Bilingual lexical memory: Towards a psycholinguistic model of adult L2 lexical acquisition, representation, and processing

Abstract

Present models of bilingual lexical processing assume common meaning representations between lexicons. The nature of these representations is such that a single meaning "node" or "set of nodes" is thought to subserve L1 and L2 translation-equivalent forms. Models of this type face two critical problems. First and foremost is the very real problem that there are very few true translation equivalents. Not only do translation "equivalents" frequently mean slightly different things, but quite often they can be used language specifically in ways the translation equivalent is unable to capture. The second problem facing these models is asymmetrical lexical performance between languages in translation priming tasks. For example, priming is robust in the L1 → L2 direction, but not in the L2 → L1 direction. Models assuming a symmetrical relationship between a common meaning node (or set of nodes) and translation-equivalent lexical forms cannot provide a straightforward account of these phenomena. In the present thesis I propose the sense model, which holds that meaning representations are comprised of distinct semantic senses, some of which may be shared across languages. A representational asymmetry is assumed between lexicons, such that, on average, L1 forms are associated with more semantic senses than L2 forms. Initially, L2 forms are associated with a restricted number of semantic senses that have been extracted from the semantic entry of the L1 translation equivalent. Later on in L2 lexical development, semantic senses specific to the L2 are incorporated into the semantic entries of L2 lexical items. The value of the sense model comes in its ability to account straightforwardly for (one) how translation "equivalents" can be used language specifically in ways not captured by its translation (the particular sense is not shared across languages); and (two) the patterns of asymmetrical lexical performance between languages. Because many of the senses represented in L2 entries are also represented in their L1 equivalent, the proportion of L2 senses activated by the L1 equivalent is large, if not complete. Conversely, because there are many senses represented in L1 entries that are not similarly represented in the L2 equivalent entry, the proportion of L1 senses activated by the L2 equivalent is very small. Hence, the translation priming asymmetry is argued to be the logical consequence of the representational asymmetry assumed by the sense model.