Text comprehension : The influence of temporal information on processing and reading rate

The research described in this thesis investigated some aspects of the process of text comprehension. A variety of experimental techniques were used including readers' decisions on paragraphing continuous prose, a self paced reading paradigm where a computer timed the speed of reading each sentence in a text, the time taken by readers to answer questions about a previously read text, and the nature of readers' continuations of short unfinished narratives. From the results of the initial experiments on paragraphing continuous prose one feature emerged as an important predictor of the points in a text where readers' chose to begin a new paragraph. This was the presence in a sentence of an explicit statement of temporal information, such as "In 1961..." or "Eighteen years later...". Subsequent experiments investigated the role of such statements in text processing. It was thought that the time statements were used as cues that a new temporal setting had begun in the text and that the preceding information in the text might be perceived by the reader as less relevant following these cues. Reading time and question answering time experiments were designed to test if such changes were reflected in the time taken to read time change statements and to process the following sentences and questions. The results of these experiments were inconclusive. The literature on text processing was studied and several potentially relevant concepts were incorporated into subsequent experiments. The concept of the 'script' or 'scenario' devised by Minsky (1975) Schank and Abelson (1977) and Sanford and Garrod (1980) seemed particularly pertinent. In the later experiments on the role of temporal information, it was hypothesized that one element of the readers' knowledge unit or scenario concerning particular stereotyped events, was the normal time course for these events. If a time change in a text exceeded this predicted range, then the readers would interpret the standard events as being completed, and the current scenario and its associated information would be considered as no longer relevant-this latter state being termed "out of focus" following, Grosz (1977). The results of the experiments in which readers continued unfinished narratives supported this hypthesis. Following a time change which was beyond the predicted range for the particular events described in the text, the characters in the text which were bound to the scenario were less liable to feature in continuations of the text. These experiments also highlighted more general features of status or relevance in a text. The characters who were not dependent on the scenario were always more liable to be mentioned in the continuations and were also the only characters liable to be referred to by pronouns. This privileged status in the text was termed foregrounding, after Chafe (1972). Reading time experiments were conducted to investigate the effect of time change statements during the comprehension process. The results of these experiments were similar to the continuation studies. Following beyond-range time changes readers took longer to retrieve information concerning the scenario dependent characters. If these scenario dependent characters were not highly predictable from the scenario title, then resolving pronoun references to them, also took longer following such time changes. The foregrounded status of the scenario-independent characters was confirmed as these characters were readily accessible for reference resolution and information retrieval in all conditions. A scenario-based model of comprehension is proposed following these results, with time range information as a useful test of the relevance of the current scenario. The results also highlight general features of comprehension such as the status of entities in the text and how the readers apportion processing capacity accordingly. The role of time change information is suggested as one device which readers use in more general processing strategies