Processing Discontinuous Chinese Separable Words: Evidence from Eye Tracking

Chinese separable words combine lexical integrity with structural separability, but how readers process discontinuous forms during sentence reading remains unclear. Thirty native Chinese speakers read sentences containing discontinuous separable words, verb–object (V–O) phrases, or quasi-separable words (QSWs) while their eye movements were recorded. Corpus-derived discontinuous-structure frequency and local contextual surprisal were included as continuous predictors. Separable words did not differ reliably from V–O phrases in acceptability judgments or target-region eye-movement measures, suggesting no stable additional processing cost relative to the V–O baseline. QSWs were accepted less often, elicited longer judgment response times, and showed greater target-region processing difficulty. Within the QSW condition, higher frequency was associated with shorter response times and greater acceptance. After frequency and surprisal were added, all adjusted pairwise comparisons for the binary fixation measures were nonsignificant. For conditional durations, reliable adjusted differences were limited to V–O phrases versus QSWs; separable words did not differ reliably from either comparison condition. Higher surprisal predicted greater second-fixation probability. Together, these findings suggest that discontinuous separable words benefit from distributional and constructional support, allowing their components to enter V–O-like integration without obligatory lexical reconstruction.