While constituting a semantically homogeneous class, cardinal numerals cross-linguistically display a rich variety of morphosyntactic properties. An approach to such variation, with the aim of identifying a limited range of morphosyntactic statuses to which these properties are related, has been taken through the application of the Role and Reference Grammar (RRG) descriptive-interpretive model. Among the different statuses are those prototypically held by major categories, like nouns and verbs. However, the non-alignment hypothesis within RRG, according to which ontological, lexical and syntactic categories do not necessarily overlap, allows the occurrence of numerals as nucleus of a reference phrase, or nucleus of a clause, to be accounted for. While pointing out elements of similarity in the morphosyntax of numerals and other word classes, the analysis highlights crucial divergences between numerals and adjectives, a word class numerals are frequently equated to in the literature. The languages mentioned here belong to diverse typological and genetic groupings.