Journal of Educational Psychology 1988. Vol. 80, No. 3, 366-380 Copyright 1988 by the American Psychological Association, Inc. 0022-0663/88/100.75 A Multifaceted Academic Self-Concept: Its Hierarchical Structure and Its Relation to Academic Achievement Herbert W. Marsh University of Sydney Sydney, New South Wales, Australia Barbara M. Byrne University of Ottawa Ottawa, Canada Richard J. Shavelson University of California, Santa Barbara Academic self-concept, originally posited by Shavelson as a single higher-order facet, was found by Marsh and Shavelson to comprise at least two higher-order academic facets (verbal and math). Marsh developed the internal/external (I/E) frame of reference model to account for the extreme separation of math and verbal self-concepts and their relations to math and verbal achievements. In our investigation, students completed the academic self-concept scales from three different instruments that were the basis of two studies. In the first study, the two higher order academic factors posited by Marsh and Shavelson fit the data substantially better than did a single higher order facet. In subsequent discussion, the Marsh/Shavelson model is more clearly defined, and directions for further research are identified. The second study provided further support for the I/E frame of reference model in that (a) verbal and math self-concepts were nearly uncorrelated, (b) verbal achievement positively affected verbal self-concept but negatively affected math self- concept (i.e., higher verbal skills led to lower math self-concepts), (c) math achievement positively affected math self-concept but negatively affected verbal self-concept, and (d) the results were consistent for each of three self-concept instruments. Both studies demonstrate that in further research at least verbal and math self-concepts, rather than a single general facet of academic self-concepts, should be considered. Shavelson, Hubner, and Stanton (1976) posited self-concept to be a multifaceted, hierarchical construct. They presented a possible representation of this hierarchy in which general self- concept appeared at the apex and was divided into academic and nonacademic self-concepts. According to this model, self- concepts in particular academic areas (math, English, etc.) combine to form a higher order academic self-concept. Shav- elson et al. based their model, in part, on conceptually similar models of ability that posit a higher order ability factor as well as more specific components of ability (e.g., Vernon, 1950). Also, achievement scores in mathematics and English typically correlate .5-.8 with each other, and academic achievements and self-concepts are posited to be substantially related. Thus Shavelson et al. posited that the different aca- demic self-concepts would be substantially correlated and could be incorporated into a single facet of academic self- concept. The Marsh/Shavelson Revision Marsh and Shavelson (1985) used responses to the Self Description Instrument (SDQ) by Australian students in The authors gratefully acknowledge funding support from the Social Sciences and Humanities Research Council of Canada, and the cooperation of teachers and students of the Carleton Board of Education. They also acknowledge Jennifer Barnes, Raymond Debus, and Bengt Muthen for their comments on earlier drafts of this article. Correspondence concerning this article should be addressed to Herbert W. Marsh, Faculty of Education, Room 83, Madsen Building, University of Sydney, Sydney, New South Wales, Australia 2006. Grades 2-5 to test Shavelson et al.'s (1976) model. Whereas their findings supported the model, the hierarchy proved to be more complicated than originally anticipated. This led to a revision of the model. Of particular relevance to our inves- tigation was that verbal and math self-concepts were nearly uncorrelated and did not combine with school self-concept to form a single, second-order academic factor. Instead, the results argued for two second-order academic factors repre- senting verbal/academic and math/academic self-concepts. In subsequent research with late-adolescent responses to the Self Description Questionnaire III (SDQIII; Marsh, 1987, in press), verbal and math self-concepts were again relatively uncorre- lated with each other and could not be adequately explained by a single higher order academic factor. The purpose of Study 1 is to apply hierarchical confirmatory factor analysis to test the generality of the Marsh/Shavelson revision across responses by North Americans to the SDQIII and to two additional self-concept instruments. The Internal/External (I/E) Frame of Reference Model The extreme separation between math and verbal self- concepts that was the basis of the Marsh/Shavelson revision also led to the development of the I/E model. In support of the construct validity of a multifaceted self-concept, research- ers have found (a) academic achievement to be more highly correlated with academic self-concept than with nonacademic and general self-concepts (Byrne, 1984) and (b) achievement in particular content areas to be more highly correlated with 366 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.