Reaction Time — Movement Time Correlations

Pierson and Rasch (8) have questioned the interpretation made by this writer and others to the effect that correlations of approximately zero between R T and M T scores signify that individual differences in R T and M T are independent and unrelated. Their position is that a correlation of zero indicates zero relationship only in the special case of a normal bivariate population, and that it has not been demonstrated that the RT-MT bivariate population is of this type. It is indeed true that normality is one of the assumptions chat become important in certain applications of r, as for example exact estimates using surface-volume tables. It should also be mentioned that peculiarities of distribution sometimes result in an r that overestimates the amount of relationship between two variates. The present issue, however, concerns the interpretation of r = 0. Rigorous mathematical derivations require assumptions that offer fuel for attacking almost any interpretation of r that is other than operational. On the other hand, experience in actual application reveals that I. is remarkably robust with respect to skewness in sample scores when the correlations are small and/or skewness is similar in both variables. In such cases, it is seldom profitable to bother with normalizing uansformations. Linearity of regression is usually a more important consideration, but departures from linearity need cause little concern when correlations are small, unless the regression is distinctly U-shaped. Nevertheless, a bivariate RT-MT sample is herewith presented and analyzed. Subsequent to Pierson's report (7 ) of substantial correlations between R T and MT, we have routinely measured these two variables in a variecy of experiments primarily concerned with other problems. A typical set of data is given in Table 1. S stood erect with his laterally extended right arm resting on the reaction key. In response to an auditory stimulus, he swung the arm forward 90' at maximal speed to pass through a vertical "pull-out" target suing placed an arm's length to his front; the distance moved was 117 un. for all Ss. The arm was kept straight with the elbow downward to avoid flexion. The experimental details were described recently in a study using 30 Ss with 40 trials per S. In that report the RT-MT correlation was .06 ( 2 ) . A new sample of 120 undergraduate college men were given 12 uials; the first 4 have been discarded and the remaining 8 averaged for each S s score. It is well known that the distributions of single trial scores of both R T and