This article provides a selective review of recent sentence processing studies that have investigated non-native speakers’ real-time comprehension of different types of discontinuous dependency in English as a second language (L2). Taken together, the results from these studies reveal a fairly systematic pattern, with L2 learners tending to show native-like processing performance for ’forwards-looking’ dependencies but not for ’backwards-looking’ ones. Forwards-looking dependencies to be considered here include different kinds of wh-dependency, and backwards-looking ones include reflexive binding, anaphoric pronouns, and antecedent-contained deletions. I will explore possible reasons for the observed differences between forwards and backwards-looking syntactic dependencies in L2 processing by considering the nature of the different types of search process involved, and the necessity of accessing or computing detailed syntactic representations for successful dependency formation.

本研究は, 日本語と韓国語との文字・語彙間距離に着目し, 韓国人日本語学習者が韓国語音韻から日本語音韻を推測する様子を調べたものである.具体的には, 学習者が韓国語の漢字音から日本語の漢字音を推測する際に用いるストラテジーの成功理由と失敗原因を明らかにすることを目的とし, 学習者が或る韓国語の漢字音に対応するいくつかの日本語の漢字音の中で１つの日本語の漢字音を選択する際に, 何を主な基準とし何を優先するかを調べた．その結果, 日本語能力試験３級以下の漢字のように日本語学習の早い段階で習った易しい漢字の音読みのうち, 韓国語の漢字音との心理的類似度が最も高い漢字音を日韓漢字音の対応規則の根拠とする傾向が見られた．そのような根拠が３級以下の易しい漢字の音読みの中から探れない状況では, 韓国語の漢字音に類似した音を根拠なく当てはめることが多くなるとわかった．

This study investigated Thai learners' abilities to detect the ungrammaticality of English sentences with null arguments. Thai, particularly in informal contexts, differs from English in that Thai allows null arguments in subject and object positions. A Grammaticality Judgment (GJ) task was employed and Thai participants were to identify ungrammatical sentences in which either subjects or objects were missing and make corrections. Based on previous L2 findings and L1 facts, we predicted that (1) null subjects would be identified and corrected at a higher rate than null objects, and (2) null subjects/objects in single clauses would be identified and corrected at a higher rate than null subjects/objects in embedded clauses. A grammaticality judgment task (GJ) was conducted with 100 Thai high school and college students. They were divided into four proficiency levels by means of an English placement test. Results from Repeated Measures ANOVAs confirmed both predictions. The results support the role of input consistency and the role of L1 null arguments in the learning of English non-null arguments. In respect of proficiency, post-hoc tests revealed significant differences between groups. This study suggests that L2 learners associate elements allowing null objects in the L1 with L2 null objects and that sensitivity to non-null arguments increases with higher proficiency in L2 acquisition.