Singulative and numeral classifying systems in English:

how a variationist approach may help understand their coexistence

Salikoko Mufwene

Sat. 9:00-10:40 C

English has typically been cast as a singulative language, in which, unlike numeral classifying languages, only mass nouns need classifiers for individuation. I assume that in English the count/mass distinction is grammatical but not lexical (Allan 1981), as supported by alternations such as how many ropes/how much rope, although nouns show countability preferences which make milk more common than a milk but many desks more common than much desk. Building on these empirical facts, I argue against the following myths: 1) English has only a handful of classifiers, compared to numeral classifying languages; and 2) it uses numeral classifiers only with mass nouns. Preliminary analysis reveals that English has more than 40 classifiers--no small number compared to numeral classifying languages. Moreover, some classifiers combine with both mass and count uses of nouns, as in pile of garbage/books, bunch of money/bananas, and trail of blood/footsteps.

This variability calls for linguistic justification of the distinction between classifiers and collectives. The term "collective" has been restricted to cases in which the complement of an otherwise classifying construction is count, as in ring of hills, book of matches, fleet of trucks, and swarm of flies However, the distinction does not account for the role of classifiers in providing some salient information (e.g., shape of natural cluster) about the mass or set of individuals denoted by the syntactic complement. Thus the "classifier"/"collective" distinction is not only unmotivated syntactically but also ill-conceived semantically. Here we may find inspiration from wholly numeral classifying languages in which all nouns pattern alike, subject to semantic/pragmatic constraints.

The data display variation that may be explained on the model of coexistent systems proposed by Mufwene (1992) and elaborated by Labov (to appear), which assumes that grammatical subsystems with non-identical functions overlap in the overall system. According to this view, alternations such as a hair/a strand of hair and an insight/a flash of insight obtain in the systems' area of over-lap. While they may to be widespread in the overall number delimitation subsystem, their distribution patterns are informative about cases in which classifying constructions do not alternate with singulative ones, showing that classifiers make important semantic contributions (having to do with mode of reference) and that the factors conditioning variation may be semantic-referential.

The data also reveal coexistence of two morphosyntactic patterns in English numeral classifying constructions. In addition to the Classifier+of+Noun pattern, there is a compounding pattern, as in cornflake, smoke ring, ice cube, and gauze pad/roll. In this respect, one finds alternations such as sand layer/layer of sand, ice block/block of ice, and gauze roll/roll of gauze. Such alternations are also worth investigating, although it is not clear what light quantitative analysis may shed on them.

Overall, this reassessment of the singulative and numeral classifying subsystems in English, in which both dominant and variable patterns can be identified, suggests that all variation is not necessarily underlain by one shared grammatical subsystem. Variability that has to do with coexisting overlapping subsystems suggests that there are cases were different subsystems may produce quantifiable alternations. Analysis of such facts may shed light on the architecture of grammar and interindividual variation in ways which traditional studies of variation have not sufficiently illuminated. By formulating several questions yet to be answered, I propose in this paper a variationist research program that should help bridge the gap between theoretical and quantitative analyses.

References

Allan, Keith. 1981. Nouns and countability. Language 56.541-567.

Labov, William, to appear. Coexistent systems in African-American English.

Mufwene, Salikoko S. 1992. Why grammars are not monolithic.