Modelling the steps of early syntax acquisition

This paper reports on empirical research concerning the early stages of syntactic development in first language acquisition, and investigates the prospects of modelling the acquisition process in an exemplar-based linguistic framework (DataOriented Parsing). We present a detailed analysis of the early stages of verbal placements in Dutch. Our analysis shows that the child invents a sequence of grammars. She starts with rigorous, but systematic simplifications. The child approaches the adult grammar in a sequence of steps. We propose that each acquisition step can be characterized by (a) a new intermediate grammar and (b) a corresponding input reduction strategy that selects the new data for the next acquisition step. We discuss in some detail the how a child acquires the complexities of Dutch word order. It turns out to be extremely implausible that this process could be simulated by a model whose syntactic operations are limited to tree-substitution.