BILINGUAL SEMANTIC AND CONCEPTUAL REPRESENTATION 1To appear in J.F. Kroll & A. M. B. De Groot (Eds.), Handbook of bilingualism: Psycholinguisticapproaches. New York, NY: Oxford University Press.Bilingual Semantic and Conceptual RepresentationWendy S. FrancisThe University of Texas at El PasoAbstractThe question of whether and to what extent semantic or conceptualrepresentations are integrated across languages in bilinguals has led cognitivepsychologists to generate over 100 empirical reports. The terms semantic andconceptual are compared, contrasted, and distinguished from other levels ofrepresentation, and terms used to describe language integration are clarified. Theexisting literature addressing bilingual episodic and semantic memory at the levelof semantic systems and at the level of the translation-equivalent word pair issummarized. This evidence strongly favors shared semantic systems, and sharedsemantic/conceptual representation for translation equivalents. Translationequivalents appear to have a different and closer cognitive status than within-language synonyms. Important directions in future cognitive research onbilingualism include neuroscientific and developmental approaches.Bilingual semantic and conceptual organization has been a topic of interest to cognitiveresearchers because of the fundamental cognitive question of redundancy versus efficiency ofrepresentation. Solutions to the redundancy/efficiency question will be important, for example,in understanding how two languages can be used competently within a single mind, and perhapsin understanding how second languages are acquired. For the purposes of this chapter, bilingualswill be considered to be all people who regularly use at least two languages (Grosjean, 1992).This definition implies spoken communicative competence, but encompasses people with abroad range of relative proficiencies in their languages. To pursue this topic, severalterminology clarifications are necessary and will be discussed in the subsequent sections beforeturning to the results of research on bilingual representation.Levels of RepresentationThe level of representation involved in a task has been a key to understanding resultspreviously thought to be discrepant. Defining these levels has proved somewhat difficult. Aspointed out previously (Francis, 1999b, 2000), cognitive research on bilingualism, falling at theBILINGUAL SEMANTIC AND CONCEPTUAL REPRESENTATION 2intersection of psychology and linguistics and including both language researchers and memoryresearchers, is inconsistent in the terminology used to address these levels. The example mostcrucial to the present discussion is that the intended relationship between the terms “semantic”and “conceptual” is ambiguous. In most articles, the relationship assumed by the authors is notspecified; in some articles, the terms are used interchangeably, but in others only one term or theother is used exclusively.To untangle this issue, let us start with something non-linguistic: the universe of possibleideas or concepts that a human can learn or understand. This must be a systems-level, possiblyhard-wired aspect of human cognition that ought to be the same across all cultures andlanguages. However, the concepts actually realized in a given person’s experience may vary,with systematic patterns across cultures. Also, in a person who is bicultural, certain conceptsmay be more relevant or more accessible in one cultural context than in the other.Now, how is language mapped onto this conceptual system? This is where semantics comeinto play. Any of the concepts a person can know ought to have the potential to be expressed inany human language. Of course, the concepts actually realized in an individual’s language inputor output will vary with systematic patterns across languages. Semantic representations may bethose concepts that are referred to by particular words or sentences. Thus, semanticrepresentations would be representations of word or sentence meaning. Word meanings, orsemantic representations of words, would be a particular type of concept. This view would beconsistent with the position of many linguists (e.g., Jackendoff, 1994). A second way to thinkabout this relationship is to consider semantic representations or word meanings as the mappingsof verbal labels to their concepts. While many of the concepts a person knows can be expressedusing individual words, there are of course many more concepts that are not associated with anyparticular word. Such concepts have the potential to be expressed as sentences (or larger units oflanguage), in which case, these concepts would be the semantic representations of the sentencesthat express them.Across different researchers in this area, some use “conceptual” or “semantic” representationexclusively to describe the focus of their research, and others use both terms interchangeably. Ifsemantic representations are considered to be a subset of the set of possible conceptualrepresentations, then all three practices seem reasonable, because the conceptual representationassociated with a particular word or sentence is its semantic representation. Some researchers(e.g., Pavlenko, 1999) have advocated separation of conceptual and semantic levels ofrepresentation. In theory, if the constructs are different, they ought to be separable. However,with a subset relationship, the separation is unlikely to be viable in experimental practice, at leastwith respect to language. To clarify, if a researcher is interested studying those concepts thatalso happen to be semantic representations of words, there is no obvious way to separate them.On the other hand, if a researcher is interested in studying concepts that are not semanticrepresentations of words or sentences, then, ironically, it is not clear how to study them usinglanguage stimuli.The central question about bilingual semantic/conceptual representations is the degree towhich they are integrated across languages. Here it is important to talk about systems of possiblerepresentations versus representations of specific words. A semantic or conceptual system canbe considered to have an innumerable set of possible semantic components, of which any wordmeaning is identified with a subset or a particular pattern of activation or “connection weights”across the entire system. Within this framework, one can consider the degree of integration at theBILINGUAL SEMANTIC AND