Nouns and verbs in Cognitive Grammar: Where is the ‘sound’ evidence?

Formalist approaches traditionally define word classes in distributional terms. By contrast, Cognitive Grammar advocates a semantic basis: nouns profile THINGS; verbs highlight PROCESSES. There is psycholinguistic support for the importance of semantics in lexical categorisation, but also for (language-particular) distributional and phonological properties. This paper focuses on phonology, whose importance is further underlined by data from language change and typology. Following a review of the psycholinguistic, historical linguistic and typological evidence, a gap in the literature is filled, i.e. an experiment involving the production of nonce nouns and verbs is conducted, providing further converging evidence for phonology. I then show how this evidence, although not currently recognised in Cognitive Grammar, can be straightforwardly accommodated as phonological sub-schemas. These sub-schemas are probably more important than the super-schemas proposed in Cognitive Grammar (which may actually be non-existent, and anyway fail to yield clear predictions vis-à-vis empirical data). I conclude that in developing the model further, a higher degree of responsibility to all the available empirical data is called for.