ABSTRACT
How children learn the meanings of words is a core puzzle in the study of language acquisition. There are commonalities in the general types of semantic notions that tend to get lexicalized across languages and those that are acquired early-for example, concepts of motion, possession, attribution, and the existence, location, and disappearance of objects (Bowerman, 1973; R. Brown, 1973; E.V. Clark, 1973; Johnston & Slobin, 1979; Slobin, 1973, 1985). There is also, however,
wide variation in how languages cut up the world into semantic categories, variation which is perhaps most systematically documented for the spatial domain (e.g., Ameka & Levinson, 2007; Levinson & Meira, 2003; Levinson & Wilkins, 2006; Majid, Eneld & van Staden, 2006). Children beginning to speak have to integrate the ways in which speakers around them use words in particular contexts with the prelinguistic categories they have already formed, in order to prune or expand these categories. They may have to create new categories, so that they can use words in the situations that call for them and not in others. They do this at an astonishing rate and from a very young age.2 The categories that they form also look language-specic from a very young age, as demonstrated by the work of Bowerman and her colleagues since the early to mid-1990s (Bowerman, 1996; Bowerman & Choi, 2001, 2004; Brown, 2001, 2008; Choi & Bowerman, 1991; de León, 2001). How children do this is still something of a mystery.
