The abnormality of the difference as a method for evaluating the magnitude of differences between pairs of Wechsler Intelligence Scale for Children-Revised subtests is discussed. Generally, abnormal differences at the .05 level of significance range from 6 to 7 scaled score points and 8 to 10 scaled score points at the .01 level. Abnormal Verbal-Performance scale IQ differences are also considered. Such differences averaged 18 IQ points at the .05 level of significance and 24 IQ points at the .01 level. The diagnostic implications of the use of the abnormality of the difference for evaluating subtest score differences are discussed.

No abstract

The reliability of subtest composites corresponding to the 46 abilities and influences presented by Kaufman (1979) was examined for the WISC-III using standardization sample data at ages 6, 11, and 16. Recategorized subtests demonstrated consistently high reliability. Increasing composite reliability was associated with the number of member subtests and the inclusion of Verbal Scale subtests in a composite. Some age fluctuations in composite reliability were noted. A table containing composite reliability for the entire standardization sample and at three age levels is provided for the clinician's reference. Partial psychometric support for use of these composites in WISC-III profile analysis was demonstrated.

This article discusses reactions by Reynolds and Silverstein to an earlier article by Piotrowski that examined the size of abnormal differences between Wechsler Intelligence Scale for Children-Revised subtest scaled scores and Verbal-Performance IQ scores. An error is acknowledged in the presentation of the Verbal-Performance IQ differences, and the validity of the abnormality of the difference statistic for examining subtest scaled scores is discussed.

Burns' (1984) attempt to explain the below average mean IQ scores reported for samples of learning disabled (LD) students through the use of the bivariate normal probability distribution was examined. Burns applied the bivariate distribution to a method of quantifying a learning disability based on a fixed achievement cut-off value. This explanation seems to have limited utility when applied to the preferred discrepancy based model of defining severe educational underachievement. Alternative explanations for the below average IQ of LD samples which focus on issues related to the referral, identification, and placement of children are presented.