no. 2.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 427 



Table 81. — Individual officers' estimates. 



' Not rated. 



Table 81 gives the distributions of ratings of the cases studied, for each individual officer. 

 It is at once apparent that we are dealing with a great diversity of conceptions of intelligence. 

 Some officers have been very conservative in assigning very high or very low ratings. One 

 officer rates nearly all of a group average or above and another rates the same men all average 

 or below. One offioer rates 85 per oent of a large group 4 or average. Another officer rates 

 only 34 per cent of the same group 4. It is therefore apparent that officers differ: 



(a) In their notions of the length of range of intelligence represented by each class. 



(6) In their notions of what constitutes "average" intelligence. 



(c) With respect to the part of the range of intelligence in which they discriminate, or 

 attempt to discriminate, different grades of intelligence. 



All of the oases were examined by alpha and beta, approximately hah of them taking alpha 

 first and the other half taking beta first. In working out correlations of tests with ratings the 

 order of the examinations has been disregarded and all cases treated together, since no effect 

 upon second examination scores of the experience of a first examination of either kind was 

 evident. Correlations were first computed for total beta and alpha scores, and scores on sepa- 

 rate beta and alpha tests with combined officers' ratings. The combination of ratings was 

 effected as follows: For the great majority of oases having three ratings, the three numerical 

 values were simply added together. Cases having only two ratings were given a combined 

 rating obtained by multiplying by 3 half the sum of the two ratings, and in the few oases 

 where only one rating was available, its numerical value was multiplied by 3. Thus the scale 

 of possible ratings extends from 3 to 21. 



For all correlations of total scores with ratings, beta scores have been classified in 5-point 

 intervals and alpha scores in 10-point intervals. All standard deviations given at the bottom 

 of correlation tables are in terms of class intervals. The standard deviation of ratings is 

 denoted by the symbol <r R and the standard deviation of scores by the symbol <r s0 . 



Tables 82 to 91 are a series of typical frequency surfaces for ratings and total (raw) alpha 



or beta scores. 



Table 82. — Alpharaw total — alpha-bef ore-beta (Meade). 



s B =1.29162 

 s BC =4. 83358 

 r=0.671 ±0.023 



