Preadolescent self-concept and self-concept/academic achievement relations: Investigating multidimensional and hierarchical structures within and across gender.

Using a confirmatory factor analytic approach, the aim of the study was to examine the construct validity of an academic expansion of the Marsh/Shavelson model of self-concept (SC; Shavelson & Marsh, 1986; Marsh, 1990d), and its relations with academic achievement (AA) within and across gender. Participants were Grades 5 and 6 girls (n = 220) and boys (n = 205) drawn primarily from middle-class suburban communities in and around Ottawa, Canada. Four nonacademic SCs (i.e., physical appearance, physical ability, parents, and peers), one general-academic SC, and 11 subject-specific academic SCs were measured with the Self-Description Questionnaires (Marsh, 1990d, 1992b) and the Self-Perception Profile for Children (Harter, 1985). Specific academic SCs were measured for four Language Arts (i.e., Listening, Reading, Writing and Speaking), two mathematics (i.e., Arithmetic and Measurement), two science (i.e., Science and Social Studies), and three non-core (i.e., Religion, Art, and Physical Education/Gym) subjects. Grades and self-reported grades in eight of these subjects served as measures of AA. The hypothesized multidimensional SC measurement and structure was supported for both girls and boys. Tests of competing hierarchical SC structures provided support for five higher-order SCs (i.e., Nonacademic, Non-core Academic, and Verbal-, Mathematics-, and Science-Academic). Gym SC was better represented hierarchically under Nonacademic, rather than under Non-core Academic SC. Gender differences in the baseline hierarchical SC models related only to Parent SC, which was defined by only Non-core Academic SC for girls, but Verbal-Academic SC for boys. Reading, Writing, and Speaking in Language Arts AAs were not distinct, and thus, were represented as a single Language Arts Skills AA construct. A multidimensional pattern of relations in which SC and AA constructs were more highly correlated for corresponding than for non-corresponding academic subject areas was supported, albeit with two exceptions. A multidimensional pattern of SC/AA relations was not consistently found for the Language Arts constructs for both girls and boys, and was not found for Art AA for girls. An hierarchical pattern of SC/AA relations in which AA constructs were more highly correlated with first- than with second-order SCs was not consistently demonstrated for either girls or boys. Tests for the gender invariance of SC measurement and structure, and of SC/AA relations were generally tenable. Gender differences dud were of the largest magnitude indicated that, in comparison to girls, Art AA was more highly related to Art SC, but less correlated with both Speaking in Language Arts and Verbal-Academic SCs for boys