Minimal brain adaptation for representational prioritization in non-native parsing: Evidence from a time-frequency analysis of recursion in Wh-dependencies in French

We report on a time-frequency analysis of electroencephalographic activity as native speakers (NSs) and non-native speakers (NNSs) resolve anaphoric reference at the recursive edge of embedded clauses in multi-clausal filler-gap dependencies. The fillers in these wh-dependencies involve lexically selected noun complements (Comp): quelle de ' cision a` propos de lui/`a son sujet 'which decision about him' or unselected NP-modifiers (Mod): quelle de ' cision le concernant 'which decision regarding him'; they also involve the lexical encoding of antecedent gender (lui/elle '3p. sing.masc/fem') or not (son '3p.sing.gen') in a pronoun. We propose that NNSs adapt to processing a non-native language by directing more resources toward computing lexically encoded properties than toward inferences that go beyond lexical information. At the recursive clause edge, a population-wide asymmetry between conditions arose at 15-16Hz in induced power: Lexical selection and gender specification led to diminished power for Comp son vs. Mod le/la and for specified lui/elle vs. unspecified son in Comps. This was also found at 17-22Hz in induced power in NSs whereas asymmetries were reversed in NNSs, with computations of lexically encoded properties demanding greater power. At 41-44Hz in induced power, group differences emerged: More power was directed at gender-specified le/la and lui/elle in NNSs, but at unspecified son in NSs. These asymmetries suggest that non-native processing seems characterized by resources directed toward lexically encoded properties in computations.