We examined the reaction time benefit that is obtained when salient features of the stimulus set and response set correspond. Components of the event-related brain potentials were used to measure the timing of stimulus-related and response-related processes in order to determine the locus of this effect. Of particular importance was the development of a new index of selective response preparation, the corrected motor asymmetry (CMA). We found no evidence for the use of preliminary, partial stimulus information in response preparation. These results suggest that the benefit is located primarily in response selection processes and probably reflects a more efficient algorithm for stimulus-response translation. Also, we found trial-to-trial variability in the duration of response selection to be the major determinant of variability in reaction time, whereas the durations of subsequent response-related processes were relatively invariant. Implications of these results for discrete and continuous models of choice reaction performance are discussed.