Abstract
Children learn new words at a prodigious rate, over nine words a day during early childhood. Further, they often converge on the adult meaning of a new word after hearing it only a few times in a single context. As many have noted, the inductive processes that underlie such efficient word learning must be highly constrained. One source of constraint derives from the fact that words in different syntactic categories differ systematically in meaning. Young children exploit information about syntactic subcategorization to inform their hypotheses about word meanings.
This paper asks to what extent such constraints reflect prelinguistic cognitive architecture, i.e., conceptual distinctions antecedently available to constrain syntax acquisition as well as word learning, and to what extent such constraints reflect language specific, culturally constructed, conceptual categories which must be induced in the course of language learning. This question is explored via a case study within the domain of noun semantics and the representation of number, five aspects of which are examined: the representations of integers in counting sequences (“one, two, three…”), quantifiers such as one, another, the criteria for individuation and identity embodied in the sortal concepts the language lexicalizes, the distinction between count and mass nouns, and the distinction between count nouns and predicates.
In this paper I sketch data from infant studies that suggest that all but the first of these (the representations of integers) are part of prelinguistic cognitive architecture. These elements of constraints on word meanings are not induced from language learning; rather language learning, including lexical learning, builds upon them.
