Numeral classifiers and plural marking

Many languages, including most Indo-European languages, have numeral noun phrases resulting from combining numerals directly with common nouns.  In English, for example, the noun ‘cow’ can directly combine with ‘one’, ‘two’, ‘three’, etc. to yield the numeral noun phrases one can use to talk about numbers of cows, such as those in ‘Alice saw one cow in the barn’, ‘Alice saw two cows in the barn’, ‘Alice saw three cows in the barn’, etc.  But there is a wide variety of languages in which the usual translations of such phrases have additional expressions of a special kind that intervene between nouns and numerals.  Those special expressions are called (numeral) classifiers, for they combine with numerals and accord with a limited group of nouns.  And languages with a substantial system of classifiers are called (numeral) classifier languages.  In this paper, I discuss some common features of classifiers of various classifier languages and examine their functions.

Usual accounts of classifiers assume that classifier languages have a radically different system of nouns than other languages (e.g., English, French, Italian).  They take all common nouns of classifier languages to be like mass nouns in that they cannot directly combine with numerals, and classifiers to be devices for specifying methods or criteria for individualizing referents of mass-like nouns.  But this view, I argue, has serious problems.  The approach conflicts with (a) the existence of optional classifier systems (e.g., Korean, Vietnamese) and (b) the compatibility of classifiers and plural marking (e.g., Itzaj Maya).  I present an alternative account of classifiers that are compatible with, and helps to explain, both features of classifiers.  The account, which I call the paranumeral account, takes classifiers to be similar to the English ‘pair’, ‘couple’, ‘brace’, and ‘yoke’ in relating to a natural number, but differ from them in relating to the number 1 rather than the number 2.  On this account, classifier language nouns with matching classifiers do not differ from count nouns of non-classifier languages in lacking criteria for individualization.  This helps to explain (a) and (b).