Parallel-Processing Mechanisms and Processing of Organized Information in Human Memory
Psychology Press eBooks · 2014
Reaction time has been used as the main dependent variable for developing and testing many models in cognitive psychology. I argue that the statistic, mean reaction time, is almost totally inadequate for such purposes (though it has proved useful in getting some areas of research started). One problem is that certain models of the recognition process, serial scanning models, can be mimicked by several other kinds of models, for example, parallel-processing models and direct-access models (based on strength theory), at the level of mean reaction time. Thus it is difficult to press any claims as to the nature of the processes involved in the tasks under study. Another problem that has arisen is that many models that are quite adequate at the level of mean reaction time have serious problems with the shape of reaction time distributions. Models of processing should account for the overall shape of reaction time distributions, that is, produce distributions that are positively skewed. If such care is not taken, then it is possible to produce models that are falsified by the data that they were designed to fit (see Ratcliff & Murdock, 1976).