The temporal consistency of the research diagnosis of DSM-III-R delusional disorder was assessed. Subsequent to their first psychiatric hospitalization, psychotic patients were diagnosed twice using a best-estimate procedure after 6- and 24-month follow-up. Only 57.1% of the 7 subjects diagnosed at the 6-month follow-up conference retained the diagnosis at the 24-month review. Conversely, 50.0% of the 8 subjects given this diagnosis at 24 months had different 6-month diagnoses. By contrast, 100% of a matched control group maintained the same diagnosis at both time points. Course and outcome in this sample varied considerably, with 2 of the subjects making severe suicide attempts. Longitudinal assessment in psychotic patients in their first episode is mandatory, and an initial diagnosis of delusional disorder has to be interpreted as provisional.