no. 2.] PSYCHOLOGICAL EXAMINING IN THE UNITED STATES ARMY. 399 



iner in giving the tests. These instructions, in mimeographed form, were sent to the camps 

 in March and were later included in the revised Examiner's Guide. (See Parti, pp. 167ff.) 



The point scale and Stanford-Binet gave general satisfaction for American-born subjects, 

 except in cases of absolute illiteracy. Some examiners even preferred to test illiterate subjects 

 by these scales, omitting those tests which presuppose a knowledge of reading and writing. 

 The non-English-speaking were, of course, always tested by the performance scale. Apart 

 from this restriction the examiner was free to use the three scales according to his preferences. 



All were used as mental-age scales in the sense that a mental age was always determined. 

 For the point scale the Yerkes-Bridges mental-age norms were used, instead of their Stanford- 

 Binet equivalents. The performance-scale scores, however, were calibrated in terms of Stan- 

 ford-Binet by equating equal percentile ranges. 



It was planned that the point scale should be used without abbreviation, but that in the 

 case of the Stanford-Binet only four tests per year group should ordinarily be given. The 

 Stanford-Binet tests recommended are indicated by asterisks in the record blank on pages 271 to 

 274 of Part I. These are not always the same as the starred tests of the record blanks supplied 

 by the Houghton-Mifflin Co. The selection for Army use was made with reference to diagnostic 

 value, suitableness for illiterates, and economy of time. More than four tests per year were 

 almost never given in the Army, and when Stanford-Binet mental ages are mentioned it should 

 be borne in mind that these always refer either to the four test per year abbreviation or to 

 another abbreviation, still more radical, to be described later. 



The reliability of the abbreviated Stanford-Binet had been determined in pre-war work at 

 Stanford University with several groups of subjects. For a group of several hundred male 

 adults, including "hoboes," prisoners, and reform-school inmates, mental age based on three 

 tests per year (four tests in year 12) correlated 0.96 with the mental ages of the same subjects 

 based on the entire scale. The mental age range of this group corresponded closely to that found 

 for soldiers, except that it had a somewhat larger proportion of low-grade cases. The corre- 

 sponding correlation (between the two halves of the scale) for several groups of public-school 

 children varied from 0.93 to 0.97. It is evident, therefore, that the reduction of the scale from 

 six to four tests per year involves but slight loss of accuracy. An index of reliability based 

 upon repeated tests had also been determined for the Stanford-Binet in pre-war investigations. 

 Tests repeated within a few days or a few months usually show a correlation, in the case of 

 fairly wide-range groups of school children, of 0.95 or better. The correlation remains con- 

 siderably above 0.90 even when the earlier and later tests are separated by a period of three 

 or four years. 



Correlations of repeated tests with the point scale are not available, but there is no reason 

 to suppose the results would differ greatly from those furnished by the Stanford-Binet, par- 

 ticularly in view of the fact that the two scales are composed largely of the same tests. The 

 reliability of point-scale measurements is indicated by the high correlation of score on the. 

 alternate tests with score in the entire scale— namely, 0.95 or better for fairly wide-range groups. 

 That which either scale measures is measured by it fairly reliably, according to present-day 

 standards of reliability in mental measurement. 



The reliability of the point scale, however, rapidly decreases in the range above 80 points — 

 that is, above a mental age of 13 to 14 years. The Stanford-Binet fails to differentiate satis- 

 factorily above the mental-age score of 17 or 18 years. In the absence of any appreciable 

 amount of Army data based on examination of the same individuals by both of these scales, 

 reference was made to the results obtained at the Laboratory of Social Hygiene on 259 delin- 

 quent women. 1 For this group the correlation between Yerkes-Bridges score (in points) and 

 Stanford-Binet intelligence quotient was 0.872 (Pearson). However, the relationship was 

 considerably curvilinear. The correlation ratios were found to be 0.894 and 0.904. In the 

 table of equivalent scores (Examiner's Guide) point scale and Stanford-Binet scores are equated 

 in terms of percentile ranks on the basis of these data. 



1 The correlation table appears in a report of findings now being prepared for publication by the Laboratory of Social Hygiene, Bedford Hills, 

 N. Y. 



