Language-specific effects on the semantic representations consulted in lexical retrieval for production. Gabriella Vigliocco & Sotaro Kita

In this talk we will first introduce architectural assumptions concerning the levels of representation consulted in lexicalising concepts and then we will discuss issues of separability and information flow between lexico-semantic representations and conceptual, syntactic and phonological information. All the arguments we present and the evidence we discuss draw upon cross-linguistic comparisons. Our aim is not to provide a new theory of lexicalisation and lexical retrieval in production but to show that language-specific properties can affect semantic representations and that these findings provide constrains on the information flow assumed by theories between levels.

A separation between conceptual and lexico-semantic representations is generally agreed upon by researchers in language production. Lexico-semantic representations are further distinguished from wordform representations and from syntactic properties of words. Models, then, greatly differ with regards to how these different levels of representations are characterised and how the information flow from one level to another is conceived. We take the Featural and Unitary Semantic System (FUSS) hypothesis (Vigliocco et al., 2004) in which conceptual representations are featurally-based, and lexico-semantic representations bind conceptual features into lexicalised concepts and provide an interface between conceptual knowledge and linguistic information (syntactic properties and phonology) as a starting point. One prime motivation in FUSS for separating conceptual and lexico-semantic representations is cross-linguistic variability in what aspects of the world are lexicalised in specific languages. Thus, whereas speakers of different languages may share conceptual representations, which bundles of features are lexicalised depends upon the specific language.

Given this basic architecture in which lexico-semantic representations interface with conceptual information on the one hand, and with syntactic and phonological information on the other hand, we address issues concerning information flow and separability at the following interfaces: (i) the interface between lexico-semantic and conceptual representations; (ii) the interface between lexico-semantic representations and syntactic information; (iii) the interface between lexico-semantic representations and phonological information. At all of these interfaces, the first crucial question is whether language-specific effects can be observed. The second question is how they come about.

Concerning the interface between lexico-semantic and conceptual representations, we will discuss the work by Kita and Özyürek (2003) that suggests language specific effects from lexico-semantic to conceptual representations. In a critical experiment, they investigated cross-linguistic differences in the spontaneous gestures that accompany speech for speakers of English, Japanese and Turkish. Co-speech gestures convey imagistic information (reflecting visuo-spatial properties of a referent, i.e., conceptual representation) but are also tightly linked to speaking as, for example, they are synchronized to co-expressive speech. Turkish, Japanese and English speakers were presented with video clips depicting motion events (for example, swinging) and were asked to describe the events. The swinging event is especially of interest because English has a straightforward way of lexicalizing the event with the verb "to swing" whereas Japanese and Turkish do not have any equivalent verb referring to change of location with an arc trajectory. It was found that spontaneous speech-accompanying gestures that depicted the swinging event reflected this cross-linguistic difference. More specifically, English speakers depicted the event with an arc gesture more often than Japanese and Turkish speakers, who often depicted the event with a straight gesture. This indicates that lexico-semantic representation influences the conceptual representation, which underlies gestures.

Concerning the interface between lexico-semantic and syntactic information, we will discuss the work by Vigliocco and colleagues (Vigliocco, Vinson & Paganelli, submitted) that suggest limited language-specific effects of a syntactic variable such as grammatical gender on the semantic representation for objects. In a critical experiment, semantic substitution errors (e.g., saying "leopard" when "tiger" is intended) were induced by asking speakers of Italian and English to name pictures of animals and tools. For both types of words, the gender of the nouns does not have a transparent link to properties of the referent. Of interest was whether the substitution errors by the Italian speakers tended to preserve gender (English was used as a baseline for controlling for other factors contributing to the substitution errors). It was found that gender had an effect on the errors for words referring to animals but not for tools. Such a tendency to preserve gender suggests that even an idiosyncratic syntactic property like gender can affect lexico-semantic representations.

Concerning the interface between lexico-semantic and phonological information, we will discuss a class of words called "mimetics" in Japanese (see Kita, 1997). Mimetics are sound symbolic in the sense that sound and meaning correspond systematically. For example, when the word starts with a voiced consonant, it refers to an event involving a large object (e.g., "goro", a large object rolling) whereas when it starts with a voiceless consonant, it refers to an event involving a small object (e.g., "koro", a small object rolling). Mimetics participate in sentential syntax as an adverbial or nominal element, and are used very frequently in Japanese conversation. Sound symbolism in mimetics opens up an interesting possibility that information may flow directly from lexico-semantic representation to phonological information (i.e., phonological features and/or metrical information) when producing these words. Although it has still to be empirically established whether mimetic words behave in a special way, theoretical implications of such possible information flow will be discussed.

These different lines of investigation, traditionally, are kept distinct. Our presentation will show, instead, how we can fruitfully consider these different language-specific effects together as they provide crucial constraints on models of lexical representation in production and on the flow of information during lexical retrieval.

References

Kita, S. (1997). Two-dimensional semantic analysis of Japanese mimetics. Linguistics, 35, 379-415.

Kita, S., & Özyürek, A. (2003). What does cross-linguistic variation in semantic coordination of speech and gesture reveal? Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language, 48, 16-32.

Vigliocco, G., Vinson, D.P, Lewis, W. & Garrett, M.F. (2004). Representing the meanings of object and action words: The featural and unitary semantic space hypothesis. Cognitive Psychology, 48, 422-488.

Vigliocco, G., Vinson, D.P., & Paganelli F. (submitted). Are giraffes girly in Italian? An exploration of language-specific effects on mental representations.