Two Dimensions for Reaction Time Distributions

Abstract

THE question of the ‘best’ statistic to use in describing the results of experiments on reaction time has received little attention, considering the long history of these experiments. Most research workers accept the arithmetic mean reaction time as the most useful and valid measure of performance, especially as its sampling characteristics are so easily calculated. The median, the mode, and the geometric mean have sometimes been used as substitutes, but the common object has been to obtain a measure of ‘average’ performance. In the many papers which followed Hick's¹ application of information theory to choice reaction times, there are few references to any other properties of reaction time data but the means. More recently, experiments in which the total time for execution of a repetitive self-paced task (such as card-sorting) is the only measure taken have become a popular, even a standard, method of investigation in this field.