342 



MEMOIRS NATIONAL ACADEMY OF SCIENCES. 



[Vol. XV, 



It was agreed by the committee that test 2 and test 10 both had their chief value in the 

 lowest ranges of intelligence. The plans for the new beta examination were expected to 

 provide for those below fourth-grade literacy and probably for all earning grades below C in 

 the alpha examination. For this reason it was not considered as important that the alpha exam- 

 ination should give accurate ratings in the lower ranges as that it should differentiate well in 

 the higher ranges. The most serious faults of examination a were in the higher ranges. Its 

 revision was guided by the belief that the new scale should allow the best individuals to dis- 

 tinguish themselves from those of moderate ability. 



It will be seen that the changes as described above were in the main not radical, except 

 for the elimination of test 2 and test 10. The time limits were not greatly altered. While the 

 time for test 8 was increased from 2 minutes to 3 minutes, and that of test 7 from 1 minute to 

 1-J- minutes, both of these tests were lengthened almost enough to offset the increase of time. 

 The time of test 5 was increased from 3 minutes to 4 minutes partly because the test was made 

 more difficult and partly in order to leave the subjects with the the feeling that they had not 

 been too much hurried. (This test was changed from fifth in the series to last.) The new 

 order of the tests was as follows: Test 1, following directions; test 2, arithmetical reasoning; 

 test 3, practical judgment; test 4, synonym-antonym; test 5, disarranged sentences; test 6, 

 number series completion; test 7, analogies; test 8, information. 



Section 13. — The alpha examination. 



After revision, examination a was called alpha to distinguish it alike from examination a 

 and from examination beta for illiterates. It was put into operation immediately and was 

 retained until the close of the war as the group examination for literates. Between April 1 and 

 December 1, 1918, it was given to approximately 1,250,000 men. 



It was decided to weight the alpha tests according to the variability of their scores, and, 

 if possible, in such a way as to make the total score comparable to that of examination a. The 

 measure of variability used was the interquartile range. 



Table 23 shows for two groups of unselected men and one group of officers the interquartile 

 range for each test, the weights adopted, and the effects of the weighting. 



Table 23. — Interquartile range of tests at three camps. 



After weighting, the variabilities of the tests are fairly well equalized except for test 4 

 (synonym-antonym), which is properly given extra value. The total score is 412, as compared 

 with 414 of examination a. 



The only changes later made in the alpha examination were the elimination or alteration 

 of unsatisfactory items and the elimination of weighting. The latter was provided for by an 

 order to the camps. The correlation of 0.993 between raw and weighted total (for 2,856 unse- 

 lected men), together with the large amount of time required for weighting the scores, unques- 

 tionably justified the substitution of raw for weighted scores. The various bases on which 

 letter ratings were assigned are set forth elsewhere (p. 421 ff). Values of the coefficient of corre- 

 lation between raw and weighted scores for examination a and for alpha are shown below. 

 Table 24 gives a scatter diagram between alpha raw and alpha weighted. 



