States, changes of state, and the Monotonicity Hypothesis
Andrew Koontz-Garboden
Ph.D. Dissertation, Department of Linguistics, Stanford University, 2007
Advisor: Beth Levin
Download in pdf or ps

Abstract
This dissertation examines the Monotonicity Hypothesis (MH), the widely assumed, but rarely discussed idea that while word formation operations can add decompositional operators to a word's lexical semantic representation, they cannot remove them. Adopting modified versions of Dowty's (1979) decompositional representations of states (e.g., red) and changes into states (e.g., redden), I observe that the MH makes two strong falsifiable predictions in this domain. First, words naming states should never be derived from words naming changes of state, as this would involve the deletion of a BECOME operator. Data from a number of languages are examined and shown to bear out the prediction, with one apparent exception. Ulwa, an endangered Misumalpan language, appears to have words naming states derived from change of state denoting roots. Detailed examination of Ulwa verbal and adjectival semantics and morphosyntax based on extensive primary fieldwork shows that this is an illusion. Given the widely held view that the semantic representation of inchoative verbs lacks the CAUSE operator present in the representation of causative verbs, a second strong prediction is that inchoatives should never be derived from causatives. This is apparently falsified by anticausativization, in which an inchoative verb is derived from a causative verb, e.g., Spanish romper `cause to become broken' versus romper se `become broken'. Building on Chierchia (2004), I argue instead for a reflexivization analysis of anticausativization, showing that it captures a wide range of facts of the phenomenon not accounted for by alternative approaches, most notably facts showing that derived inchoatives retain the CAUSE operator of the causatives from which they are derived. This analysis is consistent with the MH, since it entails no deletion of decompositional operators. Finally, I lay out several areas for future research. Formally, the relationship of the MH to the Principle of Compositionality remains to be clarified. Empirically, the MH makes many falsifiable predictions beyond the domain of states and changes of state, which suggest areas for promising future crosslinguistic investigation. Research of both kinds will shed further light on the MH and more broadly on the semantic nature of word formation operations.

Return to Andrew Koontz-Garboden's home page

eXTReMe Tracker