Online Study Of Sensitivity To Argumentative Discourse-Level Anomalies Reading And Its Influencing Factors

There are three most influential theories in information processing of text-comprehension:the constructionist theory, the minimalism hypothesis and the memory-based processing view. The divergences among the three theories occur in the inference type to choose, the text representation level to reach the coherence and the mechanism to get the global coherence. The present study is to explore whether the Chinese English learners in three different levels generate early sensitivity or delayed sensitivity to discourse-level anomalies in order to test the applicability of the three theories in the mechanism to get global coherence. In addition, the present research also explores whether the language proficiency level and topic familiarity are influential factors on sensitivity to discourse-level anomalies.This research is an on-line study to explore the mechanism of the discourse-level anomalies which took the chunk-by-chunk self-paced reading method and uses E-prime software to record the three critical times. Three groups of participants (30high level readers,30middle level readers and30low level readers) were recruited to take part in the experiment. The materials of the research included the experiment version (inconsistent version) and the control version (consistent version). The data were processed by (SPSS)13.0to make the repeated measures AVOVAs. And then the main effect was tested by the Post Hoc Tests and the interactive effect was tested by the simple effect analysis. The Pearson Correlation Analysis was taken to investigate whether there was correlation between the topic familiarity and the reading times.The final data analysis comes to the following conclusions:First, early sensitivity occurs in all the three sessions when reading the discourse-level anomalies which proves the common hypothesis of the here-and-now account and the memory-based processing theory on the the mechanism to get global coherence which claims that readers will generate instant sensitivity with the local coherence break in order to get global coherence. Second, both early sensitivity and delayed sensitivity occur in high level group. This may be because high language proficiency level readers have larger cognitive capacity and after the anomalies reading they still tend to prove their sensitivity of the anomalies. Third, as for the post-target sentence region of the anomalous version, there is no reading time penalty in the high level and middle level readers, which proves the here-and-now account. A reading time penalty occurs in the low level readers testifying the memory-based processing theory. The high level readers take positive and strategic method in discourse reading. If they could establish meaningful representation positively without inference, they tend not to make inferences. The low level readers takes a passive resonance processing that they use the continuous and autonomous resonance no matter whether inferences are essential. Fourth, early sensitivity occurs in all the three sessions, therefore, the language proficiency has no relation with the early sensitivity. Different language proficiency has influenced the reading times of the anomalous version in some degree which indicates that language proficiency level influences the discourse-level anomalies reading process. Five, the topic familiarity has significant correlation on three reading times during the normal passage reading. While in the experimental version, there is significant correlation between the topic familiarity and the reading time only in the post-target sentence region which may be because during the discourse-level anomalies inferential process, readers cost more cognitive capacity to process the context coherence, neglecting the combination of the background knowledge.