Presentation on theme: "Neural Correlates of second- language word learning: minimal instruction produces rapid change Judith McLaughlin, Lee Osterhout & Albert Kim (2004) Presented."— Presentation transcript:
Neural Correlates of second- language word learning: minimal instruction produces rapid change Judith McLaughlin, Lee Osterhout & Albert Kim (2004) Presented by: Qinghua Tang Oct 25, 2006
Background Adult second-language (L2) learning is often considered slow and laborious compared to how fast children acquire their first language. But, is it the case that all aspects of L2 learning are uniformly slow? What about the rate of L2 word learning, for example? Besides, the claim is often based on behavioral performances of L2 learners What about the neurophysiological status of the learners? Is there a gap between the behavioral performance of L2 learners and their neurophysiological status?
The goal of the study Decide how much L2 exposure is needed before an adult L2 learner’s brain activity reflects the lexical status(i.e. whether the letter string is a legitimate word or not) and meaning of L2 words?
Experimental design: technique ERP recording provides a nearly continuous sampling of the brain’s electrical activity. N400 reflects word status and word meaning native speakers of a language are supposed to show the largest N400 amplitude for pronounceable pseudowords, second largest for words preceded by semantically unrelated words, smallest for words preceded by semantically related words That is, for N400 amplitude pseudowords > unprimed words > primed words
Experimental design: process (1) Three sessions all together: Session 1: Mean 14 hours of instruction(range 5-28 hours) Session 2: mean 63 hours of instruction(range hours) session 3: mean 138 hours of instruction(range hours)
Experiment design: process (2) In each session: Participants made a lexical decision for each of the prime-target pair. Each trial started with fixation cross (500ms), followed by blank screen (500ms), prime (400ms), target (400ms), response prompt ERP was recorded at the same time at 200Hz from 13 scalp sites (three midline, five lateral pairs)
Subjects A group (n = 18, 16,13 for each session respectively) of university students taking a 9-month long introductory French course (mean age: 21.3 years) A control group (n = 8) who had never had significant exposure to French (mean age 27.6 years) All the participants had received at least 1 year of instruction in another foreign language 5 learners dropped the course All the participants were included in single session analyses, but only those who participated fully were included in multi-session comparisons.
stimuli 2 lists of 112 ‘prime- target’ pairs of letter strings, including: 40 semantically related pairs. E.g. chien - chat 40 semantically unrelated pairs. E.g. maison – soif 32 pairs in the pattern of word-pseudoword E.g. mot - nasier Target words were identical across lists but counterbalanced across prime type Each participant saw one list per session The list in session 1 was repeated in session 3
Results: behavioral Non-learners showed no sensitivity in lexical decision task in all three sessions Sensitivity is measured using a d’ representing the likelihood of a real target being recognized (d’=0 indicates no sensitivity, d’ = 4 indicates near perfect sensitivity.) Learners showed no sensitivity in session 1, but with moderate increase in sensitivity in session 2 and 3
Table 1 Proportion of words and pseudowords identified as a word in the lexical decision task, and the d' measure of sensitivity Words Related UnrelatedPseudowordsd'd' Nonlearners Session Session Session French learners Session Session Session We calculated d' using a formula provided by Miller 8 : d' = z (h) - z (fa), where d' represents the likelihood of a real target being recognized, z (h) represents the z-score that corresponds to the proportion of real words that were identified as words (hits), and z (fa) represents the z-score that corresponds to the proportion of nonwords that were identified as words (false alarms). From
Results: N400 No change observed in N400 amplitude for non- learners across all three sessions Learners showed larger N400 amplitude for pseudowords than for unprimed and primed words. The word-pseudoword difference started in session 1 and increased across all three sessions. By session 3, learners’ ERP responses were qualitatively similar to analogous native language responses. There was no significant difference between unrelated words and related words for learners in session 1 N400 effects were evenly distributed over midline sites and posteriorly distributed over lateral sites
Figure 1. Event-related potentials to target stimuli. (a) ERPs to word and pseudoword targets during the three testing sessions, for the nonlearners (n = 8; mean age: 27.6 years) and French learners (n = 18, 16 and 13 for sessions 1, 2 and 3, respectively; mean age, 21.3 years). Informed consent was obtained from all participants. Data acquired over the central midline site (Cz) are shown. The vertical calibration bar indicates target onset. Each tick mark represents 100 ms. (b) Learners' ERPs to targets, averaged over sessions. ATL/R, anterior temporal left/right; TL/R, temporal left/right; WL/R, Wernicke's area left/right. Trial sequence: fixation cross (500 ms); blank screen (500 ms); prime (400 ms); blank screen (400 ms); target (400 ms); blank screen (400 ms); response prompt. Electroencephalographic activity was sampled at 200 Hz from 13 scalp sites (three midline, five lateral pairs; 0.01−100 Hz bandpass; 3 dB cut- off; left mastoid reference). Trials contaminated by artifacts (17%) were excluded. N400 amplitude was quantified as mean voltage within a 300−500 ms window, relative to a baseline of mean voltage from 100 ms before to 50 ms after stimulus onset. We used a repeated-measures ANOVA with the Greenhouse-Geisser correction. From
Discussion (1) Learners showed a word - pseudoword N400 effect after approximately 14 hours of instruction. Did the presence of this N400 effect result from L2 exposure? Both N400 amplitude and d’ score of session 1 were regressed onto the number of instructional hours for each learner N400 amplitude difference was correlated with hours of instruction. d’ score was not correlated with hours of instruction N400 differences in session 1 were also correlated with the proportion and frequency of target words in the text material assigned to each learner before session 1 Therefore, the N400 effect was a result of L2 exposure
Figure 2 Figure 2. Session 1 N400 amplitude word/pseudoword differences and d' scores regressed onto hours of instruction before session 1. (a) N400 amplitude difference between words and pseudowords. (b) d' scores. Two subjects were excluded from the d' analysis due to technical problems. From
Discussion (2) According to Saffran, Alison & newport (1996), people rapidly extract co-occurrence statistics for letter and sound combintaions within a language. Was the N400 difference between pseudowords and words due to the smaller mean grapheme co- occurrence values for the pseudowords than for words? In other words, did the graphemes of words co- occur more often than pseudowords and therefore caused the N400 difference?
Discussion (2) continued The digram, trigram and quadragram frequencies of all the French words from the first four chapters of the textbook were computed. The words and pseudowords did not differ in bigram and trigram frequency(P>0.3), but there was a difference in quadragram frequency (t 128 =1.76, P<0.04), however, the quadragram frequency was highly correlated with target word frequency(r=0.99). Therefore, the results of the study probably reflected whole-word rather than grapheme co- occurrence frequency.
Conclusion Adult language learners quickly grasp information such as word form and word meaning (initially word form, and then word meaning) L2 learning is not as slow as it is often thought to be Some behavioral experiments might underestimate what has been learned.
Implications of the study The method used in the study can be extended to test the influence of L1-L2 similarity, instructional methods and learners’ age on L2 learning.