The structure of academic self-concept: The Marsh/Shavelson model.

New academic self-concept instruments were used to measure self-concepts in 13 (Grades 5-6) or 16 (Grades 7-10) school subjects and to test the structure of academic self-concept posited in the Marsh/Shavelson model. First-order factor analyses identified the scales each instrument was designed to measure, demonstrating that academic self-concept is remarkably subject-specific. As posited, two higher order factors were sufficient to explain relations among core academic subjects, but additional higher order factors were needed to explain other school subjects (e.g., physical education, art, and music). The hierarchy, however, was weak, and much of the variance in specific subject self-concepts was unexplained by the higher order factors. Researchers interested in self-concepts in particular subjects are advised to use self-concept scales specific to those subject areas in addition, perhaps, to other measures of academic self-concept. Prior to the 1980s, reviewers of self-concept research noted a lack of theoretical models and appropriate measurement instruments. In an attempt to address this situation, Shavelson, Hubner, and Stanton (1976) reviewed existing theory, research, and instruments and developed a multifaceted, hierarchical model of self-concept. In their model a general facet at the apex of the self-concept hierarchy is divided into academic and nonacademic components of self-concept. Academic self-concept is then divided into self-concepts in particular subject areas (e.g., mathematics, English), and nonacademic self-concept is divided into social, emotional, and physical self-concepts. The academic portion of the Shavelson et al. model, which is the focus of this study, is shown in Panel A of Figure 1. The self-concept facets and the structure proposed by Shavelson et al. were heuristic and plausible, but they were not validated by research in their review. Commenting on this problem, Byrne (1984) noted that consider this inability to attain discriminant validity among the dimensions of SC self-concept to be one of the major complexities facing SC researchers today (pp. 449-450). In contrast, more recent research based on better theoretical models and measurement instruments supports the multidimensionality of self-concept and many aspects of the Shavelson et al. model (e.g., Boersma Byrne, 1984; Dusek Fleming Halter, 1982; Marsh, 1988, in press-a; Marsh, Byrne, 1982; Soares & Soares, 1982).