Abstract: This chapter covers three main points. The first is that, with respect to null subjects in young children’s speech, the data collected thus far indicate no point at which the grammar of U.S. children speaking Standard English (henceforth, American children) clearly licenses null subjects, and no point at which IP and CP are clearly absent. In contrast, the grammars of children acquiring null subject languages do show clear evidence for null subjects, and, equally, show evidence at least for IP. This is not to say that no American child ever has an incorrect grammar, but simply that the data thus far give us no grounds for claiming an incorrect grammar for most children. The data are briefly reviewed here.
The second point is that, in order to account for the diversity as well as the commonalities in acquisition within and across languages, theories must specifically include both a competence component and a performance component, and a model of how the two interact. Each component by itself is too weak in predictive power to handle the facts. A corollary of this is that there is no metatheoretic reason to prefer competence-deficit explanations over performance-deficit explanations.
The third point is that children’s initial state is, with respect to parameters, unset. As I have argued in previous work (Valian, 1990a, 1990b), the child does not begin acquisition with one or another value preset; there is no default setting. Rather, the child entertains both options on an equal footing until sufficient evidence accrues to favor one over the other, and he or she remains with that value unless and until sufficient evidence accrues to switch to another value.