The present study was designed to investigate the correction of error reaction which was caused by the mismatch between the subject's anticipation to a forthcoming stimulus and the stimulus actually presented. The reaction times (latency of EMG : ET, latency of force-change : MT1, and latency of movement initiation : MT2) and MT1 minus ET (CT) were measured by using regularly alternating serial light stimuli (right-left). At an unforeseen trial, the same lamp as an immediately preceding trial was turned on consecutively to belie the subject's anticipation for the stimulus appearance. The second of the two consecutive stimuli was termed "feinting stimulus." The time lag from the error EMG onset to the correct EMG onset was measured as the "switching time (ST)". The results obtained are as follows: 1) The reaction times for the regularly alternating stimuli shortened in comparison with simple reaction times, indicating the anticipation for the stimulus presentation was established within the subject's mind. 2) The error reaction appreared in 22.8% of all trials with the feinting stimuli. Correct reaction for the feinting stimulus was delayed accordingly as the intensity of error reaction increased. 3) The switching tine increased according to the intensity of error reaction. When error reaction was so weak that only EMG appeared, switching time was about 100 ms. For the intense error reaction, it reached 200 ms or more. 4) The switching time was divided into two phases: one was the duration of the error EMG (P1) and the other was the silent period between error EMG vanishment and correct EMG appearance (P2). 5) P1 clearly increased in proportion to the intensity on error reaction. However, P2 ranged between 40 and 120 ms, showed no clear relation to the intensity of error reaction. 6) Even when no error reaction appeared for the feinting stimulus, the reaction times were also lengthened as compared with the reaction for the stimulus immediately before the feinting stimulus. From these results; it can be concluded that when the error reaction was induced by false anticipation for the forthcoming stimulus, at least 100 to 200 ms is necessary for correcting the error command when the error appears as the external phenomenon and at least 50 ms is necessary even when the error does not appear as external phenomenon.