Attention-deficit/hyperactivity disorder (ADHD) is a phenotypically heterogeneous and highly heritable syndrome, which commonly co-occurs with other psychiatry disorders. To assess the role of genetic influences in ADHD, we used latent class analysis (LCA) to identify subtypes of ADHD taking into account its comorbidity with separation anxiety, oppositional defiant disorder (ODD), and three major depression symptoms. A structured interview was used to collect diagnostic data from a population sample of 2904 adolescent female twins and their parents. LCA was applied to ADHD, separation anxiety, ODD symptom profiles obtained from the twins' parents, and major depression symptom profiles obtained from the twins' self-report. Odds ratios were used to test for familiality of class membership by examining the effect of zygosity on twin concordance within and between latent classes. Structural equation modeling was used to compute heritabilities for latent class membership. LCA revealed three ADHD categories of clinical interest: an inattentive subtype without comorbidity, a second inattentive subtype with increased number of ODD symptoms, and a combined inattentive/hyperactive-impulsive type with elevated levels of ODD, separation anxiety, and depressive symptoms. LCA also distinguished an ODD class and a separation anxiety class, each without increased levels of other comorbid symptoms; a second ODD class co-occurring with increased separation anxiety and depression symptoms; and a pure depression class. Odds ratios for MZ contrasted with DZ twin concordance for individual latent class membership ranged from 2·5 to 19·4. Overall, 66% of MZ pairs, but only 36% of DZ pairs, were assigned to the same latent class, consistent with a genetic hypothesis for latent class membership. Individual class membership was shown to have high heritability ranging from ·34–·85. The pattern of latent classes suggested that in the general female adolescent population, there are three highly heritable ADHD subtypes, two of which are comorbid with other disorders. These classes were consistent with a genetic hypothesis for ADHD, with each class potentially reflecting a unique genetic subtype.