Second Language Processing of Filler-Gap Dependencies: Evidence for the Role of Lexical Access

Sentences that involve direct and indirect object relative clauses contain displaced syntactic constituents (or fillers) that do not appear in their canonical object position following the verb. During real-time comprehension, when the head of such a relative clause is encountered, it has been argued that the human sentence processing mechanism will automatically project a syntactic structure that includes a trace where the displaced element originated as an object of its licensing verb (e.g., Frazier, 1987; Frazier & Fodor, 1978). The referent of the filler, after it is accessed in the lexicon, will be temporarily stored in working memory, until the moment at which this trace (or gap) is encountered later during sentence processing and at which the filler can thus be retrieved from memory and integrated into the structure (e.g., Bever & McElree, 1988; Love & Swinney, 1996; Nicol & Swinney, 1989). Thus, processing a filler-gap dependency involves not only the computation of a movement chain but also the activation (i.e., initial retrieval from the lexicon) and maintenance (e.g., storage in working memory) of the referent of the filler. Furthermore, there is a crucial interaction between the two systems: Encountering the trace of a moved wh-expression triggers reactivation of the referent. Empirical research on the processing of filler-gap dependencies in a native language (L1) has revealed response patterns that are highly suggestive of mental reactivation of the filler at the structurally defined gap site from which it originated (e.g., Gibson & Warren, 2004; Love, 2007; Love & Swinney, 1996; Roberts, Marinis, Felser, & Clahsen, 2007). Among native speakers (NSs), evidence for this trace-induced reactivation is observed in longer reading times on segments that include a syntactic gap (e.g., Gibson & Warren, 2004) or in faster decision times in categorizing, at crucial moments, visually presented probes (images or words) that are related to the filler (e.g., Love, 2007). The question of whether the same sentence processing mechanisms are available in adult second language (L2) acquisition has received much recent attention. A substantial body of research has shown that many of the same sources of information that NSs are able to deploy during online comprehension are also used in L2 sentence processing (e.g., Dekydtspotter, Donaldson, Edmonds, Liljestrand-Fultz, & Petrush, 2008; Frenck-Mestre, 2005; Frenck-Mestre & Pynte, 1997; Juffs, 1998, 2006; Juffs & Harrington, 1995; Omaki & Schulz, 2011; Williams, Möbius, & Kim, 2001). Although learner results may differ from NS results on some tasks, this does not necessarily point to more fundamental underlying differences between L1 and L2 processing (Dekydtspotter, Schwartz, & Sprouse, 2006). However, many others have argued that, in contrast to NSs—who show a clear syntactic reflex during processing—(even advanced) L2 learners seem to rely more heavily on nonsyntactic information (Clahsen & Felser, 2006a, 2006b; Felser & Roberts, 2007; Felser, Roberts, Marinis, & Gross, 2003; Marinis, Roberts, Felser & Clahsen, 2005). Clahsen and Felser’s (2006a, 2006b) shallow structure hypothesis maintains that— due to the incomplete and imperfect nature of the L2 grammar, which cannot be called upon during online processing— L2 learners tend to overrely on semantic cues, contextual information, and real-world knowledge to assign meaning to linguistic input processed in real time. For proponents of the shallow structure hypothesis, impaired grammatical representations are thus the main source of the supposed L1-L2 processing differences. This model fails to consider another plausible source of differing L1-L2 response patterns in experimental research—namely, the possible effects of underroutinized L2 lexical retrieval, which has been shown to be lacking in automaticity