Syntactic processing in second language acquisition of English

Abstract

The present study investigates relative clause (RC) attachment preferences of end-state Turkish L2 speakers of English in their L2 and compares their sentence processing strategies to those of monolingual Turkish speakers and English native speakers in sentences including a complex genitive noun phrase (NP) modified by an RC such as the following: Someone shot [NPnon-local the servant] of [NPlocal the actress] [RC who was on the balcony]. Although such constructions have been investigated in English in several previous studies, data from Turkish, a language with a relatively free SOV word order, can shed light on the issue in the exploration of cross-linguistic variation in RC attachment preferences. In Turkish, possession is realized through genitive-possessive constructions that are similar to the Saxon genitive in English. However, unlike English, it does not have an NP-PP-NP construction. [RC Balkon-da dur-an] [NPlocal aktis-in] [NPnon-local hizmetçi-si]-ni Balcony-LOC stand-SubjPART actress-GEN servant-3SG.POSS-ACC vur-du-lar. shoot-PAST-3PL They shot the actress’s servant who was standing on the balcony. They shot the servant of the actress who was standing on the balcony. Investigating RC attachment preferences of end-state Turkish learners of L2 English can also be revealing in the sense that it allows to test the role of first language (L1) in sentence processing strategies as well as availability of Universal Parser (UP)/Universal Grammar (UG) in L2 acquisition. A set of online and offline sentence processing tasks was given to the participants. The results suggest that both native speakers of English and native speakers of Turkish prefer to attach the RC to the local NP in their respective languages while reading online and offline. The data confirm that both the L1 Turkish and L1 English groups go through serial processing while comprehending sentences consisting of RCs with complex genitive antecedents. They make an initial commitment to a specific attachment site, and they reanalyze their initial interpretations when the initial analysis proves to be incorrect. However, this initial preference is not solely determined by syntactical factors. The results suggest that native Turkish and English speakers use lexical information such as the animacy of the noun as well as the syntax in their initial judgments. Turkish learners of L2 English, on the other hand, do not show a clear attachment preference for either local or non-local NP in the absence of the lexical information while reading online and they seem to use the lexical information differently from the L1 speakers of both languages. The results of the L1 data seem to confirm the predictions of the Unrestricted Race Model proposed by Van Gompel, Pickering, and Traxler (2000; 2001). The model suggests that the human parser makes and commits itself to a single analysis at a time. However, this analysis is not necessarily determined by syntax alone. The results of the study are compatible with the Unrestricted Race Model as L1 Turkish and L1 English v groups used the lexical information (i.e., animacy of the noun) during their initial analysis. Furthermore, there is evidence that the participants made a reanalysis when the first analysis did not prove to be correct. The results of the L2 data are consistent with the Shallow Structure Hypothesis (SSH) of Clahsen and Felser (2006a, 2006b). The SSH suggests that unlike child L2 learners and adult native speakers, adult L2 speakers do not make syntactically detailed computations. Instead, they rely on the lexical-semantic sources of information while making analysis. The findings of this study confirm the prediction as the L2 speakers relied heavily on the lexical information in their analysis. The animacy information carried by the NP guided RC attachment preferences of L2 speakers. It is suggested that L2 speakers’ reliance on lexical information results from a deficit in their implicit grammar, which seems to suggest that the UP is not available to L2 speakers as they cannot access UG (Clahsen & Felser, 2006b). For several reasons, some researchers suggested that the investigation of RC attachment ambiguities do not address the universality of parsing principles (Carreiras & Clifton, 1993; Cuetos, Mitchell & Corly, 1996; Frazier & Clifton, 1996; Gibson, Pearlmutter, Cansseco-Gonzalez, and Hickok, 1996; Gilboy, Sopena, Clifton, & Frazier, 1995; Hemforth, Konieczny, Scheepers, and Strube, 1998; Pearlmutter & Gibson, 2001). Therefore, the results of this study should be interpreted considering this argument. Nevertheless, findings can still be revealing to understand cross-linguistic similarities and differences in RC ambiguity resolution.