The stimulus evaluation view on P3 latency holds that P3 latency mainly reflects stimulus-processing time, in contrast to response-processing time. A review of the experimental evidence, however, leads to the conclusion that P3 is not a sensitive tool for separating between stimulus- and response-related processes. Rather, it appears that P3 latency is a sensitive index of any response-time changes when response times in the fast condition are brief, with its sensitivity decreasing when response times in the fast condition get longer. This regularity was confirmed by a detailed analysis of the published evidence from Sternberg's task and was not attributable to speed-accuracy trade-off or to different methods of parametrization. The structures generating the scalp P3b are involved both in stimulus processing and in response selection. Response selection may exert its effect on P3 in one of two ways; either directly, fully delaying P3 latency, or affecting a second P3 component (P-CR) only, thus having an attenuated effect on P3 latency.