Learning Disabilities Practice

Regrouping Wechsler Intelligence Scale for Children-Third Edition (WISC-III) subtests into Bannatyne’s spatial, conceptual, and sequential patterns has been thought by many to identify children with learning disabilities (LD). This study investigated the prevalence and diagnostic utility of WISC-III Bannatyne patterns by comparing 1,302 children with LD to 2,158 children in the WISC-III normative sample. Further analysis was conducted on a subsample of students with specific reading disabilities. Results indicated that the presence of the Bannatyne WISC-III pattern would not lead to decisions that are useful in differentiating children with LD from children without LD. For example, receiver operating characteristic (ROC) analysis, measured by the area under the curve (AUC), indicated that the Bannatyne WISC-III pattern exhibited low diagnostic utility (AUC = 0.54–0.55). Due to its inaccuracy, use of the Bannatyne WISC-III pattern is not recommended. More than 250 million standardized tests are administered to public school children each year in the United States (Salvia & Ysseldyke, 1998). Although much school-based testing is accomplished in groups, a substantial number of standardized tests are also employed in individual evaluations. For example, millions of children served in special education programs have participated in individual psychoeducational evaluations (U.S. Department of Education, 2001). Over 50 percent of the students enrolled in special education programs are diagnosed as learning disabled (LD). Given that LD diagnoses have commonly rested Requests for reprints should be sent to Marley Watkins, Pennsylvania State University, Department of Educational and School Psychology and Special Education, 125 CEDAR Building, University Park, PA 16802. Electronic inquiries may be sent to [email protected]. on an ability-achievement discrepancy (Mercer, Jordan, Allsopp, & Mercer, 1996), individual evaluations to examine special education eligibility often include a standardized measure of intellectual functioning. Of the available individual intelligence tests, the Wechsler Intelligence Scale for Children-Third Edition (WISC-III; Wechsler, 1991) is the most frequently used (Kaufman & Lichtenberger, 2000) and has become an integral part of psychological assessment in the schools (Politano & Finch, 1996). Given the popularity of the WISC-III, much attention has been focused on its usefulness in discriminating average and exceptional children and in detecting specific areas of cognitive strength and weakness (Saklofske, Schmidt, & Yackulic, 1984). Typically, interpretation of the WISC-III is based on a hierarchical, top-down model that first considers global IQ scores. Next, to extract more information from the WISC-III, distinct patterns or profiles of WISC-III subtest scores that are presumed to be associated with intellectual or educational disabilities are analyzed. This practice of interpreting the pattern of subtest scores attained by children on individual measures of intelligence is known as profile analysis (Sattler, 1992). More than 75 different Wechsler subtest patterns have been identified. One of the most popular was developed by Bannatyne (1968), who recategorized the subtest scores of the Wechsler Intelligence Scale for Children (WISC; Wechsler, 1949) to identify children with learning disabilities. Bannatyne believed that it did not serve a constructive purpose to divide the WISC performance of children with reading disabilities into verbal and performance IQs. Instead, he attempted to reanalyze the scaled scores by grouping them into three logical categories: spatial, conceptual, and sequential. According to Bannatyne, subtests in the spatial category (Block Design, Object Assembly, and Picture Completion) require the ability to manipulate objects in multidimensional space without sequencing, subtests in the conceptual category (Similarities, Vocabulary, and 50 LEARNING DISABILITIES PRACTICE Comprehension) involve the use of concepts and abstract reasoning, and subtests in the sequential category (Digit Span, Picture Arrangement, and Coding) engage the ability to remember sequences of visual or auditory stimuli. Bannatyne (1971) reported that disabled readers had their highest scores in the spatial category, intermediate scores in the conceptual category, and lowest scores in the sequential category (spatial > conceptual > sequential). Rugel (1974) reviewed 25 studies that reported WISC subtest scores of disabled readers and identified the Bannatyne pattern across 22 of the samples for which complete recategorization of the subtest scores was possible. Although Rugel found that children with reading disabilities demonstrated a clear deficit in the sequential category, they did not perform significantly lower than the general population on the Picture Arrangement subtest. In addition, children with reading disabilities scored lower on the Arithmetic subtest, which was not a part of Bannatyne’s (1968) original model. Subsequently, Bannatyne (1974) acknowledged that the Picture Arrangement subtest was erroneously included in the sequential category and substituted the Arithmetic subtest for the Picture Arrangement subtest to modify the sequential category. Thus, the revised Bannatyne pattern included the spatial category (Block Design, Object Assembly, and Picture Completion), conceptual category (Similarities, Vocabulary, and Comprehension), and sequential category (Digit Span, Arithmetic, and Coding). Following this revision, the Bannatyne pattern was applied to the Wechsler Intelligence Scale for ChildrenRevised (WISC-R; Wechsler, 1974). Initial investigations frequently found the spatial > conceptual > sequential pattern among children with learning and reading disabilities (Smith, Coleman, Dokecki, & Davis, 1977; Vance & Singer, 1979). Although later studies generally agreed that the pattern existed among some children with learning disabilities, it was not clear whether the pattern was useful in discriminating between children with different disabilities or those of different ethnic backgrounds (Clarizio & Bernard, 1981; D’Angiulli & Siegel, 2003; Dundon, Sewell, Manni, & Goldstein, 1986; Gutkin, 1979; Henry & Wittman, 1981; Kavale & Forness, 1984; Moore & Wilson, 1987; Zarske & Moore, 1982). Research on the Bannatyne pattern extended to the WISC-III soon after its publication. Prifitera and Dersh (1993) compared the Bannatyne WISC-III pattern of children with LD, children with ADHD, and children without disabilities. Base rates of the Bannatyne WISCIII pattern in each sample were used to estimate the probability of LD given the presence of the pattern. The base rate for children with a LD was 33 percent, while the base rate for children with ADHD was 47 percent and for children without a disability was 14 percent. Given these relative proportions, Prifitera and Dersh suggested that the Bannatyne WISC-III pattern is useful for diagnostic purposes and recommended that “the presence of a pattern or patterns would suggest strongly that the disorder is present” (1993, p. 53). Although Prifitera and Dersh (1993) recognized the possibility of misclassification if the Bannatyne pattern was used, the magnitude of this problem was not explicated. For example, the Bannatyne pattern correctly recognized only 33 of their sample of 99 children with LD. In contrast, it incorrectly identified 293 of the 2,158 children in the WISC-III normative sample as LD. Thus, only 33 of the 326 children marked by the WISC-III Bannatyne pattern were actually enrolled in LD programs. Conversely, two-thirds of the children with LD were missed by the Bannatyne pattern. Most studies of the Bannatyne pattern utilized analyses that are useful in identifying group differences, but are not as informative for differential diagnosis of individuals within those groups. Typically, statistically significant differences between regular and special education groups have been interpreted as evidence of diagnostic accuracy for individuals. This illustrates reliance on classical validity methods instead of the more appropriate clinical utility approach (Wiggins, 1988). As noted by Elwood, “significance alone does not reflect the size of the group differences nor does it imply the test can discriminate subjects with sufficient accuracy for clinical use” (1993, p. 409). Little attention has been paid to the overlap in score distributions between regular and exceptional groups, although its importance has been known for decades (Meehl, 1973). In sum, group separation is necessary, but not sufficient, for accurate decisions about individuals. To date, Prifitera and Dersh (1993) are the only researchers who have examined the Bannatyne WISCIII pattern in children with LD. Nevertheless, the Bannatyne pattern is still common in psychological training and practice. School psychologists report a reliance on subtest interpretations when analyzing intelligence tests (Pfeiffer, Reddy, Kletzel, Schmelzer, & Boyer, 2000; Watkins, 2000). Some assessment texts illustrate calculation of Bannatyne recategorized WISC-III scores (e.g., Cooper, 1995) or imply that the Bannatyne pattern has diagnostic significance (e.g., Aiken, 1996). Additionally, a common automated scoring program for the WISC-III computes Bannatyne scores (SAWS, 1995). Beyond its widespread use in the United States, the Bannatyne pattern has also been applied with non-English speaking populations (Alm & Kaufman, 2002; Chen, Yang, & Tang, 2002; Morad & Mahmoud, 2001). Given that decisions based on the WISC-III Bannatyne profile may have a major impact on children, additional research regarding its validity is necessary. Consequently, the purpose of the present study was to investigate the prevalence of the Bannatyne WISCIII pattern in a large sample of children with learning disabilities. Additionally, appropriate diagnostic utility statistics were applied to determine whether the presence of the Bannatyne WISC-III pattern was useful in differentiating children with disabilities from children without disabilities. SMITH AND WATKINS: BANNATYNE WISC-III PATTERN 51