During the training phase, 36 subjects received (a) EMG biofeedback from multiple muscle sites, (b) EMG biofeedback from the frontal site, or (c) no biofeedback. Results indicated that neither biofeedback procedure reduced self-reports of anxiety, but that multiple-site biofeedback was effective in reducing several indices of autonomic arousal (pulse rate, finger pulse volume, and skin temperature) while frontal biofeedback was not. During the generalization/stress phase, all subjects were threatened with and received electric shocks and were told to apply the relaxation techniques they learned during the training phase even though no additional biofeedback would be provided. Results indicated that multiple-site biofeedback was effective in reducing self-reports of anxiety and autonomic arousal but that frontal biofeedback was not. These results confirm previous data indicating that frontal biofeedback is not an effective procedure for controlling stress, but suggest that EMG biofeedback can be effective in reducing self-reported anxiety and autonomic arousal if a multiple muscle-site feedback procedure is employed.