To examine mechanisms of response activation, we asked subjects to respond differentially to the central letter of one of four arrays--HHHHH, SSHSS, SSSSS, and HHSHH--and measured event-related brain potentials (ERPs) and electromyographic activity (EMG). For very fast responses, accuracy was at chance level for all arrays, suggesting that subjects were guessing. For intermediate latency responses, accuracy was above chance if the noise was compatible with the targets and below chance if it was incompatible, suggesting that these responses were based on partial stimulus analysis. For slow responses, accuracy was above chance for all arrays, suggesting that these responses were based on complete stimulus analysis. The occurrence and accuracy of fast responses could be predicted by examining motor potentials preceding the presentation of the array. Measures of the motor potentials in the period following the presentation of the array suggested that partial analysis of stimulus information could activate responses and that the level of response activation at the time of the EMG response was constant for trials with different response latencies. The data are discussed in terms of a response channel conception.