On the relation between syntactic theory and sentence processing, and a theory of island phenomena
Abstract: There is currently a seemingly intractable gulf among different syntactic theories. Generative syntactic theories in the Minimalist Program are insightful in providing a theory of the objects of language, tree structures. However, Minimalism does a poor job at explaining real-time sentence processing, child language acquisition, and neuroimaging and neuropsychological data. On the contrary, “lexicalist” grammatical theories (e.g., TAG, Construction grammar, Unification) do a much better job at connecting with real-time sentence processing, language acquisition, and neuroscience. However, lexicalist approaches are devoid of insight into the objects of language: where do these stored structures come from, and why do they have the properties that they do? In this talk I propose a reconciliation between the two approaches along the lines of Frank (2002), by positing a Minimalist grammar as a theory of how structures are generated, and TAG as a theory of the use of these structures during sentence production and comprehension. By making these connections more explicit, it is also possible to incorporate recent insights into the nature of working memory into understanding how the structure of sentences is created in real-time. I argue that this integrated approach provides more successful insights into the nature of island phenomena than traditional grammatical and processing accounts: namely, the unacceptability of island structures can be attributed to violations of grammatical principles like subcategorization and the theta criterion, and subjacency can be eliminated from the grammar.