Twenty-four migraine patients were randomly assigned to one of four conditions: (a) self-monitoring of headache activity (waiting list), (b) frontalis EMG biofeedback, (c) digit temperature biofeedback, and (d) digit temperature biofeedback plus Rational-Emotive Therapy (RET). Bidirectional control over the target physiological response was assessed through a reversal design in each session. Following at least a four-week baseline, the three biofeedback groups received 8 to 10, 30-minute sessions of bidirectional biofeedback training, scheduled twice a week. Subjects in the combined digit temperature biofeedback plus RET group received three 40-minute sessions of RET as an addition to the third, fifth, and seventh biofeedback sessions. Records of daily home practice were kept throughout treatment and three-month followup. Subjects on the waiting list monitored headaches for at least five months, corresponding to "baseline", "treatment", and three-month followup. Digit temperature biofeedback alone and in conjunction with RET did not prove to be more effective than the control conditions. All the EMG subjects reduced headache activity to two-thirds or less of the baseline level by the third month of followup. Bidirectional digit temperature performance did not improve with training, was demonstrated in only 33% of the biofeedback sessions, was not maintained over time, and was unrelated to improvement in headache activity. EMG subjects reported biofeedback performance to be an easier task and met the performance criterion on 85% of the sessions. The frequency of home practice contributed over 55% of the variance in retrospective estimates of headache improvement but was not related to changes in daily records of headache activity.