The Semantic Content of the General Classifier in Mandarin

A basic question in the study of nominal classifier systems concerns the relative roles of lexical semantics and grammatical rules. In this paper we address this issue by studying whether the Mandarin general classifier ge can be selected by a lexicon-independent default rule. Evidence comes first from a corpus analysis, which found no semantically coherent set of privileged lexical exemplars for ge. This finding was confirmed in an experiment in which native speakers of Taiwan Mandarin were given classifiers and asked to list as many appropriate nouns as they could; the type/token ratio for ge was significantly lower than that for most other classifiers tested, again suggesting that ge has no privileged exemplars. A second experiment was conducted to determine if the choice of ge can be forced by the presence of specific lexical semantic features. The experiment showed that it cannot; no significant difference in rate of ge use was found between nonce words given specific semantic features and nonce words given no meaning at all. Together the evidence implies that ge does not have semantic content, and therefore is selected by grammatical rule independent of the lexicon.