Discriminant Validity of the Stanford-Benet Intelligence Scale: Fourth Edition

Abstract

Performance on the Stanford-Binet Intelligence Scale: Fourth Edition (SB:FE) was examined in a Canadian clinic sample (N=1220), age 2 through 23 years, with a range of demographic characteristics and ability levels. Data were anaiyzed for the samples 2-6-14, 7-11-11, and 12-23- 11. SB:FE subtest, Reasoniig Area, and Composite Standard Age Scores (SAS’s) decreased significantly (p < .05) with increasing age. Within each age group, the intercorrelations among subtests, the four Reasoning Area, and the Composite SAS’s supported the four cognitive ability areas posited by Thorndike et al. (1986b). Performance of subjects on the SB:FE full battery and SB:FE Genera! Purpose Abbreviated Battery (GPAB) were compared. Significant differences (p < .05), attributabie to the large sample sizes, were found between means and variances in Reasoning Area and Composite SAS’s. Uncorrected correlation coefficients among the two measures were significant (p < .01) and close to unity for the Verbal, Quantitative, Short Term Memory, and Composite SAS’s. The correlations between Abstract/Visual SAS’s, while significant, were somewhat lower. Also, similar and significant (p < .05) correlations were observed among the two versions of the SB:FE and the Wide Range Achievement Test-Revised (WRAT-R). Next, internally valid, reiable, and replicable groups displaying differences in profile elevation and/or shape were obtained through application of hierarchical agglomerative and iterative partitioning clustering procedures to SB:FE GPAB data. For the age sample 2-4-11,a two cluster solution, with high average and average groups was optimal. For the samples 5-6-11, 7-11-11, and 12-23 11, a three cluster solution comprising high, average, and low scoring groups was optimal. Mean WRAT-R subtest scores of the groups in all ages samples were significantly different (p < .01). However, when cluster solutions were compared with clinically derived a priori learning disability models, clusters were more similar with respect to Composite SAS's or profile elevation, than educational diagnosis. In general, results suggest the SB:FE is rnost appropriately used as an index of global ability. Caution is needed interpreting Reasoning Area SAS's, although the GPAB may provide a reasonable representation of the full battery