A study of the process of listeners' and readers' generation and verification of expectations about spoken and written discourse presented to them examined the possible interactions between surface form and cognitive constraints, to establish baseline measures of the effectiveness of different sentence structures in constraining the production of subsequent utterances. The subjects, 32 native English-speaking university students, were assigned randomly to four groups, each using as stimuli a series of 48 sentences with a different sentence construction type but similar content for each group. The subjects were asked to write a continuation sentence for each stimuli sentence in the style of the original sentence. The continuation sentences were scored according to the entity in the stimulus sentence most strongly linked to the continuation sentence. Results show no evidence that readers automatically assume the first noun phrase to be the current discourse topic, but they do suggest that the choice of topic for an ongoing discourse can be affected considerably by the choice of syntactic construction, with readers relying on cues that mark discourse entities as potential topics. Further research on the relative importance of the variables investigated is recommended. (MSE)