Towards the Structuring of First Parse Processes

Abstract

This article discusses the issue of initial structure building in sentences processing. To understand a sentence it is necessary to integrate single words into a syntactic structure. I will show how this initial integration might be pursued by the parsing system. The investigation on initial syntactic processing is restricted to limited local relationships such as verb-complement combinations. Taken these relations as an example I will propose that an initial parsing process can be separated into two phases. The first phase uses a general structuring rule (X-bar) to project from a head to a phrase. Thus, it pursues the first integration of a word onto the sentence level by using the category information of the head. This computation is independent from other lexically listed information and frequency considerations. Although these assumptions can be found in the garden-path theory (see Frazier, 1990, for a detailed description) I will point out that the initial processing of a word does not dependent on attachment decisions initiated by other words. The second phase provides information retrieved from a mental lexicon about the subcategorization properties of the complementation. It can be classified as a checking mechanism (Mitchell, 1987) which guarantees the grammatical interpretation of the so far computed structure. As a matter of fact the proposed separation of the initial structure building process into two phases can be seen as a contribution to the discussion of syntactic and lexical computation in sentence comprehension (Frazier & Clifton, 1996; MacDonald, Pearlmutter, & Seidenberg, 1994). Additionally, I will provide experimental evidence for the first and second phase in initial structure building. Furthermore, I claim that both processes are not dependent on each other generally. However, some rare cases exist where the second can influence the first phase. This interaction can occur under restricted circumstances only. All assumptions about processing stages are pursued in a manner which separates syntactic and lexical information use as well as the usage of grammar and mental lexicon. This exploration should be taken as a proposal which needs further evaluation by experimental investigations.