Abstract
There are two types of numeral classifiers in Japanese: those which cannot be used as a noun as -nin in 3-nin 'three (persons)' and those which can be used as a noun as -setai in 3-setai 'three (families)'. The author names the latter as nominal classifiers (noun-type numeral classifiers) in this article and examines their lexical characteristics based on two viewpoints: (i) productivity as numeral classifiers, i.e., whether they can be used with the interrogative form nan-'how many' or not, and (ii) semantic fields in the Bunrui Goihyo (Word List by Semantic Principles) they belong. The examples for analysis were extracted from a newspaper corpus. Nominal classifiers are classified into two types: container-type and non-container-type. Numeral classifiers of container-type are used to measure the volume with reference to a noun as a container, e.g., -sara in 3-sara 'three (dishes)'. Numeral classifiers of non-container-type consist of two subgroups: those which can be used with nan-'how many' (quasi-classifier) and those which cannot (pseudo-classifier), e.g., 3-setai 'three (families)'/nan-setai 'how many (families)' (quasi-classifier), 3-kakuryo 'three members of the cabinet'/*nan-kakuryo (pseudo-classifier). Quasi-classifiers are used to count numbers by accumulating them. Many of them denote abstract relations and are used to supplement the numeral classifier system in Japanese. Pseudo-classifiers, on the other hand, are temporarily used with numerals in a sentence. They tend to be composed of nouns whose concreteness is high such as persons and organizations, and have a function of endophora. The classification based on the productivity as numeral classifiers can be a perspective for interpreting classifications proposed in previous studies in an appropriate way.
