Grammatical Gender and Meaning

Grammatical Gender and Meaning Gabriella Vigliocco (g.Vigliocco@ucl.ac.uk) Department of Psychology, University College London, Gower Street London WC1H 6BT, England David P. Vinson (d.vinson@ucl.ac.uk) Department of Psychology, University College London, Gower Street London WC1H 6BT, England Federica Paganelli (f.paganell@ucl.ac.uk) Department of Psychology, University College London, Gower Street London WC1H 6BT, England Abstract Two experiments assessed whether grammatical gender of Italian nouns referring to animals and tools affects conceptual representations of the corresponding objects, comparing re- sults from Italian and English. In the first experiment, we elic- ited semantic substitution errors (e.g., saying “hammer” when “axe” is intended), finding language-specific gender effects (more errors in Italian than English for words sharing gender) for words referring to animals but not for words referring to tools. In the second experiment, words sharing gender were judged as more similar in meaning by Italian speakers than English speakers, again only for animals and not for tools. Moreover, no such gender effect was observed for pictures of the same animals. Introduction As Roman Jakobson (1959) put it: Languages differ essen- tially in what they must convey and not in what they may convey (p.236). That is, languages differ in which concep- tual or formal properties must be realized in sentential form. For example, in English the word friend does not indicate the sex of the friend, while in Italian the corresponding word is differentially inflected for a man ( amico ) or a woman ( amica ). In English, adjectives used as predicates (e.g., tall in The boy is tall and “The girl is tall”) do not agree in gender with the subject of the sentence, while they must in Italian (e.g., Il ragazzo e’ alto or “La ragazza e’ alta”). Such differences in obligatory expression may imply that speakers of different languages pay more or less attention to those dimensions of meaning. For example, Italian speakers may pay more attention to the sex of referents than English speakers. By extension, Italian speakers may tend to think of objects in the world as more male- or female-like on the basis of the words’ grammatical gender (as suggested by the work of, e.g. Boroditsky, Schmidt & Phillips, 2003; Sera, Elieff, Forbes, Burch, Rodriguez, & Dubois, 2002). But how strong and pervasive can these effects be? Here we present experiments investigating the conditions under which effects of a language-specific property (gram- matical gender of Italian nouns) are present, contrasting performance by Italian and English speakers on translation- equivalent nouns. Grammatical gender allows a conserva- tive test of language-specific effects on cognition because it is largely arbitrarily linked to meaning (although see Foun- dalis, 2002). How could grammatical gender affect conceptual represen- tations for objects? Effects of grammatical gender could arise as a consequence of general language-learning mecha- nisms based on similarity. According to this hypothesis (to which we will refer as “Similarity and Gender”), words that are similar to each other on any linguistic dimension (in- cluding but not limited to grammatical gender) may become more semantically similar as a consequence of the fact that words of the same syntactic class (e.g., same gender, same grammatical class, etc.) appear in the same syntactic con- texts. For example, in languages with grammatical gender, nouns are used in sentences along with gender-marked de- terminers and adjectives, whether the nouns refer to sexuated entities or not. Sensitivity to shared sentence con- text could allow children to bootstrap properties of similar- ity in meaning from the syntactic contexts in which the words occur during language acquisition (Landauer & Du- mais, 1997). This hypothesis does not require any explicit associations between grammatical gender and sex of human referents; instead it predicts that any effects of grammatical gender on semantic representations should be found in any gendered language (no matter how many gender classes are in the language), and that they should be found for all words (whether the referents are sexuated or not). However, mechanisms mediating such effects may be more specific and limited. According to this other hypothesis (to which we will refer as “Sex and Gender”), effects of gram- matical gender could arise because children would treat all grammatical categories as revealing specific semantic prop- erties (Boroditsky, et al., 2003). In the case of grammatical gender, these effects would require linking the grammatical gender of nouns referring to humans to the sex of referents. Across languages there is a core correspondence between grammatical marking of gender and biological sex (Corbett, 1991), although the consistency of this mapping differs across languages. According to this view, children learning a gendered language would first notice the core correspon- dence between the gender of nouns and male/female seman- tic properties of human referents (and some animals). They would then generalize this correspondence to other nouns for which there is no clear conceptual foundation of gender,