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



385 



should bo thus evaluated, since it was the least standardized type of examination of those inclu- 

 ded in the army series. Data were later obtained from other groups regarding the relationship 

 of each of these examinations with officers' estimates of their men (see pp. 425 ff), but it seemed 

 impracticable to attempt to secure these at this time. 



One point of special practical importance in connection with this relationship is the ques- 

 tion of the dividing line between D and E grades in beta. Because of the rule to recall E 

 beta cases for individual examination, it was important that cases recalled by beta should in- 

 clude those likely to be rated as very low grade by individual examination but not those likely 

 to be rated above the questionable range. It was also a matter of practical importance that 

 too large a number should not be recalled. 



Relationship between beta and alpha. — The correlation table 1 for beta and alpha, total 

 weighted scores, was plotted and the Pearson product-moment coefficient found to be 0.S06 ± 0.009. 

 Inspection of the table showed the relationship to be clearly curvilinear. The correlation ratio 

 which would represent the relationship more fairly, was not, however, computed for the present 

 data; later it was figured for the correlation between alpha and beta raw scores, where it was 

 found to be only slightly higher than r (see p. 392). 



Inspection of the table showed that the curvilinear relationship was partially due to "jam- 

 ming" of alpha cases at the lower end. There is evidently a marked tendency in alpha for the 

 piling up about the lower limit, of cases, many of whom would trail off toward a lower point if 

 the scale extended further down. This tendency shows more clearly for this group than for the 

 alpha groups ordinarily reported from the camps, because of the fact that we include here some 

 cases of the type which the camps have endeavored to exclude from alpha, even though we have 

 not included any non-English-speaking cases. The account of the development of segregation 

 methods (see p. 347) makes it clear that the purpose of segregation became largely the very 

 practical one of preventing call of a man for two examinations. Accordingly the attempt was to 

 exclude from alpha as many as possible of the men who could be expected to make such low scores 

 that they would need to be recalled for beta, whether this failure was due to illiteracy proper or 

 to low-grade intelligence. Such cases have not been thrown out from the present groups. 



Reference to the distribution showed also a marked tendency on the part of beta toward 

 massing cases at the upper part of the scale. This does not show actual jamming against the 

 perfect score, as does the alpha distribution against zero score; the piling up appears to be 

 accounted for, in part at least, by the large number making perfect scores in the separate tests. 

 (For distributions in the separate beta tests see table 52). This fact is important since it re- 

 sults in making beta a less satisfactory test than alpha for high-grade cases in spite of its supe- 

 riority to alpha for low-grade cases. 



For practical purposes the type of discrepancy between alpha and beta scores, noted above, 

 was not necessarily of great importance since assignment of letter grades might be made in 

 such a way as to correct for these differences. If given grades mean the same whether assigned 

 for alpha or for beta the main requirement has been met, since the grades and not the scores 

 were reported to army officials. The following figures show the percentages of the group of 

 653 English-speaking cases making various grades by the two scales. 2 



