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



321 



extent as to create grave prejudice among commanding officers against the work. Besides, it 

 was the general opinion of the psychological examiners that the benefits derived from the 

 repeated examination were small in comparison with the time and labor involved. At Camp 

 Devens only men who had made a score of less than 70 on examination a were recalled for 

 examination b. Those who made a total weighted score on examination b of 70 (tests 1 and 2 

 omitted) were passed without individual examination. This saved nearly half of the individual 

 examinations, but whether it furnished a more reliable measure of the subject's intelligence 

 than examination a is doubtful. 



The probable error of a score secured by giving the group examination twice on different 

 days and averaging the results is undoubtedly somewhat less than the probable error of a score 

 resulting from a single test. The latter is not far from 10 points, weighted score, for examina- 

 tion a. This is only about one-eighth of the standard deviation of the distribution of scores for 

 unselected literate men. Repeating the test might conceivably reduce the probable error to 

 one-ninth or one-tenth of the standard deviation, but the practical gain for Army purposes 

 would not be worth the time and labor required. 



As a means of favoring the exceptionally slow, repetition of the examination with extended 

 time was wholly disappointing. (See p. 480f.) An investigation at Camp Lee showed 

 that, while on the whole there was a slight gain in score with extended time, the relative stand- 

 ing of a man was not often materially altered. For example, examination a given twice to a 

 group of 51 unselected men gave a correlation of 0.943. The correlation of examination a 

 with examination b for 380 unselected men was but little less — namely, 0.913. The data from 

 more than 500 men in seven companies who took both the a and b examinations indicated that 

 the relation of a man's score in b to his score in a was more influenced by the use of different 

 "forms" in the two examinations than by the difference in time limits. The facts are as 

 follows : 



Groups given examinations a and 6. 



Form used 

 in examin- 

 ation a. 



Form used 

 in examin- 

 ation 6. 



Average gain or loss. 



Three groups, 203 men 

 Two groups, 200 men. . 



One group, 63 men 



Onegroup 



5 points, gain. 



For one group, 1 point gain; lor one group, 1 point loss . 



24 points, gam. 



11.5 points, loss. 



It is shown elsewhere (p. 338) that an extension of time sufficient to permit 50 per cent of 

 the men to complete each test (instead of the usual 5 or 10 per cent) does not decrease appre- 

 ciably the percentage of zero scores for most of the tests ; also that doubling the time allotment 

 does not demonstrably improve the correlation of the tests with officers' ratings of men or with 

 Binet mental age in the case of school children (p. 339). The expectation that many men of 

 fair intelligence would make low scores in examination a because of slow mental processes 

 looked reasonable enough, but it is not borne out by the facts. 



Section 3. — Group test for illiterates — Stenquist. 



The Stenquist test was given to 14,610 men during the initial experiment in four canton- 

 ments. It was far from satisfactory, but it was regularly used because nothing else was avail- 

 able for the group examination of illiterates and foreigners. The chief objection to it was its 

 low value as a measure of intelligence. Even with unselected literate men it correlated with 

 examination a only to the extent of 0.45 to 0.55. Its correlations with officers' ratings of their 

 men were very much lower than this, usually not more than 0.20 or 0.30. The letter grades 

 which it yielded, therefore, were not comparable with those based upon examination a. The 

 following data bearing on the validity of the Stenquist test were received from the camps: 



Camp Taylor, 109 men, unselected group, r for examination a and Stenquist, 0.323. 



