Integrating Neurolinguistics into Second Language Acguisition Research

DAVID SINGLETON

(Trinity College, University of Dublin, Ireland)

1.INTRODUCTION

The cognitive neuroscience of second language acquisition (SLA) has been described in a much-cited book on this topic as “a rapidly growing field of study, touching on a range of theoretical questions” and as “being associated with ‘excitement’ and a ‘sense of momentum’” (Indefrey and Gullberg 2006: 7).Indeed, there is absolutely no doubt but that neurolinguistics is one of the areas in linguistics which is attracting most interest at the present time.With regard, specifically, to second language acquisition research, neuroscientists-from Penfield onwards—have always volunteered insights of relevance to this area, and these insights have sometimes found their way into mainstream discussion.However, the recent advent of brain-imaging technology, with its promise of the possibility of finally making the link between the psychological and the physical dimensions of cognition, has now rendered the appeal of the neurosciences all but irresistible to researchers in the language acquisition/processing domain, and has made it virtually impossible to ignore the neuroscientific perspective.

On this last point, two major technologies are used to image the brain in the study of language processing: Positron Emission Tomography (PET) and Functional Magnetic Resonance Imaging (fMRI).Both technologies image the brain in a dynamic way, thereby making it possible to detect changes over time.Almost all right-handed and most left-handed subjects are shown by brain-imaging to exhibit language-related activation that is strongly lateralized to the left hemisphere (Matthewsetal.2003).However, interpreting the images generated by these technologies is fraught with difficulty, as experimental tasks typically activate many different areas of the brain simultaneously.Moreover, different modes of language use—e.g.speaking versus reading-seem to involve the activation of different areas of the language centres (Scottetal.2000).Complex neural networks are activated by tasks involving even single words; when words are combined into phrases and sentences the networks become still more complex, owing to the activation of many pragmatic and affective as well as linguistic areas.

Many researchers who would themselves lay absolutely no claim to being neuroscientists have nevertheless had occasion in their exploration of various issues relating to language acquisition and processing to examine the pronouncements of neuroscientists, and, it has to be said that, for various reasons, sometimes such pronouncements have been found wanting.While certainly not denying the enormous potential of the contribution of neuroscience in this area, this paper will argue that we need to be careful not to be bedazzled by neuroscientific research; that, whatever our degree of technical expertise, we need always to be attentive to the necessity of inspecting the detail of arguments and evidential support adduced to support claims made by neurolinguists.

The paper will begin by citing perspectives from researchers whose point of departure was that of being convinced of the value of what the neurosciences have to offer to language acquisition/processing research, but who have urged caution with respect to claims made on the basis of neuroscientific data.The need for such caution will then be illustrated by reference to three domains of second language acquisition research in which neuroscientists have played a role: research concerning the ‘modularity hypothesis’, research relating to the age factor and research with respect to the organization of the mental lexicon.

2.SOME CAUTIONARY WORDS FROM NEUROSCIENCE-FRIENDLY RESEARCHERS

A good place to start in this exploration is perhaps Obler and Gjerlow’s (1999) textbook on language and the brain, whose preface contains the statement: ‘[i]n this book we hope to share our enthusiasm for the field and spark readers’ interests in its multiple facets’ (p.xvi).Despite such a ringing endorsement of the research domain, Obler and Gjerlow are perfectly prepared to admit to its fault lines and limitations.They refer, for instance, in this context (pp.9-12) to the long-running and the continuing debate among neurolinguists between the ‘localizationalists’ who talk about ‘language centres’ in the brain in a fairly traditional sense and ‘connectionists’ and ‘interactionists’ who see the functioning of the brain more holistically and who see patterns of connections as more important than location in the cortical ‘map’ delivered by current research.

Obler and Gjerlow are also perfectly clear about the uncertainties and deficits of the current state of neurolinguistic research:

Surely our language ‘map’ of the brain is more correct that the phrenological map of Gall, but its shape is still shifting and the labels are not yet fixed.Moreover the two-dimensional notion ofmapwill not suffice as an analogy for the future, as cortical topography is at best the surface component of a multidimensional set of systems—cortical linked with subcortical—that enable us to use language.(Obler and Gjerlow 1999: 168)

What this implies is that we should beware of over-interpreting topographical evidence.Obler and Gjerlow’s general comment regard to neurolinguistics is that “overarching truths may not appear in our lifetime” (ibid.; cf.Posner and Pavese 1998).

Let’s now consider the views of De Bot, whose credentials as a second language acquisition researcher with a neuroscience-friendly starting point would seem unimpeachable.In a review article published in 2000, for example, he cites with approval a passage from a research booklet produced by the Max-Planck-Institut (MPI):

...there is a growing awareness among neuroscientists that they should construct models of cognitive functions in which neurobiological constraints are taken seriously...The rapidly developing field of cognitive neuroscience is therefore based on the conviction that findings at the neurobiological level of analysis should have real consequences for the psychological analysis...(MPI 1998; cited in De Bot 2000: 231)

De Bot is by no means, however, inclined to the view that neuroscience is now the only show in town! Interestingly, his quotation from the MPI booklet ends with the statement that “similarly, the results at the psychological level should have substantial implications for our understanding of the neurobiological system” (ibid.).He goes on to note that “[n]euro-imaging...is still in its infancy, and even for ‘big questions’, such as the neural substrates of individual languages, the more refined techniques have not yet led to real conclusions” (ibid.).He also (ibid.) refers to Paradis’s (1997) warning against ‘simplified overgeneralizations,’ noting Paradis’s suggestion that differences in proficiency and strategy use may result in the activation of different parts of the brain.

In later papers (e.g.2004, 2008), De Bot offers an even more critical appraisal.Thus, in his 2004 paper (p.20) he talks about the ‘often conflicting’ data offered by neuroscience and describes brain-imaging research as “clearly still in a ‘wild’ phase in which dramatic but often non-replicable findings find their way into the research literature.” He cites, by way of illustration, an fMRI (Functional Magnetic Resonance Imaging) study (Marianetal.2002) based on a very small number of subjects whose findings show variability both within and between subjects on both behavioural and anatomical levels and no coherence between processing models and the neuro-imaging data—which present, in other words, enormous problems of interpretation.In his 2008 article De Bot argues (p.129) that the return on investment in neuro-imaging with respect specifically to multilingual processing is low, and that “[b]oth researchers and funders may want to consider to what extent an increase in [neuro-imaging] research is warranted”.

A further source of interest in the current context is the work of the Fabbro, a neurolinguist who has taken a particular interest in bilingualism and second language acquisition (see e.g.Fabbro 1999, 2002).He notes that there are important limitations to the methodology of neuro-imaging having to do with time factors, interpretation factors and task-related factors:

(a) the time needed to study the cerebral representation of a function is expressed in seconds, whereas language processes are expressed in milliseconds;

(b) subtractive comparisons between the two tasks [general cognitive and linguistic] are often difficult to interpret;

(c) often results of neuroimaging studies do not correspond to clinical neuropsychological findings;

(d) neuroimaging techniques do not allow one to determine whether activation of a structure depends either on an increase in activation processes or in neurophysiological inhibition processes; and, lastly,

(e) brain activation was studied with tasks that ...generally simultaneously activate many linguistic, pragmatic and affective structures, thus making it difficult to interpret data.(Fabbro 2002: 209f.)

The foregoing demonstrates clearly that many of those whose research is firmly planted within the context of neurolinguistics and depends on the technology and techniques of the neurosciences are conscious of the need for great care in the use of the relevant procedures and the interpretation of results.The following quotation from Carreiras and Clifton (2004) encapsulates this perspective and may serve as a coda to this part of the discussion:

Each of the techniques...has its own strengths and weaknesses.Clever experimentation is needed to use these techniques properly, as is careful and explicit theorizing, including specification of the presumed links between cognitive processes and the data that can be observed in each technique.(Carreiras and Clifton 2004: 8)

Someexamplesofincautioususesofneurolinguisticdata

Unfortunately, not all neurolinguists have heeded the kind of cautionary counsel illustrated in the last section.The discussion that follows will focus on some studies which appear to exemplify the way in which neurolinguists should not proceed.The studies in question deal with, respectively, the modularity question, the age factor question and the question of the organization of the mental lexicon.

Modularity

Much of the foregoing dwelt on the potential and problematicity of brain-imaging techniques, but there is more to neurolinguistics than neural imaging.Neurolinguists have often based their conclusions on a close analysis of data from subjects suffering from various kinds of aphasia.One very widely cited study of this kind is Linebarger’s (1989) review of studies of a variety of aphasias with a view to casting light on the modularity hypothesis (cf.Singleton 1998).

This hypothesis states that the mind is “not a seamless, unitary whole whose functions merge continuously into each other” but rather contains, perhaps in addition to some general-purpose structures, “a number of distinct, specialized, structurally idiosyncratic modules that communicate with other cognitive structures in only very limited ways” (Garfield 1987: 1).Chomsky (e.g.1980, 1988), Fodor (e.g.1983, 1989) and others have postulated a module dedicated to language, which is taken to be the encapsulated locus and scene of operations of bio-endowed Universal Grammar (UG).This hypothesis is of great interest to second language acquisition research—especially perhaps in relation to the issue of the extent to which the mechanisms the so-called language module are seen as remaining available to later second language learners (cf.Mitchell and Myles 2004: 78f.).Both Chomsky’s conception of the language module and Fodor’s place the formal operations of syntax inside the module but the processing of meaning outside it.Linebarger’s article essentially interprets the aphasia data it addresses in terms of a neat divide between semantics and syntax, and thus as evidence for the Chomskyan/Fodorian language module.

Thus, for example, Linebarger presents the following data from a patient suffering from Wernicke’s aphasia, which she characterizes as combining syntactic well-formedness with semantic anomalousness:

His wife saw the wonting to wofin to a house with the umbledor.Then he left the wonding then he too to the womin and to the umbrella upstairs.

In fact, the data do not support the idea of a neat syntactic/semantic divide.The semantic disruption is accompanied by quite serious grammatical problems (deployment of inappropriate grammatical categories, omission of required grammatical categories, misconceived argument structure, etc.).

Other data referred to by Linebarger come from a subject with Broca’s aphasia, which she describes as a syndrome in which syntactic impairment is combined with the capacity to use content words appropriately.The ‘agrammatic’ data are extracted from the Broca’s aphasic’s attempt to describe a picture of a girl beside a boy attempting to steal cookies behind the back of a woman washing dishes.

A mother...a dish...drying....Plate...a faucet...running...a boy, eating cookies...eating the cookies...girl.

To describe such data as ‘agrammatic’ is a wild exaggeration.There is plenty of ‘grammar’ here: a high proportion of the word order is unexceptionable, for example, a faucet running, a boy eating cookies, and in most cases where an article is required one is supplied (amother,adish,afaucet,aboy,thecookies).

Similarly with the rest of the data discussed by Linebarger.Whatever may turn out to be the truth regarding the degree of distinctness and encapsulation of the neurological processing of syntax with respect to semantics, Linebarger’s evidence fails to demonstrate any hard and fast separation.Linebarger’s commentary on the evidence she presents appears to be an instance of an attempt to squeeze the data into a particular theoreticalpartipris, a bad strategy whether its perpetrator is a neuroscientist or not! Interestingly, other neurolinguists read the current evidence as not offering support for the notion that there is a physiological correlate for a putative autonomous language module.Schumann, for example, has the following to say on the matter, with specific reference to second language acquisition:

The neural equipment for every aspect of the SLA process is available in general learning systems distributed throughout the brain....The UG claim is a neurobiological claim; it is a theory of the brain, but...the neural mechanisms to subserve the acquisition of a second language exist to subserve learning in general.(Schumann 2004: 177f.)

Theagefactor

As is well known, there has been a long history of attributing age effects in language acquisition to neurological factors.Even within this particular strand of research into maturational constraints on language acquisition, it should be noted, there is a vast array of explanations on offer (see e.g.Muoz and Singleton 2011; Singleton 2005; Singleton and Muoz 2011; Singleton and Ryan 2004).As Hyltenstam and Abrahamsson put it (2003: 563), with wonderful understatement, “there is certainly no consensus” with regard to the neurobiology of age effects.

Some brain-imaging research has addressed the question of whether early second language acquisition results in different spatial representations in the brain from late acquisition.For instance, Kimetal.(1997) used magnetic resonance imaging to investigate the spatial representation of L1 and L2 in the cerebral cortex of early and late bilinguals during a sentence-generation task.The results revealed little or no age-related separation of activity in Wernicke’s area, but differences did emerge in respect of activity in Broca’s area.Among the late bilinguals two distinct but adjacent centres of activation were revealed for L1 and L2, whereas in the early bilinguals a single area of activation for both languages emerged.This looks like evidence of different kinds of brain organization in early and late bilinguals.

Wattendorf, Westermann, Zappatore, Franceschini, Lüdi, Radü and Nitsch (2001) attempted to replicate Kimetal.’s research for multilinguals, some of whom (early bilinguals) who had been exposed to two languages before age 3 and to a third after age 10, and others of whom (late multilinguals) had experienced only one language up to age 10 and had then been exposed to two further languages.Basically, the results of this study show that in the early bilinguals the neural substrate in Broca’s area activated by the use of the first two languages in a narration task overlapped, whereas in the late multilinguals the pattern of activation was more diffuse.Evidence from the above two studies has been interpreted as indicating different kinds of brain organization in early and late bilinguals in relation to syntax but not in relation to the lexicon.

However, there are reasons to treat such evidence with caution.As Marinova-Toddetal.(2000) note, in Kimetal.’s study there is no control of the proficiency level of the later beginners.Accordingly there is the possibility “that the adult learners assessed...were poorly selected and do not represent highly proficient adult bilinguals” (Marinova-Toddetal.2000: 17-18).If this were so, then the neurological differences observed might simply reflect different proficiency levels.Proficiency rather than age may also have been the critical factor in relation to the late trilinguals in the Wattendorfetal.study; they are reported to have all used their three languages in their daily lives, but this hardly constitutes a detailed evaluation of their proficiency levels.Other studies have indeed suggested that proficiency is a “critical factor in shaping the functional brain organization of languages” (Abutalebi, Cappa and Perani 2001: 187) and that it may outweigh age of onset as the crucial factor in this connection (see Muoz and Singleton 2011: 20-25 ).

To summarize, some studies which have interpreted data from brain-imaging experiments with bi- and multilinguals as indicating differences in brain organization relating to the age of onset of the languages in question have failed to take account of the very important variable of level of proficiency.This is matter of poor experimental design.The fact that it was not immediately spotted (one of the papers concerned was actually published inNature) speaks volumes for the mesmerizing powers of the new neural imaging technologies—powers against which, I would suggest, we should be constantly on our guard.

Thementallexicon

One of the major issues in research and theorizing relative to the mental lexicon is the extent to which the lexicon associated with each of the languages known to an individual is separate from or integrated with the lexicon associated with each of the other languages.Thus, the second language mental lexicon has sometimes been represented as qualitatively different and, by implication, distinct from the first language mental lexicon (see e.g.Meara 1984).Arguments against such a qualitative difference are not hard to find (see, e.g., Joannopulou 2002; Singleton 1999; Singleton and Little 1991; Wolter 2001), and some researchers have indeed have gone to the opposite extreme, claiming that the mental lexicon is unitary, no matter how many languages are involved (for further discussion see e.g.Singleton 2003, 2006, 2007).

Franceschinietal.(2003) address this issue from a neurolinguistic standpoint.Their conclusion from their review of a range of brain-imaging studies is that lexical-semantic aspects of the processing of all languages known to an individual are subserved by essentially the same areas of the brain.This certainly suggests very close connections between lexical operations relating to the languages in question, but there are grounds for treating with some skepticism Franceschinietal.’s inference that lexico-semantic processing draws on a common system across languages:

...the two languages of a bilingual access a common semantic system independent of age of acquisition and attained proficiency.(Franceschinietal.2003: 162)

Semantics, and in particular the lexicon, appear to be based on more common ground among languages [as compared with syntax and phonology].(ibid.: 164)

For one thing, we need to bear in mind what has been said regarding the limitations of brain-imaging technology.As Obler and Gjerlow note (see above), there is a very great deal that it does not allow us to see.Indeed, Franceschinietal.acknowledge this in the following qualification of their conclusion:

Perhaps at the lexical level the differences among multilingual processes do exist, but cannot be revealed with the [earlier described] macroscopic techniques.(ibid.: 164)

We should also recall the old linguistic truism—namely, the one which recognizes that every language articulates the world differently in terms of its lexical structure, and that the concepts and configurations of concepts that are lexicalized vary from language to language.In the published version of his highly influential doctoral thesis, John Lyons put it this way some forty-four years ago:

It is now commonly accepted by linguists that a ‘structural approach’ of the kind long practised in the phonological and grammatical analysis of languages is required also for their semantic description: each language must be thought of as having its own semantic structure, just as it has its own phonological and grammatical structure....It is not so much that one language draws a greater or less number of semantic distinctions ...It is rather that these distinctions are made in completely different places.(Lyons 1963: 37f.)

What the above implies is that, to the extent that multiple language users draw on the lexico-semantic systems specific to each of their languages, they must draw on systems which are unique and radically differentiated from each other.This is an implication which clearly ought to figure in Franceschinietal.’s weighing of the research findings they consider, but does not.

3.CONCLUDING REMARKS

What has been argued for in this article is an approach to neuroscientific contributions to our field which recognizes the potential of such contributions, but which also recognizes the problems and uncertainties which currently attend neurolinguistic research.From the present critique of a number of specific neurolinguistic studies, have been derived some comments relating to the over-interpretation of evidence in the service of particular theories, the failure to design studies in such a way as to take account of key variables, and the ignoring of facts about language that have been established for decades.

The essential point is that, if neuroscientific research is to be integrated into second language acquisition studies, it has to operate on the basis of the same ground-rules as more ‘traditional’ research-with reference to the (1) acknowledgment of methodological and technical limitations, (2) the restriction of conclusions to what is licensed by the data, (3) the controlling for possible confounding variables, and (4) the incorporation of established linguistic and psycholinguistic facts into the analysis of findings.

REFERENCES

Abutalebi,J.,S.Cappa, andD.Peran.2001.‘The bilingual brain as revealed by functional neuroimaging,’Bilingualism:LanguageandCognition4/2: 179-90.

Carreiras,M.andC.Clifton.2004.‘On the on-line study of sentence comprehension’ in M.Carreiras and C.Clifton (eds.):TheOn-lineStudyofSentenceComprehension:Eyetracking,ERPs,andBeyond.Hove: Psychology Press, pp.119-137.

Chomsky,N.1980.RulesandRepresentations.Oxford: Blackwell.

Chomsky,N.1988.LanguageandProblemsofKnowledge:TheNicaraguanLectures.Cambridge, MA: MIT Press.

DeBot,K.2000.‘Psycholinguistics in applied linguistics,’AnnualReviewofAppliedLinguistics20: 224-37.

DeBot,K.2008.‘The imaging of what in the multilingual mind?’SecondLanguageResearch24/1: 111-33.

Fabbro,F.1999.TheNeurolinguisticsofBilingualism:AnIntroduction.Hove: Psychology Press.

Fabbro,F.2002.‘The neurolinguistics of L2 users’ in V.Cook (ed.):PortraitsoftheL2User.Clevedon: Multilingual Matters, pp.199-218

Fodor,J.1983.TheModularityofMind:AnEssayonFacultyPsychology.Cambridge, MA: MIT Press.

Fodor,J.1989.‘Why should the mind be modular?’ in A.George (ed.):ReflectionsonChomsky.Oxford: Blackwell, pp.179-202.

Franceschini,R.,D.Zappatore, andC.Nitsch.2003.‘Lexicon in the brain: What neurobiology has to say about languages’ in J.Cenoz, U.Jessner and B.Hufeisen (eds.):TheMultilingualLexicon.Dordrecht: Kluwer, pp.153-66.

Garfield,J.1987.‘Introduction: Carving the mind at its joints’ in J.Garfield (ed.):ModularityinKnowledgeRepresentationandNatural-languageUnderstanding.Hillsdale, NJ: Erlbaum, pp.1-13.

Hyltenstam,K.andN.Abrahamsson.2003.‘Maturational constraints in SLA’ in C.Doughty and M.H.Long (eds.):TheHandbookofSecondLanguageAcquisition.Malden, MA: Blackwell, pp.539-88.

Indefrey,P.andGullberg,M.2006.‘Introduction’ in M.Gullberg and P.Indefrey (eds.):TheCognitiveNeuroscienceofSecondLanguageAcquisition.Oxford: Blackwell, pp.1-8.

Joannopoulou,M.2002.‘Form and meaning in the second language lexicon: Some evidence from Greek advanced learners of English,’JournalofAppliedLinguistics18: 29-42.

Kim,K.H.S.,N.R.Relkin,L.Kyoung-Min, andJ.Hirsch.1997.‘Distinct cortical areas associated with native and second languages,’Nature388: 171-74.

Linebarger,M.1989.‘Neuropsychological evidence for linguistic modularity’ in G.Carlson and M.Tanenhaus (eds.):LinguisticStructuresinLanguageProcessing.Dordrecht: Kluwer, pp.197-238.

Lyons,J.1963.StructuralSemantics:AnAnalysisofPartoftheVocabularyofPlato.Oxford: Blackwell/The Philological Society.

Marian,V.,M.Spivey, andJ.Hirsch.2002.‘Shared and separate systems in bilingual language processing: Converging evidence from eye-tracking and brain imaging,’BrainandLanguage86/1: 70-82.

Marinova-Todd,S.,D.Marshall, andC.Snow.2000.‘Three misconceptions about age and L2 learning,’TESOLQuarterly34/1: 9-34.

Matthews,P.,J.Adcock,Y.Chen,S.Fu,J.Devlin,M.Rushworth,S.Smith,C.Beckmann, andS.Iversen.2003.‘Towards understanding language organization in the brain using fMRI,’HumanBrainImaging18: 239-247.

Meara,P.1984.‘The study of lexis in interlanguage’ in A.Davies, C.Criper and A.P.R.Howatt (eds.):Interlanguage.Edinburgh: Edinburgh University Press, pp.225-35.

MitchellR.andF.Myles.2004.SecondLanguageLearningTheories(2ndedition).London: Edward Arnold.

MPI.1998.BookletforPsycholinguistics:NeurocognitionofLanguageProcessing.Nijmegen: Max-Planck-Institut für Psycholinguistik.

Muoz,C.andD.Singleton.2011.‘A critical review of age-related research on L2 ultimate attainment,’LanguageTeaching44/1: 1-35.

Obler,L.K.andK.Gjerlow.1999.LanguageandtheBrain.Cambridge University Press.

Paradis,M.1997.‘The cognitive neuropsychology of bilingualism’ in A.de Groot and J.Kroll (eds.):TutorialsinBilingualism.Mahwah, NJ: Erlbaum, pp.331-54.

Perani,D.,E.Paulesu,N.S.Galles,E.Dupoux,S.Dehaene,V.Bettinardi,S.F.Cappa,F.Fazio, andJ.Mehler.1998.‘The bilingual brain: proficiency and age of acquisition of the second language,’Brain121/10: 1841-52.

Posner,M.andA.Pavese.1998.‘Anatomy of word and sentence meaning,’ProceedingsoftheNationalAcademyofSciences95: 899-905.

Schumann,J.H.2004.‘Conclusion’ in J.H.Schumann, S.Crowell, N.E.Jones, N.Lee, S.A.Schuchert and L.E.Wood (eds.):TheNeurobiologyofLearning:PerspectivesfromSecondLanguageAcquisition.Mahwah, NJ: Erlbaum, pp.175-80.

Scott,S.,C.Blank,S.Rosen, andR.Wise.2000.‘Identification of a pathway for intelligible speech in the left temporal lobe,’Brain123: 2400-06.

Singleton,D.1998.‘Lexicalprocessingandthe‘languagemodule’,CLCSOccasionalPaperNo 53.

Singleton,D.1999.ExploringtheSecondLanguageMentalLexicon.Cambridge: Cambridge University Press.

Singleton,D.2003.‘Perspectives on the multilingual lexicon: A critical synthesis’ in J.Cenoz, U.Jessner and B.Hufeisen (eds.):TheMultilingualLexicon.Dordrecht: Kluwer, pp.167-76.

Singleton,D.2005.‘The Critical Period Hypothesis: A coat of many colours,’InternationalReviewofAppliedLinguisticsinLanguageTeaching(IRAL) 43/4: 269-85.

Singleton,D.2006.‘Lexical transfer: Interlexical or intralexical?’ in J.Arabski (ed.):Cross-linguisticInfluencesintheSecondLanguageLexicon.Clevedon: Multilingual Matters, pp.130-43.

Singleton,D.2007.‘How integrated is the integrated mental lexicon?’ in Z.Lengyel and J.Navracsics (eds.):SecondLanguageLexicalProcesses.Clevedon: Multilingual Matters, pp.10-29.

Singleton,D.‘Multilingual lexical operations: Keeping it all together ...and apart’ in S.Flynn and J.Rothman (eds.):ThirdLanguageAcquisitioninAdulthood(in press).Amsterdam, John Benjamins.

Singleton,D.andD.Little.1991.‘The second language lexicon: Some evidence from university-level learners of French and German,’SecondLanguageResearch7/1: 61-82.

Singleton,D.andC.Muoz.2011.‘Around and beyond the Critical Period Hypothesis’ in E.Hinkel (ed.):HandbookofResearchinSecondLanguageTeachingandLearning.Volume Ⅱ.New York: Routledge, pp.407-25.

Singleton,D.andL.Ryan.2004.LanguageAcquisition:TheAgeFactor(2ndedition).Clevedon: Multilingual Matters.

Ullman,M.2003.‘A neurocognitive perspective on second language acquisition and processing: The Declarative/Procedural Model’.Contribution to theLanguageLearningRoundtable on the Cognitive Neuroscience of Second Language Acquisition, 13thAnnual Conference of the European Second Language Assocation (EUROSLA), Edinburgh.

Wattendorf,E.,B.Westermann,D.Zappatore,R.Franceschini,G.Lüdi,E.-W.Radü, andC.Nitsch.2001.‘Different languages activate different subfields in Broca’s area,’NeuroImage13/6: 624.

Wolter,B.2001.‘Comparing the L1 and L2 mental lexicon: A depth of individual word knowledge model,’StudiesinSecondLanguageAcquisition23/1: 41-69.