The construct validity of a multidimensional adolescent self-concept was estimated separately for males and females, and the results were compared across gender. The application of four multitrait-multimethod (MTMM) approaches to estimating construct validity was demonstrated, and the scope and consistency of the findings derived from each were compared. Methodological approaches included: (1) the Campbell-Fiske model of D. T. Campbell and D. W. Fiske (1959); (2) the analysis of variance model of M. J. Kavanagh et al. (1971); (3) the confirmatory factor analysis model of K. G. Joreskog (1969); and (4) the composite direct product model of M. W. Browne (1984). These procedures were applied to data comprising four self-concept traits (general, academic, English, and mathematics) as measured by three measurement scales (Likert, semantic differential, and Guttman). The sample consisted of 412 males and 420 females in grades 11 and 12 from two suburban high schools in Ottawa (Ontario). Results derived from the composite direct product model are less than optimal. Other results demonstrate that adolescent males and females can differ with respect to various aspects of construct validity related to multidimensional facets of self-concept. Seven tables present study data. (SLD)