Binding in Finnish and the language-cognition interface

Binding conditions are usually modelled as a mapping from syntactic structures into sets of coreference relations expressed and represented by narrow syntactic devices such as indices and/or formal operations such as Agree. Here we consider an alternative based on binding data from Finnish and English in which the mappings are generated dynamically during left-to-right comprehension of an arbitrary number of sentences (“conversations”) at the language-cognition interface. The model assumes that binding regulates semantic assignment management at the language-cognition interface by blanking out portions of the transient discourse available for coreference computations at the hearer’s end. The hypothesis is tested by using the computational generative grammar methodology.