A Study Of Text-based Variable Of On-line Predictive Inference Generation Among Chinese EFL Learners

The dissertation presents a study of text-based variable of on-line predictive inference generation among Chinese EFL learners. Researches have become increasingly aware of the critical role of predictive inference in reading. Both Constructionism and Minimalist hypotheses have approved the possibility of on-line generation of predictive inference. A considerable amount of research has been conducted during the past few decades investigating the inference-making processes. However, surprisingly little is known about predictive inferences in L2 reading process. Most of the inference models and researches have been limited in L1 reading field, especially English as the native language. The current investigation was designed to fill this gap by considering Chinese EFL readers'on-line processes during predictive inference generation.The process of generating predictive inferences is constrained by numerous factors. These factors may be divided into two major categories: text-based variable and reader-based variable. Text-based variable, without doubt, is one of the most important factors. This study, based on previous researches, will dig deeper the effect of context's support on the generation of predictive inference. The author proposes that, if a context's situation is active and high-supportive, it will be possible and easier for reader to generate predictive inference on-line; otherwise, the probability will be weaker.In order to test the hypothesis, one experiment was designed. This experiment was implemented among fourth-year English major college students at Ocean University of China. The method of time-measuring was adopted. With controlled context information of situation, this research tends to specialize its consequence towards predictive inference, and probe into the on-line generation of predictive inference. The experimental materials are divided into high and low supportive situations, varying only in several words. During reading process, while consciously maintaining the integrative coherence of the text, readers will use causal consequence to comprehend the whole text. In this case, if the situation of context is highly supportive to the consequence, readers will generate predictive inference during reading process, but if there is too much difficulty ahead, readers may give up. The results of this investigation support the contention that context's support to causal consequence is an important factor during on-line generation of predictive inference.This thesis consists of five parts. Starting with an introduction, it provides the research orientation, objective and the outline of this study. Chapter One, theoretically and empirically, reviews relevant literature. Minimalist hypotheses and Constructionist theory have also been expatiated in this chapter. Related with the models, the author focused on the review and discussion of two topics: whether predictive inference can be generated or not during reading process, and what are those factors that can affect the generation of inference. Chapter Two introduces an experiment, exploring the effects of high and low context situation on the generation of predictive inference in L2 reading process. Detailed descriptions of methods and procedures for the experiment are presented in this chapter. Next, chapter Three provides the data analysis and a general discussion of all results about this experiment. A summarization of the major findings, implications of the study, and suggestions for future research are explicated in the last chapter.