Introduction

A numeral classifier is a member of a paradigm of words or morphemes which form a binary construction with a cardinal numeral, called classifier phrase. The classifier phrase, in turn, forms a binary construction with an individual noun (or count noun) designating what is being counted.

A mensurative (measure word) is a member of a paradigm of words which form a binary construction with a cardinal numeral, called measure phrase. The measure phrase, in turn, forms a binary construction with a mass noun designating what is being measured.

Most languages have mensuratives (Cabécar [Chibchan] has none). In many languages (including English), they number in the hundreds.

Some languages have numeral classifiers in addition to mensuratives (Cabécar has them although lacking mensuratives). There are never more than a few dozens of them.

The cover term for numeral classifiers and mensuratives is enumerative.

While the construction with the cardinal numeral is crucial for the numeral classifier, in some languages it combines with other words or morphemes, too. In Yucatec Maya and Cabécar, e.g., it also combines with the interrogative pronoun ‘how many’. In Mandarin Chinese, it also combines with demonstrative pronouns.

The distribution of the numeral classifiers and of the numerals in a language may be such that one does not occur without the other. In that case, it is difficult to ascertain their degree of structural autonomy. It is often assumed that the numeral must be a word, while the classifier is an affix. As a general default, such an assumption appears eurocentric. In Yucatec Maya, it turns out that the numeral never occurs freely, while there is at least one construction in which the numeral classifier is found without being in a binary construction with a numeral. Here, the numerals must be bound morphemes. Moreover, since many numeral classifiers stem, etymologically, from nouns or verbs, their word status may be expected in many languages.

How to tell numeral classifiers and mensuratives apart

The two classes of items may be distinguished by distributional criteria:

Sometimes, individual nouns and mass nouns may be distinguished by distributional criteria. In English, for instance, individual nouns occur in the context many __, while mass nouns occur in the context much __. If these two classes of nouns may be kept apart on structural grounds, then those enumeratives that occur with numerals combining with individual nouns are numeral classifiers, while those enumeratives that occur with numerals combining with mass nouns are mensuratives.
Sometimes the syntagmatic relation between a measure phrase and the mass noun is a possessive relationship (e.g. in Mandarin with optional attributor de), while the syntagmatic relation between a classifier phrase and the counted noun is apposition.
Many mensuratives obey the distributional restriction that their mensurative phrase does not combine with any (mass) noun. This is typically true for time units such as hours, days or times (of iteration). – If the classifier phrases of a language do at all form a construction with a counted noun,¹ then every numeral classifier in the set can do so.
Some mensuratives may behave like individual nouns in combining with a classifier phrase. – The opposite – a mensurative phrase combining with a numeral classifier being used as a mass noun – never occurs.
In many languages, mensuratives combine not only with numerals, but also freely with quantifiers like ‘some’, ‘many’, while numeral classifiers strictly require a numeral.

Sometimes there are no structural criteria to tell numeral classifiers and mensuratives apart. Then semantic criteria such as the following may be used:

In the construction with a given noun, one mensurative often contrasts with other members of its paradigm. Then different amounts of a mass are being designated, while the nature of the mass itself remains unmodified. – In the construction with a given noun, often a given classifier cannot be replaced by another member of its paradigm. If it can, then a different kind of object is being designated.
The theoretical basis of this criterion is the fact that the numeral classifier is typically predictable insofar as the semantic class of a noun is entailed by its more specific semantic features, while the specific mensurative is alien to the meaning of a mass noun.
Both numeral classifiers and mensuratives form paradigms which may be hierarchically organized. However, the classifier hierarchy is a taxonomy, while the mensurative hierarchy may be a meronymy to the extent that it is standardized. Thus, sentences of the form ‘one M₁ of X is less than one M₂ of X’, e.g. a pinch of salt is less than a spoonful of salt, or even ‘ten M₁ (of X) make one M₂ (of X)’, e.g. three feet (of a string) make one yard (of a string), make sense if M_i is a mensurative, while there are no such sentences for numeral classifiers.

¹ Incipient numeral classifiers, such as German Stück and Mann, do not fulfill this condition.