CNL Lunch

Jeff Lidz
Northwestern University

Where to find UG

Thursday February 28th, 12:30pm, 3416 Marie Mount Hall

In recent years, there has been a resurgence of interest in statistical approaches to langauge acquisition. This research is driven by the observation that distributional evidence for a given analysis is often available in the input and can be taken advantage of by a learner. In this talk I consider two cases in which young children and infants seem to be reaching grammatical conclusions that are unsupported by distributional analysis of the input that they receive. In the first case, the acquisition of the syntax of causation in 3 year-olds, I show that children seem to be ignoring the input in favor of universal grammar, viewed as a kind of filter on the input. In the second case, the acquisition of anaphoric "one" and the structure of NP in 18-month-olds, I show that there is no evidence at all in the input that would support the structures that these infants seem to possess. Thus, children's presuppositions about what the input is likely to be like are better predictors of their behavior than what the input actually is like.