Cross-validation of the WISC-III factor structure in two samples of children with learning disabilities

Scores on the WISC-III from 280 children with learning disabilities were an alyzed with confirmatory factor analysis. Six competing models reported in the literature were tested and then cross-validated in an independent sampl e of 240 children with learning disabilities. Results revealed that a 4-fac tor model offered the most accurate and stable explanation of the data when the Symbol Search subtest was included. A 3-factor model also fit the data closely and behaved reliably when the Symbol Search subtest was excluded f rom the analysis. Both models left much of the variance for each subtest un explained, however, and correlations between the subscale (factor) scores a nd their respective factors indicated that the Freedom from Distractibility subscale was saturated with unidentified sources of variation in both mode ls. The practical implications of these results were discussed.