The latent structure and stability of 10 common mental disorders were examined in a birth cohort at ages 18 and 21. A 2-factor model, in which some disorders were presumed to reflect internalizing problems and others were presumed to reflect externalizing problems, provided a more optimal fit to the data than either a 1- or a 4-factor model. To a significant extent, persons in the sample retained their relative positions on the latent factors across the 3-year period from age 18 to age 21. Results offer potential clarification of the meaning of comorbidity in psychopathology research by suggesting that comorbidity may results from common mental disorders being reliable, covariant indicators of stable, underlying "core psychopathological processes."