When speakers hear the fundamental frequency (F0) of their voice altered, they shift their F0 in the direction opposite the perturbation. The current study used ERPs to examine sensory processing of short feedback perturbations during an ongoing utterance. In one session, participants produced a vowel at an F0 of their own choosing. In another session, participants matched the F0 of a cue voice. An F0 perturbation of 0, 25, 50, 100, or 200 cents was introduced for 100 ms. A mismatch negativity (MMN) was observed. Differences between sessions were only found for 200-cent perturbations. Reduced compensation when speakers experienced the 200-cent perturbations suggests that this larger perturbation was perceived as externally generated. The presence of an MMN, and no earlier (N100) response suggests that the underlying sensory process used to identify and compensate for errors in mid-utterance may differ from feedback monitoring at utterance onset.