Confirmatory Factor Analyses of the Stanford-Binet Intelligence Scale, Fourth Edition

Abstract

This study presents results of confirmatory factor analyses of the recently published Fourth Edition of the Stanford-Binet Intelligence Scale. Four possible models of the scale's factor structure were tested with confirmatory factor analysis via LISREL at five age levels and for the standardization sample as a whole. Results of these analyses show that although the four-factor model presented by the scale's authors fits the general standardization data fairly well, it did not fit similarly well with data from several age groups. Overall, however, models based on the authors' theoretical structure were usually significantly better fitting to empirical data than were models with fewer factors. Results are discussed in terms of their implications for interpretation and use of the Stanford-Binet, Fourth Edition with young children.Thorndike, Hagen, and Sattler (1986) recently have published a major revision of the well-known Stanford-Binet Intelligence Scale. With test batteries like the Stanford-Binet, exploratory factor analysis often provides information useful in interpreting scores derived from them. Until recently, however, the age-level construction of the Stanford-Binet did not allow investigation of its underlying ability structure via exploratory factor analysis. The new edition of the Binet, its fourth (SB-IV), provides standardized scores for each of the measures' subtests so that their intercorrelations can be calculated. The resulting correlation matrix then can be studied through factor analysis.Just such an analysis is reported by Reynolds, Kamphaus, and Rosenthal (1987). Through traditional principal components and principal factors solutions, they found most frequently either a one-or two-factor structure in the Binet at the various age levels presented in the Binet's Technical manual (Thorndike et al., 1986). Results presented by Reynolds et al. provided little support for the model of cognitive abilities that underlies the new Binet. They report that the number of factors obtained varied across age levels and ranged from one to three.Reynolds et al. also observe that the meaning of the factors obtained is unclear and also may vary across age levels. Because the division of the Binet's scores into four subdomains is based on the notion that each reflects a distinct ability domain, the failure to find such factors in exploratory analyses calls the-overall organization of the scale into question. As Reynolds et al. conclude, &dquo;The previously cited results indicate that the multiple scale division of the Stanford-Binet, Fourth Edition, is illusory&dquo; (1987, p. 16). The sort of exploratory factor analysis presented by Reynolds et al., however, does not allow for tests of alternate factor models for the scale. A variant of factor

Review of validity research on the Stanford-Binet Intelligence Scale: Fourth Edition.

“…Significant caveats exist, however, concerning the validity of this model with children below age 6 years. Among this group of children, the Short-Term Memory factor does not appear to be valid (Keith et al, 1988;Ownby & Carmin, 1988). The findings from the Keith et al study also call into question the second level of the R. L. Thorndike et al (1986aThorndike et al ( , 1986b hierarchical model and the notion that crystallized intelligence is measured by the Verbal Reasoning and Quantitative Reasoning factors.…”

“…Several confirmatory factor-analytic studies of the SB:FE have been undertaken since its publication (Keith, Cool, Novak, White, & Pottebaum, 1988;Kline, 1989;Ownby & Carmin, 1988). These confirmatory factor analyses have used the intercorrelation matrices provided by R. L. Thorndike et al (1986b) in the Technical Manual for the standardization sample and some version of the LISREL VI computer program (JSreskog & Sorbom, 1986) to test the four-factor theory underlying the SB:FE (i.e., Verbal Reasoning, Quantitative Reasoning, Abstract/Visual Reasoning, Short-Term Memory).…”

“…A second confirmatory factor analysis using four models was conducted by Ownby and Carmin (1988). Model I was a g-only model similar to that examined by Keith et al (1988).…”

“…Finally, Kline (1989) conducted confirmatory factor analyses to determine whether R. L. Thorndike et al's (1986aThorndike et al's ( , 1986b four-factor model or Sattler 's (1988) two-or three-factor models provided the best goodness of fit for the standardization sample. Similar to Ownby and Carmin (1988), Kline did not specify whether a strict or more relaxed model was used in his analyses. He also did not conduct an analysis of the standardization sample as a whole.…”

The relation of the Stanford-Binet: Fourth edition to measures of achievement: A Concurrent validity study

“…However, the line of investigation addressing the factor structure of the measure and the utility and stability of the area scores has been less than supportive of the instrument's model. Exploratory and confirmatory factor analyses have questioned whether the four area scores are capable of measuring distinct aspects of ability (see, for example, Ownby & Carmin, 1988;Reynolds, Kamphaus, & Rosenthal, 1987). In fact, Sattler (1988) cautioned against the use of the area scores in providing interpretive statements about patterns of individual performance.…”