Conceptual organisation of the Chinese-English bilingual mental lexicon: investigations of cross-language priming

The majority of research on the organisation of bilinguals’ lexical memory has focused on alphabetic languages with shared etymological roots and scripts. Theories based on such evidence may not generalise to noncognate languages with different scripts, such as Chinese and English. This thesis reports a systematic series of experiments designed to investigate the organisation of lexical and conceptual knowledge for bilinguals’ first (L1) and second (L2) language in late L1-dominant Chinese-English bilinguals using the classical cross-language priming paradigm. It aims to investigate how such bilinguals store the meanings of Chinese and English words. It also aims to identify the similarities and discrepancies in the conceptual organisation between noncognate languages with different scripts, i.e., Chinese and English, and to investigate how the lexical representations of a bilingual’s two languages interact with each other and with the conceptual representation. The introductory chapter reviews early theoretical formulations of bilingualism, and evaluates more recent models of bilingual memory. The empirical chapters present three series comprising eight experiments which directly compared cross-language translation priming and semantic priming in both L1-L2 and L2-L1 language directions under conditions designed to tap automatic semantic processes using the same relatively short stimulus onset asynchrony (SOA) of 200 ms but different priming paradigms and task contexts. Series 1 (Experiments 1A and 1B) compared repetition/translation priming and semantic priming within and between languages for various semantic relations using an unmasked priming paradigm in lexical decision and word naming tasks. Both tasks produced similar patterns of unmasked translation priming in both L1-L2 and L2-L1 directions, although the priming effects in naming were of a smaller magnitude. Both tasks also showed significant unmasked semantic priming effects for English word targets in the L1-L2 and L2-L2 conditions, but there was little evidence of semantic priming for L1 word targets in the L1-L1 and L2-L1 conditions. Neither task yielded any semantic priming in the within-language L1-L1 condition. Series 2 (Experiments 2A, 2B and 3A, 3B) reported two pairs of semantic categorisation and lexical decision tasks designed to test the predictions of the Sense Model (Finkbeiner, Forster, Nicol, & Nakamura, 2004). The experiments replicated Finkbeiner et al.’s finding that L2-L1 priming is somewhat stronger in semantic categorisation than lexical decision, selectively for category exemplars. However, the direct comparison of L1-L2 and L2-L1 translation priming failed to confirm the Sense Model’s central prediction that translation priming asymmetry is significantly reduced in semantic categorisation. The findings therefore did not support the category filtering account of translation priming asymmetry proposed by the Sense Model but were consistent with semantic feedback (e.g., Hoshino, Midgley, Holcomb, & Grainger, 2010; Midgley, Holcomb, & Grainger, 2009) accounts of cross-script L2-L1 translation priming and suggested that pre-activation of relevant semantic features by a category cue compensates for the weak connections between L2 lexical forms and their conceptual referents. Series 3 (Experiments 4A and 4B) directly compared masked translation and cross-language semantic priming for moderately semantically related pairs with no associative relationships, in semantic categorisation and lexical decision tasks. Both tasks showed similar asymmetrical patterns of masked translation and cross-language semantic priming, characterised by larger priming effects from L1 to L2 than from the reverse. The masked translation priming data fully replicated the findings obtained in Series 2. Masked semantic priming was significant in the L1-L2 but not in the L2-L1 direction, and of smaller magnitude than masked translation priming in both directions. Neither experiment found masked L2-L1 semantic priming. These data can be accommodated by a modified version of the Revised Hierarchical Model (RHM, Kroll & Stewart, 1994) based on Duyck and Brysbaert’s (2004) proposal for alphabetic languages in combination with the semantic feedback account. The data are also consistent with the DevLex-II model (Li & Zhao, 2013; Li, Zhao, & MacWhinney, 2007; Zhao & Li, 2010, 2013) regarding the graded relationships between translation and cross-language semantic priming. The findings of this research clearly demonstrated both shared and independent aspects of L1 and L2 semantic representations in unbalanced Chinese-English bilinguals. They are compatible with the cognitive architecture of the RHM combined with the representational assumptions of the Distributed Conceptual Feature Model (De Groot, 1992a, 1992b, 1995; Van Hell & De Groot, 1998)