认知与发展心理学研究室知识库

在语篇理解过程巾，整合和预期是两种最为基本的信息加工过程。整合和预期都是基于语境信息的，然而，不同距离的语境信息如何影响整合和预期加工仍然是不清楚的。此外，在不同的实验任务下，读者会采用不同的阅读策略来理解语篇。那么，实验任务又是如何调控不同距离的整合和预期加工呢?本论文采用EEG技术，通过两个研究四个实验来考查语篇理解中不同距离的信息加工以及实验任务的调控作用。

研究一旨在考查在阅读理解任务(实验1)和合理性评价任务(实验2)中，语篇中不同距离的语境信息如何影响语义整合加工。我们通过改变语篇中不同位置的信息产生不同距离的语义违反(短距离、中等距离、长距离)。结果发现，在实验1中，随着整合距离的增大，语义违反诱发的p600效应越来越弱，表明在阅读理解任务中，整合距离越长，语义违反探查加工越困难。在实验2中，随着整合距离的增大，语义违反诱发的早期正波效应越来越强，表明在合理性评价任务中，随着整合距离的增大，语义违反的探查需要付出更多的认知努力。实验1和实验2的结果表明，在不同的实验任务一下，读者以不同的方式来整合信息:阅读理解任务下，短距离整合加工占优势，而在合理性评价任务中，长距离加工则更具有优势。

研究二旨在考查在阅读理解任务(实验3)和合理性评价任务(实验4)中，不同距离的的语境信息如何影响预期加工。我们通过改变语篇中不同位置的内容来操纵语境限制性，从而使得关键词可以被预期到或者不能被预期到，共形成四种实验条件:短距离一低预期，短距离一高预期，长距离一低预期，长距离-高预期。结果显示，在实验3中，与低预期条件相比，不同距离高预期条件下的关键词均使得N400波幅降低，而且不同距离条件下的N400效应没有显著差异，但是在晚期窗口，只有在短距离条件下，低预期比高预期条件诱发了更大的晚期正波，而长距离条件下则没有这种效应。这一结果表明，在阅读理解任务中，尽管不同距离的语境信息都能够快速地促进高预期词汇的语义加工，但是只有短距离语境信息能够支持词汇水平的预期，在信息不符合词汇预期时启动抑制加工。在实验4中，两种距离的高预期条件都比相应的低预期条件诱发了更小的N400波幅，而且长距离一高预期比短距离一高预期条件诱发的N400波幅更小，表明在合理性评价任务中，长距离信息对于高预期词汇语义加工的促进作用大于短距离信息。实验3和实验4的结果表明，预期力}a工同样受到实验任务的调控:阅读理解任务中，短距离的语境信息所起的作用更大，而在合理性评价任务中，长距离信息所起的作用更大。

研究一和研究二的结果表明，不同距离的语境信息在语篇加工中所起的作用并不相同，会受到实验任务的调控。在阅读理解任务中，短距离语境信息要比长距离语境信息具有更大的加工优势，而在合理性评价任务中，长距离语境信息反而比短距离信息更具优势。语境信息在工作记忆中的可及性，和实验任务诱发的自上而下的认知控制状态，共同决定了语篇理解的方式和结果。

Integration and prediction are two kinds of fundamental processes in discourse comprehension, and are based on preceding context. However, how context across different distances influences integration and prediction remains unclear. In addition different tasks lead to differences in the way of comprehension. Then, how task modulates integration and prediction across different distances? Using electroencephalogram (EEG) technique, the present research aimed at investigating the effects of contexts across different distances on integration and prediction, and their modulation by task demand.

Study 1 aimed to investigate how context across different distances influences integration in reading comprehension task (Experiment 1) and acceptability evaluation task (Experiment 2). Semantic violations across short, medium and long distances were constructed by changing the content at different positions in discourses. The results showed that in Experiment 1，P600 effect decreased as distance increased, indicating that in reading comprehension task, the detection of violation became increasingly difficult as distance increased. In Experiment 2, early positivity effect increased as distance increased, suggesting that in acceptability evaluation task, more cognitive efforts were needed to detect the violation across long distance than that across short distances Taken together9 the results in Study 1 indicated that task modulates integration across different distances: integration across short distance takes the priority in reading comprehension, while integration across long distance takes the priority in acceptability evaluation.

Study 2 aimed to investigate how context across different distances influences prediction in reading comprehension task (Experiment 3) and acceptability evaluation task (Experiment 4). The context across short or long distance was changed to make a target word predictable or unpredictable.’Thus9 four conditions were constructed: short-unpredictable，short-predictable, long-unpredictable, long-predictable. The results showed that in Experiment 3} both short-predictable and long-predictable words reduced the N400 relative to unpredictable counterparts respectively, and a larger later positivity was found for unpredictable words compared to predictable ones only in the short condition, but not in the long condition. The results suggested that although both the short and long contexts facilitated semantic processing of predictable words, only the short context could support specific lexical prediction. In Experiment 4, the N400 effect was observed in both short and long conditions, and was larger for the long condition than for the short condition, suggesting that in acceptability evaluation task, the long-distance context 11ad brought more benefits for semantic processing of predictable words than the short-distance context. Over a11, the results in Study 2 suggested that task modulates prediction across different distances: prediction based on short-distance context takes the priority in reading comprehension, while prediction based on long-distance context takes the priority in acceptability evaluation.

Taken together, the present research demonstrated that contexts across different distances have different effects on integration and prediction, which was modulated by task demand at hand. In reading comprehension task, information processing across short distance was taken preferentially. Whereas, in acceptability evaluation task, the context across long distance was considered more, and have stronger effect on information processing relative to the context across short distance. The accessibility of context in working memory and the cognition control state induced by task demand together determined the way of integration and prediction.