Generation Of Causal Predictive Inferences In Discourse Comprehension

The field of discourse processing has grown tremendously during the past two decades. The currency of psychology rises to the extent that disourse psychologist can improve reading, text design, complex learning, and social interaction. The focus of this article is on the causal predictive inferences, which is an important process during reading. It is proposed that the standards for coherence that a reader adopt and textual constraints together determine whether causal predictive inferences are likely to be generated, and whether the inferences will be integrated into the long-term memory representation of the text. There are four experiments, 11 sub-experiments in the study.The aim of Experiment 1 is to explore whether causal predictive inferences are generated on-line. The effect of the support from contextual features contained within the situation model is studied. The reading time for a target sentence that including predictive information is measured. It is found that the mean reading time for the target sentence in the high-support context condition is faster than in the low-support context condition. It suggests that predictive inferences can be generated on the basis of contextual features contained within the situation model and with the foregrouding devices, no matter how long the text distance between the context and the inference-evoking point.Experiment 2 is designed to explore whether causal predictive inferences are generated on-line when there is another situation model. The results of experiment 2a and 2b show that when there is another situation model, which is also about the protagonist's goal but can not be integrated into the supportive context, embedded between the context and inference-evoking point, predictive inferences are not likely generated on-line in this case. Experiment 2c and 2d are conducted to examine whether the order of thetwo-situation model in long-term memory would make the availability of them different. In Experiment 2c, a target sentence as in Experiment 1 is used to test whether the high-support context being before or after the distributive situation would lead to different possibility of predictive information generated on-line. The results show that predictive inferring does not occur in both of cases. In Experiment 2c, a new task, word recognizing, is introduced, which is presumed to hit directly the availability processes of contextual information. No difference between the two conditions is found. Thus, all results in Experiment 2 show that predictive inferences are not drawn on-line if there is a distributive situation, despite its location in the text. Mqybe it iw because that when two separate situation models stored in the long-term memory, readers should active both of them and access which is necessary for coming text. So it is more difficult for readers to predictive what is coming next.Experiment 3 is conducted to study whether causal predictive inferences are generated on-line when improving the importance of high-support context in the case that there are two situation models in the long-term memory. The results show that if the embedded situation model does not bear on the source of events in the story, it has no passive effect on the generation of predictive inferences. When giving the cues that predictive outcome is possibly important, or elaborating the high-support context, predictive inferences will be encoded on-line, even when there is a distributive situation. So it is true that a reader expects that subsequent event will be somehow causal coherence with the current event.The persistent of predictive inferences is examined in Experiment 4, including 3 sub-experiment. Most of the discourse psychologists assumed that forward inferring occurs during reading but that predictive inferences are quickly deactivated. In this study, it is found that if readers aware that predictive information is necessary to achieve coherence, they can maintain them. Three methods are used to improve the importance of predictive information, further sup...