통계적 학습과 언어: 영어 관계절과 수일치 구문 중심으로

Abstract

This study investigates whether sensitivity to nonadjacent and adjacent dependency rules in an artificial grammar learning paradigm (i.e., statistical learning ability) predicts L2 learners’ performance in comprehending two different types of English structures. A total of 110 L1-Korean participants completed the four tasks: two artificial grammar learning tasks, involving nonadjacent and adjacent dependency rules, and two self-paced English reading tasks, involving English relative clauses and adjacent number agreement in noun phrases. The results show that their statistical learning scores predicted the L2 learners’ performance only in processing object relative clauses: the ability to discover the nonadjacent pattern in artificial grammar learning is correlated with comprehension accuracy on object relative clauses, but not on adjacent number agreement. The findings of this study indicate that the statistical learning ability to discover a particular pattern (a nonadjacent dependency) can be considered a predictor in comprehending long-distance dependency patterns in the L2. These also illuminate the role of implicit learning processes of language learning, which will have pedagogical implications for second language instruction as well.