The intercorrelations among the Alpha and Beta subtests that were obtained for the sample of men who first took the Alpha and then were referred for Beta testing ranged from a low of .36 for Beta test 2 (Cube Analysis) with Alpha test 3 (Practical Judgment) to a high of .68 for Beta test 4 (Digit Symbol) with Alpha test 2 (Arithmetical Problems) (Yerkes, 1921, p. 634, Table 155). These are quite a bit lower than the correlations in the range of .59 to .86 for intercorrelations among the Alpha tests given above.

The correlation of Beta total scores with schooling ranged from .45 for a sample of over 11,000 native born men with education levels ranging from none to college, to .67 for a sample of 653 native-loom draftees (Yerkes, 1921, p. 781, tables 327, 328). For a sample of 5,803 foreign born the correlation of Beta total score with schooling was .50. In general, then, the correlations of Beta scores with years of schooling were lower than the correlations of Alpha scores and education (.75) when the full range of Alpha test takers (including those subsequently sent for Beta testing) was included.

When the Alpha and Beta test total scores (excluding non-English speakers) were correlated with mental age scores on the Stanford-Binet individually administered intelligence test, the resulting coefficients were .81 and .73, respectively. Since the StanfordBinet is essentially an auding test, in which the administrator speaks of questions and the given information, it is perhaps to be expected that the correlation between the heavily language-laden Alpha and Stanford-Binet tests would be greater than the very low language-based Beta test with the Stanford- Binet. For a sample of 653 recruits, the correlation of Stanford-Binet with years of schooling was .65 (Yerkes, 1921, p. 782, Table 330).

Reference

Robert M. Yerkes (1921). Psychological Examining in the United States Army. Memoirs of the National Academy of Sciences, Vol. XV. Washington, DC: U.S. Government Printing Office.