Diagnostic Validity of Wechsler Subtest Scatter

It has long been speculated that cognitive subtest scatter or variability is a pathognomonic sign. Early researchers hypothesized that subtest scatter would predict scholastic potential or membership in exceptional groups (Harris & Shakow, 1937). In fact, Binet suggested that scattered passes and failures on the Binet-Simon scale were diagnostically significant (Matarazzo, 1985), and Wechsler (1941) conjectured that subtests discrepant from the mean of the Wechsler Bellevue scale might be useful in differential diagnosis. Similar propositions have been proffered for each successive revision of the Wechsler and Stanford-Binet scales (Watkins, 2003). Across time, subtest scatter has been quantified in four ways. The first method was the range (i.e., the difference between an examinee’s highest and lowest subtest scaled scores). Variance, the second quantification of subtest scatter, was thought to be more useful than range because it utilizes information from all subtests, rather than just the highest and lowest (Wilcox, 1996). Labeled the profile variability index (PVI) by Plake, Reynolds, and Gutkin (1981), this scatter index is computed by applying the sample variance formula to the subtest scores of an individual examinee. Third, Mahalanobis distance (D2), a multivariate method, has been recommended for its statistical superiority over univariate measures such as range and variance (Huba, 1985). Given that subtest scores are correlated, overall subtest scatter will be lower than if the subtests were uncorrelated (Burgess, 1991). This is corrected by D2, which takes into account the means and variances of each subtest and the correlations and covariance between subtests. Finally, researchers have looked at the number of