Chinese Verbal Classifier Processing in Verb-Phrase Naming: An ERP Study

Current language production models have disputes regarding the activation, selection, and encoding of to-be-produced word, with ongoing debate on whether the syntactic features mediate between semantic and phonological access. This study focused on the lexical selection stage in speech production, investigating the grammatical feature retrieval of Chinese verbal classifiers, a topic with limited exploration despite its lexico-syntactic properties resembling nominal classifiers. Three research questions were therefore addressed: i) Do traceable effects exist for classifier congruency and semantic interference? ii) Can mismatched trials induce a more negative N400 effect? iii) As a lexical-syntactic feature, how are verbal classifiers activated and selected during speech production?
A 2x2 within-subject design was employed, with classifier congruence (congruent vs. incongruent) and semantic relatedness (related vs. unrelated) as factors. Forty-one healthy Mandarin Chinese native speakers voluntarily participated in the verbal classifier phrase naming (verb+one+classifier) experiment under the PWI paradigm. Behavioral and EEG data from 33 participants underwent analysis using a single-trial modeling approach and followed a top-down modeling process. The chosen 17 electrodes (F3, Fz, F4, FC5, FC1, FC2, FC6, C3, Cz, C4, CP5, CP1, CP2, CP6, P3, Pz, P4) and time window (350ms-500ms) were determined through permutation tests.
Behavioral results showed a significant semantic interference effect in naming latency and a marginal classifier congruency effect was found in naming accuracy. In contrast, EEG data revealed more negative amplitudes for classifiers incongruent trials compared to congruent trials. Though not reaching significance, semantically unrelated trials tended to be more negative than related ones. Topographical maps indicated a distribution of the classifier N400 effect from frontal to parietal regions, while the marginal semantic N400 effect was concentrated in medial frontal and fronto-central areas, exhibiting a tendency toward the right hemisphere. Presumably, verbal classifier feature was automatically activated, with competitive selection occurring between target and distractor classifiers. The null classifier congruency effect in behavioral data could be attributed to the opaque and one-to-various collocational relationship between verbs and verbal classifiers, which highly dependent on semantic information. Additionally, the limitations of the PWI paradigm may make it insensitive to detect competition processes for non-initial elements.