Generality of test-session observations to kindergarteners' classroom behavior

Test observations are integral components in the comprehensive assessment of children's academic talents and personal adjustment. They are used to validate standardized test scores obtained during test sessions and to draw inferences about children's general pattern of learning and social adjustment. Unfortunately, little empirical justification exists for extending inferences beyond the confines of test sessions. This issue was examined with 155 kindergarten children using the Stanford Binet Observation Schedule (SBOS), a popular instrument for the summative integration of test-session observations. Children's ratings on the SBOS were used to estimate teacher-observed behavior on the Guide to the Child's Learning Style (GCLS). SBOS and GCLS scores were subjected to independent principal components analyses with varimax rotation and, thereafter, to a joint canonical variance analysis. Weighted factor composites from two SBOS dimensions constituted the predictor data set and three GCLS dimensions the criterion data set. Although multivariate statistical significance was achieved (lambda = .834, F(6,200) = 3.18, p less than .005), cannonical redundancy showed merely 9.4% of classroom behavior predictable from test-session. Results are examined in light of research on the situational specificity of behavior.