ASM18210.1177/107319 3931061110393106Holdnack et al.Assessment © The Author(s) 2011 Reprints and permission: http://www. sagepub.com/journalsPermissions.nav Assessment 18(2) 178 –191 Confirmatory Factor Analysis © The Author(s) 2011 Reprints and permission: http://www. sagepub.com/journalsPermissions.nav of the WAIS-IV/WMS-IV DOI: 10.1177/1073191110393106 http://asm.sagepub.com James A. Holdnack1, Xiaobin Zhou1, Glenn J. Larrabee2, Scott R. Millis3, and Timothy A. Salthouse4 Abstract The Wechsler Adult Intelligence Scale–fourth edition (WAIS-IV) and the Wechsler Memory Scale–fourth edition (WMS-IV) were co-developed to be used individually or as a combined battery of tests. The independent factor structure of each of the tests has been identified; however, the combined factor structure has yet to be determined. Confirmatory factor analysis was applied to the WAIS-IV/WMS-IV Adult battery (i.e., age 16-69 years) co-norming sample (n = 900) to test 13 measurement models. The results indicated that two models fit the data equally well. One model is a seven-factor solution without a hierarchical general ability factor: Verbal Comprehension, Perceptual Reasoning, Processing Speed, Auditory Working Memory, Visual Working Memory, Auditory Memory, and Visual Memory. The second model is a five-factor model composed of Verbal Comprehension, Perceptual Reasoning, Processing Speed, Working Memory, and Memory with a hierarchical general ability factor. Interpretative implications for each model are discussed. Keywords WAIS-IV, WMS-IV, joint factor analysis, confirmatory factor analysis, memory, working memory, intelligence The factor structure of the Wechsler intelligence scales, (Wechsler, 1997). Subsequent reanalysis of the factor struc- adult, child, and international versions, has been the subject ture of WMS-III found that a five-factor model yielded of numerous research studies. For the editions of these tests inadmissible parameter estimates (i.e., correlations >1.0) published in the past two decades, there is strong evidence because of the high correlation between immediate and that a four-factor model—Verbal Comprehension, Percep- delayed memory measures (Millis, Malina, Bowers, & tual Reasoning, Working Memory, and Processing Speed— Ricker, 1999). Additionally, low communality estimates for best describes the latent structure of the intelligence Faces contributed to model specification errors for the scales (Bowden, Lissner, McCarthy, Weiss, & Holdnack, visual memory factor (Millis et al., 1999). This finding was 2007; Bowden, Weiss, Holdnack, Bardenhagen, & Cook, confirmed by Price, Tulsky, Millis, & Weiss (2002). For 2008; Dickinson, Iannone, & Gold, 2002; Donders, & special groups, a two-factor model consisting of working Warschausky, 1996; Taub, McGrew, & Witta, 2004; memory and general memory was found in a large left and Wechsler, 2003, 2008); although some studies disagree right temporal lobe epilepsy sample (Wilde et al., 2003); with such conclusion (Kaufmann, Lichtenberger, & McLean, whereas four factors—Auditory, Visual, Working Memory, 2001; Keith, Fine, Taub, Reynolds, & Kranzler, 2006). and Learning—were reported for a mixed clinical and Although there is a substantial body of research evaluating control sample (Burton, Ryan, Axelrod, Schellenberger, & the factor structure of the Wechsler intelligence scales, the Richards, 2003). Inconsistent findings suggest that the rela- study on the Wechsler memory scales has not been as exten- tionship among memory scales may be moderated by meth- sive, and obtaining consistent factor structure across studies odological and cognitive factors beyond memory itself. and test editions has been elusive (Wechsler, 2009). Factor analytic studies of previous versions of the 1Pearson, San Antonio, TX, USA Wechsler memory scale have yielded inconsistent results. 2Sarasota, FL, USA Initial factor analytic results published for the Wechsler 3Wayne State University, Detroit, MI, USA 4 Memory Scale–third edition (WMS-III) indicated that University of Virginia, Charlottesville, VA, USA a five-factor model (i.e., Auditory Immediate, Auditory Corresponding Author: Delayed, Visual Immediate, Visual Delayed, and Working James A. Holdnack, 5 Rose Hill Drive, Bear, DE 19701, USA Memory) best fit the data from the normative sample Email: [email protected] Downloaded from asm.sagepub.com at UNIV OF VIRGINIA on May 12, 2011 Holdnack et al. 179 Memory functioning cannot be directly assessed; rather, In a series of articles employing confirmatory factor memory is measured in the context of other cognitive skills analysis, Bowden and colleagues initially found that a such as visual spatial or verbal abilities (e.g., story recall five-factor model: Verbal Comprehension, Perceptual requires intact language functioning). Heilbronner (1992) Organization, Attention–Concentration/Working Memory, details the nature of additional cognitive functions that Verbal Memory, and Visual Memory, best accounted for the affect performance on visual memory tests and concludes joint factor structure of the WAIS-R and WMS-R in a non- that the concept of pure visual memory measures is unten- clinical sample (Bowden, Carstairs, & Shores, 1999) and a able. The methodology used to assess memory (e.g., free substance abuse sample (Bowden et al., 2001). In a later recall, learning, and recognition) could also affect the psy- study, however, a six-factor model that also included chometric properties of the test and its relationship to other Processing Speed was supported in clinical and community memory measures (Compton, Sherer, & Adams, 1992). samples (Bowden, Cook, Bardenhagen, Shores, & Carstairs, Therefore, performance on the memory tests could vary by 2004). The WAIS-R and WMS-R studies indicate that five the nature of additional cognitive skills measured and the or six factors are supported with auditory and visual memory methodology employed, which may facilitate or inhibit factors identified from the memory scale. efficient memory storage and retrieval and, in turn, attenu- The Wechsler intelligence and memory scales were ate or facilitate the correlation between two memory mea- co-normed for the first time for the third edition of the tests sures. These variations among memory measures result in (Tulsky & Ledbetter, 2000). Tulsky and Price (2003) com- memory tests being only moderately correlated with each pleted an extensive evaluation of the joint factor structure of other (Compton et al., 1992; Golden, White, Combs, the WAIS-III and WMS-III. They reported that a six-factor Morgan, & McLane, 1999). model, Verbal Comprehension, Perceptual Organization, The first combined factor analysis of the Wechsler Adult Working Memory, Processing Speed, Visual Memory, and Intelligence Scale (WAIS) and the original WMS clarified Auditory Memory, was best supported by the data. Model the construct validity of the tests comprising the WMS by specification errors were observed when immediate and analyzing the factor structure in the context of other measures delayed memory factors were evaluated in the same factor of verbal and visuospatial skills, attention, and processing analysis; additionally, Faces was observed to have poor fit speed (Larrabee, Kane, & Schuck, 1983). This exploratory on any factor. The model fit was best when subtests were factor analysis yielded a memory factor defined by Logical allowed to load on multiple factors specifically for Arithmetic, Memory and Paired Associate Learning, an information/ Spatial Span, Picture Arrangement, and Visual Reproduction orientation factor, and an attention/concentration factor (Tulsky, Ivnik, Price, & Wilkins, 2003). defined by Digit Span, Arithmetic, and Mental Control. Of The WAIS-IV has a very similar subtest and index struc- particular interest, WMS Visual Reproduction loaded more ture as its predecessor, the WAIS-III (Wechsler, 2008). The strongly with a WAIS visuospatial ability factor (.70) than primary differences are the inclusion of a new subtest, with the memory factor (.09). Subsequently, Larrabee, Visual Puzzles, on the Perceptual Reasoning index and the Kane, Schuck, and Francis (1985) factored immediate recall addition of a sequencing procedure to the Digit Span sub- and delayed recall Logical Memory and Visual Reproduction test. The four-factor model of Verbal Comprehension, subtests in separate factor analyses, thereby controlling for Perceptual Reasoning, Working Memory, and Processing method variance (Larrabee, 2003), and found that Immediate speed was confirmed by factor analysis (Wechsler, 2008). Visual Reproduction loaded more strongly with a visuospa- The WMS-IV, however, was substantially revised from the tial ability factor defined by Block Design and Object previous edition (Wechsler, 2009). At the subtest level, Assembly than with a memory factor defined by Logical there is a new visual memory subtest, Designs, and two new Memory and Paired Associate Learning. When delayed visual working memory subtests: Spatial Addition and recall scores (e.g., Delayed Logical Memory and Visual Symbol Span. The index structure includes Verbal Memory, Reproduction) were factored, the loading pattern of Delayed Visual Memory, Visual Working Memory, Immediate Visual Reproduction shifted so that the primary loading Memory, and Delayed Memory. The Visual Memory, was with the memory factor, with a secondary loading on Auditory Memory, and Visual Working Memory indexes the visuospatial ability factor. This basic finding (stronger were confirmed by factor analysis