386 MEMOIRS NATIONAL ACADEMY OF SCIENCES. [Vol.sv, 



Discrepancies in proportions given E grades by the two scales may be disregarded, since 

 alpha was understood to be unreliable for this portion of the scale. The table shows, however, 

 that there is marked discrepancy throughout the range, and that, except for grade A, the dis- 

 crepancy is always in the direction of greater leniency for beta. Summarizing the data, it 

 appears that 15.3 per cent make A or B on beta as compared with 11.6 per cent on alpha; that 

 68.4 per cent make C or better on beta as compared with 43.7 per cent on alpha; and that 78.6 

 per cent make C — or better on beta as compared with 53.3 per cent on alpha. The outstanding 

 fact from this evidence is that grades on alpha and on beta were far from meaning the same 

 thing at this time, and that the probability was that a man rated by beta would get a higher 

 standing than one rated by alpha. Information received regarding regular draft quotas in- 

 dicated that there was error in both scales; that alpha was too severe and beta too lenient. 

 Alterations to accomplish a double adjustment — made in August in connection with additional 

 changes in beta — will be discussed at a later point. A readjustment of alpha grades, made in 

 July, reduced the degree of discrepancy slightly. Figures are not presented for this com- 

 parison, since the August changes represent the final ruling on grade standards. 



Comparison of beta with Stanford-Binet. — The fact has been noted that beta apparently 

 makes distinctions within a part of the scale which alpha entirely fails to reach. It is evident, 

 therefore, that the adequacy of these distinctions must be judged by reference to some measure 

 other than alpha. Comparison with Stanford-Binet is therefore of special importance for this 

 portion of the scale, which is, for practical purposes, the part of beta which it is most important 

 to have reliable. 



The Pearson product-moment coefficient for the correlation between beta weighted score 

 and Stanford-Binet mental age was found to be 0.731 ±0.012. (It may be noted that correla- 

 tions figured for four of the separate camp groups were found to be as follows : Dix, 0.740 ± 0.032 ; 

 Jackson, 0.646±0.039: Lee, 0. 766±0. 027; Meade, 0. 659±0.040.) It is evident that these 

 coefficients are not high enough to justify anticipation of a high degree of reliability of pre- 

 diction. The table 1 shows a large amount of scatter at all points. In so far, therefore, as the 

 intention was to secure, in beta, an instrument of group measurement which should produce 

 results closely similar to those which the individual examination would have given, with great 

 reduction in time, the results are disappointing. 



The statement regarding the lack of certainty of prediction has special importance for 

 its bearing on the dividing line between D and E in beta. Recognizing that an individual exam- 

 ination — in case of English-speaking men, either the Stanford-Binet or the Yerkes-Bridges 

 point scale — furnished the main basis of decision with reference to low-grade cases, the following 

 assumptions seemed justified: 



1 . It is important to recall all cases who would test under 8 years mental age if given indi- 

 vidual examination. 



2. It is desirable to recall the majority of cases who would test between 8 and 9 years. 



3. It is desirable, if pressure of camp conditions be not too great, to recall at least a small 

 number of cases who would test between 9 and 10 years. 



4. It is desirable that few or no cases testing over 10 years be recalled. 



In the present group of 653 cases there are 4 testing under 8 years, 22 testing between 8 and 

 9, 62 testing between 9 and 10, and 565 testing over 10 years. The dividing line of 40 (beta 

 weighted score) recalls the following numbers: 



Below 8 years, 3 casea (75 per cent of all cases below 8 years). 



8 to 9 years, 10 cases (45. 5 per cent of all cases between 8 and 9). 



9 to 10 years, 9 cases (14.5 per cent of all cases between 9 and 10). 

 Above 10 years, 14 cases (2.5 per cent of all cases above 10 years). 



Total number recalled, 36 cases (5.5 per cent of total English-speaking group). 



i This table is not presented, because of its close similarity to table 52, which shows the correlation of beta raw score with Stanford-Binet 

 mental age (q. v.). 



