Distinct ERP signatures for singular "they" and gender violations

The English pronoun system is undergoing rapid change, with singular "they" increasingly used to refer to specific individuals. A key question is how this innovative usage is processed online, particularly when paired with gendered antecedents. The critical issue is whether singular "they" incurs processing costs similar to canonical gender mismatches or instead engages a distinct interpretive profile that avoids the neural signatures typically elicited by incongruent uses of "he" and "she".

We addressed this question using event-related potentials (ERPs) in readers with high familiarity with singular "they". Participants listened to sentences containing reflexive pronouns whose antecedents varied in gender specificity (gendered vs. non-gendered) and referential status (referential vs. bound-variable).

Across both the P600 (500-800 ms) and N400/Nref (300-500 ms) time windows, gender-incongruent reflexives elicited robust neural signatures of processing difficulty that depended on antecedent type. Referential mismatches with proper names (e.g., "John…herself") produced frontal negativities consistent with referential integration difficulty, whereas bound-variable mismatches with quantified antecedents (e.g., "every woman…himself") elicited large P600 effects characteristic of morphosyntactic repair. In contrast, singular "themselves" patterned differently from these gender mismatch responses in gendered antecedent contexts. In referential contexts, the N400/Nref response to singular "themselves" was intermediate between those for gender-congruent and gender-incongruent forms, consistent with a gradient of predictability. In bound-variable contexts, singular "themselves" patterned closely with gender-congruent reflexives, showing no evidence of a violation response.

These findings suggest that, for speakers with substantial exposure, singular "themselves" does not behave like a canonical agreement mismatch, even in contexts that strongly penalize gender-incongruent forms. This pattern counters the intuition that gender-neutral reflexives necessarily impose a processing cost when paired with gendered antecedents.