Simple reaction time has been measured using various distributions of interstimulus intervals (ISIs), which differed both in the shape of their envelope and in their degree of discretization. For each of 7 such conditions, and for three mean ISIs (2.0, 3.3, and 5.8s) 600 responses of two subjects have been accumulated.Reaction times depend significantly on the shape of ISIdistributions but also on their degree of discretization. Within an experimental run they depend on single ISIs back to the third before reaction, and on sequences of ISIs. The latter effects are again influenced by the discretization of ISIs. Finally it was found that some learning of the distribution pattern (not the mere mean value) of ISIs takes place.Our results shed some light on existing hypotheses of RT. Some inconsistencies of earlier experimental results can probably be explained by insufficient consideration of the discretization and the learning effects.