Abstract
Negation is ubiquitous in natural language and a fundamental logical operator in formal semantics, yet it remains difficult to capture with distributional semantic models. \r In this thesis, we examine negation of verbal predicates by creating distributional semantic models of a medium-sized text corpus. We show that the effect negation has on vectors of negated verbal predicates is similar to that of dimensionality reduction, suggesting that negation in context, or conversational negation, functions as a kind of pragmatic salience signaling. We also introduce an annotated data set of alternative plausibility ratings for negated verbal predicates and find that antonyms and distributionally similar words are plausible alternatives for negated verbs.