Friday, April 17, 2009
University of Rochester, BCS department
A unified model of lexical and syntactic processing?
A key question in human language processing is the amount of lexical detail in the representations of syntactic knowledge, and to what extent lexical and syntactic access are the same process. There is increasing evidence that people have knowledge of probability distributions of sequences and structures beyond the word level (Bell et al 2003, Aylett and Turk 2004, Gahl and Garnsey 2004, Levy 2008), just as they have for words. Such evidence has led to the proposal that there may be no strict difference between the lexical and syntactic levels (Jurafsky 1996, MacDonald et al 1994), so similar processing effects should be found for words and multi-word structures. Evidence will be presented from several corpus studies and experiments that, like lexical access, syntactic production and comprehension are sensitive to probability distributions of fine-grained representations, as well as fine-grained similarity to recent previous instances of language use. However, unlike lexical access, syntactic production appears to not be sensitive to several measures of the density of the syntactic representational neighborhood. The implications for theories of syntactic representation and production models will be discussed.