ABSTRACT
This chapter talks about the fundamental problem of getting the child started in forming the correct types of rules for natural languages. This "bootstrapping problem" is the first problem one must solve in designing models of language acquisition. The chapter reviews a solution that worked well enough to lie at the foundation of an explicit theory of language acquisition I outlined in Pinker, 1984a Language Learnability and Language Development (LLLD), henceforth called the "LLLD"theory. Then it sketches out an alternative class of language acquisition mechanisms, based on "constraint satisfaction" models taken from artificial intelligence research, and shows how these models might solve the bootstrapping problem. This chapter also spells out the implications of this class of models for the debate over language bootstrapping. Prosody is surely important in language bootstrapping, but we need explicit proposals about the psychophysics that the child would exploit; Cooper and Paccia-Cooper's findings are not enough.
