Retour vers liste

Détail de la contribution

Auteur: Viviane DÉPREZ

Co-Auteur(s): Asya Achimova, Rutgers U., USA.

Specificity matters even when definiteness transfers

Abstract/Résumé: For Ionnin&als, definiteness and specificity are parametrizable lexico-semantic features. On this view, while article-misuse by L2-learners with determiner-less-L1 reflects fluctuation between specificity/definiteness UG-parameter settings, article acquisition by L2-learners with article-based-L1 is predicted to manifest feature transfer. Recent studies of predicted transfers, however, reveal conflicting results. To re-assess the role of specificity in L2 in clear expected transfer cases, 97 English L2 learners of French (3 levels) completed an on-line forced-choice elicitation task with gender and number kept constant and definite, indefinite, partitive and blank as choices that featured 87 dialogues, 52 critically manipulating specificity and definiteness. French natives served as controls. Since both English and French lexically encode definiteness but not specificity, transfer predictions are clear: only definiteness should influence L2 determiner choice. We ran a linear-mixed-model analysis to assess whether contextual definiteness and specificity affected the proportion of errors in determiner choice. Definiteness was shown to influence the amount of errors, with a significant main effect p<.01, as did Level p<.01 and Specificity p<.05. In contexts targeting definites, learners overused the indefinite article more in [-specific] than in [+specific] contexts, p<.01. Similarly, in contexts targeting indefinites, they overused the definite article significantly more in [+specific] than in [-specific] contexts, p<.01. These results show that specificity influences L2 determiner-choice even when only definiteness transfer is expected. Although not predicted on Ionnin’s account, our results make sense with specificity conceived as an L1 independent pragmatic notion rather than as a parameterized semantic feature. On this alternative view, overuse of definites in [-def+spec] contexts point to L2-learners’ over-reliance on the pragmatic notion speaker identifiable (Kagan 89, De Cat 11,12) for article choice, rather than to parametric fluctuations between article-systems, while the overuse of indefinites underscores reliance on the complementary notion ‘not-speaker-identifiable’. Together, both types of misuse highlight the L1-independent importance of a speaker-centered perspective in reference computation as a source of L2 determiner errors rather than a parametrically governed transfer of lexico-semantic features. As psychologists showed (Keysar&als), adults a resort to a speaker-centered (egocentric) perspective when confronted with increased processing costs, conditions that are clearly relevant in early stages of adult L2-acquisition.