no. 3.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 



625 



curve. But allowing for the probable distribution of cases in the terminal class intervals (i. e., 

 the distribution of these cases had they been as adequately measured as were all other cases), 

 it is apparent that at least a moderately good fit by some type of Pearson curve 1 could be 

 obtained for all distributions except that for beta 3. Such frequency curves, however, would 

 be descriptive not of the distribution of class marks but of the distribution of the variate back 

 of the class mark, i. e., of the distribution of intelligence as measured by different methods and 



400-, 



550- 



500- 



250- 



200- 



150- 



100 



50- 



250 - 



200- 



1.50- 



100- 



50- 



TE.ST 4, 

 ALPHA 



TE5T 7. 

 ALPHA 



lol I I I liol I I I feoi I I I bd I I i 



° l i ' I I ho 1 ' ' I W ' ' ' So 1 I ' ' W 



Fig. 3b. Histograms of scores in alpha tests. Special experimental Group X. 1,047 cases. Horizontal scale is score in test; vertical scale is number 



of cases. 



quantitatively interpreted in different units. It is again evident that a distribution of total 

 scores when such scores are merely sums of class marks can not be descriptive of the distribu- 

 tion of intelligence as measured by a linear, homogeneous scale. 



Inspection of the histograms of the alpha tests (fig. 3) indicates how much the various 

 tests differ in the length of range of ability measured. Tests 1 and 2 appear to be almost negli- 



1 Pearson, K., Skew Variation in Homogeneous Material, Phil. Trans., Series A, vol. 186, 1895, pp. 253fl. 



