ABSTRACT
A central theme running through several of the chapters in this book concerns the question of whether syntactic and semantic processing are interactive or not. The chapters by Frazier and by Perfetti start from the conventional assumption that there are distinct syntactic and semantic processors. The question of interest then concerns how these two processes communicate with each other. The whole question of interaction has generally been based on the assumption that there are separate processors responsible for syntactic and semantic analyses. However, the strongest possible claim one could make about interaction between syntactic and semantic processing would be to argue that they were so tightly coupled that they should properly be thought of as a single integrated processing system. This is the view taken in the chapter by Taraban and McClelland. They consider that comprehension is a single, highly interactive process, which is not decomposable into subcomponents. There are no separate processes of parsing or semantic interpretation. Comprehension consists of a single process in which multiple constraints from different sources are weighed together in parallel to produce a final interpretation of the sentence which minimizes the discrepancies between the various constraints. This parallel and highly interactive constraint satisfaction mechanism probably takes the form of a large connectionist network of the sort proposed by St. John and McClelland (in press). In this model, a single, undifferentiated network takes responsibility for processing operations that we would normally think of as being performed by separate, but possibly interacting, processes of parsing and semantic interpretation.
