Previous research suggests that patterns of comorbidity of a limited number of anxiety, depressive, substance use and antisocial personality (ASP) disorders among adults are best reflected by a hierarchical three-factor structure with two correlated factors ('anxious-misery' and 'fear') that are summarized in a second-order 'internalizing' factor and one 'externalizing' factor. It has not been examined whether this structure is developmentally stable and robust against additions of more diagnoses. Using data from a prospective-longitudinal community study of adolescents and young adults we re-evaluate the three-factor model originally proposed by Krueger (Archives of General Psychiatry, 1999; 56, 921-926). Using confirmatory factor analysis with identical conventions as in Krueger's original work we found that the three factor model did not fit robustly across age or a wider range of diagnoses. Using explanatory factor analysis we examined alternative structures. We found various clinically meaningful patterns with good fit that go substantially beyond the original three-factor structure. However, again, there is little consistency in findings when different age groups or different diagnoses are considered. Our findings suggest that psychopathology cannot be reduced to any simple structure.