LINGUIST List 14.3079

Tue Nov 11 2003

Diss: Psycholing: Sabourin: 'Grammatical...'

Editor for this issue: Takako Matsui <takolinguistlist.org>

Directory

  • lsabourin, Grammatical Gender and Second Language Processing

    Message 1: Grammatical Gender and Second Language Processing

     Date: Tue, 11 Nov 2003 00:56:00 +0000
    From: lsabourin <lsabourinpsych.ubc.ca>
    Subject: Grammatical Gender and Second Language Processing


    Institution: University of Groningen Program: Center for Language and Cognition Dissertation Status: Completed Degree Date: 2003

    Author: Laura Sabourin

    Dissertation Title: Grammatical Gender and Second Language Processing: An ERP Study

    Linguistic Field: Psycholinguistics, Neurolinguistics, Language Acquisition

    Dissertation Director 1: Ger G. J. de Haan Dissertation Director 2: Laurie A. Stowe

    Dissertation Abstract:

    Second language (L2) speakers were tested on their ability with the Dutch gender system. The goals of the research was to determine L2 speakers' overall ability to retrieve a noun's gender, to see if they could apply rules for gender agreement, to determine how L2 speakers process grammatical gender information and whether an effect of the native language (L1) could be found. These goals were investigated by having participants do either off-line or on-line grammaticality judgment tasks. L2 speakers of from three language backgrounds were studied: German (similar gender system to Dutch), Romance (gender system not similar to Dutch), or English (no gender system).

    Off-line Tasks: In a control task looking at the L2 verb system it was found that all L2 groups performed at a native speaker level. However, while acquiring grammatical gender seems to be possible at the lexical level, all participants showed problems with gender agreement (with no group performing at a native level). A hierarchy of performance can be seen with the German group performing the best followed by the Romance speakers and then the English group, who only performed at chance. These results map directly onto a strong transfer effect. The German group has the closest gender system to the Dutch system and performed the best of the L2 group. The Romance group has a gender system in their L1 and although it is not related to the Dutch system it seems to be the case that having gender in the L1 is enough to give a group of participants an advantage in acquiring a new gender system. The English group, with no L1 gender system, has no information that can help them tap into the L2 gender system.

    On-Line Tasks: The ERP results for the Dutch speakers showed a P600 to all sentence types (control and gender sentences). This is considered to reflect the fact that violations of grammatical gender agreement are syntactic in nature. The ERP data for all L2 groups show a delayed P600 with a limited scalp distribution. This suggests that for these sentence types only a quantitative difference is present between L1 and L2 speakers. The ERP patterns for grammatical gender processing show very different results. Only the German group showed any patterns similar at all to the native speakers. However, even the German group, only showed similar processing for gender violations where the information was present in the form of a definite determiner. This is likely because when the determiner cues access to gender they are able to directly map onto their L1 system. Both the Romance and English groups performed poorly on the on-line task. Both groups did show, for some gender conditions, significant differences between the grammatical and ungrammatical sentences. However, neither group showed a clear P600 pattern.

    L2 learners can be found who can acquire and process parts of an L2 grammar with native-like ability. However, there are limitations to this ability that are greatly influenced by the L1 of the learners. Within the verb domain where there are overt markers for both subject-verb agreement and for finiteness all the L2 groups tested here were able to show both native-like knowledge and processing. Not surprisingly their processing was slower, but nonetheless, the processing is qualitatively very similar to the native speaker processing patterns.

    Performance on gender showed a strong effect of transfer. The ability to acquire gender knowledge was modulated by the presence of the abstract category gender in the L1: both the German and Romance groups had an advantage over the English group. On the other hand the ability to process L2 gender similarly to L1 was modulated by the presence of very similar surface features in the L1, an abstract category appears not to be enough.