no. 2] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 317 



Table A. 



In our opinion it is doubtful whether any change is worth while beyond changing the multipliers for tests 4 and 5 

 from 3 and 2 to 2 and 1, respectively. It would probably aid the prophecy somewhat to use 4 and 3 instead of 3 and 

 2 for tests 1 and 2, respectively, but this means larger numbers to be added. The too high weight of test 9 can not be 

 reduced properly without adding much to the task of transmuting the scores. 



However, the tests all intercorrelate so closely that the revised weighting would not produce a much better result. 

 It might raise the correlation from 0.50 to 0.55 and lower the probable error from 0.15 to 0.14. 



Retests were made of 380 individuals for the purpose of determining the probable error 

 of the score. The facts are as follows : 



Regular Army 76 cases, P. E. = 1 1 points. 



Brooklyn group. 127 cases, P. E. = 19 points. 



Feeble-minded 40 cases, P. E, = 9 points. 



Delinquent women 50 cases, P. E. = 16 points. 



Graduate students 87 cases, P. E. = 6 points. 



Thorndike summarizes these results by saying: 



The P. E. of one 45-minute trial with the test seems to be about 15 points. The judgment from a man's score 

 should be thought of as fairly safely within the limits of 30 up and 30 down, and as almost certainly within the limits 

 of 60 up and 60 down. These determinations include the effects of changing interest, attitude, and the like. If a 

 man really does his best the P. E. will be much lower, say not over 10. 



The correlation between total score in the test and reported length of schooling was found 

 to be as shown in table 3. Only records bearing English or Irish names were used. Thorn- 

 dike comments as follows: 



The correlation (0.67) would be higher if perfect measures of educational advantages instead of mere reports of 

 grades were recorded, and if four or five tests instead of one had been used, probably about 0.75. Even so, it is clear 

 that the test does not measure simply amount of schooling. Since length of schooling is itself caused in part by inborn 

 ability, I think it fair to claim that a man's difference from his fellows in the test is determined as much by inherent 

 ability as by circumstances. This inherent ability is, however, obviously specialized for words, numbers, and other 

 abstract and symbolic content. Hence the need for tests of ability with things and people before decidine the fate 

 of a recruit. 



