What are categories? Adjective-like and noun-like semi-lexical numerals in Polish

In many languages, numerals appear to straddle the boundary between adjectives and nouns, sometimes behaving like adjectives, sometimes like nouns, and sometimes showing a mix of behaviors (Corbett 1978). The intermediate status of numerals presents a problem for theories of categories: how can something be simultaneously adjectival and nominal? In this paper I address the issue from the perspective of Polish numerals and argue that categories are derived notions, definable (in part) through the phi-features of an element. Elements with full sets of valued phi-features are nouns and elements with full sets of unvalued phi-features are adjectives. Building on the notion of semi-lexicality in Emonds (1985) and Corver and van Riemsdijk (2001), I suggest that numerals represent a new sort of semi-lexicality, involving incomplete phi-feature sets or feature sets with a mix of valued and unvalued phi-features. Such semi-lexical elements are predicted to show idiosyncratic behaviors which may or may not resemble adjectives and nouns. This is the approach taken for the Polish numeral system. Numeral 1 is defined as an adjective (unvalued gender, unvalued number) and numeral 1000 as a noun (valued gender, valued number). Numerals 2, 3, and 4 are treated as adjectival nominals (valued number, unvalued gender) and numerals 5+ (5–10, 100) as deficient nominals (valued number, missing gender). With these assumptions in hand, the case and agreement facts of Polish numerals can be captured. The implication of this study is that categories are not syntactic primitives; by viewing them as derived notions, we have the flexibility to deal with those elements like numerals which normally defy categorization.