Grammatical constraints in second language sentence processing

A central issue in L2 research concerns the nature of grammatical representations that late L2 learners come to develop in the L2. Previous work suggests that L2 learners sometimes underuse morpho-syntactic information during online processing of L2 sentences, leading to an ongoing debate about how they represent and process structural information in sentence processing. Some researchers propose that L2 sentence processing is qualitatively different from L1 sentence processing in that the former characteristically involves ‘shallow’ structural analysis (Clahsen & Felser, 2006a), whereas other researchers suggest that the differences between L1 and L2 processing are attributable to quantitative factors such as the amount of language experience, the proficiency level in the target language or the availability of processing resources (Frenck-Mestre, 2002; Hopp, 2006; McDonald, 2006). The present dissertation seeks to further our understanding of adult L2 syntactic processing by examining L2 learners’ sensitivity to ‘island constraints’ (Ross, 1976) in the course of online processing of long-distance wh-dependencies, using plausibility judgments and eye-movement monitoring techniques. In Experiment 1, a stop-making-sense task was conducted to investigate L2 learners’ sensitivity to the subject/relative clause island constraint in online plausibility judgments. The native speakers showed immediate sensitivity to island constraints, as evidenced by the fact that although they interpreted a wh-dependency at the earliest possible gap site when it is grammatically licit, they suspended the immediate gap postulation within a syntactic island. The L2 learners were not as efficient as native speakers in suppressing active gap search, but they ultimately ruled out an illegal dependency, in accordance with the island constraint. Experiment 2 employed eye-movement monitoring techniques to examine the way native speakers and L2 learners apply the subject/relative clause island constraint when processing filler-gap dependencies under a more natural reading situation. Working memory capacity of the participants was also measured in an attempt to capture potential individual differences in filler-gap processing and grammar application. The results indicate that even native speakers failed to immediately suppress automatic active gap creation inside an island in this natural reading situation, and both groups applied island constraints at a later stage. There was also suggestive evidence that readers with larger working memory capacity applied island constraints earlier than those with smaller working memory capacity, in both native speakers and L2 learners, suggesting that more processing resources may allow a more rapid and efficient application of grammatical constraints. Experiment 3 investigated whether L2 learners are sensitive to a more subtle grammatical distinction--differential distribution of parasitic gaps within two kinds of extraction islands (i.e., subjects with an infinitival complement vs. subjects with a finite relative clause modifier). The native speakers’ eye movement patterns showed evidence for a rapid distinction between the two types of islands, indicating active gap search only in the island that allows a parasitic gap. Some of the L2 learners showed a similar pattern of sensitivity to this subtle grammatical constraint, though the effect appeared in a later region as compared to the native speakers. The three experiments showed additional quantitative differences between the native speakers and L2 learners, including differences in reanalyzing the incorrect initial wh-dependency and in speed of constructing a complex phrase. Taken together, these findings suggest that L2 learners are sensitive to an array of important linguistic and cognitive factors in ways that are qualitatively comparable to native speakers in online sentence processing, and their comprehension is based on detailed, grammatically-correct syntactic parse of L2 sentences. The L1-L2 differences in sentence processing can be largely attributed to quantitative factors such as processing speech/efficiency or working memory capacity differences. The findings are discussed in terms of the theoretical debate on the nature of the L2 sentence processing mechanisms