The assessment of the multitrait-multimethod matrix is placed in the context of generalizability theory for the purpose of demonstrating the convergent and discriminant validity of gender-related constructs. Once convergent validity is established with the traditional generalizability/reliability approach, a two-facet (traits and methods) generalizability design is proposed, with the levels of the trait facet comprised of scores on three separate measures. In this latter design, emphasis is placed on examination of the subject x trait and the subject x method interactions, with equations presented to aid in interpreting the degree to which constructs account for unique variance in the measurement of sex-role attitudes. (Author/RL)