by patients with left or right speech dominance as determined by carotid amytal tests. Neuropsychologia, 31: 127–136. Kovelman, I., Baker, S.A., & Petitto, L.-A. 2008a. Bilingual and monolingual brains compared: A functional magnetic resonance imaging investigation of syntactic processing and possible “neural signature” of bilingualism. Journal of Cognitive Neuroscience, 20: 153–169. Kovelman, I., Shalinsky, M.H., Berens, M.S., & Petitto, L.-A. 2008b. Shining new light on the brain’s “bilingual signature”:  A functional Near Infrared Spectroscopy investigation of semantic processing. NeuroImage, 39: 1457–1471. Krashen, S. 1973. Lateralization, language learning and the critical period: Some new evidence. Language Learning, 23: 63–74. Krashen, S. 1981. Second language acquisition and second language learning. Oxford: Pergamon Press. Krashen, S. 1982. Principles and practice in second language acquisition. New York: Pergamon Press. Kuhl, P.K., Williams, K.A., Lacerda, F., Stevens, K.N., & Lindblom, B. 1992. Linguistic experience alters phonetic perception in infants by 6 months of age. Science, 255: 606–609. Lai, C., Fisher, S., Hurst, J., Vargha-Khadem, F., & Monaco, A. 2001. A forkhead-domain gene is mutated in a severe speech and language disorder. Nature, 413: 519–523. Lalor, E., & Kirsner, K. 2001. The role of cognates in bilingual aphasia: Implications for assessment and treatment. Aphasiology, 15: 1047–1056. Lambert W.E., & Fillenbaum, S. 1959. A pilot study of aphasia among bilinguals. Canadian Journal of Psychology, 13: 28–34. Lamendella, J.T. 1977. General principles of neurofunctional organization and their manifestation in primary and secondary language acquisition. Language Learning, 27: 155–196. Lamendella, J.T. 1979. The neurofunctional basis of pattern practice. TESOL Quarterly, 13: 5–19. Lamme, V.A.F. 2006a. Towards a true neural stance on consciousness. Trends in Cognitive Sciences, 10: 494–501. Lamme, V.A.F. 2006b. Zap! Magnetic tricks on conscious and unconscious vision. Trends in Cognitive neuroscience, 10: 193–195. Laski, E., & Taleporos, E. 1977. Anticholinergic psychosis in a bilingual: A case study. American Journal of Psychiatry, 134: 1038–1040. Lau, H.C., Rogers, R.D., Ramnani, N., & Passingham, R.E. 2004. Willed action and attention to the selection of action. NeuroImage, 21: 1407–1415.



References 

Lebrun, Y. 1978. Evaluation of language-impaired children. In F. Peng & W. von Raffler-Engel (Eds), Language acquisition and developmental kinesics (pp. 182–193). Hiroshima: Bunka Hyokon. Lebrun, Y. 1997. Adult-onset stuttering. In Y.  Lebrun (Ed.), From the brain to the mouth (pp. 105–138). Dordrecht: Kluwer. Lebrun, Y. 2002. Implicit competence and explicit knowledge. In F. Fabbro (Ed.), Advances in the neurolinguistics of bilingualism (pp. 299–313). Udine: Forum-Udine University Press. Leeman, B., Laganaro, M., Schwitter, V., & Schnider, A. 2007. Paradoxical switching to a barelymastered second language by an aphasic patient. Neurocase, 13: 209–213. Lenneberg, E.H. 1967. Biological foundations of language. New York: John Wiley & Sons. Lewicki, P., Czyzewska, M., & Hill, T. 1997. Nonconscious information and personality. In D. Berry (Ed.), How implicit is implicit learning? (pp. 48–72). Oxford: Oxford University Press. Libben, G. 2006. How language learners comprehend and produce language in real time. Applied Psycholinguistics, 27: 72–74. Liegeois, F., Baldeweg, T., Connelly, A., Gadian, D.G., Mishkin, M., & Vargha-Khadem, F. 2003. Language fMRI abnormalities associated with FOXP2 gene mutation. Nature Neuroscience, 6: 1230–1237. Locke, J. 1690. An essay concerning human understanding. London: Thomas Basset. Lubrano, V., Roux, F.E., & Demonet, J.F. 2004. Writing-specific sites in frontal areas: A cortical stimulation study. Journal of Neurosurgery, 101: 787–798. Lucas, T.H., McKhann, G.M., & Ojemann, G. 2004. Functional separation of languages in the bilingual brain: A comparison of electrical stimulation language mapping in 25 bilingual patients and 117 monolingual control patients. Journal of Neurosurgery, 101: 449–457. Luke, K.K., Liu, H.L., Wai, Y.Y., Wan, Y.L., & Tan, L.H. 2002. Functional anatomy of syntactic and semantic processing in language comprehension. Human Brain Mapping, 16: 133–145. Luria, A.R. 1973a. The working brain: An introduction to neuropsychology. New York: Penguin Education. Luria, A.R. 1973b. Two basic kinds of aphasia disorders. Linguistics, 115: 57–66. Luu, P., Kelley, J.M., & Levtin, D.J. 2001. Consciousness: A preparatory and comparative process. In P.G. Grossenbacher (Ed.), Finding consciousness in the brain (pp. 247–275). Amsterdam: John Benjamins. Lyster, R. 1990. The role of analytic language teaching in French immersion programs. Canadian Modern Language Review, 47: 159–176. Lyster, R. 2004. Research on form-focused instruction in immersion classrooms: Implications for theory and practice. Journal of French Language Studies, 14: 321–341. MacDermot, K.D., Bonora, E., Sykes, N., Coupe, A.M., Lai, C.S., Vernes, S.C., Vargha-Khadem, F., McKenzie, F., Smith, R.L., Monaco, A.P., & Fisher, S.E. 2005. Identification of FOXP2 truncation as a novel cause of developmental speech and language deficits. American Journal of Human Genetics, 76: 1074–1080. Mack, A., & Rock, I. 1998. Inattentional blindness. Cambridge, MA: MIT Press. Mack, M.A. 1984. Early bilinguals:  How monolingual-like are they? In M.  Paradis & Y.  Lebrun (Eds), Early bilingualism and child development (pp. 161–173). Lisse, ND: Swets & Zeitlinger. Mack, M.A. 1986. A study of semantic and syntactic processing in monolinguals and fluent early bilinguals. Journal of Psycholinguistic Research, 15: 463–488. Mack, M.A. 2003. The phonetic systems of bilinguals. In M.T. Banich & M.A. Mack (Eds), Mind, brain, and language (pp. 309–349). Mahwah, NJ.: Lawrence Erlbaum Associates.

 Declarative and procedural determinants of second languages MacWhinney, B. 2005. A unified model of language acquisition. In J.F. Kroll & A.M.B. de Groot (Eds), Handbook of bilingualism: Psycholinguistic perspectives (pp. 49–67). Oxford: Oxford University Press. Mahendra, N., Plante, E., Magloire, J., Milman, L., & Trouard, T.P. 2003. fMRI variability and the localization of languages in the bilingual brain. NeuroReport, 14: 1225–1228. Marcos, L.R. 1976. Bilinguals in psychotherapy: Language as an emotional barrier. American Journal of Psychotherapy, 30: 553–560. Marcos, L.R., & Urcuyo, L. 1979. Dynamic psychotherapy with the bilingual patient. American Journal of Psychotherapy, 33: 331–338. Marcus, G.F., & Fisher, S.E. 2003. FOXP2 in focus: What can genes tell us about speech and language? Trends in Cognitive Sciences, 7: 257–262. Marian, V., Shildkrot, Y., Blumenfeld, H.K., Kaushanskaya, M., Faroqi-Shah, Y., & Hirsch, J. 2007. Cortical activation during bilinguals:  Similarities and word processing in late differences as revealed by functional magnetic resonance imaging. Journal of Clinical and Experimental Neuropsychology, 29: 247–265. Mariën, P., Abutalebi, J., Engelborghs, S., & De Deyn, P.P. 2005. Pathophysiology of language switching and mixing in an bilingual child with subcortical aphasia. Neurocase, 11: 385–398. Mariën, P., Engelborghs, S., Fabbro, F., & De Deyn, P.P. 2001. The lateralized linguistic cerebellum: A review and a new hypothesis. Brain and Language, 79: 580–600. Marini, A., Tavano, A., & Fabbro, F. 2008. Assessment of linguistic abilities in Italian children with Specific Language Impairment. Neuropsychologia, 46: 2816–2823. Marinova-Todd, S. 2003. Comprehensive analysis of ultimate attainment in adult second language acquisition. Doctoral thesis, Graduate School of Education, Harvard University. Marinova-Todd, S., Marshall, D.B., & Snow, C.E. 2000. Three misconceptions about age and second-language learning. TESOL Quarterly, 34: 9–34. Marinova-Todd, S., Marshall, D.B., & Snow, C.E. 2001. Missing the point:  A response to Hyltenstam and Abrahamsson. TESOL Quarterly, 35: 171–176. Marshall, J., Atkinson, J., Woll, B., & Thacker, A. 2005. Aphasia in a bilingual user of British sign language and English:  Effects of cross-linguistic cues. Cognitive Neuropsychology, 22: 719–736. Masters, R.S.W. 1992. Knowledge, knerves, and know-how: The role of explicit versus implicit knowledge in the breakdown of a complex motor skill under pressure. British Journal of Psychology, 83: 343–358. Mayberry, R.I. 1993. First-language acquisition after childhood differs from second-language acquisition: The case of American Sign Language. Journal of Speech and Hearing Research, 36: 1258–1270. Mayberry, R.I. 2006. Grammatical processing in American Sign language: Age of first-language acquisition effects in relation to syntactic structure. Language and Cognitive Processes, 21: 608–635. Mayberry, R.I., & Lock, E. 2003. Age constraints on first and second language acquisition:  Evidence from linguistic plasticity and epigenesis. Brain and Language, 87: 369–383. Mayberry, R.I., Lock, E., & Kazmi, H. 2002. Linguistic ability and early language exposure. Nature, 417: 38. Mayberry, M., Taylor, M., & Obrien-Malone, A. 1995. Implicit learning—Sensitive to age but not IQ. Australian Journal of Psychology, 47: 8–17. McDonald, J.L. 2006. Beyond the critical period: Processing-based explanations for poor grammaticality judgment performance by late second language learners. Journal of Memory and Language, 55: 381–401.



References 

McGeorge, P., Crawford, J.R., & Kelly, S.W. 1997. The relationship between psychometric intelligence and learning in an explicit and an implicit task. Journal of Experimental Psychology: Learning, Memory and Cognition, 23: 239–245. McLaughlin, B. 1987. Theories of second language learning. London: Arnold. Meguro, K., Senaha, M., Caramelli, P., Ishizaki, J., Chubacci, R., Meguro, M., Ambo, H., Nitrini, R., & Yamadori, A. 2003. Language deterioration in four Japanese-Portuguese bilingual patients with Alzheimer’s disease: A transcultural study of Japanese elderly immigrants in Brazil. Psychogeriatrics, 3: 63–68. Merikle, P.M., Smilek, D., & Eastwood, J.D. 2001. Perception without awareness: Perspectives from cognitive psychology. Cognition, 79: 115–134. Meschyan, G., & Hernandez, A.E. 2006. Executive function is necessary to enhance lexical processing in a less proficient L2: Evidence from fMR1 during picture naming. Bilingualism: Language and Cognition, 9: 177–188. Miller, E.K., & Cohen, J.D. 2001. An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24:167–202. Miller, G.A. (Ed.), 2007. WordNet: An electronic lexical database. Cambridge, MA: MIT Press. Minkowski, M. 1927. Klinischer Beitrag zur Aphasie bei Polyglotten, speziell im Hinblick aufs Schweizerdeutsche. Schweizer Archiv für Neurologie und Psychiatrie, 21: 43–72. Translated in Paradis (ed.) (1983), 205–232. Minkowski, M. 1928. Sur un cas d’aphasie chez un polyglotte. Revue Neurologique, 49: 361–366. Translated in Paradis (ed.) (1983), 274–279. Minkowski, M. 1933. Sur un trouble aphasique particulier chez un polyglotte. Revue neurologique, 59: 1185–1189. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A., & Wager, T. 2000. The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41: 49–100. Möhring, A. 2001. The acquisition of French by German pre-school children:  An empirical investigation of gender assignment and gender agreement. In S.H. Foster-Cohen & A. Nizegorodcew (Eds), EUROSLA Yearbook (pp. 171–193). Amsterdam: John Benjamins. Montrul, S., Foote, R., Perpiñán, S., Thornhill, D., & Vidal, S. 2006. Full access and age effects in adult bilingualism:  An investigation of Spanish accusative clitics and word order. In N. Sagarra & J. Toribio (Eds), Selected Proceedings of the 9th Hispanic Linguistic Symposium (pp. 217–228). Somerville, MA: Cascadilla Proceedings Project. Mooney, R.  Prather, J., & Roberts, T. 2008. Neurophysiology of birdsong learning. In J.H. Byrne (Ed.), Learning and memory: A comprehensive reference, Vol. 3. (Ch. 28). Oxford: Academic Press. Moretti, R., Bava, A., Torre, P., Antonello, R., Zorzon, M., Zivadinov, R., & Gazzoto, G. 2001. Bilingual aphasia and subcortical-cortical lesions. Perceptual and Motor Skills, 92: 803–814. Morford, J.P. 2003. Grammatical development in adolescent first-language learners. Linguistics, 41: 681–721. Morgan, G., & Kegl, J. 2006. Nicaraguan Sign Language and theory of mind: The issue of critical periods and abilities. Journal of Child Psychology and Psychiatry, 47: 811–819. Moscovitch, M. 2004. Models of consciousness and memory. In M. Gazzaniga (Ed.), The cognitive neurosciences III (pp. 1341–1356). Cambridge, MA: MIT Press. Moyer, A. 1999. Ultimate attainment in L2 phonology. Studies in Second Language Acquisition, 21: 81–108. Moyer, A. 2004. Age, accent and experience in second language acquisition. Clevedon, UK: Multilingual Matters.

 Declarative and procedural determinants of second languages Munro, M., & Mann, V. 2005. Age of immersion as a predictor of foreign accent. Applied Psycholinguistics, 26: 311–341. Murphy, S. 2003. Second language transfer during third language acquisition. Working Papers in TESOL & Applied Linguistics, 3: 1–21. Nakai, T., Matsuo, K., Kato, C., Matsuzawa, M., Okada, T., Glover, H., Moriya, T., & Inui, T. 1999. A functional magnetic resonance imaging study of listening comprehension of languages in human at 3 tesla-comprehension level and activation of the language areas. Neuroscience Letters, 263: 33–36. Neufeld, G.G. 1977. Language learning ability in adults: A study of the acquisition of prosodic and articulatory features. Working Papers on Bilingualism, 12: 45–60. Neufeld, G.G. 1979. Towards a theory of language learning ability. Language Learning, 29: 227–241. Neufeld, G.G. 1980. On the adult’s ability to acquire phonology. TESOL Quarterly, 14: 285–298. Neville, H.J., Mills, D.L., & Lawson, D.S. 1992. Fractionating language: Different natural subsystems with different sensitive periods. Cerebral Cortex, 2: 244–258. Newman, A.J., Bavelier, D., Corina, D., Jezzard, P., & Neville, H.J. 2002. A critical period for right hemisphere recruitment in American Sign Language processing. Nature Neuroscience, 5: 76–80. Newport, E.L., Bavelier, D., & Neville, H.J. 2001. Critical thinking about critical periods: Perspectives on a critical period for language acquisition. In E. Dupoux (Ed.), Language, brain, and cognitive development (pp. 481–502). Cambridge, MA: The MIT Press. Nicholas, M., Tabor-Connor, L., Obler, L.K., & Albert, M.L. 1998. Aging, language, and language disorders.” In M.  Taylor Sarno (Ed.), Acquired aphasia, 3rd edition (pp. 413–449). New York: Academic Press. Nieuwenhuis, S., & Yeung, N. 2005. Neural mechanisms of attention and control: Losing our inhibitions? Nature Neuroscience, 8: 1631–1633. Nikolov, M., 2000. The CPH reconsidered: Successful adult learners of Hungarian and English. International Review of Applied Linguistics, 38: 109–124. Nikolov, M., & Mihaljevic Djigunovic, J. 2006. Recent research on age, second language acquisition, and early foreign language learning. Annual Review of Applied Linguistics, 26: 234–260. Obler, L.K., & Albert, M.L. 1981. Language in the elderly aphasic and the demented patient. In M. Taylor Sarno (Ed.), Acquired aphasia (pp. 385–398). New York: Academic Press. Obler, L.K., & Albert, M.L. 1984. Language in aging. In M.L. Albert (Ed.), Clinical neurology of aging (pp. 245–253). New York: Oxford University Press. Obler, L.K., Hyun, J.M., Conner, P.S., O’Connor, B., & Anema, I. 2007. Brain organization of language in bilinguals. In A. Ardila & E. Ramos (Eds), Speech and language disorders in bilinguals (pp. 21–46). Hauppauge, NY: Nova Science Publishers. Ojemann, G.A., & Whitaker, H.A. 1978. The bilingual brain. Archives of Neurology, 35: 409–412. Ojemann, J.G., Ojemann, G.A., & Lettich, E. 2002. Cortical stimulation mapping of language cortex by using a verb generation task:  Effects of learning and comparison to mapping based on object naming. Journal of Neurosurgery, 97: 33–38. Osaka, N. 2003. Working memory-based consciousness: An individual difference approach. In N. Osaka (Ed.), Neural basis of consciousness (pp. 27–44). Amsterdam: John Benjamins. Palij, M. 1990. Acquiring English at different ages: The English displacement effect and other findings. Journal of Psycholinguistic Research, 19: 57–70. Panksepp, J. 2007. Affective neuroscience and the ancestral sources of human feelings. In H.  Cohen & B.  Stemmer (Eds), Consciousness and cognition (pp. 173–188). London: Academic Press.



References 

Paradis, M. 1969. Pierre Bayle et la Théodicée de Leibnitz. Ph.D. thesis. McGill University. Paradis, M. 1977. Bilingualism and aphasia. In H. Whitaker & H.A. Whitaker (Eds), Studies in neurolinguistics, vol. 3 (pp. 65–121). New York: Academic Press. Paradis, M. 1981. Neurolinguistic organization of a bilingual’s two languages. The Seventh LACUS Forum (pp. 486–494). Columbia, SC: Hornbeam Press. Paradis, M. 1983. Epilogue. In M. Paradis (Ed.), Readings on aphasia in bilinguals and polyglots (pp. 802–813). Montreal: Didier. Paradis, M. 1987. Neurolinguistic perspectives on bilingualism. In M. Paradis & G. Libben, The assessment of bilingual aphasia (pp. 1–17). Hillsdale, NJ: Lawrence Erlbaum Associates. Paradis, M. 1989. Bilingual and polyglot aphasia. In F. Boller & J. Grafman (Eds), Handbook of neuropsychology, Vol. 2 (pp. 117–140). Amsterdam: Elsevier. Paradis, M. 1990. Language lateralization in bilinguals: Enough already! Brain and Language, 39: 576–588. Paradis, M. 1991. Implication de mécanismes cérébraux différents selon les méthodes d’apprentisage. Paper presented at the International Congress on Memory and Memorization in Acquiring and Learning Languages, Brussels, 21–23 November. Published in M-T Claes & J. Chapelle (Eds), Proceedings (pp. 205–224). Louvain: C.C.L. (1993). Paradis, M. 1994a. Neurolinguistic aspects of implicit and explicit memory: Implications for bilingualism. In N. Ellis (Ed.), Implicit and explicit learning of second languages (pp. 393–419). London: Academic Press. Paradis, M. 1994b. Toward a neurolinguistic theory of simultaneous translation:  The framework. International Journal of Psycholinguistics, 10: 319–335. Paradis, M. 1998a. Neurolinguistic aspects of the native speaker. In R. Singh (Ed.), The native speaker (pp. 205–219). Beverly Hills, CA: Sage. Paradis, M. 1998b. The other side of language. Journal of Neurolinguistics, 11: 1–10. Paradis, M. 2004. A neurolinguistic theory of bilingualism. Amsterdam: John Benjamins. Paradis, M. 2006. More Belles infidèles—or why do so many bilingual studies speak with forked tongue? Journal of Neurolinguistics, 19: 195–208. Paradis, M. 2007a. L1 attrition features predicted by a neurolinguistic theory of bilingualism. In B.  Köpke & B.  Schmid (Eds), First language attrition (pp. 121–134). Amsterdam: John Benjamins. Paradis, M. 2007b. The neurofunctional components of the bilingual cognitive system. In I. Keczkes (Ed.), Cognitive aspects of bilingualism (pp. 3–28). New York: Springer. Paradis, M. 2007c. Why single-word experiments do not address language representation. In J.  Arabski (Ed.), Challenging tasks for psycholinguistics in the new century (pp. 22–31). Katowice, Poland: University of Silesia Press. Paradis, M. 2008a. Bilingualism and neuropsychiatric disorders. Journal of Neurolinguistics, 21: 199–230. Paradis, M. 2008b. Language and communication disorders in multilinguals. In B. Stemmer & H. Whitaker (Eds), Handbook of the neuroscience of language (pp. 341–350). Amsterdam: Elsevier Science. Paradis, M. 2008c. Bilingual effects are not unique, only more salient. Bilingualism: Language and Cognition, 11: 181–183. Paradis, M. 2008d. Bilingual laterality: Unfounded claim of validity. A comment on Hull and Vaid (2007). Neuropsychologia, 46: 1588–1590. Paradis, M., & Goldblum, M.-C. 1989. Selective crossed aphasia in one of a trilingual’s languages followed by antagonistic recovery. Brain and Language, 36: 62–75.

 Declarative and procedural determinants of second languages Paradis, M., Goldblum, M.-C, & Abidi, R. 1982. Alternate antagonism with paradoxical translation behavior in two bilingual aphasic patients. Brain and Language, 15: 55–69. Paradis, M., & Gopnik, M. 1994. Compensatory strategies in familial language impairment. McGill Working Papers in Linguistics, 10: 142–149. Paradis, M., & Gopnik, M. 1997. Compensatory strategies in genetic dysphasia:  Declarative memory. Journal of Neurolinguistics, 10: 173–185. Paradis, M., Hagiwara, H., & Hildebrandt, H. 1985. Neurolinguistic aspects of the Japanese writing system. New York: Academic Press. Paradis, M. & Lecours, A.R. 1979. L’ Aphasie chez les bilingues et les polyglottes. In A.R. Lecours, F. Lhermitte et al., L’ aphasie. Paris: Flammarion. Parker, S.P. (Ed.), 2005. McGraw Hill encyclopedia of science and technology. New York: McGraw Hill. Pascual-Leone, A., & Walsh, V. 2001. Fast backprojections from the motion to the primary visual area necessary for visual awareness. Science, 292: 510–512. Patkowski, M.S. 1990. Age and accent in a second language: A reply to James Emil Flege. Applied Linguistics, 11: 73–89. Peck, E. 1973. The relationship of disease and other stress on second language. International Journal of Social Psychiatry, 20: 128–133. Perani, D., & Abutalebi, J. 2005. The neural basis of first and second language processing. Current Opinion in Neurobiology, 15: 202–205. Perani, D., Abutalebi, J., Paulesu, E., Brambati, S., Scifo, P., Cappa, S.F., & Fazio, F. 2003. The role of age of acquisition and language usage in early, high-proficient bilinguals: An fMRI study during verbal fluency. Human Brain Mapping, 19: 170–182. Perani, D., Dehaene, S., Grassi, F., Cohen, L., Cappa, S., Paulesu, E., Dupoux, E., Fazio, F., & Mehler, J. 1996. Brain processing of native and foreign languages. NeuroReport, 7: 2439–2444. Perani, D., Paulesu, E., Sebastián-Gallés, N., Dupoux, E., Dehaene, D., Bettinardi, V., Cappa, S., Fazio, F., & Mehler, J. 1998. The bilingual brain: Proficiency and age of acquisition of the second language. Brain, 121: 1841–1852. Piggott, G.L., & Kessler Robb, M. 1999. Prosodic features of familial language impairment:  Constraints on stress assignment. Folia Phoniatrica et Logopaedica, 51: 55–69. Pillai, J.J., Allison, J.D., Sethuraman, S., Araque, J.M., Thiruvaiyaru, D., Ison, C.B., Loring, D.W., & Lavin, T. 2004. Functional MR imaging study of language-related differences in bilingual cerebellar activation. American Journal of Neuroradiology, 25: 523–532. Piller, I. 2002. Passing for a native speaker: Identity and success in second language learning. Journal of Sociolinguistics, 6: 179–206. Pinker, S., & Ullman M.T. 2003. Beyond one model per phenomenon  – Response. Trends in Cognitive Neurosciences, 7: 108–109. Pitres, A. 1895. Etude sur l’aphasie chez les polyglottes. Revue de Médecine, 15: 873–899. Pitta, P., Marcos, L.C., & Alpert, M. 1978. Language switching as a treatment strategy with bilingual patients. The American Journal of Psychoanalysis, 38: 255–258. Plihal, W., & Born, J. 1997. Effects of early and late nocturnal sleep on declarative and procedural memory. The Journal of Cognitive Neuroscience, 9: 534–547. Price, C., Green, D., & von Studnitz, R. 1999. A functional imaging study of translation and language switching. Brain, 122: 2221–2235. Poldrack, R.A., Clark, J., Paré-Blagoev, E.J., Shohami, D., Moyano, J.C., Myers, C., & Gluck, M.A. 2001. Interactive memory systems in the human brain. Nature, 414: 546–550. Poldrack, R.A., Prabhakaran, V., Seger, C.A., & Gabrieli, D.E. 1999. Striatal activation of a cognitive skill. Neuropsychology, 13: 564–574.



References 

Posner, M.I., & Rothbart, M.K. 1992. Attentional mechanisms and conscious experience. In A.D. Milner & M.D. Rugg (Eds), The neuropsychology of consciousness (pp. 91–111). London: Academic Press. Pu, Y.L., Liu, H.L., Spinks, J.A., Mahankali, S., Xiong, J.H., Feng, C.M., Tan, L.H., Fox, P.T., & Gao, J.H. 2001. Cerebral hemodynamic response in Chinese (first) and English (second) language processing revealed by event-related functional MRI. Magnetic Resonance Imaging, 19: 643–647. Rajah, M.N., Ames, B., & D’Esposito, M. 2008. Prefrontal contributions to domain-general executive control processes during temporal context retrieval. Neuropsychologia, 46: 1088–1103. Rajah, M.N., & McIntosh, A.R. 2006. Dissociating prefrontal contributions during a recency memory task. Neuropsychologia, 44: 350–64. Rangamani, G.N. 1989. Aphasia and multilingualism: Clinical evidence towards the cerebral organization of language. Unpublished Ph.D. dissertation, The University of Mysore. Rapport, R.L., Tan, C.T., & Whitaker, H. 1983. Language function and dysfunction among Chinese- and English-speaking polyglots: Cortical stimulation, Wada testing and clinical studies. Brain and Language, 18: 342–366. Rauch, S.L., Whalen, P.J., Savage, C.R., Curran, T., Kendrick, A., Brown, H.D., Bush, G., Breiter, H.C., & Rosen, R.R. 1997. Striatal recruitment during an implicit sequence learning task as measured by functional magnetic resonance imaging. Human Brain Mapping, 5: 124–132. Ravel, S., & Richmond, B.J. 2005. Where did the time go? Nature Neuroscience, 8: 705–707. Reber, A.S. 1976. Implicit learning of synthetic languages: The role of instructional set. Journal of Experimental Psychology: Human Learning and Memory, 2: 88–94. Reber, A.S. 1997. Implicit ruminations. Psychonomic Bulletin and Review, 4: 49–55. Reber, A.S., Walkenfeld, F.F., & Hernstadt, R. 1991. Implicit and explicit learning:  Individual differences and IQ. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17: 888–896. Riccardi, A., Fabbro, F., & Obler, L.K. 2004. Pragmatically appropriate code-switching in a quadrilingual with Wernicke’s aphasia. Brain and Language, 91: 54–55. Rieber, R.W., & Vetter, H. 1995. Psychopathology and the polyglot. In R.W. Rieber & H. Vetter, The psychopathology of language and cognition (Ch. 8, pp. 151–178). New York: Plenum Press. Riese, W. 1928. Po povodu odnogo slushaia poliglotishnoi afasii. Deutsch-Russische Zeitschrift, 6: 262–265. Translated in Paradis (ed.) (1983), 280–284. Robinson, P. 2002. Effects of individual differences in intelligence, aptitude and working memory on adult incidental SLA: A replication and extension of Reber, Walkenfeld and Hernstadt (1991). In P.  Robinson (Ed.), Individual differences and instructed language learning (pp. 211–265). Amsterdam: John Benjamins. Rodriguez-Fornells, A., van der Lugt, A., Rotte, M., Britti, B., Heinze, H.J., & Munte, T.F. 2005. Second language interferes with word production in fluent bilinguals: Brain potential and functional imaging evidence. Journal of Cognitive Neuroscience, 17: 422–433. Roehr, K. 2008a. Linguistic and metalinguistic categories in second language learning. Cognitive Linguistics, 19: 67–106. Roehr, K. 2008b. Metalinguistic knowledge and language-analytic ability in university-level L2 learners. Applied Linguistics, 29: 173–199. Rönnberg, J., Rudner, M., & Ingvar, M. 2004. Neural correlates of working memory for sign language. Cognitive Brain Research, 20: 165–182. Ross, D.S., & Bever, T.G. 2004. The time course for language acquisition in biologically distinct populations: Evidence from deaf individuals. Brain and Language, 89: 115–121.

 Declarative and procedural determinants of second languages Rossi, E., Denes, G., & Bastiaanse, R. 2003. A single case study of pathological mixing in a polyglot aphasic. Brain and Language, 87: 46–47. Roux, F.-E., & Lubrano, V. 2006. Language studied by electrostimulation mapping in bilingual patients. In F.J. Chen (Ed.), Brain mapping and language (pp. 111–133). New York: Nova Science Publishers. Roux, F.-E., Lubrano, V., Lauwers-Cances, V., Trémoulet, M., Mascott, C.R., & Demonet, J.F. 2004. Intra-operative mapping of cortical areas involved in reading in mono- and bilingual patients. Brain, 127: 1796–1810. Roux, F.-E., & Trémoulet, M. 2002. Organization of language areas in bilingual patients: a cortical stimulation study. Journal of Neurosurgery, 97: 857–864. Ruiz, C., Paredes, N., Macizo, P., & Bajo, M.T. 2008. Activation of lexical and syntactic target language properties in translation. Acta Psychologica, 128: 490–500. Rüschemeyer, S.A., Fiebach, C.J., Kempe, V., & Friederici, A.D. 2005. Processing lexical semantic and syntactic information in first and second language: fMRI evidence from German and Russian. Human Brain Mapping, 25: 266–286. Sabourin, L. 2006. Does the shallow structures proposal account for qualitative differences in first and second language processing? Applied Psycholinguistics, 27: 81–84. Salameh, E.K. 2006. Linguistic and cultural factors affecting assessment and intervention for bilingual children with specific language impairment. Keynote address, 6th European CPLOL Congress, Berlin, Germany, 16 September. Sanders, L.D., & Neville, H.J. 2003. An ERP study of continuous speech processing II. Segmentation, semantics, and syntax in non-native speakers. Cognitive Brain Research, 15: 214–227. Santiago-Rivera, A.L., & Altarriba, J. 2002. The role of language in therapy with the SpanishEnglish bilingual client. Professional Psychology Research and Practice, 33: 30–38. Schachter, D.L., Badgaiyan, R.D., & Alpert, N.M. 1999. Visual stem completion priming within and across modalities: A PET study. NeuroReport, 10: 2061–2065. Scherag, A., Demuth, L., Rösler, F., Neville, H.J., & Röder, B. 2004. The effects of late acquisition of L2 and the consequences of immigration on L1 for semantic and morpho-syntactic language aspects. Cognition, 93: B97-B108. Scherer, L.C., Giroux, F., Lesage, F., Senhadji, N., Benali, H., & Ansaldo, A.I. 2006. An optical imaging study of semantic and syntactic processing by bilinguals. Brain and Language, 99: 197–198. Schmidt, R. 1993. Awareness and second language acquisition. Annual Review of Applied Linguistics, 13: 206–226. Schmidt, R. 1994. Deconstructing consciousness in search of useful definitions for applied linguistics. AILA Review, 11: 11–26. Schmidt, R. 2001. Attention. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3–32). Cambridge: Cambridge University Press. Schumann, J.H. 1998. The neurobiology of affect in language. Language Learning, 48: Supplement 1. Scott Kelso, J.A. 2002. The complementary nature of coordination dynamics: Self-organization and agency. Nonlinear Phenomena in Complex Systems, 5: 364–371. Sebastián-Gallés, N., & Bosch, L. 2001. On becoming and being bilingual. In E. Dupoux (Ed.), Language, mind and cognitive development (pp. 379–393). Cambridge, MA: MIT Press. Segalowitz, N. 2003. Automaticity and second languages. In C.J. Doughty & M.H. Long (Eds), The handbook of second language acquisition (pp. 381–408). Oxford: Blackwell.



References 

Segalowitz, N. & Segalowitz, S. 1993. Skilled performance, practice, and the differentiation of speed-up from automatization effects: Evidence from second language word recognition. Applied Psycholinguistics, 14: 369–385. Segalowitz, S.J. 1986. Validity and reliability of noninvasive lateralization measures. In J.E. Obrzut & G.W. Hynd (Eds), Child neuropsychology, vol. 1 (pp. 191–208). Orlando, FL: Academic Press. Segalowitz, S.J., & Bryden, M.P. 1983. Individual differences in hemispheric representation of language. In S.J. Segalowitz (Ed.), Language functions and brain organization (pp. 341–372). New York: Academic Press. Segalowitz, S., Segalowitz, N., & Wood, A. 1998. Assessing the development of automaticity in second language word recognition. Applied Psycholinguistics, 19: 53–67. Seger, C.A., & Cincotta, C.M. 2005. The roles of the caudate nucleus in human classification learning. The Journal of Neuroscience, 25: 2941–2951. Seidenberg, M.S., & Zevin, J.D. 2006. Connectionist models in developmental cognitive neuroscience: Critical periods and the paradox of success. In Y. Munakata & M.H. Johnson (Eds), Processes of change in brain and cognitive development (pp. 585–612). Oxford: Oxford University Press. Senghas, A., & Coppola, M. 2001. Children creating language: How Nicaragua Sign Language acquired a spatial grammar. Psychological Science, 12: 323–328. Sharwood Smith, M. 1978. Applied linguistics and the psychology of instruction:  A case for transfusion. Studies in Second Language Acquisition, 1: 91–117. Sharwood Smith, M. 1981. Consciousness raising and the second language learner. Applied Linguistics, 2: 159–168. Sidtis, J.J. 2007a. Activation: Will hot spots get the hot field in hot water? Brain and Language, 102: 127–129. Sidtis, J.J. 2007b. Some problems for representations of brain organization based on activation in functional imaging. Brain and Language, 102: 130–140. Silveri, M.C., Leggio, M.G., & Molinari, M. 1994. The cerebellum contributes to linguistic production:  A case of agrammatic speech following a right cerebellar lesion. Neurology, 44: 2047–2050. Simos, O., Castillo, E., Francis, J.  et al. 2001. Mapping receptive language cortex in bilingual volunteers by using magnetic source imaging. Journal of Neurosurgery, 95: 76–81. Singleton, D. 1989. Language acquisition: The age factor. Clevedon, UK: Multi-lingual Matters. Singleton, D. 2001. Age and second-language acquisition. Annual Review of Applied Linguistics, 21: 77–89. Slabakova, R. 2006. Is there a critical period for semantics? Second Language Research, 22: 302–338. Smith N., & Tsimpli, I.M. 1995. The mind of a savant. Language learning and modularity. Oxford: Blackwell. Solin, D. 1989. The systematic misrepresentation of bilingual crossed aphasia data and its consequences. Brain and Language, 36: 92–116. Spada, N. 1997. Form-focused instruction and second language acquisition: A review of classroom and laboratory research. Language Teaching, 29: 1–15. Sperry, R.W. 1974. Lateral specialization in surgically separated hemispheres. In F.O. Schmitt & F.G. Worden (Eds), The neurosciences, third study program (pp. 5–19). Cambridge, MA: MIT Press. Stewart, C. 2002. Memory. In E. Perry, H. Ashton, & A. Young (Ed.), Neurochemistry of consciousness (pp. 65–81). Amsterdam: John Benjamins.

 Declarative and procedural determinants of second languages Stowe, L.A., & Sabourin, L. 2005. Imaging the processing of a second language: Effects of maturation and proficiency on the neural processes involved. IRAL, 43: 329–353. Suh, S., Yoon, H.W., Lee S, Chung, J-Y, Cho, Z-H., & Park, H-W. 2007. Effects of syntactic complexity in L1 and L2; an fMRI study of Korean-English bilinguals. Brain Research, 1136: 178–189. Sundara, M., & Polka, L. 2008. Discrimination of coronal stops by bilingual adults: The timing and nature of language interaction. Cognition, 106: 234–258. Tabor-Connor, L., DeShazo Brady, T., Snyder, A.Z., Lewis, C., Blasi, V., & Corbetta, M. 2006. Cerebellar activity switches hemispheres with cerebral recovery in aphasia. Neuropsychologia, 44: 171–177. Tan, L.H., Feng, C.M., Liu, H.L., Shen, G., & Gao, J.H. 2001. Semantic representations in the bilingual brain. NeuroImage, 13 (6 Part 2): S611. Tan, L.H., Spinks, J.A., Feng, C.M., Siok, W.T., Perfetti, C.A., Xiong, J.H., Fox, P.T., & Gao, J.H. 2003. Neural systems of second language reading are shaped by native language. Human Brain Mapping, 18: 158–166. Tatsuno, Y., & Sakai, K.L. 2005. Language-related activations in the left prefrontal regions are differentially modulated by age, proficiency, and task demands. The Journal of Neuroscience, 25: 1637–1644. Tavano, A., De Fabritiis, P., & Fabbro, F. 2005. Contributo alla valutazione standardizzata dell’eloquio narrativo nei bambini. Giornale di Neuropsichiatria dell’Età Evolutiva, 25: 48–64. Teichmann, M., Dupoux, E., Kouider, S., Brugières, P., Boissé, M.F., Baudic, S., Cesaro, P., Peschanski, M., & Bachoud-Lévi, A.C. 2005. The role of the striatum in rule application: The model of Huntington’s disease at early stage. Brain, 128: 1155–1167. Tham, W.P., Rixkard Liow, S.J., Rajapakse, J., Leong, T.C., Ng, S.E.S., Lim, W.E.H., & Ho, L.G. 2005. Phonological processing in Chinese-English bilingual biscriptals:  An fMRI study. NeuroImage, 28: 579–587. Tomioka, N. 2002. How to distinguish procedural and declarative memory use in L2 processing with a behavioral test. Paper presented at the International Linguistics Association 47th Annual Conference, York University, Toronto, 6 April. Tomlin, R., & Villa, V. 1993. Attention in cognitive science and SLA. Technical report No 93–12, University of Oregon, Institute of Cognitive and Decision Science. (Cited by Schmidt, 1994.) Toppelberg, C.O. 1997. Biopsychosocial model. American Journal of Psychiatry, 154: 1328. Toribio, A.J. 2001. On the emergence of bilingual code-switching competence. Bilingualism:  Language and Cognition, 4: 203–231. Tremblay, P., & Gracco, V.L. 2006. Contribution of the frontal lobe to externally and internally specified verbal responses: fMRI evidence. NeuroImage, 33: 947–957. Ullman, M.T. 2001. The neural basis of lexicon and grammar in first and second language: The declarative/procedural model. Bilingualism: Language and Cognition, 4: 105–122. Ullman, M.T. 2004. Contributions of memory circuits to language: The declarative/procedural model. Cognition, 92: 231–270. Ullman, M.T. 2005. A cognitive neuroscience perspective on second language acquisition: The declarative/procedural model. In Sanz, C. (Ed.), Mind and context in adult second language acquisition:  Methods, theory, and practice (pp. 141–178). Washington, DC:  Georgetown University Press. Ullman, M.T. 2006a. The declarative/procedural model and the shallow structure hypothesis. Applied Psycholinguistics, 27: 97–105.



References 

Ullman, M.T. 2006b. Is Broca’s area part of a basal ganglia thalamocortical circuit? Cortex, 42: 480–485. Ullman, M.T., & Gopnik, M. 1994. Past tense production: Regular, irregular and nonsense verbs. McGill Working Papers in Linguistics, 10: 81–118. Ullman, M., & Gopnik, M. 1999. Inflectional morphology in a family with inherited specific language impairment. Applied Psycholinguistics, 20: 51–117. Ullman, M.T., Pancheva, R., Love, T., Yee, E., Swinney, D., & Hickok, G. 2005. Neural correlates of lexicon and grammar: Evidence from the production, reading, and judgment of inflection in aphasia. Brain and Language, 93: 185–238. Unsworth, S. 2004. On the syntax-semantics interface in Dutch:  Adult and child acquisition compared. International Review of Applied Linguistics, 42: 173–187. Urbanik, A., Binder, M., Sobiecka, B., & Kozub, J. 2001. fMRI study of sentence generation by early bilinguals differing in proficiency level. Rivista di Neuroradiologia, 14: 11–16. Urponen, M.I. 2004. Ultimate attainment in postpuberty second language acquisition. Unpublished doctoral dissertation, Boston University. Vaid, J. 1983. Bilingualism and brain lateralization. In S. Segalowitz (Ed.), Language functions and brain organization (pp. 315–339). New York: Academic Press. Vaid, J. & Genesee, F. 1980. Neuropsychological approaches to bilingualism: A critical review. Canadian Journal of Psychology. 34: 417–445. Vaid, J. & Hall, D.G. 1991. Neuropsychological perspectives on bilingualism:  Right, left, and center. In A. Reynolds (Ed.), Bilingualism, multiculturalism and second language learning (pp. 81–112). Hillsdale, NJ: Lawrence Erlbaum Associates. Van der Linden, M. 1993. L’ apprentissage de nouvelles connaissances sémantiques chez le patient amnésique. In M.-T. Claes & J. Chapelle (Eds), Proceedings of the 1st International Congress on Memory and memorization in acquiring and learning languages (pp. 349–360). Louvain: C.C.L. Van der Linden, M., Meulemans, T., & Lorrain, D. 1994. Acquisition of new concepts by 2 amnesic patients. Cortex, 30: 305–317. van Gorp, W.G., Altshuler, L., Theberge, D.C., & Mintz, J. 1999. Declarative and procedural memory in bipolar disorder. Biological Psychiatry, 46: 525–31. van Hout, R., Hulk, A., & Kuiken, F. 2003. The interface. Concluding remarks. In R. van Hout, A.  Hulk, F.  Kuiken, & R.  Towell (Eds), The lexicon-syntax interface in second language acquisition (pp. 219–226). Amsterdam: John Benjamins. VanPatten, B. 2002. Processing instruction: An update. Language Learning, 52: 755–804. Vargha-Khadem, F., Gadian, D.G., Copp, A., & Mishkin, M. 2005. FOXP2 and the neuroanatomy of speech and language. Nature Reviews Neuroscience 6: 131–138. Veyrac, G-J. 1931. Etude de l’aphasie chez les sujets polyglottes. Thèse pour le doctorat en médecine, Paris. Translated in Paradis (ed.) (1983), 320–338. Vingerhoets, G., Van Borsel, J., Tesink, C., van den Noort, M., Deblaere, K., Seurinck, R., Vandemaele, P., & Achten, E. 2003. Multilingualism:  An fMRI study. NeuroImage, 20: 2181–2196. Vygotsky, L.S. 1978. Mind in society. The development of higher psychological processes. Cambridge, MA : Harvard University Press. Walker, J.A., Quiñones-Hinojosa, A., & Berger, M.S. 2004. Interoperative speech mapping in 17 bilingual patients undergoing resection of a mass lesion. Neurosurgery, 54: 113–118. Walter, H. 1988. Le français dans tous les sens. Paris: Robert Laffont.

 Declarative and procedural determinants of second languages Walters, J. 2005. Bilingualism:  The sociopragmatic-psycholinguistic interface. Mahwah, NJ: Lawrence Erlbaum Associates. Wang, Y., Sereno, J.A., Jongman, A., & Hirsch, J. 2003. fMRI evidence for cortical modification during learning of Mandarin lexical tone. Journal of Cognitive Neuroscience, 15: 1019–1027. Wang, Y., Xue, G., Chuansheng, C., Xue, F., & Dong, Q. 2007. Neural bases of asymmetric language switching in second-language learners:  An ER-fMRI study. NeuroImage, 35: 862–870. Ward, M.E. & Marshall, J.C. 1999. “Speaking in tongues”: Paradoxical fixation on a non-native language following anaesthesia. Anaesthesia, 54: 1201–1203. Watkins, K.E., Vargha-Khadem, F., Ashburner, J., Passingham, R.E., Connelly, A., Friston, K.J., Frackowiak, R.S.J., Mishkin, M., & Gadian, D.G. 2002. MRI analysis of an inherited speech and language disorder: Structural brain abnormalities. Brain, 125: 465–478. Weber-Fox, C.M., & Neville, H.J. 1996. Maturational constraints on functional specializations for language processing:  ERP and behavioral evidence in bilingual speakers. Journal of Cognitive Neuroscience, 8: 231–256. Weber-Fox, C.M., & Neville, H.J. 2001. Sensitive periods differentiate processing of open- and closed-class words: An ERP study of bilinguals. Journal of Speech, Language, and Hearing Research, 44: 1338–1353. Weisenburg, T.H., & McBride, K.E. 1935. Aphasia, a clinical and psychological study [Case 4, pp. 160–182]. New York: Commonwealth Fund. White, L. 1992. On triggering data in L2 acquisition: A reply to Schwartz and Gubala-Ryzak. Second Language Research, 8: 120–137. White, S.A., Fisher, S.E., Geschwind, D.H., Scharff, C., & Holy, T.E. 2006. Singing mice, songbirds, and more:  Models for FOXP2 function and dysfunction in human speech and language. The Journal of Neuroscience, 26: 10376–10379. Williams, J.N. 2004. Implicit learning of form-meaning connections. In B. VanPatten, J. Williams, S. Rott, & M. Overstreet (Eds), Form-meaning connections in second language acquisition (pp. 203–218). Mahwah, NJ: Lawrence Erlbaum Associates. Williams, J.N. 2005. Learning without awareness. SSLA, 27: 269–304. Wilson, S.M., & Saygin, A.P. 2004. Grammaticality judgment in aphasia: Deficits are not specific to syntactic structures, aphasic syndromes, or lesion sites. Journal of Cognitive Neuroscience, 16: 238–252. Xue, G., Dong, Q., Jin, Z., & Chen, C. 2004a. Mapping of verbal working memory in nonfluent Chinese-English bilinguals with functional MRI. NeuroImage, 22: 1–10. Xue, G., Dong, Q., Jin, Z., & Wang, Y. 2004b. An fMRI study with semantic access in low proficiency second language learners. NeuroReport, 15: 791–796. Yazaki-Sugiyama, Y., Kushner, J., Hessler, N.A., & Hensch, T.K. 2007. Early GABA function regulates sensory critical period during birdsong learning. Neuroscience Research, 58: S173. Yetkin, O., Yetkin, Z., Haughton, V., & Cox, R.W. 1996. Use of functional MR to map language in multilingual volunteers. American Journal of Neuroradiology, 17: 473–477. Yokoyama, S., Okamoto, H., Miyamoto T, Yoshimoto, K., Kim, J., Iwata, K., Jeong, H., Uchida, S., Ikuta, N., Sassa, Y., Nakamura, W., Horie, K., Sato, S., & Kawashima, R. 2006. Cortical activation in the processing of passive sentences in L1 and L2: An fMRI study. NeuroImage, 30: 570–579. Young, G.B., & Pigott, S.E. 1999. Neurobiological basis of consciousness. Archives of Neurology, 56: 153–157.



References 

Yuan, F., & Ellis, R. 2003. The effects of pre-task planning and on-line planning on fluency, complexity and accuracy in L2 production. Applied Linguistics, 24: 1–27. Zanini, S., Tavano, A., Vorano, L., Schiavo, F., Gigli, G.L., Aglioti, S.M., & Fabbro, F. 2004. Greater syntactic impairments in native language in bilingual Parkinsonian patients. Journal of Neurology, Neurosurgery and Psychiatry, 75: 1678–1681. Zeman, A. 2005. What in the world is consciousness? In S.  Laureys (Ed.), The boundaries of consciousness: Neurobiology and neuropathology (pp. 1–10). Amsterdam: Elsevier. Zettin, M., Cappa, S.F., D’Amico, A., Rago, R., Perino, C., Perani, D., Fazio, F. 1997. Agrammatic speech production after right cerebellar hemorrhage. Neurocase, 3: 375–380.

Subject index

A accuracy  6, 7, 32, 38, 116, 119, 123, 124, 174, 189 acquisition  1, 4, 8, 10, 11, 13, 14, 16, 20, 21, 23, 25–27, 31, 34, 37, 39, 40, 49, 51–53, 55–57, 59, 60, 62–64, 66, 67, 69, 73–75, 78–80, 82, 83–89, 93, 94–102, 104–106, 109, 111, 113–125, 127, 130, 131–135, 137, 138, 143, 150, 152, 153, 169, 170, 173–175, 178, 179, 183, 186, 187, 189, 190 affect  3, 11, 13, 35, 43, 59, 61, 67, 72, 93, 96, 98, 101, 104, 111, 120, 135, 152, 171, 178, 186, 189 Alzheimer  111, 174, 175 appropriation  1, 4, 6, 13, 21, 22, 25, 33, 37, 40, 63, 65, 83, 96, 98, 101, 103, 110, 116, 119, 120, 122, 123–125, 129, 132, 134, 135, 142, 169, 183, 186, 189 aphasia  139, 158, 161, 170–174, 176, 178, 184 attention  4, 10, 23, 27, 29, 30, 35, 37–39, 41, 42, 45, 50, 51–55, 60, 67, 69, 73, 78, 89, 91, 96, 97, 99, 100, 106, 111, 112, 121, 125, 130, 132, 141, 142, 150, 151, 153, 154, 156–158, 160, 162, 163, 167–171, 185–187, 189 attrition  134, 135 automatic  4, 5–9, 12, 19, 22, 23–27, 30–32, 35, 38, 39, 44, 48, 56, 58, 60, 65, 69, 70–72, 76, 81, 86, 87–89, 93, 94, 97, 98, 100, 102–104, 111, 113, 119,

120–122, 124–126, 130, 133–135, 142, 143, 147–151, 154–160, 162–166, 168, 169, 173, 175, 182, 185, 186–188 awareness  4, 14, 27, 37–41, 43, 45–51, 55, 59, 68, 74, 75, 77, 89, 90–92, 97, 99, 100, 102, 106, 126, 161, 166, 168, 187 B basal ganglia  7, 11, 30, 70, 94, 114, 134, 140, 145, 150, 162, 163, 168, 169, 171, 174, 185, 187, 188 borrowing  20, 156, 157, 160–162, 164–166, 169 C cerebellum  11, 94, 114, 126, 134, 140, 145, 147, 149, 150, 163, 171, 185, 187, 188 cingulate  7, 19, 94, 120, 123, 140, 142, 145, 147, 149, 151, 155, 158, 159–161, 163, 167, 168, 179, 185, 187, 189 compensatory  114, 133, 134, 150, 153, 186 comprehension  4, 5, 8, 12, 25, 29, 30, 56, 63, 65, 90, 98, 110, 116, 119, 123, 125, 146, 149, 155, 164, 167, 174, 180–182, 188 computer  61, 62, 67, 188 concept  1, 11, 15, 45, 78, 104, 112, 133 conscious  1, 2, 6, 8, 12, 14, 15, 17, 19, 21, 26, 27, 30–32, 34, 37–51, 53, 55, 56, 58–60, 62–66, 68, 70–76, 78, 80, 81, 83–87, 89, 90–96, 98–100, 102, 103–105, 107, 110, 112–114, 117, 118, 120,

121, 125, 131, 134, 145, 149, 154–164, 167, 168–170, 172, 176, 177, 179, 181, 184, 185–189 continuum  9, 26–28, 35, 68, 69 control  2, 6, 8, 19, 26, 27, 29, 30, 32, 37–39, 41, 48, 59, 68, 70, 71, 74, 76, 86, 87, 95, 100, 109, 111–115, 117, 119–121, 123–125, 128–130, 141–143, 147, 149–151, 153, 154, 157–171, 174, 175, 177, 180, 182, 185, 186, 188, 189 critical period  113, 127, 129 (see also Optimal period) D deliberate  4, 8, 10, 26, 27, 35, 60, 70, 155–157, 159–161, 163, 164–166, 168, 169, 186 E entrenchment  119, 133–135 explicit  1, 4, 5–9, 11, 12–14, 16, 17–19, 21, 22–26, 28–32, 34, 35, 37, 38, 40, 41–53, 56–80, 82, 83–101, 103–107, 109–111, 113, 114, 116, 118, 119–123, 129, 130, 132–134, 137, 143, 145, 146–148, 151, 153, 154, 157, 159, 161, 164, 166–170, 173, 175, 177, 178, 183–189 F fluency  6, 7, 34, 68, 85, 119, 121, 123, 124, 141, 144, 150, 190 FOXP2  10, 11, 113–115, 118, 126, 127, 189 G genetic dysphasia  8, 10, 18, 21, 126, 133, 176

 Subject index H hippocampal system  7, 9, 140, 142, 145, 147, 172, 185, 187, 188 Huntington’s  20, 175 I implicit (linguistic) competence  ix–xi, 1–9, 11, 12–17, 19, 20, 22–26, 28–35, 37, 39–45, 48–50, 52, 53, 55–57, 59, 61–82, 84–94, 96–99, 101–106, 110–115, 117–119, 123, 124, 126, 127, 129–136, 142, 143, 145–147, 149, 151–155, 157, 162–164, 166, 179–172, 174, 176–179, 181, 184–190 incidental  4, 8, 21, 26, 27, 33, 34, 51, 53, 60, 99, 105, 110, 113, 123, 127, 130–132, 134, 189 inference  142, 148, 151 inhibition  44, 45, 157, 158, 160, 162–166, 169, 180, 183 input  5, 11, 37–39, 42–44, 47–57, 59, 60, 62, 66, 67, 71, 73, 78, 80, 83–85, 90, 94, 100, 101, 105, 106, 114, 124, 125, 152, 155, 187, 189 intake  5, 37, 39, 51, 53–57, 59, 60, 67, 71, 78, 80, 83, 84, 90, 94, 96, 105, 187 interface as indirect influence  61, 63, 64, 67, 68, 76, 78, 80, 96–99, 101, 102, 106 dynamic  61, 65, 66–68, 88, 89, 104 in consciousness  68, 71, 73, 74, 76, 82, 98, 104, 105, 187 interference dynamic  173 static  16 L laterality  126, 137, 138–140, 148, 170, 184, 185 learning  1, 4, 5, 8, 10, 13, 14, 20, 21–23, 26, 27, 31, 34, 39, 40, 49, 51–53, 55, 56, 60, 63, 64, 66, 67, 69,

70, 74, 75, 78, 80–88, 93–95, 98, 99–101, 105, 106, 110, 113, 114–118, 120, 121–123, 127, 128–134, 138, 145, 146, 150, 153, 160, 170, 173, 174, 176, 177, 183, 187, 189 lexicon  5, 9, 12, 14–18, 20–22, 34, 35, 51, 61, 112, 114, 118, 139, 148, 172, 179, 182, 184, 185, 188, 190 M memory declarative  xi, xii, 1, 8, 9–19, 21–27, 30, 31, 34, 35, 37, 47, 51, 55, 58, 66, 71, 86, 90, 94, 95, 104, 106, 110, 111, 113, 114, 118, 120, 122, 123, 125, 126–132, 134, 135, 139, 140, 145, 146, 149, 152, 153, 155, 157–161, 172, 173–177, 179–183, 185, 188, 189 episodic  120, 147 implicit  4, 9, 10, 24, 34, 50, 58, 66, 67, 70, 95, 103, 173, 176, 184 procedural  xi, xii, 1, 3, 9, 10–14, 16, 21, 22–24, 26, 28, 30, 31, 33–35, 38–40, 47, 58, 70, 95, 110, 113, 118–120, 122, 123, 126, 127, 129–133, 135, 137, 139, 147, 152, 155–157, 160, 164, 172, 175, 181, 182–184, 186 working  23, 49, 50, 70, 85-87, 99, 116, 118, 120, 121, 130-132, 135, 146, 151, 153, 167, 168, 182, 189 mixing  155, 156–158, 161, 162, 164, 167, 169, 186 modularity  58, 101 module  3, 21, 24, 30, 72, 76, 90, 132, 164, 175 monitor  70, 79, 80, 82, 94, 97–99, 145, 182 N native speaker  24, 103, 111, 117, 125, 182 neuroimaging  ix, xi, 2, 7, 11, 19, 20, 33, 70, 137, 140, 141,

143, 145–147, 152, 153, 155, 157, 158–160, 162, 166, 169, 184–186, 189 neuropsychiatric  169 O optimal period  24, 52, 109, 113, 114, 116, 117, 124, 126, 129, 132, 133–135, 189 (see also critical period) output  29, 31, 40, 43, 44, 46, 47–49, 52, 61, 62, 65, 67, 68, 70, 73, 74, 78, 79, 85, 86, 90, 93–95, 97–99, 101, 104, 115, 117, 123, 133, 134, 136, 145, 152, 157, 168, 182, 189, 190 P parameter  96, 127, 173 Parkinson’s  20, 175 pragmatics  3, 7, 24, 29, 34, 73, 81, 82, 111, 112, 126, 132, 143, 145, 150, 152–156, 158, 161, 162, 178, 179, 185 priming  3, 18, 46, 122, 184 S sensitive period  110, 117 single word  18, 141, 150, 158 skill  24, 33, 74, 78, 86, 112, 128 SLI  10, 176, 177 speeded-up  6–8, 25, 26, 30, 40, 97, 99, 103, 114, 119, 124, 135, 136, 142, 147, 189, 190 striatum  xi, 70, 140, 145, 163, 168, 173, 187 subsystems  xii, 15, 33, 49, 58, 120–122, 146, 152, 156, 164–166, 177–180, 182 switching  26, 33, 38, 68, 99, 135, 155–169, 183, 186, 188, 189 systematicity  7, 35, 183, 186 T tag  165, 166 teaching  2, 22, 23, 100, 102, 137 thought  10, 13, 26, 47, 109, 128, 168

Subject index  threshold  ix, x, xii, 34, 41–47, 59, 90, 92, 134, 150, 151, 153, 155–157, 161, 162, 164–167, 180, 187 translation  17, 72, 163, 164, 180–182, 186 U universal grammar  51, 115 

utterance  3, 4, 19, 29, 31, 54, 56, 57, 70, 78, 81, 85, 94, 96, 100, 104, 112, 125, 145, 159, 161, 162, 179, 181, 186 V variability  7, 30, 113, 115, 118, 125, 130, 131, 135, 147, 153, 183, 186, 189

variation  7, 10, 30, 46, 118, 119, 147, 183 verbal communication  3, 23, 24, 29, 39, 49, 50, 60, 109, 126, 143, 152, 170, 187 vocabulary  5–7, 9, 12, 14–18, 20–22, 31, 33, 35, 51, 95, 109, 110, 114–116, 133, 134, 148, 171, 179, 182, 184, 185

In the series Studies in Bilingualism (SiBil) the following titles have been published thus far or are scheduled for publication: 40 Paradis, Michel: Declarative and Procedural Determinants of Second Languages. 2009. xii, 219 pp. 39 Montrul, Silvina A.: Incomplete Acquisition in Bilingualism. Re-examining the Age Factor. 2008. x, 312 pp. 38 Plaza-Pust, Carolina and Esperanza Morales-López (eds.): Sign Bilingualism. Language development, interaction, and maintenance in sign language contact situations. 2008. xvi, 389 pp. 37 Niño-Murcia, Mercedes and Jason Rothman (eds.): Bilingualism and Identity. Spanish at the crossroads with other languages. 2008. vii, 365 pp. 36 Hansen Edwards, Jette G. and Mary L. Zampini (eds.): Phonology and Second Language Acquisition. 2008. vi, 380 pp. 35 Rocca, Sonia: Child Second Language Acquisition. A bi-directional study of English and Italian tenseaspect morphology. 2007. xvi, 240 pp. 34 Koven, Michèle: Selves in Two Languages. Bilinguals' verbal enactments of identity in French and Portuguese. 2007. xi, 327 pp. 33 Köpke, Barbara, Monika S. Schmid, Merel Keijzer and Susan Dostert (eds.): Language Attrition. Theoretical perspectives. 2007. viii, 258 pp. 32 Kondo-Brown, Kimi (ed.): Heritage Language Development. Focus on East Asian Immigrants. 2006. x, 282 pp. 31 Baptista, Barbara O. and Michael Alan Watkins (eds.): English with a Latin Beat. Studies in Portuguese/Spanish – English Interphonology. 2006. vi, 214 pp. 30 Pienemann, Manfred (ed.): Cross-Linguistic Aspects of Processability Theory. 2005. xiv, 303 pp. 29 Ayoun, Dalila and M. Rafael Salaberry (eds.): Tense and Aspect in Romance Languages. Theoretical and applied perspectives. 2005. x, 318 pp. 28 Schmid, Monika S., Barbara Köpke, Merel Keijzer and Lina Weilemar (eds.): First Language Attrition. Interdisciplinary perspectives on methodological issues. 2004. x, 378 pp. 27 Callahan, Laura: Spanish/English Codeswitching in a Written Corpus. 2004. viii, 183 pp. 26 Dimroth, Christine and Marianne Starren (eds.): Information Structure and the Dynamics of Language Acquisition. 2003. vi, 361 pp. 25 Piller, Ingrid: Bilingual Couples Talk. The discursive construction of hybridity. 2002. xii, 315 pp. 24 Schmid, Monika S.: First Language Attrition, Use and Maintenance. The case of German Jews in anglophone countries. 2002. xiv, 259 pp. (incl. CD-rom). 23 Verhoeven, Ludo and Sven Strömqvist (eds.): Narrative Development in a Multilingual Context. 2001. viii, 431 pp. 22 Salaberry, M. Rafael: The Development of Past Tense Morphology in L2 Spanish. 2001. xii, 211 pp. 21 Döpke, Susanne (ed.): Cross-Linguistic Structures in Simultaneous Bilingualism. 2001. x, 258 pp. 20 Poulisse, Nanda: Slips of the Tongue. Speech errors in first and second language production. 1999. xvi, 257 pp. 19 Amara, Muhammad Hasan: Politics and Sociolinguistic Reflexes. Palestinian border villages. 1999. xx, 261 pp. 18 Paradis, Michel: A Neurolinguistic Theory of Bilingualism. 2004. viii, 299 pp. 17 Ellis, Rod: Learning a Second Language through Interaction. 1999. x, 285 pp. 16 Huebner, Thom and Kathryn A. Davis (eds.): Sociopolitical Perspectives on Language Policy and Planning in the USA. With the assistance of Joseph Lo Bianco. 1999. xvi, 365 pp. 15 Pienemann, Manfred: Language Processing and Second Language Development. Processability theory. 1998. xviii, 367 pp. 14 Young, Richard and Agnes Weiyun He (eds.): Talking and Testing. Discourse approaches to the assessment of oral proficiency. 1998. x, 395 pp. 13 Holloway, Charles E.: Dialect Death. The case of Brule Spanish. 1997. x, 220 pp. 12 Halmari, Helena: Government and Codeswitching. Explaining American Finnish. 1997. xvi, 276 pp. 11 Becker, Angelika and Mary Carroll: The Acquisition of Spatial Relations in a Second Language. In cooperation with Jorge Giacobbe, Clive Perdue and Rémi Porquiez. 1997. xii, 212 pp.

10 Bayley, Robert and Dennis R. Preston (eds.): Second Language Acquisition and Linguistic Variation. 1996. xix, 317 pp. 9 Freed, Barbara F. (ed.): Second Language Acquisition in a Study Abroad Context. 1995. xiv, 345 pp. 8 Davis, Kathryn A.: Language Planning in Multilingual Contexts. Policies, communities, and schools in Luxembourg. 1994. xix, 220 pp. 7 Dietrich, Rainer, Wolfgang Klein and Colette Noyau: The Acquisition of Temporality in a Second Language. In cooperation with Josée Coenen, Beatriz Dorriots, Korrie van Helvert, Henriette Hendriks, Et-Tayeb Houdaïfa, Clive Perdue, Sören Sjöström, Marie-Thérèse Vasseur and Kaarlo Voionmaa. 1995. xii, 288 pp. 6 Schreuder, Robert and Bert Weltens (eds.): The Bilingual Lexicon. 1993. viii, 307 pp. 5 Klein, Wolfgang and Clive Perdue: Utterance Structure. Developing grammars again. In cooperation with Mary Carroll, Josée Coenen, José Deulofeu, Thom Huebner and Anne Trévise. 1992. xvi, 354 pp. 4 Paulston, Christina Bratt: Linguistic Minorities in Multilingual Settings. Implications for language policies. 1994. xi, 136 pp. 3 Döpke, Susanne: One Parent – One Language. An interactional approach. 1992. xviii, 213 pp. 2 Bot, Kees de, Ralph B. Ginsberg and Claire Kramsch (eds.): Foreign Language Research in CrossCultural Perspective. 1991. xii, 275 pp. 1 Fase, Willem, Koen Jaspaert and Sjaak Kroon (eds.): Maintenance and Loss of Minority Languages. 1992. xii, 403 pp.