Processing filler-gap dependencies in an L2: An Event-Related Potential study

The present study investigated second language (L2) learners’ processing of filler-gap dependencies and tested the implications of Clahsen and Felser’s (2006) shallow structure hypothesis according to which L2 learners underuse syntactic information during the processing of these sentence types. Advanced L2 learners of English (native Swedish speakers) listened to English sentences with object-relative, subject-relative and finite complement clauses. Event-related potentials were recorded at the word that, the embedded verb and the following word: a prepositional phrase, a semantically congruent or incongruent noun phrase or an extra noun phrase ungrammatically occupying the canonical object position after the verb in object-relative clauses. Compared to correct sentences, the extra noun phrase elicited a late positivity with a central-posterior distribution (P600). No increased negativity indicating problems with semantic processing (N400) was observed, suggesting that the extra noun phrase was initially interpreted as the object of the verb, and the integration of the clause-initial noun phrase (the filler) with the verb was delayed. The syntactic violation involved in this interpretation led to syntactic repair processes reflected in the P600 effect. The relative pronoun that in object-relative clauses gave rise to an enhanced negativity over anterior and central sites probably indexing storage of the filler in working memory. Finally, semantically anomalous nouns yielded a P600 but no N400 effect in grammatically correct complement clauses, which was interpreted to reflect the conflict between the relations dictated by the syntactic structure and the alternative thematic relationships that might be established to make sense of these sentences. The observation that participants relied on both syntactic and semantic information during sentence comprehension is not consistent with Clahsen and Felser’s proposal. The results of this study suggest that learners might differ from native speakers in terms of processing speed, which potentially affects the relative order and nature of interaction of different subprocesses involved in sentence comprehension