OBJECTIVE The study explored the phenomenology and prognostic significance of delusions in major depressive disorder. METHOD From 452 patients with DSM-III major depression, we selected those with at least one belief fulfilling both DSM-III prerequisites for a delusion (i.e., being of "delusional proportions" and being maintained with "delusional intensity"). These patients were compared to the others with respect to demographic, historical, and index episode features; time spent in a depressive episode during a prospective observation period; and 10-year outcome. The same comparisons were made between patients with mood-incongruent delusions and those with mood-congruent delusions only. The study covered the period between January 1, 1978, and December 31, 2005. RESULTS About 20% of patients had at least one delusion in their index episode. An additional 5.3% had a belief fulfilling only one of the DSM-III prerequisites for a delusion. In about one quarter of delusional patients, the index episode was not "severe." Almost 10% of delusional patients had both mood-congruent and mood-incongruent delusions. In patients with delusions, time to syndromal recovery from index episode was longer and antipsychotic medication was more frequently used (both p < .0001). The presence of delusions predicted a higher depressive morbidity during the prospective observation period (p < .05), but not a poorer 10-year outcome. No variable discriminated patients with mood-incongruent delusions from those with mood-congruent delusions only. CONCLUSION The presence of delusions in a major depressive episode has significant therapeutic and short-term prognostic implications. However, the boundary between delusions and nondelusional sustained preoccupations is somewhat fuzzy, and some DSM-IV assumptions concerning psychotic depression (i.e., that this depression is always "severe"; that in an individual patient, delusions will be either all congruent or all incongruent with depressed mood; and that mood-incongruent delusions are associated with a poorer prognosis) may be unwarranted.