To examine the necessary and sufficient status of delusions and hallucinations as clinical features of psychotic (delusional) depression, we studied a consecutive sample of 137 patients meeting DSM-III, RDC and our clinical criteria for endogenous depression/melancholia, of whom 35 had delusions and/or hallucinations, and represented our putative 'psychotic depressives' (PDs). The PDs were contrasted with the remaining 'endogenous depressives' (the EDs), and an age- and sex-matched subsample of the latter, the MEDs. Univariate and multivariate analyses of clinical features established that, in addition to the presence of delusions and/or hallucinations, the PDs could be distinguished in particular by severe psychomotor disturbance, as well as by sustained and unvarying depressive content, the absence of any diurnal mood variation and by constipation. Latent class analyses suggested that overt psychotic features (such as delusions and hallucinations) were sufficient but not necessary for a subject to be assigned to the 'psychotic' latent class, and a subsequent chart review suggested that, in some PDs actual psychotic features may not be able to be elicited because of severe psychomotor change, suggesting that clinical reliance on eliciting delusions or hallucinations may result in a number of 'masked psychotic depressives' escaping valid diagnosis.