Successful language comprehension requires the computation and integration of several conceptually distinct types of information. Two exigencies are the recovery of linguistic information from long-term lexical memory and the computation of a structured representation. This program of research investigates the essential nature of these component processes as they apply to language comprehension in normal, intact populations. The primary aim of the project is to specify the computational structure of lexical and syntactic processes, a necessary precursor to the development of well-specified computational models of natural language understanding. Using procedures that tap the time-course of processing and procedures that measure processing resources, we investigate whether these processes have an intrinsic serial or parallel capacity. We examine whether lexical retrieval processes recovery of the exhaustive set of syntactic structures associated with key elements in a sentence. Results in hand indicate that lexical retrieval is parallel but short of exhaustive, suggesting that limited retrieval constrains subsequent syntactic and interpretative processes. To better understand these constraints, experiments are designed to identify factors that control the probability of retrieving syntactic information from lexical representations. Concerning parsing operations, we investigate whether more than syntactic representation is computed during on-line processing. One set of experiments addresses the issue of parsing capacity at the level of immediate structure building, while other experiments examine how parsing routines respond to ambiguous material. Preliminary results suggest that several lexical syntactic structures are applied in parallel in the initial phase of structure building. Experiments are proposed to explore this finding further and to examine whether contextual information can restrict initial structure building operations. An additional set of experiments investigates how readers respond to temporarily ambiguous material, examining whether more than one representation is computed and maintained over an ambiguous region. Finally, a third set of experiments investigates how lexical and interpretative processes successfully converge on the correct interpretation of syntactically unambiguous but indeterminate structures. We argue that lexical and syntactic processes must be supplemented by a type of enriched interpretative processing.
