Semi-lexicality refers to lexical items which show both lexical and functional properties. Numerals, for instance, very often show some nominal or adjectival properties (lexical), despite having a quantificational function (functional). Such semi-lexical elements challenge theories of categories which assume discrete distinctions between nouns, verbs, and adjectives. If a numeral shares some, but not all properties with a noun, should we call it a noun instead of a numeral?
This dissertation addresses such questions, and explores the concept of semi-lexicality. It focuses on phenomena in the nominal domain which largely involve idiosyncratic patterns of case and agreement. This thesis aims to identify how semi-lexicality arises in the syntax and what consequences it has for conceptualizations of categories. Semi-lexicality is explored via three case studies: Polish numerals, English quantificational nouns (lot, ton, bunch, number), and English kind-words (kind, type, sort). These case studies illustrate a variety of ways of being semi-lexical, and motivate an approach to semi-lexicality which decomposes basic notions of categories. By developing a precise theory of what it means to be lexical, functional, and semi-lexical, this thesis shows that it is possible to capture the idiosyncrasies of semi-lexicality in a predictable way.