things dogs or tables are ... This is a fundamental human capacity that differs from the ability to tell one thing apart from another... As [infants] begin to encounter ani-

Even this brief account of this small subset of Nelson's empirical studies demonstrates that some aspects of early language acquisition can only be explained with explicit reference to the notion of function-both in the sense that underlying concepts have a functional core and in the sense that linguistic categories are formed on the basis of communicative-discourse function. These are both ideas on the cutting edge of current thinking about the relation of language to cognitive development (e.g., see the recent work of Slobin, 1991; Bowerman, 1993, on "thinking for speaking"; and Tomasello, 1992a, 1999, on grammatical category formation as "functionally based distributional analysis").