Additivity of the effects of experimental manipulations on mean reaction time (RT) has been taken to suggest that the underlying mechanism can be divided into independently changeable, serially arranged operations (Stage Model, additive-factor method; Sternberg, 1969). We consider two other explanations of the same additive pattern of means: a model with independently changeable alternate pathways, and the McClelland-Ashby Cascade Model. In all three models, additivity results from selectivity of the influence of experimental factors on mental operations. Among the tests we develop are comparisons of entire RT distributions. Applied to the results of four diverse experiments (overlapping tasks, classification, identification, and detection), analyses of distributions and variances support the Stage Model and contradict the Cascade and Alternate-Pathways models. One set of distributional analyses, based on a suggestion by Ashby and Townsend (1980), supports the Stage Model remarkably well.
Table of Contents