The relationship of "awareness of muscle tension" to depth of relaxation was explored. In one experiment, accuracy of forearm flexor control was assessed using the psychophysical method of magnitude production, and depth of flexor relaxation was measured using the integrated EMG before and after EMG biofeedback training. No consistent relationship between motor-control accuracy and depth of relaxation was found. A second, similar experiment with frontalis showed increased accuracy of frontalis control with deeper relaxation. Accuracy of passive, verbal judgments of spontaneous frontalis tension fluctuation exhibited no clear relationship with depth of relaxation. It was concluded that forearm flexor and frontalis may be under the control of distinct mechanisms, and that afferent information probably contributes to the control of neither muscle. Three structural theories of the control mechanisms were considered, and one depending on the central monitoring of efferent outflow (rather than afferent inflow) seemed most compatible with the frontalis data. Both flexor and frontalis data could be accounted for by a two-phase scheme combining central outflow monitoring with the monitoring of mental contents for arousal value at very low muscle tension levels.