The status of nominal sub-categories: Exploring frequency densities of plural -s

Abstract

Abstract Plurale-tantum nouns (scissors, leggings, glasses) are an example of the systematic lack of an unmarked form of a lexeme. In contrast to singulare-tantum nouns, most notably mass nouns, this systematicity is mostly restricted to individual lexemes and analogously related ones (trousers, pants, knickers). It remains an open question whether there is any functionally motivated nominal subclass that goes beyond smaller lexical fields. The main goal of this paper is to estimate whether such extreme proportions in the absence or presence of inflectional markers cause distinctly high concentrations of lexemes, i.e. nominal subclasses. In a first step, the probabilities for a lemma to occur with plural -s were bootstrapped with replacement. Secondly, the bootstrapped data was equally split into 10 strata at varying inflection probabilities. Homonyms and polysemes that differ in their probability to be inflected are thus disambiguated. For each stratum, type frequencies were extrapolated by means of LNRE models. The same process was repeated for reference data sets containing verbal -ed and -ing. The bootstrapped data showed that frequency and proportion of inflection reveal clusters likely to represent different polysemes or homonyms. The type frequencies of the partially disambiguated singulare-tantum nouns turned out to be clearly distinct. However, for the plurale-tantum nouns, the extrapolated type frequencies were only marginally higher than those of the other suffixes, which are not usually thought to have a tantum-like subcategory.