- 認知科学 (ISSN:13417924)
- vol.19, no.1, pp.100-121, 2012 (Released:2013-12-27)
In the area of educational psychology, much research has suggested that habitual reading of texts improves people’s vocabulary (e.g., Stanovich, Cunningham, & West, 1998). However, there are few cognitive approaches addressing why reading texts changes readers’ vocabulary. Our study examined this issue by word association task and simulations through Latent Semantic Analysis (LSA; Landauer & Dumais, 1997). One thousand, one hundred and eighty-nine Japanese adults participated in our research through an internet website. Participants answered two questions about their reading habits of newspapers and novels, and then performed the word association task. We constructed three LSA semantic spaces drawn from corpora of newspapers, novels, and a corpus that includes both newspapers and novels. Participants were divided into four groups according to their reading habits-newspaper and novel readers, newspaper and novelonly readers, and non-readers. In each group, from each stimulus and associated word pair that participants generated, an association strength value, a newspaper-based LSA similarity value, a novel-based LSA similarity value, and newspaper/novel-based LSA similarity value were derived. We conducted two analyses: one between stimulus words and all associated words, and another between stimulus words and most frequently associated words. In both analyses, association strength value in the novelonly reading group was predicted best by the novel-corpora-constructed LSA. In addition, especially in the latter analysis, association strength value in the group who often read both newspapers and novels was predicted best by the LSA constructed from the newspaper/novel corpora. This data suggests that one reason why reading text affects readers’ vocabulary is that readers acquire knowledge by usage-based learning processes, such as LSA.