Lemhofer, K. & Herbert, S.
Radboud University Nijmegen
Learning word gender in a second language (L2) is difficult, especially in languages where a noun’s form does not predict its gender (e.g., Dutch). Recently, we have shown that the high incidence of gender errors in L2 - often resulting from inappropriate L1-L2 transfer - persists even after repeated feedback. One possible reason for the limited ‘learnability’ of L2 gender is a lack of sensitivity to gender cues in L2 input. We investigated this claim in an ERP-experiment with German learners of Dutch, who read Dutch sentences that contained a singular target word preceded by its correct or incorrect (gender-marked) definite determiner. A second type of determiner violations were plural targets preceded by the correct plural determiner 'de' or the singular determiner 'het' (i.e., number agreement violations). For a greater resemblance with natural L2 input, no additional task was given apart from reading for comprehension. After the experiment, participants performed an offline gender assignment task on all targets from the singular condition. The data from this task enabled us to re-assign the singular sentences per participant into subjectively correct and incorrect sentences. The results show that first, incorrect singular determiners for plural targets produced both a N400 and a P600 effect, showing that this type of violations was easily detected. Second, objective incorrectness in the singular condition did not produce any effects. However, when re-sorted according to individual gender assignments, subjective incorrectness resulted in a P600 effect. These results show that L2 learners were sensitive to gender cues in L2 input, but only when ‘correctness’ was based on the individual’s gender representations. This finding thereby demonstrates that investigations of the processing of grammatical violations in L2 should take each individual’s performance on the respective grammatical feature into account.