Abstract
Several recent studies have indicated the characteristics of error and error correcting reaction times in choice reaction tasks involving ungraded keyboard tasks. At the same time, models of error detection and correction have been proposed, none of which can exclusively account for all the results. In this paper a detailed analysis was made of errors and their correction which were found in a discrete tracking task where the movements were graded and performed with one anatomical unit. Records were obtained of displacement, acceleration and muscle activity underlying the responses. In common with the keyboard results, the errors were typified by relatively short reaction times. They were corrected in the absence of any specific error feedback and the stimulus to which the response should originally have been made. As a result of the sensitivity for the acceleration measures, so-called error correcting reaction times were observed which were considerably shorter than those previously reported. A mean value of 64 msec was obtained. This indicates that under these conditions monitoring of performance, most probably by central feedback loops, can be achieved with much delay.