A Study Of The Cognitive Mechanisms And Neural Basis Of Statistical Learning For Chinese-English Bilinguals

Statistical learning refers to the process by which individuals gradually discover statistical patterns in a continuous stream from the environment and was first proposed in Saffran et al.’s(1996)classical infant speech segmentation study.Specially,individual differences in statistical learning were reflected in their sensitivity to statistical structure such as frequency and probability.Previous studies have confirmed that statistical learning have a great effect on cognitive activities such as language.Recently,the relationship between statistical learning and bilingual experience has received considerable empirical attention.However,the research on the mechanism of bilingual statistical learning advantage is not deep enough,failing to consider the occurrence conditions of bilingual advantage and the neural mechanism in the process of statistical learning.At present,there are two main approaches to interpret the bilingual advantage in statistical learning:bilinguals may have strong cognitive ability to a certain extent,that is,bilinguals have better executive function.Thus,bilinguals may have enhanced executive functions to support and facilitate statistical learning(cognitive advantage).Another explanation is that bilingual experience may make learners more sensitive to changes in the underlying structure in perception when processing language input in a novel environment.The experience of dealing with multiple languages in a bilingual environment may make bilinguals more sensitive to changes in different language inputs and more likely to form predictions about changes in language inputs(representational advantage).The main research goal of this dissertation is to explore how the statistical learning advantages of bilinguals are reflected in different statistical cues,and to clarify the internal mechanism between bilingual experience and statistical learning performance.In the current study,according to the theory of bilingual advantage,we conducted three behavioral experiments to explore the effects of single and multiple statistical cues on the statistical learning for speech segmentation in Chinese monolinguals and Chinese(L1)-English(L2)bilinguals,focusing on whether the advantage of bilinguals in statistical learning tasks stems from cognitive advantage or representation advantage.Moreover,we also did an ERP to reveal the neural basis of the potential advantage in statistical learning.The main questions to be answered are as follows:(1)Compared with Chinese monolinguals,do Chinese-English bilinguals have an advantage in single-cue statistical learning?(Verification of cognitive advantages)(2)Compared with Chinese monolinguals,do Chinese-English bilinguals have an advantage in the statistical learning of pauses(a suprasegmental feature shared by languages)and syllable statistical cues?(Verification of cognitive advantage and representational advantages of linguisticgeneral cues)(3)Compared with Chinese monolinguals,whether Chinese-English bilinguals have advantages in the statistical learning of tone(specific suprasegmental features of tone language)and syllable statistical cues?(Verificaiton of cognitive advantages and representational advantages of linguisticspecific cues)(4)What is the neural mechanism behind the different performances of Chinese monolinguals and Chinese-English bilinguals?According to the research purpose and the research questions of this dissertation,four experiments were carried out in this paper.Experiments 1,2,and 3 all used behavioral experiments,and Experiment 4 used event-related potential(ERP)technology to explore the neural activities involved in the process of statistical learning.This dissertation explored first the statistical learning perfotmacne of bilinguals in single cues and multiple cues,and then explore the neural basis of statistical learning of bilinguals,aiming to reveal the cognitive rules and neural basis of statistical learning of bilinguals.Experiment 1 tested the statistical learning of Chinese monolinguals and Chinese-English bilinguals in a single-cue monotone artificial language.The test materials in Experiment 1 did not contain any language-related acoustic cues and but only statistical relations between monotonic syllables.Compared with monolinguals,whether bilinguals have cognitive advantages rather than representational advantage for mastering multiple languages and clarify the features of learners’ statistical learning of single clues.The results found that Chinese monolinguals and Chinese-English Bilinguals can successfully perform word segmentation,and the statistical learning performance of Chinese-English bilinguals in monotonic artificial language was better than that of Chinese monolinguals.Importantly,in the regression analysis,it was found that the accuracy of 2-back and digit span tasks can significantly predict the accuracy rate of word segmentation in the statistical learning task to a certain extent,which suggested that working memory can more significantly predict word segmentation in statistical learning tasks.Therefore,Experiment 1 shows that the advantages of bilinguals in statistical learning tasks may be derived from cognitive advantages rather than representational advantages.Experiment 2 examined multiple cues(i.e.,pause and syllable cues consistent or inconsistent)statistical learning performance between Chinese monolinguals and Chinese(L1)-English(L2)bilinguals.In Experiment 2,two segmentation cues,syllables with statistical relations and pauses(pause is the common segmentation clue of language),are provided.The purpose is to investigate the statistical learning of bilinguals in the pause cues shared by languages,verify the representational advantages of bilinguals,and explore the mechnism of bilinguals’statistical learning of multiple consistent and inconsistent cues.The results found that(1)In the condition of consistent pauses and syllable cues,there was no significant difference in performance between Chinese monolinguals and Chinese-English bilinguals;in the condition of inconsistent pauses and syllable cues,there was a significant difference in performance between Chinese monolinguals and ChineseEnglish bilinguals.The statistical learning performance of Chinese-English bilinguals is better than that of Chinese monolinguals.(2)The regression analysis found that in the consistent condition,working memory was somewhat predictive of statistical learning performance,while in the inconsistent condition,learners who can successfully segment the artificial language(using any available cues:pauses or syllables with statistical information)tend to respond faster in Simon tasks,while those with an accuracy of about 50%(without learning either of the two segmentation cues)tend to respond slower in Simon tasks.Therefore,the results of Experiment 2 suggested that the advantages of bilinguals in statistical learning tasks may derived from cognitive advantages rather than representational advantages of pauses,a linguisitic-general cue in languages.Experiment 3 examined the statistical learning performance of bilinguals in the tone and syllable cue consistent or inconsistent conditions.In Experiment 3,two segmentation cues,syllable with statistical relations and Mandarin lexical tones,were provided.The main purpose was to investigate the characteristics of bilingual statistical learning in language-specific tone cues and verify the representation advantages of bilinguals.The results found that(1)Chinese monolingual speakers and Chinese-English bilingual speakers had no significant difference in the condition of consistent tone and syllable cues;There were significant differences in performance between Chinese monolinguals and Chinese-English bilinguals in the pause and syllable cue inconsistent conditions.(2)The results of regression analysis found that in consistent conditions,working memory had a predictive effect on the performance of statistical learning,while in inconsistent conditions,performance in Simon task can significantly predict the performance of learners’ statistical learning.Learners with high accuracy in word segmentation tasks tend to respnd faster in Simon tasks,while those with low accuracy tend to respond slower in Simon tasks.Therefore,the advantages of bilinguals in statistical learning tasks may derived from cognitive advantages rather than representational advantages of pauses,a linguistic-specific cue.Experiment 4 did an ERP study to measure the neural mechanisms of online statistical learning processing in bilinguals,and the study revealed an interesting phenomenon:Chinese-English bilinguals elicited larger N100 and N400 waveforms during statistical learning.The N100 and N400 amplitudes of ChineseEnglish bilinguals correlated with the transitional probability in the artificial language,i.e.,syllables located between words elicited larger wave amplitudes of N100 and N400 wave amplitudes than syllables within words.The N400 component differed significantly between between Chinese monolinguals and Chinese-English bilinguals.,whereas the N100 component did not differ significantly between Chinese monolinguals and Chinese-English bilinguals.On this basis,we propose that:First,Chinese-English bilinguals have statistical learning advantages in the condition of a single cue or multiple inconsistent cues;Second,The performance of Chinese monolinguals and ChineseEnglish in statistical learning with multiple consistent cues was related to working memory,and multiple consistent statistical cues can help promote statistical learning;Third,the statistical learning of inconsistent cues of Chinese monolinguals and Chinese-English bilinguals was related to the Simon effect,that is,the stronger the inhibitory control ability,the better the statistical learning word segmentation performance in inconsistent cues;Fourth,ERP research shows that the advantages of bilinguals still exist in the process of online statistical learning,and the the N400 amplitude induced between words can predict learners’ statistical learning performance..To sum up,this paper draws the following conclusions:(1)The superior performance of bilinguals in statistical learning may be due to better cognitive advantages,that is,better working memory and inhibitory control ability of bilinguals;(2)Bilinguals have different brain mechanisms of statistical learning from monolinguals,that is,in the process of statistical learning,bilinguals can induce wider and larger N400 components;(3)Individual differences in language experience and cognitive ability as well as statistical stimuli with different features will restrict the performance of statistical learning.