中文摘要：

采用移动窗口技术探讨读者在记叙文阅读中是否对人物第一印象进行即时更新。实验1探讨读者在阅读过程中是否会根据第一印象建构人物特征情境模型。实验2探讨已经形成的人物特征情境模型是否会受最近印象的影响得以更新。实验3探讨在明确描述人物特征的情况下，读者是否仍然会更新人物特征情境模型。研究结果表明，记叙文阅读中，读者会对人物第一印象进行即时更新，表现出首因效应向近因效应的转化。

英文摘要：

The primacy effect and the recency effect are important issues in memory research （Wu Yanhong ＆ Zhu Yin, 1999; 1998）. Many studies verified the primacy effect on people＇ s behavior and memory. In the text comprehension, Rapp, Gerrig ＆ Prentice （2001） explored whether the first information would change the comprehension of the trait-based models of characters. Malone （2005） further studied the recent effect which affected the understanding of the character＇s behavior, if the later information denied the first information about the character＇ s trait. The study explored whether the readers maintained the primacy impression when they read the sentences which were inconsistent with the earlier sentences and their reading manifested the primacy effect, or constructed the trait-based situation models of the characters in narrative comprehension and would update the models by later information. The moving window display technique was used in Experiment 1, Experiment 2 and Experiment 3. Experiment 1 tested whether the readers constructed the trait-based models of characters according to the first impression during reading. The experiment adopted the within-subjects design and manipulated the trait consistency. There were three levels of consistency, consistent, inconsistent, and neutral. The independent variable was the reading time of the target sentences. The reading time mode of the target sentences of three versions was inconsistent〉 neutral〉 consistent, p〈 .05. The result showed that readers made use of the primary information to construct the trait-based situation models of the characters. Experiment 2 was to inspect whether the existing trait- based model of character is updated due to the recent effect. These stories were the same as those in Experiment 1 except for having two episodes and two target sentences added to every story. The reading time mode of the target sentences 1 of three versions was the same as in Experiment 1. But there was no difference in target sentences 2, 3, and