In phase one, 56 subjects either did (biofeedback training) or did not (no biofeedback training) receive frontal EMG biofeedback training, and either were (stress) or were not (no stress) threatened with electric shocks during the training periods. Results indicated that the threat of shock was effective in increasing arousal (as measured by self-report and multiple physiological indices), and that the biofeedback training provided under both stressful and non-stressful conditions was effective in reducing frontal EMG levels and the extent to which subjects worried about the shock, but was not effective in reducing physiological arousal, self-report of arousal, or the perceived painfulness of the shock. In phase two, the subjects from phase one were exposed to another threat of shock situation, but no additional biofeedback was provided. Fourteen additional subjects were not threatened with electric shock and served as a control group. Results indicated that the threat of shock was again effective in increasing arousal, and that subjects who had previously received biofeedback training under either stressful or non-stressful conditions were able to reduce their levels of frontal EMG and self-report of arousal, but not their levels of physiological arousal. The results provide evidence that EMG biofeedback training can reduce EMG levels under stressful as well as non-stressful conditions, and that this effect will transfer to a stressful non-training situation. However, the fact that the EMG reductions were generally not accompanied by reductions in arousal suggests that EMG biofeedback may have limited effectiveness as a general procedure for coping with stress.