A functional theory of gender paradigms

A central goal of typological research is to characterize linguistic features in terms of their functional role in a language. One longstanding puzzle for typologists concerns why certain languages employ grammatical gender, which assigns nouns to distinct classes. From a taxonomic perspective, gender specification can appear arbitrary, with little obvious correspondence between semantics and noun class. Gender has thus long been viewed as a useless ornament with no apparent rhyme or reason. However, there is an accumulating body of evidence that native speakers use determiners to guide lexical access. Here, we investigate whether an information theoretic perspective might shed some light on the communicative function of noun classification in German. We hypothesize that the system works to efficiently smooth information over discourse, making nouns both more predictable — and more equally predictable — in context. In line with these predictions, a large-scale corpus analysis reveals that German gender markers systematically reduce nominal entropy, facilitating the use of a more diverse (and more informative) set of nouns. Moreover, the structure of the gender system mirrors that of other subsystems of language, in that it provides systematic support for lower frequency forms. Thus, it is only from a taxonomic standpoint that gender's purpose can appear opaque: Our findings indicate that German gender classes conform to a tight 'discriminative' logic, employing a structure system of semantic clusters and contracts to facilitate lexical processing.