Neuroimaging Study Of Semantic Processing In Chinese-English Bilinguals Based On Multivariate Analyses

The neural mechanism of bilingualism has always been a hot topic of neuroimaging research.In this thesis,functional magnetic resonance imaging(f MRI)was combined with traditional univariate analysis and multivariate analyses to study neural activity of language processing in the bilingual brain.In this study,twenty-six Chinese-English bilinguals were recruited and their functional magnetic resonance data were collected when they completed both Chinese and English sentence comprehension tasks.Univariate analysis showed that both native language(L1)and second language(L2)tasks activated bilateral occipital cortex,bilateral motor area,left middle and inferior temporal cortex,left parietal cortex,bilateral supraorbital cortex and so on.Direct contrast between neural signals of the two languages did not find any voxel showing significant activation,indicating that there was no significant difference in activation levels between L1 and L2 based on univariate analysis at the group level.We then performed three multivariate analyses—multivariate pattern analysis(MVPA),representational similarity analysis(RSA),and a connectome-based predictive modeling(CPM)to study similarity and differences between L1 and L2 neural activity from multiple perspectives,as well as relationship between neural activation and L2 learning.MVPA takes the relationship of multiple voxels into consideration in the study of the differences between the two language neural activity patterns.RSA quantified the difference in neural activity patterns between L1 and L2,and further studied the relationship between the neural pattern differences and L2 acquisition age(Ao A)and L2 proficiency.CPM establishes a functional connection model for predicting behavioral metrics such as proficiency in L2,and used this model to investigate which connections were most closely related to L2 proficiency from a functional connection perspective.The mean accuracy of classifiers trained in MVPA were more than 99% for classifying Chinese and English conditions and the accuracy of Chinese and English classification for all subjects was 58.89%.The results of MVPA show that the activation patterns of the two tasks are highly distinct in multivariate mode.In the RSA analysis,the differences in neural activity patterns between L1 and L2 were negatively correlated with Ao A in MFG(Middle frontal gyri),IFG(Inferior frontal gyri),ITG(Inferior temporal gyri)and LOC(Lateral occipital cortex)and positively correlated with second language proficiency in ITG and MFG.The results showed that the earlier the L2 were acquired and the higher the proficiency was,the higher the degree of pattern difference between L1 and L2 in these brain regions.This may be due to the fact that as Ao A increase,bilinguals may prefer to use the mature and stable language nervous system,i.e.,the L1 neural network,to process L2,and this degree of propensity leads to changes in the differences in neural activity between the two languages.In the CPM analysis,there were differences in functional connectivity between Chinese and English tasks,but there was no significant correlation between the predicted CPM value of L2 Ao A and proficiency and the test value.In sum,this thesis used univariate analysis and multivariate analysis methods to study brain activity pattern in Chinese-English bilinguals.The results revealed the differences in multivariate neural patterns between the two languages and the relationship between these differences and L2 Ao A/proficiency.The results provided new evidence for revealing the brain mechanism of bilingual semantic processing,and also provided insight for neuroscience-based bilingual education.