In natural languages, closed class items (e.g., the) determine the structure of a sentence. With Elissa Newport, I am studying how language learners analyze these items. We have discovered that learning is biased: people treat closed class items as more likely than open class items to occur in a predictable context. Now, we are asking how this bias shapes learners' mental representations of language.

Every language has its own set of rules for how to form sentences. To learn those rules, children have to search for patterns in the sentences they hear. This project asks: What kinds of patterns do learners search for? How do they organize their knowledge of these patterns? And how does knowledge of patterns develop into hierarchically structured representations of language?

A mystery in language acquisition is how children acquire language despite the Poverty of the Stimulus. For example, want to and wanna are not always interchangeable, and this seems impossible to learn given the types of sentences children hear. But I have argued that the input may only appear impoverished, because we are only considering a single learning approach. There may not be enough input for that specific learning approach to work, but there might be another approach for which input actually is adequate. I argued that is the case for wanna.