Sequential processes and the shapes of reaction time distributions

Abstract

It is sometimes suggested that reaction time (RT) distributions have the same shape across conditions or groups. In this note we show that this is highly unlikely if the RT is the sum of the stochastically independent durations of 2 or more stages (sequential processes) (a) that are influenced selectively by different factors, or (b) 1 of which is influenced selectively by some factor. We provide an example of substantial shape differences in RT data from a flash-detection experiment, data that have been shown to satisfy requirement (a). Ignoring these requirements, we also note that in a large range of instances reviewed by Matzke and Wagenmakers (2009) in which the ex-Gaussian distribution was fitted to RT data from different conditions in the same experiment, most sets of distributions fail to satisfy even a weak requirement for shape invariance. In the Appendix we describe the Summation Test for selectively influenced stages with independent durations (Roberts & Sternberg, 1993), and provide an example of its application. (PsycINFO Database Record