This study takes up the Generative Lexicon model of Pustejovsky (1995). Key in this model is the notion of a generative lexical mechanism, which helps select the appropriate sense of a word in context. I ask: keeping Pustejovsky's levels of representation (qualia, event, and argument structure) constant, is the nominal classification described in Pustejovsky (1995) the only classification that can emerge from Pustejovsky's levels of representation? Is there some way that generative lexical mechanisms can give insight into noun classification in addition to their role in sense selection? To answer these questions, I rely on the methodology in Levin (1993), replacing verbal alternations with generative lexical mechanisms. For example, in the sentences

        (1)  Sally began the Pepsi.
                --> (Sally began drinking the Pepsi.)
        (2) *Sam began the knife.
                --> (Sam began cutting with the knife.)

participation in the True Complement Coercion with "begin" indicates that "the Pepsi" is a member of Nouns with Inceptions while "the knife" is not. I claim that such a methodology yields a motivated system of noun classes that repairs emerging conflicts between linguistic and encyclopedic world knowledge in the Generative Lexicon.

Levin, Beth. English Verb Classes and Alternations. The University of Chicago Press. 1993.
Pustejovsky, James. The Generative Lexicon. The MIT Press. 1995.