FREE ASSOCIATION. 123 



CORRELATIONS BETWEEN THE VARIOUS MEASUREMENTS. 



Owing to the exceptional opportunity, certain measurements of 

 correlation were made, though they do not bear directly on the alcohol 

 question. First, with reference to whether those who have more usual 

 responses also have shorter or longer reaction times. This was the 

 only definite relationship observed, the Pearson r's 1 being in the four 

 experiments respectively 0.75, 0.50, 0.53, and 0.33. That is, 

 the person who gives usual responses seems also likely to have shorter 

 reaction-times, as is not difficult to understand. It does not follow 

 so pronouncedly that the usual reactions of a certain individual are 

 quicker than his unusual ones, though there is a tendency in this 

 direction. There appeared no significant correlation between frequency 

 and pulse-change, or between pulse-change and reaction-time. Even 

 when extreme cases only are considered, there was no special tendency 

 for the longer reaction-times to be accompanied by larger pulse-changes, 

 as table 18 shows. 



In table 18 we attempt to indicate the correlation of the pulse-change 

 and reaction- times, by comparing the average latency of the 5 reactions 

 that had the largest pulse-change with the average latency of the 5 

 reactions that had the least pulse-change. These values, together with 

 their mean variations, are entered under the appropriate legend for each 

 subject. A comparison of the average reactions is entered for each 

 subject in the extreme right-hand column as the average excess of 

 reaction-times with high pulse-change over those with low pulse-change. 

 The extreme irregularity of these results shows an absence of correlation 

 between pulse-change and reaction- time. This is interesting, in view of 

 the fact that both increased pulse-rate and longtime-reactions have been 

 suggested as indicators of the same thing, i. e., an emotional complex. 

 (Coriat. 2 ) No correlation between the pulse-change and the galvano- 

 meter readings taken in connection with the so-called psycho-galvanic 

 reflex was observed. 



Table 19 gives the relationships that were calculated between the 

 reaction-time and the frequency of response, between pulse-changes 

 and frequency, and between reaction-time and pulse-changes, series D'. 



1 Pearson's coefficient of correlation, r, was computed according to the familiar formula: 

 r = in which the x'a are the series of deviations from the median in the first group of data 



71 (Tl (72 



and the y'a are the deviations from the median in the second group; ffi is the standard deviation 

 of the first, ffz is the standard deviation of the second group ; and n is the number of cases. Zero 

 resulting from such a computation would show that the values in the two groups have no correla- 

 tion. Plus values in the coefficient of correlation show that the values are positively correlated ; 

 i. e., increase in one group is more or less regularly accompanied by increase in the other group. 

 Minus values in the coefficient show that the values are negatively correlated, i. c., that an 

 increase in the one is accompanied by a decrease in the other. Absolute positive or negative 

 correlation is indicated by +1 and 1 respectively. Our results indicate that in our experiments 

 there was considerable negative correlation between the duration of the reaction and the "fre- 

 quency" of the response; that is, the longer reaction-times tended to correspond with the less 

 common associates. 

 2 Coriat, Journ. Abnormal Psychol., 1909, 4, pp. 1 and 261. 



