no. 2.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES AEMY. 



315 



Table 1. — Distributions of scores for various groups. 



The score distributions above are such as would be expected if the tests were reasonably 

 valid measures of general intelligence. Officers and university students on the one hand, and 

 feeble-minded subjects on the other, make scores which deviate widely from the scores of 

 enlisted men. 



Another indication of the validity of the tests is the correlations between scores and officers' 

 ratings of their men for intelligence. Such ratings were secured by E. K. Strong for 313 men 

 of three national guard companies. The correlations of these ratings with the separate tests 

 were as follows: 



Correlation 

 with offi- 

 cers 'rating. 



Probable 

 error. 



Test. 



Correlation 

 with offi- 

 cers'rating. 



Probable 

 error. 



1. Oral directions 



2. Memory span 



3. Disarranged sentences. 



4. Arithmetical problems 



5. Information 



0.47 

 .34 

 .48 

 .46 

 .54 



0.019 

 .032 

 .018 

 .024 

 .036 



6. Synonym— antonym 



7. Practical judgment ... 



8. Number series completion 



9. Analogies 



10. Number comparison 



0.51 

 .41 

 .42 

 .36 



.47 



0.017 

 .029 

 .013 

 .048 

 .016 



For single brief tests (time limits, 1 to 5 minutes) the above correlations must be regarded as 

 very satisfactory. The correlations are fairly uniform except for the appreciably lower coeffi- 

 cients for tests 2 and 9, and the relatively high coefficients for tests 5 and 6. Correlations between 

 officers' ratings and total score in the ten tests were obtained for groups in each of the four 

 stations. The correlations were as follows: 1 Syracuse, 218 cases, r=.5; Brooklyn, 168 cases, 

 r=.2; Indiana (excluding officers' training camp), 634 cases, r=.3; Nashville, 770 cases, 

 r=.65; officers' training camp, 113 cases, r=.15. 



The methods by which the officers' ratings were made, their reliability as indicated by 

 correlations between ratings of the same men by two or more officers, and the tabular data 

 from which the correlations were derived were not reported by the statistical unit. Thorndike 

 states, relative to these data: 



For enlisted men the general tendency is thus to a correlation of 0.5. I should think it safe to assume that the score 

 in the group-test when given twice, once with short and once with abundant time, would correlate with an omniscient 



1 This statement calls for the following comment: The coefficient of correlation between ratings and tests depends upon the heterogeneity of 

 the group in question. The low correlation for officers' training camp groups is due largely to the fact that such groups are highly selected. 

 The correlations between ratings and tests are low for any narrow-range group, whether superior, average, or inferior. 



