The purpose of this study was to clarify the influence of the no-feedback muscle activity on awarenss of the muscle tension as well as the performance during tension c6ntrol training with the use of electromyographic (EMG) feedback. To relax the agonist (biceps) muscles and to contract the antagonist (triceps) during arm flexion, ten undergraduate male students were tested before, during and after continuous visual feedback from each muscle activity. In addition, the no-feedback muscle (deltoideus and supinator) activity was recorded in the same way with the feedback muscle activity. The experiment consisted of three sessions: pretraining (10 trials), feedback training (4 blocks of 10 trials), and testing (10 trials). Immediately after each trial in the pretraining and test sessions, the subject was asked to guess whether the control level of both muscle tensions increased or decreased, relative to the immediate preceding trial. The following results were obtained: 1) In the mean EMG of the triceps and the mean proficiency of muscle control, significant differences among blocks were found in the feedback training session, while the test session showed a significantly greater value than that of the pretraining session. 2) In the no feedback muscle, though there were no significant EMG differences among blocks of the training session and between the pretraining and test sessions, the EMG proved a significantly greater correlation not with the proficiency of muscle control but with the feedback muscle EMG. 3) The probability of correct guessing between adjacent trials in the test session was significantly greater than that in the pretraining session. 4) A significant interaction was found between overall subjective gussing and the actual activities of the no-feedback muscle in the test session. These results were discussed with the implications of the application for EMG feedback training to the motor task.