Retour vers liste

Détail de la contribution

Auteur: Claudia FELSER

Binding constraints in non-native language processing

Abstract/Résumé: REVISED TITLE: Binding constraints in non-native language processing I will report the results from a series of eye-movement monitoring experiments that investigated the timing of binding constraints during native and non-native reading comprehension. Within the 'primitives of binding' (POB) framework developed by Reuland and colleagues (e.g. Reinhart & Reuland, 1993; Reuland, 2011), binding condition A, which constrains the interpretation of reflexives, is thought to be a reflex of syntactic structure building and as such is assumed to be the easiest type of antecedent-pronoun dependency to compute. Non-reflexive pronouns, on the other hand, are always technically ambiguous and need not be syntactically bound. They are constrained by binding condition B, the application of which requires a comparison of alternative semantic representations within the POB framework (compare Grodzinsky & Reinhart's, 1993, "Rule I"), which is assumed to be computationally more costly. In four reading-time experiments we examined the online application of binding conditions A and B in English. Participants included adult German-speaking, proficient learners of English as an L2 and native English-speaking controls. Unlike the native speakers, our L2 speakers initially violated binding condition A during processing, which shows that condition A does not necessarily filter out syntactically illicit antecedents from the initial set of candidate antecedents for reflexives in L2 processing. The condition B experiments yielded a somewhat different picture. Here we found that L2 comprehenders showed a preference for linking non-reflexive pronouns to the non-local subject, which is consistent with the application of condition B. However, their matrix subject preference also extended to so-called 'short-distance' pronouns, which are known to be exempt from condition B. The native controls, in contrast, differentiated between syntactic environments that allowed local binding and those that did not. Together, these findings indicate that during processing, non-native comprehenders are less sensitive than native ones to structural cues to anaphor resolution – even if they demonstrated native-like knowledge of these in corresponding offline tasks. The results from the non-native speakers also fail to support the POB prediction that reflexive binding should be easier to compute than pronoun binding or coreference relationships. Instead, it appears that L2 comprehenders initially try to resolve both types of pronominal on the basis of semantic or pragmatic, rather than syntactic, information.