A Study on the Development of Grammatical Competence of Japanese EFL Learners : The Application of Item Response Theory and Differential Item Functioning

虽然在过去的几十年里，第二语言习得的理论有了很大的发展，但许多研究都集中在单一的语法特征上，并且假设特征之间的相互关系是由给定的语言理论所赋予的。作为二语习得中最常用的测量工具，语法判断任务的可靠性是其缺点之一。由于从不同研究中获得的数据是样本和测试相关的，因此很难比较研究结果。为了弥补这一不足，我们采用IRT（项目反应理论）将不同语法判断测试的数据等同起来；由于IRT模型的参数为等分提供了理论依据，因此同一量表上不同语法项的习得比较具有较高的信度。大约1200名在日本学习英语的成年日语母语人士参与了我们的研究。这些参与者进行了语法判断测试，以检查各种英语语法特征的习得情况，如非否定动词、非宾格动词、心理动词、关系从句结构、wh-疑问句结构、to不定式和加格变换。在此基础上，我们讨论了在二语习得研究中使用IRT的适宜性和益处，并探讨了这些语法特征在参与者习得过程中的相互关系。

Although the last couple of decades have seen great theoretical developments in SLA (Second Language Acquisition), many studies have focused on single grammatical features, and have assumed interrelationships among features are granted by a given linguistic theory. One of the weaknesses of grammaticality judgment task, the most common measurement tool in SLA, is its reliability. Since the data obtained from different studies are sample-and test-dependent, it is difficult to compare the findings. In order to compensate for this weakness, we employed IRT (Item Response Theory) to equate the data from different grammaticality judgment tests; since the parameters of IRT models provide a theoretical justification for equating scores, comparison with great reliability is possible of acquisition among different grammatical items on the same scale. Approximately 1,200 adult native speakers of Japanese learning EFL in Japan participated in our study. These participants were given grammaticality judgment tests in order to examine the acquisition of various English grammatical features such as unergative verbs, unaccusative verbs, psych verbs, relative clause constructions, wh-question constructions, to infinitives, and dative alternation. On the basis of our findings, we discussed the appropriateness and benefits of using IRT in SLA research, and explored interrelationships among these grammatical features in participants' acquisition process.