Farmer et al. (2006) found that how typical a words phonology is of other words in its lexical category influences the reading times of nouns and verbs in predictive contexts. When a preceding context generated a strong expectation for a noun, target noun-like nouns were read faster than verb-like nouns, along with a similar effect for verbs. Further, Dikker et al. (2010) found that the magnitude of the M100 response (sensitive to category-expectation violations) was modulated by phonological typicality: when a mismatch existed between whether context predicted a noun and the nouns degree of typicality, a heightened M100 response occurred in visual cortex. Here, using lexical decision and a self-paced reading, we examine whether these word-form effects occur in all cases where a noun or verb is possible, or whether they are more robust when expectations are strong, and thus visual word-form information may be enough to facilitate categorization.