To the Editor: We appreciate the opportunity to reply to each point made by Dr. Quitkin and his colleagues. Our overview did note the latent class analysis study by Kendler et al. (1996) but without comment as to whether it supported atypical depression as an entity or not. Those authors noted the intrinsic limitations of latent class analysis ("It cannot prove that such discrete classes exist") and that their identified classes might "reflect only differing points on a single underlying continuum of severity." Dr. Quitkin et al. are correct that items will be independent within classes in latent class analysis. This could be expected to hold in the atypical latent class defined by DSM-IV criteria, provided these reflect all key items from the original studies. If DSM-IV criteria, however, define something closer to a syndrome, then independence is not to be expected. More pertinently, we found independence (very low associations) both in the subjects with atypical depression and in the entire group, indicating that not much would be accounted for by any latent variable.