| This dissertation concerns the processing of syntactic ambiguities. The serial, parallel and delay parsing models are described, and predictions about the presence and location of increased processing loads are made for each model. Data were collected using timed continuous syntactic and lexical decisions. Experimental materials contained two types of structural ambiguity: main verb/passive participle ambiguities (garden path sentences) or direct object/complement clause ambiguities. In the direct object/complement clause materials, length of the ambiguity was manipulated, as well as animacy and plausibility of the noun within the ambiguous region. Evidence indicated that subjects parsed the main verb/participle ambiguities serially, and appeared to delay processing of the direct object/complement clause ambiguities for several words. If the sentence was not disambiguated within about three words, then a preferred analysis was adopted, and the pattern at the disambiguating word was consistent with the serial model. Neither animacy nor plausibility of the nouns affected parsing strategy, as predicted by the modularity thesis. Verbs were tested for structural preference with a sentence completion task, and then sorted. Although a garden path effect at disambiguation appeared to be stronger among the transitively-biased verbs than among those with a complement-bias, this apparent difference was not statistically significant. In a post-hoc test, reading fluency was also found to affect parsing strategy. A case for the parser's flexibility is made. |
