To examine whether Korean-English bilinguals rely on first language (L1) or second language (L2) semantic space during lexical processing in L2 as a function of their L2 proficiency, a semantic priming categorization task was conducted. The semantic similarity between the prime and target in L2 was manipulated (strong vs. weak) based on word embedding vectors derived from either L1 or L2, representing the respective L1- and L2-based semantic spaces. While participants were asked to judge whether each target word belonged to a given category, preceded by overtly processed primes, their EEG responses were recorded. Results showed a significant P600 effect when semantic similarity was defined by L2 semantic space, only in bilinguals with high L2 proficiency. This study demonstrated, using a novel approach that leveraged a distributional semantic model for capturing bilingual semantic representations, ERP evidence that L2 proficiency modulates semantic processing.
Moon et al. (Tue,) studied this question.