Do classifiers make the syntactic count/mass distinction? Insights from ERPs in classifier processing in Japanese

• The Japanese classifier system does not elicit a grammatical distinction between object names and substance names.
• The violation of the noun–classifier agreements elicited the N400 instead of the LAN.
• The processing of Japanese numeral classifiers is primarily semantic-based.

It has long been assumed that classifier languages lack syntactic distinctions of objects and substances (i.e., count/mass distinction; Allan, 1977; Chierchia, 1998; Krifka, 1995; Lucy, 1992; Quine, 1969). Several linguists, however, claimed that classifier languages also make the syntactic count/mass distinction through the selective use of count classifiers (i.e., sortal classifiers) and mass classifiers (e.g., Cheng & Sybesma, 1998, 1999). The present study examined whether Japanese speakers make a syntactic count/mass distinction using the classifier system. The electroencephalogram (EEG) was recorded while Japanese speakers read word pairs (Experiment 1) or sentences (Experiment 2) in which noun–classifier agreement relations were manipulated. We examined event-related potentials (ERPs) elicited by the within-count/mass-category violation (e.g., an object name accompanied by a classifier for other objects) and the across-count/mass-category violation (e.g., an object name accompanied by a classifier for substances). In both experiments, the violation of the noun–classifier agreements elicited the N400, regardless of whether the noun–classifier disagreement was made within or across the ontological object/substance boundary. The across-count/mass-category violation did not recruit a syntactic process in the brain, suggesting that the Japanese classifier system does not highlight the distinction between objects and substances. The results also indicated that the processing of Japanese numeral classifiers is primarily semantic-based.