Self-Concept Validation Research: Some Current Complexities

Univariate statistical treatments in studies of complex personality constructs like self-concept can seriously mislead both practitioners and personality researchers. In Study 1, 349 students in grades 4 through 6 responded to the Sears Self-Concept Inventory and the STEP Science Achievement scale in January and in April. Multiple regression analyses predicting achievement from factor-defined self-concept subscales highlighted interpretive problems arising from multicollinearity and suppression effects among measures of achievement, and academic and physical facets of self-concept. These multivariate analyses also revealed stronger relations between several facets of self-concept and achievement than would be obtained with univariate procedures, emphasizing the interdependence of the two variables. In Study 2, 488 fourth-through sixth-grade students responded to the Gordon, Piers-Harris, and Sears self-concept inventories. Canonical variate analyses using factor-defined cluster scores revealed bipolar dimensions of self-concept, hidden in previous studies using univariate methods, that hold significant import for future theorizing about the construct.