Fake Feature and Valuation From Context

Abstract

This thesis offers a new account of a persisting puzzle in the theory of ellipsis and association with focus: the fact that φ-featural content on full DPs and on bound pronouns can sometimes be ignored in focus alternatives and in calculating identity for ellipsis (‘Fake Features’). I present new data about gender and number mismatch in ellipsis which proves difficult to model on existing approaches to fake features (e.g. Sauerland 2013; Sudo & Spathas 2020). The heart of the proposal is a derivational theory of contentful φ-features: they do not, as usually assumed, enter a derivation from the lexicon with listed meanings (presuppositions) that constrain the denotation of their host DP; rather, they are inserted late in the derivation towards PF by a process called “Valuation from Context” (Kucerov ˇ a´ 2018): the features are inserted based on the meaning of the DP in the (local or global) context of evaluation. Ellipsis identity and focus alternatives are computed off of the feature-less representation. The theory assumes that the construction of local contexts for embedded constituents (Schlenker 2009) is blind to information encoded in focus alternatives. The account supports an architecture of grammar in which representations that are submitted to semantic interpretation (meaning-in-context) feed morphological valuation processes. It also implies that there is no substantial difference between “interpreted” and “uninterpreted” φ-features; in a sense, both are uninterpreted, the distinction being whether they are valued from context or from pieces in the structure.