A Study On Models Of Bilingual Mental Lexicon Representation And Access

As a research focus in the psycholinguistics, study on models of bilingual mental lexicon representation has attracted many researchers'attention in the past few decades. Starting with the main methods of empirical research concerning bilingual mental lexicon representation, this dissertation has made a review and critical comments on various models of bilingual mental lexicon representation.From the author's point of view, though word association model and concept mediation model suggest bilinguals'L2 proficiency will determine the L2 word processing route so that low proficiency leads to word association and high proficiency results in concept mediation, these two models cannot explain some research findings that even nonproficient bilinguals process L2 words through both its L1 translation equivalent (word association) and concept mediation. The revised hierarchical model has reconciled the word association model and concept mediation model into one, whereas it is found weak to explain some research results contending both directions of translation are affected by meaning variables. Although distributed models have done an excellent job in accounting for word concreteness effect, they have not made claims about how lexical form and meaning might interact during word recognition.While the above models focus on representation at only lexical and conceptual levels, the BIA model and BIA+ model go one step further into sublexical level, incorporating orthography, phonology and semantics in bilingual word identification system as is held by BIA+ model that orthographic and phonological representations at the lexical and sublexical levels of the two languages of bilinguals are stored in an integrated lexicon. Moreover, these two models deal with lexical access and both models are in support of language non-selective access. However, these two models have their limitations in explaining languages using different alphabets or non-alphabetic languages such as Chinese and English because they are only restricted to languages sharing orthography to some extent.Based on all the previous studies, this dissertation specially considers the mental lexicon representation structure of Chinese-English bilinguals and proposes two representation models: the distributed asymmetrical model and distributed symmetrical model. This dissertation puts an emphasis on the distributed asymmetrical model for non proficient (unbalanced) Chinese-English bilinguals.The distributed asymmetrical model holds that two languages of Chinese-English unbalanced bilinguals are connected at both lexical level and conceptual level in an asymmetrical way. At the lexical level, the link from L2 to L1 is stronger than that from L1 to L2, and at the conceptual level, L1 words have stronger link to the concept system than L2 words. The distributed feature at the conceptual level shows the degree of conceptual overlap across languages and suggests translation equivalents across language may share some concepts in common as overlapping nodes as well as represent language unique features separately for each language.In order to test the distributed asymmetrical model, a word association experiment was conducted using a group of 65 Chinese college students as subjects. The results indicate that these Chinese EFL learners have partly shared conceptual representation of Chinese and English lexicons, and the patterns of responses in Chinese are different from that in English. These findings suggest that Chinese EFL learners (i.e. low proficient Chinese-English bilinguals) have an asymmetrical representation of their mental lexicon in which L1 words have stronger link with the concept system than L2 words and L1 lexicon is larger than L2 lexicon. Moreover, conceptual overlapping across languages as well as language-specific representation patterns reveal the distributed feature at the conceptual level which suggests translation equivalents across languages may share some concepts in common as overlapping nodes and also represent language unique features separately for each language. Thus the experiment supports the distributed asymmetric model in its distributed and asymmetric features at the conceptual level and asymmetric lexicon sizes. However, the asymmetric feature at the lexical level (i.e. the link from L2 to L1 is stronger than that from L1 to L2) as predicted by the asymmetric distributed model is beyond the testing of the experiment.