Various types of noncontingent feedback have been used as control procedures in EMG training; however, their effects on such training have received little attention. Experiment 1 in the present study examined the effects of noncontingent feedback on EMG training, and Experiment 2 assessed the effects of feedback characteristics on EMG responses. In Experiment 1, three noncontingent feedback groups (yoked control, randomly fluctuating tones, and decreasing tones) and one contingent group underwent 20 minutes of training for frontal EMG decreases. Procedures in Experiment 2 were identical to those in Experiment 1 except that subjects were instructed merely to listen to the feedback tones. Results of Experiment 1 indicated that contingent and noncontingent fluctuating feedback groups achieved significantly lower EMG levels than noncontingent decreasing and yoked control groups. In Experiment 2, however, no differences in EMG activity were found among groups. In both experiments, groups did not differ in terms of subjective variables such as frustration, suspiciousness about the tone, or length of time attending to the tone. Results of these two experiments suggest that differences in EMG responses to various types of noncontingent feedback result from interactions between characteristics of the feedback stimulus and instructions to decrease the stimulus.