Category formation and lexical encoding of a new contrast.

Abstract

The question whether category formation is a prerequisite for U.S.‐English learners of French to encode a non‐native contrast in lexical representations is investigated, looking at front [y‐œ] and back [u‐ɔ] rounded vowels. An ABX categorization experiment revealed no group difference between advanced (N=18) and inexperienced learners (N=18). Both made significantly more errors than French controls (p<0.01) on the [u‐y] contrast, despite a good global discrimination (15% error). The possibility that minimal pairs of difficult contrasts (e.g., sourd [sur] deaf vs sur [syr] sure) are stored as homophones was tested in a lexical decision task with repetition priming. French words and non‐words were paired with either themselves (repetition) or a minimal‐pair‐counterpart (minimal pair) in a 260 item list. Correct RTs were measured for each item. Given a comparable RT‐advantage on the repetition and the minimal pair condition, merged lexical representations were assumed. Advanced learners, like native speakers, showed no RT‐advantage for minimal pair conditions; inexperienced learners displayed significant facilitation for [u‐y] and [ɔ‐œ] minimal pairs (but not [i‐y] control condition). This suggests that successful lexical contrast is possible for advanced English‐French users despite persistent perception errors—the hallmark of an insecure category establishment—presenting an argument for the dissociation of both mechanisms.