The Role of Individual Differences in Working Memory Capacity in Processing of Filler-gap Dependency Structures by ESL Learners

One of the central questions in recent second language (L2) processing research is whether the types of parsing heuristics and linguistic resources learners compute during online processing are qualitatively similar or different from those used by native speakers of the target language. In this line of research, the shallow structure hypothesis (SSH) proposed by Clahsen and Felser (2006a, 2006b) suggests that the nature of L2 processing is fundamentally and qualitatively different from L1 processing. According to the SSH, the type of syntactic representations adult learners utilize during online processing are structurally shallower and hierarchically less detailed, compared to those used by native speakers for two possible reasons: inadequate interlanguage grammatical representations and/or insufficient processing abilities to make use of relevant information in real time. The SSH further states that this is, by and large, the likely case regardless of learners’ proficiency or working memory capacities (WMC). This study sought to test the validity of these claims by investigating how proficient ESL learners process relative clause (RC) island structures while reading sentences containing filler-gap dependency constructions in real time. A total of 49 advanced ESL learners varying of age of arrival (AOA)—28 adult (AOA between 18-31) and 21 early learners (AOA between 2-9)— as well as 24 native English speaker controls, participated in an eye-tracking reading experiment and two different types of computerized non-verbal working memory span tests (operation-span and symmetry-span) adapted from Oswald, McAbee, Redick, & Hambrick, 2015). Developed based on the materials used in Omaki and Schulz (2011), each target sentence in the reading task had four experimental conditions in a 2 x 2 Latin square design, with plausibility (i.e., plausible and implausible) and island (non-island and island) manipulation, but participants received only one of those four conditions for each sentence. Results suggested that while all participants made use of active filler strategies to fill the gap as early as possible in the non-island environment, both the English control and ESL groups appeared to have rapidly deployed relevant syntactic information of island constraints regardless of their WMC, avoiding illicit filler-gap formation inside the relative clause islands. Some WM effects were found for adult the ESL learners only in the later region at the ultimate gap (i.e., canonical) where the parser is expected to perform a filler-gap reanalysis. At this region, the early ESL learners and native English speaker controls showed sensitivity to structural cues and gap identifications fairly early, initiating filler-gap reanalysis processes from earlier stages of processing. However, it was the only those with higher WMC among the adult ESL learners that presented such immediate filler-gap reanalysis effects during early stages of processing. In contrast, the adult learners with lower WMC showed sensitivity (i.e., reanalysis effect) only during the later stages of processing. These findings are further discussed in the light of the SSH and second language development.