The algebraic structure of morphosyntactic features

Abstract

The most common way of separating homophony from syncretism — which is a basic challenge for any inflectional analysis: to distinguish between accidental and systematic form-identity — is attributing only the latter to a coherent feature combination instantiating a natural class. Features predetermine which form-identities can or cannot be analyzed as natural-class syncretism. Hence, they are crucial for the restrictiveness and predictions of morphological grammar. However, most current theoretical frameworks (e.g.] Anderson 1992, Corbett & Fraser 1993, Halle & Marantz 1993, Stump 2001) do not make explicit their assumptions regarding the formal status of features. They miss out on state-of-the-art formalisms to introduce feature notations like Formal Concept Analysis (FCA, going back to Wille 1982, Ganter & Wille 1999) which provides a formal model of conceptualization in general. In this paper, I will show how FCA provides an all-embracing terminology to reproduce, visualize, and compare feature systems from different morphological frameworks, enables more precise and consistent morphological analyses, and crucially serves to rule out excessively powerful notations where the feature combinatorics are decoupled from the distributional facts they represent.