An fMRI study examined how speakers inspect their own speech for errors. In a word production task, we observed enhanced involvement of the right posterior cerebellum for trials that were correct, but on which participants were more likely to make a word-as compared to a non-word error. Furthermore, comparing errors to correctly produced utterances, we observed increased activation of the same cerebellar region, in addition to temporal and medial frontal regions. Within the framework associating the cerebellum to forward modelling of upcoming actions, this indicates that forward models of verbal actions contain information about word representations used for error monitoring even before articulation (internal monitoring). Additional resources relying on speech perception and conflict monitoring are deployed during articulation to detect overt errors (external monitoring). In summary, speech monitoring seems to recruit a network of brain regions serving domain general purposes, even for abstract levels of processing.