Category labels such as dog' andgreen' appear to induce more categorical representations--highlighting category diagnostic features and helping to distinguish category members from non-members. Here we investigate whether covert language use has a similar effect by taking advantage of natural variation in people's reported use of inner speech. To measure categoricality, we use a repetition-priming task in which people make a semantic judgment of repeated images. We find a robust repetition effect of categories such that people are faster to respond to a cat if they have seen a previous image of a cat. These differences in inner speech are not associated with differences in repetition priming, but interacted in complex ways with differences in visual imagery and susceptibility to a verbal interference task.