University of Arizona
I'll begin my talk by outlining three currently viable approaches to language acquisition: deduction from universal linguistic constraints, constrained associative learning, and constrained hypothesis selection. I will provide some data against the first approach by demonstrating that infants are able to learn a language-like principle that is not typical of real human languages. I will also provide some data against the second approach by demonstrating that infants appear to entertain multiple possible generalizations for a set of input. Complementing the language data with studies of infants learning musical patterns, I will suggest that, across cognitive domains, infants initially entertain a broad class of hypotheses about the input and that they narrow their hypothesis space through exposure to domain-relevant input statistics.
Reception to follow in 1413 Marie Mount Hall.