1. It has been suggested that the long-latency "event-related" cerebral evoked potentials (ERPs) reflect certain aspects of the neural processing underlying sensory discrimination in a two-choice reaction time task. The present paper examines the hypothesis that the coupling of these ERPs to sensory discrimination is variable and that the discrimination process is completed at different points during the course of cerebral processing, depending on the actual requirements of the task. 2. We recorded the cerebral evoked potentials and electromyogram (EMG) of the responding muscle in five different reaction time tasks, each requiring sensory discrimination and response selection of varying complexity. In the Choice condition two stimuli were presented, and two separate responses were required. In the two Go-No Go conditions two stimuli were presented, but a response was required to only one or the other of the stimuli. In the two Simple conditions only one stimulus was presented, and one response was required. 3. Under both Choice and Go-No Go conditions, the frequency histogram of the onset latency of the compound muscle action potential for the response to the frequent tone showed a bimodal distribution without overlap, suggesting that there were two distinct types of responder: fast and slow. The comparable histograms for the onset latency of the response to the rare tone also showed a bimodal distribution, but the mean onset latency was prolonged relative to the response to the frequent tone, and the mean separation was less so that the two distributions overlapped each other. 4. Despite the marked difference in response latencies between the fast and slow responders, there was no appreciable difference in cerebral evoked responses between the two groups. Moreover, in response to the frequent tone, all slow responders and, likewise, all fast responders had similar onset latencies of the averaged EMG activity regardless of condition. Nonetheless, fast or slow responders to the frequent tone on one block of trials were also fast or slow responders, respectively, to the rare tone on the same block of trials. 5. These results suggest that two separate sensory discrimination processes are occurring; the first relating to the presence or absence of the frequent (expected) stimulus, and the second relating to the presence or absence of the rare stimulus. The response to each tone can either be generated immediately after that tone is positively identified or, when accuracy is required, can be delayed until both stimuli have been successfully discriminated by the subject.