Cognitive factors: Working memory and lexical development Alan Juffs Support National Science Foundation Thanks to RSAs: Jenifer Larson-Hall Greg Mizera Jessica Giesler Sean Coyan Vivian Chen SBR-9709152 Publications Dekeyser, R and A. Juffs. (2005). Cognitive considerations in L2 learning. Handbook of Research in Second Language Teaching and Learning. 437454. Mahwah, NJ: Erlbaum. Juffs, A. (2004). Representation, processing and working memory in a second language. Transactions of the Philological Society, 102, 199-226. Juffs, A. (2005). Some effects of first language and working memory in the processing of long distance wh- questions. Second Language Research 21, 121-151. In press a. Processing reduced relative vs. main verb ambiguity in English as a Second Language: a replication study with working memory. A festschrift for XXXX. Amsterdam: John Benjamins. Structure of talk Sketch of working memory models Brief Sketch of sentence processing Experiment in working memory and sentence processing in English as a second language Memory, aptitude, and low educated learners Baddeley’s Working Memory Model 1 Figure 1. Standa rd Working Memory Model, Baddeley (2000a). Central control Slave system Visual Semantics Central Execu tive Visuo spatial sketchpad Phono logical loop Episodic LTM Language Shaded area: Τcrystalli zed cogn iti ve systems capable of accumulating long-term know ledgeΥ Baddeley’s Working Memory Model 2 Figure2. Further Development of the Working Memory Model, Baddeley (2000a). Central control Slave system Visuo spatial sketchpad Visual Semantics Central Execu tive Episodic buffer Episodic LTM Phono logical loop Language Shaded area: Τcrystalli zed cogn iti ve systems capable of accumulating long-term know ledgeΥ Behavioural Measures Central executive Reading Span Task (Daneman and Carpenter, 1980) What does the RST claim to measure? Relative clause types and WM 1. Animacy effects in reduced relative clauses • The evidence [inanimate] examined by the lawyer was convincing. • The witness [animate] examined by the lawyer was convincing. 2. Subject and object asymmetry in relative clauses. The reporter that the senator attacked ____ The reporter that ___ attacked the senator 3. regretted the error. regretted the error. Reduced relatives and cue strength. The bad boys seen during the game were playing in the park. - no ambiguity; good cue for ambiguity resolution The bad boys watched almost every day were playing in the park. ambiguity + bad cue for ambiguity resolution. Phonological Loop Non-word span, digit span What does this measure? acquisition of new words, and does not reflect the knowledge base. Gathercole, Baddeley, & Papagno (1998, p. 159, Table 1) in partial correlations for 3 year-olds, non-word repetition is more strongly correlated with vocabulary measures than digit span (0.31 vs. 0.16 (ns), whereas for 8 year-olds neither span is correlates (0.22 (ns) vs. 0.23 (ns)). The data they report for 13 year olds, simple digit span is related to vocabulary measures (r= .46, p = .05). Phonological loop in adults May be important in ability to learn new words in adults, but it has not been implicated in studies of on-line ambiguity resolution. These ‘now you see it, now you don’t’ effects of PSTM in L1 learning are not reflected in L2 reviews of the literature. Issue and controversies Does the reading span tap general or specifically linguistic capacity? Does WM reflect experience? Which test is a better test of WM? What is the role of the phonological loop? The role of memory as a key component of aptitude Recent L2 WM research Myle s et al. 1998 PSTM 1999 Engl is h-speaking learners of French Production data 2001 Waters and Caplan RST Robinson 2002 Osaka & Osaka RST Engl is h-speaking learners of Spanish and French 17 Japanese learners of Samoan in a lab setting Translation to and from words in the L2 Erga tives Incorpo ration Locatives Kroll et al Supe rior abili ty in chunk ing rela ted to highe r WM. Later better chunke rs better at creative u se. Fail ed to find a reli able relationsh ip between WM and translation processing. WM memory, and no t gene ral i ntelli gen ce found to be best predictor, but only sho rt-term, ΤeasyΥ structures. Problem wit h many correlations . Relia ble correlations only on GJ li stening. I.e. Of 24 correlations with WM reported, only 4 reli able. None above .52. Amount of variance exp lained not clear. No attempts at regr ession . Recent L2 research Mackey et al 2002 Plaus ible-non plausible versions of RST; Non word recall Willi ams 2003 Non-word & Lova tt PSTM test based on target vo cab. 30 Japanese learners of ESL WM and interactiona l feedback 1. 20 Eng li sh1. speaking learne rs Laboratory study of 2. 21 Eng li shdetermi ners speaking learne rs in Itali an. 2. Inven ted lang based on Japanese, determi ners aga in. Variable correlations between RSTs and non-word recall . Composite scores deve loped because of the correla tions. No reli able relationsh ip found between WM and noticing. Other factors at work ? p. 202. Non word repetiti on d id NOT correlate wit h L2 li stening. p. 209. Exp. 1Prior language expe rienc e most related to success, Language background was. PSTM More strongly related to RATE of learning than ultim ate leve l. Exp. 2. Few correlations , specifically none between PSTM and vocabu lary. Effect of language background NOT mediated by memory measures. Learning that occurred was exp li cit . Can not be assumed that rules emerge from memory representations of input sequenc es. (Contra Ell is) The grammar and the parser Crain and Fodor (1985, p. 126) suggested: Frazier & Clifton (1996, 24-25): a theory of grammar that will be useful to a theory of parsing is one that is compatible with the on-line application of constraints. Licensing grammars, based on current versions of GB theory, may be developed so that they provide attractive alternatives [to head projection models] Chomsky (2000, p. 91) ‘ the major problem is to discover the principles and parameters … and to proceed beyond, to the study of use, acquisition, pathology, cellular mechanisms, …’ Hence Chomsky includes ‘use’ in the MP? Second Language Acquisition Development Projectionist accounts (Principles and Parameters) constructionist accounts (Goldberg, 1995) Process of the L2 lexicon: ‘what’: of acquisition: ‘how’ Processing break down Accumulation of chunks/structures L2 vocabulary: Nation 1990 1. The spoken form of a word 2. The written form of a word 3. The grammatical behaviour of a word 4. The collocational behaviour of a word 5. How frequent the word is 6. The stylistic register constraints on a word 7. The conceptual meaning of a word 8. The associations a word has with other related words Experiment - Questions Do measures of working memory correlate in the L1 and L2? Can individual differences in working memory account for individual differences in sentence processing based on verb meaning? What is the effect of the L1 on L2 processing? Method -1 Proficiency measure Measure of Reading Span in L1 and L2 Measure of Word Span in L1 and L2 Method 2 Data from on-line reading: record word by word reading times Method The ‘moving window’ paradigm Without her contributions would be Impossible Possible or not possible? Participants 30 Chinese 28 Japanese 46 Spanish 21 English speakers Table 1. Michigan Test Results: Raw Scores. Michigan Chinese M* Vocabulary 28.33 a 7.67 20.39 a 6.21 26.65 7.58 Grammar 29.8 b 6.0 25.07 b 5.28 26.89 Total 58.03 c 12.59 45.46 c d 10.32 53.45 d SD Japanese M SD Spanish M *Means that are co-superscripted are reliably different. SD F df p 9.6 2,102 .0002 7.26 4.042 2,102 .0205 13.96 7.29 2,102 .0011 Results Working memory Sentence processing Chinese-speaking learners L1 Word Span L2 Word Span L1 Reading Span L1 Word Span 1 L2 Word Span 0.34* 1 L1 Reading 0.02 0.05 1 -0.18 0.17 0.62*** L2 Reading Span Span L2 Reading Span 1 Japanese-speaking learners L1 Word Span L2 Word Span L1 Reading Span L1 Word Span 1 L2 Word Span 0.28 1 L1 Reading 0.41** 0.54** 1 0.30 0.44** 0.56*** L2 Reading Span Span L2 Reading Span 1 Spanish-speaking Learners L1 Word Span L2 Word Span L1 Reading Span L1 Word Span 1 L2 Word Span 0.48* 1 L1 Reading 0.44** 0.28 1 0.24 0.09 0.46** L2 Reading Span Span L2 Reading Span 1 Chinese WM & Proficiency L1 Word Span Vocabu la ry 0.29 Grammar 0.27 L2 Word Span 0.27 0.35* L1 Reading Span 0.07 0.10 L2 Reading Span 0.04 0.02 Japanese WM & Proficiency L1 Word Span Vocabu la ry 0.11 Grammar 0.09 L2 Word Span 0.26 0.04 L1 Reading Span 0.22 0.28 L2 Reading Span 0.08 0.06 Spanish WM & Proficiency L1 Word Span Vocabu la ry 0.22 Grammar 0.20 L2 Word Span 0.11 0.24 L1 Reading Span 0.30* 0.31* L2 Reading Span 0.28* 0.29* Sentences that impose processing load Garden Path sentences After the children cleaned the house looked neat and tidy The doctor knew the nurses liked the man from England Unconscious GP Processing Unconscious Garden Path 1200 RT in milliseconds 1000 800 600 400 the doctor knew the nurses liked Word by Word the man from England Garden Path Processing- L1 Garden Path Sentence 1200 Chines e English Japanes e Spanis h RT in milliseconds 1000 800 600 400 After the children cleaned the hous e W ords by W ord looked very neat and tidy Garden Path Processing - WM Working Memory and Parsing 1200 1000 RT Milliseconds 800 HI-GP LO-GP HI-Non-GP 600 LO-Non-GP 400 200 0 After children Det VERB Structure/word neat tidy Transitivity and cue type (1) a. The experienced soldiers warned about the dangers conducted the midnight raid. b. The experienced soldiers chosen for their skills conducted the midnight raid. 2. a. The bad boys criticized during the morning were playing in the park. b. The bad boys criticized almost every day were playing in the park. 6 Sentence types Unambiguous good and bad cues Two way ambiguous, good and bad cues Three way ambiguous, good and bad cues Easiest: unambiguous, good cue The bad boys chosen during the game were playing in the park. Most difficult: three way ambiguous, bad cue The bad boys watched almost every day were playing in the park. Processing reduced relatives The bad boys XXX were playing Main verb mean processing time 850 800 750 RT MSEC 700 Chinese Japanese Spanish English 650 600 550 500 450 400 Unambig-G Unambig-B Ambig2-G Ambig2-B Ambiguity and Cue Type Ambig3-G Ambg3-B Working memory and reduced relatives No correlations with WM and processing at key point for any of the groups at any point in parsing except early on All weak correlations, suggesting much of the variance can be explained by other factors Main effects for language robust Points to remember L1 a better predictor of performance than WM WM does not correlate with individual differences in processing L2 speakers show reading profiles analagous to natives in many cases Use of WM tests need to be fully justified in L2 research Overemphasis of WM when results don’t support it More points to remember More careful regression analyses Clearer acknowledgement of the role of prior linguistic knowledge is necessary. Role of the ‘new’ link proposed by Baddeley between visual spatial ability and the PL and language needs to be looked at Aptitude and ultimate attainment DeKeyser 2000 Replication of Johnson and Newport 1989 Added MLAT measure 58 Hungarian-speaking learners of ESL Findings: replicated Johnson and Newport The only adults who succeed are those who score high on the aptitude battery Cf. Bialystok’s commentary and reply http://www.pitt.edu/~rdk1/ Skehan 2001 Aptitude: speed or ultimate attainment? DeKeyser (2000, p. 518) aptitude has a role in ultimate attainment Skehan (2001, p. 93) points out that the MLAT was designed to predict RATE and not ultimate attainment, contra (?) DeKeyser 2000) Does the MLAT measure communicative competence? Or an ability on discrete point items? Mackey et al. 2001 RST and WM test Combined measure: • Low WM tended to notice less at lower developmental stages than High WM • High WM - more development in delayed post-test • High WM tended to notice more Robinson 2001 Implicit learning: not related to higher IQ or aptitude measures? Incidental learning: unintentional and uncontrolled? Explicit learning: does relate to higher IQ measures? Dual system for implicit/explicit knowledge? Robinson 2001 Japanese learners of Samoan Relationship between IQ and explicit learning confirmed Surprising: low IQ scores outperform high IQ scores on implicit learning GJ judgements and production are also unrelated to individual differences learning of locatives, and may be incorporation, but not ergatives. Learning clearer in production tasks compared to GJ tasks Concluding remarks Research on cognitive abilities is deeply divided between those who maintain access to UG in some form (dual system, encapsulated) and those who believe in critical period/general learning. Aptitude measures do seem to predict performance on SOME discrete point item tests of the Johnson and Newport type Conclusions Evidence suggests that the L1 exerts the greatest influence on L2 processing Lexical learning and processing shows that verb transitivity (a highly complex system) is acquired and affects L2 reading and processing and is NOT predicted by IDs in working memory Unlikely that this is ‘generalized’ knowledge Conclusions Therefore it is PREMATURE to conclude that adults are unable to master details of a linguistic system unless they have some higher aptitude: this is because the learners in these studies showed that they can use complex information in millisecond by millisecond parsing decisions. Finally For ‘low-educated’ learners, this is an important issue because it means that low aptitude/IQ/education does not preclude successful language learning (= achievement of communicative competence) given exposure and motivation and cultural conditions Selected References • • • • • • • • • • • • • BADDELEY, ALAN, 2000. ‘Short-term and working memory,’ in Endel Tulving & Fergus Craik (eds.), The Oxford Handbook of Memory, New York: Oxford University Press, 77-92. BADDELEY, ALAN, GATHERCOLE, SUSAN & PAPAGNO, COSTANZA, 1998. ‘The phonological loop as a language learning device,’ The Psychological Review 105, 158-73. BERQUIST, BRETT, 1997. ‘Individual differences in working memory span and L2 proficiency: capacity or processing capacity?,’ Paper presented at Proceedings of the GALA ‘97 Conference on Language Acquisition, Edinburgh, UK. CARPENTER, PATRICIA, JUST, MARCEL Adam & REICHLE, ERIC D., 2000. ‘Working memory and executive function,’ Current Opinion in Neurobiology 10, 195-99. DANEMAN, Meredith & CARPENTER, PATRICIA, 1980. ‘Individual differences in working memory and reading,’ Journal of Verbal Learning and Verbal Behavior 19, 450-66. ELLIS, NICK C., 1996. ‘Sequencing in SLA: phonological memory, chunking and points of order,’ Studies in Second Language Acquisition 18, 91-126. ELLIS, NICK C., 2002. ‘Frequency effects and language processing: investigating formulaic use and input in future expression,’ Studies in Second Language Acquisition 24, 143-88. GIBSON, EDWARD & SCHÜTZE, CARSON T, 1999. ‘Disambiguation preferences in noun phrase conjunction do not mirror corpus frequency,’ Journal of Memory and Language 40, 263-79. HARRINGTON, MICHAEL W, & SAWYER, MARK, 1992. ‘L2 working memory capacity and L2 reading skills,’ Studies in Second Language Acquisition 14, 25-38. JUFFS, ALAN, 1998. ‘Main verb vs. reduced relative clause ambiguity resolution in second language sentence processing,’ Language Learning 48, 107-47. JUST, MARCEL Adam, CARPENTER, PATRICIA A & WOOLLEY, JACQUELINE D., 1982. ‘Paradigms and processes and in reading comprehension,’ Journal of Experimental Psychology: General 3, 228-38. JUST, MARCEL Adam, CARPENTER, PATRICIA & KELLER, Timothy, 1996. ‘The capacity theory of comprehension: new frontiers of evidence and arguments,’ The Psychological Review 103, 773-80. JUST, MARCEL ADAM & VARMA, SHASHANK, 2002. ‘A hybrid architecture for working memory: Reply to MacDonald and Christianson 2002,’ Psychological Review 109, 55-65. Selected References • • • • • • • • • MACDONALD, MARYELLEN C & CHRISTIANSEN, MORTEN H, 2002. ‘Reassessing working memory: comment on Just and Carpenter 1992 and Waters and Caplan 1996,’ Psychological Review 109, 35-54. MACDONALD, MARYELLEN C, 1994. ‘Probablistic constraints and syntactic ambiguity resolution,’ Language and Cognitive Processes 9, 157-201. MACDONALD, MARYELLEN, JUST, MARCEL & CARPENTER, PATRICIA, 1992. ‘Working memory constraints on the processing of syntactic ambiguity,’ Cognitive Psychology 24, 56-98. MACKEY, ALISON, PHILP, JENEFER, EGI, TAKAKO, FUJII, AKIKO & TATSUMI, TOMOAKI, 2002. ‘Individual differences in working memory, noticing of interactional feedback and L2 development,’ in Peter Robinson (eds.) Individual Differences And Instructed Language Learning, Philadelphia: John Benjamins, 181209. MYLES, FLORENCE, HOOPER, JANET & MITCHELL, ROSAMOND, 1998. ‘Rote or rule? Exploring the role of formulaic language in the foreign language classroom,’ Language Learning 48, 323-64. MYLES, FLORENCE, MITCHELL, ROSAMOND & HOOPER, JANET, 1999. ‘Interrogative chunks in French L2: A basis for creative construction?’ Studies in Second Language Acquisition 21, 49-80. OSAKA, MARIKO & OSAKA, NAOYUKI, 1992. ‘Language independent working memory as measured by Japanese and English reading span tests,’ Bulletin of the Psychonomic Society 30, 287-89. PRITCHETT, BRADLEY LOUIS, 1988. ‘Garden path phenomena and the grammatical basis of language processing,’ Language 64, 539-76. PRITCHETT, BRADLEY LOUIS,1992. Grammatical Competence And Parsing Performance. Chicago: Chicago University Press. Selected References • ROBERTS, ROSE & GIBSON, EDWARD, 2003. ‘Individual differences in sentence memory,’ Journal of Psycholinguistic Research 31, 573-98. • ROBINSON, PETER, 2002a. ‘Effects of individual differences in intelligence, aptitude and working memory on incidental SLA,’ in Peter Robinson (ed.), Individual Differences And Instructed Language Learning, Philadelphia: John Benjamins, 211-51. • WATERS, GLORIA S. & CAPLAN, DAVID, 1996a. ‘Processing resource capacity and the comprehension of garden path sentences,’ Memory and Cognition 24, 342-55. • WATERS, GLORIA S. & CAPLAN, DAVID, 1996b. ‘The measurement of verbal working memory capacity and its relation to reading comprehension,’ Quarterly Journal of Experimental Psychology- Human Experimental Psychology, 49A, 51-79. • WEINBERG, AMY, 1999. ‘A minimalist theory of human sentence processing,’ in Sam Epstein &Norbert Hornstein (eds.) Working Minimalism, Cambridge, MA: MIT Press, 287-315. • WHITE, LYDIA, 2003. Second Language Acquisition and Universal Grammar. New York: Cambridge University Press. • WILLIAMS, JOHN N, MÖBIUS, PETER & KIM, CHOONKYONG, 2001. ‘Native and nonnative processing of English wh- questions: parsing strategies and plausibility constraints,’ Applied Psycholinguistics, 22, 509-40.