基于语境的预期推理构建的实验研究

两个实验用移动视窗技术,用目标句阅读时间和命名探测词两种指标来考察当文本中有两个情境模型时,读者是否会即时进行预期推理。结果表明,如果在支持性语境与引发推理的信息之间插入一个与支持性语境所描述的情境模型不同的,且与主人公目标有关的因果链上的情境模型的信息,则不管这个插入的情境模型在文本中的位置如何,也不管干扰信息的干扰水平高低,都会降低因果性预期推理在阅读中产生的可能性。

The Effect of Discourse Information on the Generation of Causal Predictive Inferences

Two experiments were designed to explore whether causal predictive inferences were generated on-line when there were two situation models in a discourse. The results of experiment 1 showed that when except the high-support context, there was another situation model, which was also about the protagonist's goal but could not be integrated into the supportive context, embedded between the context and inference-evoking point, predictive inferences were not likely generated on-line in this case. Experiment 1 also showed that predictive inferring did not occur, despite the location of distributive situation in a text. Experiment 2 showed that the contributing degree of the contributing infor mation did not affect the generation of causal predictive inferences. Maybe it was because that when two sepa rate situation models stored in the long-term memory, readers should activate both of them and access which was necessary for coming text. So it was more difficult for readers to predict what would be coming next.