Abstract:
A multivariate, multilevel model meta-analysis was conducted to study the effectiveness of interventions for students in enhancing various self-concept domains. In support of a construct validation approach to self-concept research, self-concept outcomes directly related to the aims of the intervention yielded higher effect sizes than those that were of secondary or incidental relevance (e.g. math self-concept in a math self-concept intervention); this finding concurrently supports multidimensional perspectives in self-concept research. Beyond this conceptually significant finding, findings are reported that have methodological repercussions in terms of both research methodology and intervention design. The ramifications for research methodology include the finding that higher effect sizes were observed for studies using random assignment to treatment/control conditions. In terms of significance for self-concept intervention design, an interesting finding emerged suggesting that praise/feedback treatments were more effective in enhancing self-concept than other techniques, including discussion groups and counseling. In addition to these exciting findings pertaining to self-concept interventions, the current paper contrasts the commonly practiced meta-analytic method of random effects models with the revolutionary multilevel modeling approach to meta-analysis. The result is a comprehensive paper that has implications for self-concept theorists, intervention designers, intervention evaluators, and meta-analysts.