An ERP Study On Filler-Gap Processing In English Relative Clauses By Chinese English Learners

Language as a kind of complicated psychological cognitive activities is integrated within milliseconds. ERP technique can describe the time course of successive real-time language processing and remedy the limitations of traditional behavioral studies which can only report synthesized results of language processing by means of reaction time and accuracy. Hence, the ERP technology is applied to the studies on language processing. In the present study, therefore, the ERP technique is selected as the research instrument to overcome the limitations of traditional research methodologies on language processing.Currently, ERP studies on simple sentence processing receive much attention with fruitful results while ERP studies on complex sentence processing are still in progress. Relative clauses as a type of complex sentence structures are of language universality. In this view, exploring relative clause processing is of great significance. To investigate relative clause processing, the core lies in the smooth resolution of filler-gap dependencies. In terms of the first language (L1), the interpretations of filler-gap processing tend to be convergent while those in the second language (L2) tend to be divergent. Thus, the study on filler-gap processing by Chinese English learners is of practical significance.The present study aims to explore the processing mechanism of filler-gap dependencies, utilizing ERP technique to tentatively interpret the real-time parsing process of filler-gap dependencies by Chinese English learners with the violation paradigm. To reach this end, the present experiment has selected 31 subjects among college students in different proficiency levels, namely,16 subjects in high proficiency group (Group H) and 15 ones in lower proficiency group (Group L). Participants in Group H are English major postgraduates having passed TEM-8 and those in Group L are non-English major freshmen or sophomores without passing CET-4. Comparisons are made between the two groups under conditions of non-word gap anomaly, filled gap anomaly and grammatical object to reveal filler-gap processing mechanism with the influence of language proficiency. Through analyzing and discussing the data acquired, the current study concludes some major findings as follows:1. In the early time course, early negativities are detected for both groups across all conditions. The amplitude becomes largest in the anterior region. This seems to reflect the initial syntactic building based on lexical category information, which could be corresponding to the first phase of syntactic parsing model (Friederici,1995,2002). Though they cannot distinguish significantly experimental and control conditions, the early-phase anterior negativities seem to manifest structural priming prior to semantic priming due to the parser's preliminary awareness of the syntactic structure.2. Under the condition of non-word anomaly, compared with the grammatical object condition, widely-distributed P600 is detected in Group H while broadly-distributed N400 is detected in Group L. This appears to indicate their different parsing mechanism. On encountering the non-words ungrammatically filled in the gap, subjects in Group H tend to syntactic parsing prior to semantic parsing and their parser attempts to resolve filler-gap dependencies by repair and reanalysis. However, subjects in Group L are inclined to semantic processing prior to syntactic processing and their parser only detects semantic mismatch with previous sentential context but lacks filler-gap integration. In the framework of syntactic parsing model by Friederici (1995,2002), N400 detected in Group L could be corresponding to the second phase of syntactic processing while P600 detected in Group H could be corresponding to the third phase.3. Under the condition of filled gap anomaly, LAN component is detected in both groups but the amplitude in Group H appears larger and has wider distribution. This seemingly reflects their detection of morphosyntactic disagreement (Friederici,1995,2002). The absence of N400 roughly indicates the successful theta-role assignment of the noun phrase filled in the gap to the embedded verb. Semantic access seems to attenuate the ungrammaticality of filler-gap structure.4. The experimental results tend to partly support Shallow Structure Hypothesis in that subjects in Group L underuse their syntactic knowledge for deeper parsing. Also, the experimental results partly reject this hypothesis in that subjects in Group H seem to process language with deeper syntactic information, which is similar to L1 speakers. The present research seems to support Trace Reactivation Hypothesis but not Direct Association Hypothesis in that the resolution of filler-gap dependencies is mediated by structurally expected gap but not directly by theta-role assignment.