434 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



[Vol. XV, 



coefficients to the same degree of individual variability for each group, we have, as an extreme 

 case, the following contrasted results: 



If all the ratings of the men in each group had been made by one officer, the conclusion 

 would be justifiable that the two officers had taken into account different sets of personal 

 characteristics of the men they were rating, or, what is nearly the same thing, that they differed 

 appreciably in their notions of what constitutes "intelligence." Since these two groups are 

 actually made up of smaller groups rated by different officers, so that ratings from 28 different 

 officers are thrown together in the Beauregard-MacArthur material, and at least three sets of 

 ratings are combined in the alpha before a group, smaller groups, the members of which were 

 all rated by one officer were studied to determine to what extent the divergent results from 

 mixtures of different sets of ratings may be regarded as due to qualitative differences of ratings. 



In the MacArthur material one group of 154 cases was found in which each had been rated 

 independently by three different officers. The results obtained from this material are inter- 

 esting in several respects. 



The contingency of rating and score for each officer separately is shown in tables 96, 97, 

 and 98. 



Table 96. — Company H, Fifth Battalion, Infantry Replacement Camp, MacArthur — Ratings by officer 1. — Alpha raw total. 



It is evident that, although Table 96 might be treated by ordinary product-moment 

 methods, Table 98 can not, owing to the small number of rows, and to the concentration of over 

 80 per cent of the total frequency into one row. Table 97 is intermediate in these respects. 



