A Fundamental Study on English L2 Learners' Collocational Knowledge

本研究的目的是利用语料库、反应时和近红外光谱（NIRS）三种数据分析英语学习者和英语本族语者在搭配知识方面的差异。我们收集了日本英语学习者和英语母语者的英语作文，并获得了版权许可：每组200篇。在此基础上，我们编制了名古屋中介语英语语料库（NICE）。NICE包括两个子语料库：NICE-NNS（non-native speaker data）和NICE-NS（native speaker data）。此外，我们还通过n-gram程序提取了搭配表达式，并创建了用于以下实验的表达式列表。2007年，我们分析了非母语数据（NICE-NNS）和母语数据（NICE-NS）之间显著差异的语言学特征。从随后的判别分析中，我们发现了明显的差异。我们的反应时分析显示，英语学习者确实有一定的搭配知识，但这些知识的数量在不同的水平之间存在差异。近红外光谱实验结果表明，本族语者只使用大脑中的语言区来处理搭配表达，认知加工负荷最小。学习者，然而，使用各种领域以外的语言领域，没有特定的，共同的神经激活模式。我们的学习者语料库，NICE，目前向公众开放，并在互联网上提供。我们的最后报告也以PDF文件的形式在互联网上提供。

The purpose of this research was to analyze the differences in collocational knowledge between EFL learners and native English speakers, utilizing three kinds of data: corpora, reaction time, and near-infrared spectroscopy(NIRS).In 2005, we designed and compiled a new learner corpus. We collected English essays, with copyright permission, written by both Japanese EFL learners and native English speakers: 200 essays were collected from each group. Based on the essay data, we compiled the Nagoya Interlanguage Corpus of English (NICE). The NICE consists of two subcorpora: NICE-NNS (non-native speaker data) and NICE-NS (native speaker data).In 2006, we added native English speakers' corrections to the learners' essay data in order to carry out error analysis. In addition, we picked up collocational expressions by means of an n-gram program, and created the list of expressions used for the following experiments.In 2007, we analyzed the linguistic features of distinctive differences between non-native speaker data (NICE-NNS) and native speaker data (NICE-NS). From the subsequent discriminant analysis, we found distinctive differences. Our reaction time analysis disclosed that the EFL learners did have some collocational knowledge, but that the amount of such knowledge varied among the proficiency levels. Our NIRS experiment revealed that the native speakers used only the linguistic area in the brain to process the collocational expressions, with minimum cognitive processing load. The learners, however, used various areas other than the linguistic area, with no particular, common neural activation patterns.Our learner's corpus, NICE, is currently open to the public, and available on the Internet. Our final report is also available on the Internet in the form of PDF files.