Wechsler scales of intelligence are arguably among the most frequently used individual measures of cognitive abilities world–wide (Georgas, van de Vijver, Weiss, & Saklofske, 2003; Lichtenberger & Kaufman, 2009) with several translations, adaptations, and standardizations. Interpretation of Wechsler test scores is dependent on evidence beyond that presented in the respective technical manuals. This symposium presents a collection of four papers detailing psychometric evaluations of the WISC–IV, WISC–IV Spanish, WISC–V, and WAIS–IV with implications for ethical interpretation of scores. The first paper presents independent EFA and CFA analyses of the 16 primary and secondary subtests of the WISC–V standardization sample. Results address the viability of the five–factor model posited by the publisher as well as relative importance of the general versus lower–order factors. The second paper reports on independent assessment of the latent factor structure of the WISC–IV Spanish by examining hierarchical structure with methods used in the first paper. The third paper presents an investigation of the structural validity and incremental predictive validity of the WAIS–IV with a large university sample of students seeking disability evaluations. The final paper presents an assessment of diagnostic utility of the WISC–IV GAI > CPI Profile for correctly identifying children with Autism. This collection of papers helps shed light on the validity and diagnostic utility of Wechsler scales from a variety of methods and supplements and compliments previous research. Importantly, these studies challenge many of the recommendations for interpretation promulgated by the publisher and various textbooks and interpretive guidebooks. The importance of such studies is illustrated and emphasized for consideration of ethical interpretation of tests such as Wechsler scales.

 

Paper 1: What does the WISC–V Measure? EFA and CFA with 16 Primary and Secondary Subtests
Gary L. Canivez, Marley W. Watkins, Stefan C. Dombrowski

Abstract: The Wechsler Intelligence Scale for Children–Fifth Edition (WISC–V; Wechsler, 2014a) was recently published purporting to measure five–factors.  When an instrument is revised and reformulated, its structure would benefit from an examination using both exploratory and confirmatory methods that present variance partitions, latent factor–based reliability coefficients, and a comparison of all plausible structural models. Unfortunately, this information was omitted from the Technical and Interpretative Manual (Wechsler, 2014b).  Further, CFA analyses reported in the WISC–V Technical and Interpretative Manual were not adequately informative (Boomsma, 2000). No explanation for choice of weighted least squares estimation was provided and numerous threats to model viability included standardized path of 1.0 between general intelligence and Fluid Reasoning, cross–loading Arithmetic on three factors, failure to test rival bifactor models, and failure to provide decomposed variance estimates. This paper presents independent EFA and CFA of the 16 primary and secondary subtests for the WISC–V standardization sample including analyses and information lacking in the Technical and Interpretative Manual. Hierarchical EFA supported four first–order factors (no separate Fluid Reasoning) and the second–order general intelligence. CFA found none of the proposed higher–order WISC–V models that included five latent first–order factors were acceptable (including the publisher’s preferred final WISC–V model) as the Fluid Reasoning (FR) factor produced negative variance estimates and model specification errors. Bifactor representation of four latent first–order factors was superior to the higher–order representation. Decomposed variance estimates for the bifactor representation showed the dominance of g in capturing subtest variance (except for Processing Speed [PS] subtests) and omega–hierarchical indicated interpretability of the latent g–factor, but small variance portions and low omega–subscale for group factors indicated they are likely not interpretable independent of g.  Implications for WISC–V interpretation are discussed.

Paper 2: Structural Validity of the WISC–IV Spanish
Ryan J. McGill & Gary L. Canivez

Abstract: The Wechsler Intelligence Scale for Children-Fourth Edition Spanish (WISC–IV Spanish) is a translation and adaptation of the WISC–IV for use with Spanish speaking children and adolescents aged 6–16 years. It features 10 core and 4 supplemental subtests, and provides four index scores as well as an overall Full Scale score (FSIQ), that represents general intellectual ability. Inexplicably, evidence to support higher–order structure is absent from the Technical Manual despite the implied influence of the CHC model. In order to verify and elucidate the structure of the WISC–IV Spanish, the present study examined the latent structure of the 14 WISC–IV Spanish subtests using exploratory and confirmatory procedures recommended specifically for multi-level structure. That is, EFA, using multiple criteria to support factor extraction, followed by application of the Schmid and Leiman (1957) procedure to enable a more thorough examination of variance partitions. With the exception of publisher theory, none of the extraction criteria suggested the presence of four factors. EFA results from a forced four–factor extraction found that subtests were associated with their theoretically hypothesized factors. Still, larger portions of common and total variance are associated with the g factor compared with specific group factors. Using CFA, four– (Wechsler) and five– (CHC) factor oblique first–order models were examined with the CHC model providing statistically significant better fit across multiple fit metrics (², CFI, RMSEA, AIC). Additionally, direct (bifactor) and indirect (higher-order) hierarchical models were explicated to account for the high correlations between first–order factors. Examination of higher–order structure supported the retention of a CHC–based bifactor measurement model. Whereas, decomposed variance estimates and latent factor reliability estimates provided strong support for general intelligence, support for the CHC group factors was less consistent. Implications for clinical interpretation of WISC–IV Spanish scores are discussed.

Paper 3: Structural and Incremental Validity of the WAIS-IV in College Student Disability Evaluations
Jason M. Nelson, Gary L. Canivez, & Marley W. Watkins

Abstract: The latent factor structure of the 10 core Wechsler Adult Intelligence Scale Fourth Edition (WAIS–IV) subtests and incremental prediction of WAIS–IV factor index scores from evaluations of 574 college students referred for learning difficulties is reported. A core purpose was to investigate WAIS-IV Technical and Interpretive Manual recommendation that clinicians focus on the four index scores “as the primary level of clinical interpretation†(Wechsler, 2008, p. 127). One, two, three, and four oblique first-order factor models were examined along with higher-order and bifactor models through confirmatory factor analysis. The higher-order (indirect hierarchical) model was compared to a bifactor (direct hierarchical) model to determine the superordinate versus breadth aspects of psychometric g. Hierarchical multiple regression analyses (HMRA) were used to investigate the incremental validity of the four first-order WAIS-IV index scores beyond the FSIQ when predicting academic achievement. The current work extends results of Nelson, Canivez, and Watkins (2013) with a larger sample. Preliminary analyses found the oblique four–factor model fit best, but high first–order factor correlations required explication of a hierarchical structure and higher-order and bifactor models were compared. The bifactor model produced negative variance estimate for BD so judged inadequate, thus the higher-order model is preferred. Decomposed variance estimates from the higher-order model showed dominance of g. Omega hierarchical coefficient was substantial for g but inadequate for the four group factors, with the possible exception of PS. HMRA showed that the WAIS–IV FSIQ predicted 9.0%–50.1% of WJ-III subtest and NDRT score variance and 21.1%–56.0% of WJ-III composite score variance. After accounting for FSIQ prediction, the combined effect of WAIS-IV Factor Index scores accounted for an additional .5%–29.1% of WJ-III subtest and NDRT score variance and 4.0%–20.6% of WJ-III composite score variance. Implications for interpretation of WAIS-IV scores will be discussed.

Paper 4: Diagnostic Utility of WISC–IV GAI > CPI Profile for a Referred Sample of Students with Autism
Kara M. Styck, Marley W. Watkins, & Laura L. Pendergast

Abstract: The present study evaluated the diagnostic utility of the General Abilities Index (GAI) > Cognitive Proficiency Index (CPI) cognitive profile of the Wechsler Intelligence Scale for Children—Fourth Edition for children and adolescents diagnosed with autism spectrum disorder (ASD). Participants included 302 children and adolescents aged 6 to 16 years who were administered psychoeducational evaluations and diagnosed with ASD (n = 79) or found ineligible for special education services by school multidisciplinary teams (n = 223). First, participants’ Perceptual Reasoning Index scores and Verbal Comprehension Index scores were combined to form the GAI and participants’ Working Memory Index scores and Processing Speed Index scores were combined to form the CPI. Next, the GAI-CPI difference score was computed to form a continuous outcome variable. Diagnostic status (i.e., ASD or ineligible) was cross-referenced with participants’ GAI-CPI difference scores to compute true positive (i.e., profile present and ASD diagnostic status) and false positive rates (i.e., profile absent and ASD diagnostic status), which were plotted on a receiver operating characteristic curve. The area under the curve was analyzed using a nonparametric statistical approach. Results indicated that the GAI > CPI profile could not accurately distinguish between these two samples of students. Specifically, a randomly selected student from the ASD participant group had a 60.3% 95% CI [52.4%, 68.3%] probability of displaying a higher GAI-CPI difference score than a randomly selected individual from the referred comparison group. Results of the present study indicate that the hypothesized GAI > CPI cognitive profile may not be unique to individuals with ASD. Moreover, these results suggest that previous research emphasizing group mean differences between ASD and non-clinical samples as evidence in support of the presence of the GAI > CPI profile in individuals with ASD may be misleading.