It is recognized that different associations, or no associations, might have been found if unhospitalized cases had been added to the study. Insofar as this may be true, the identification of characteristics associated with hospitalized cases may not contribute to the epidemiology of old-age psychoses.It is concluded, however, that little would be gained in the epidemiology of old-age psychoses if it were found that the true distribution of the disease resembles that of hospitalized cases. In the first place, the evidence suggests that the associated social and economic conditions might be the results rather than the causes of the disease. In the second place, it is suggested that the initial stages of the disease might have been present at an early age and exercised an influence on the "environment" many years before diagnosis. As long as there is a possibility that the disease itself may select the environment, as suggested in parts of this study, the conditions associated with the disease at the time of diagnosis cannot be assumed to be the same as those that might have been found prior to onset.For this reason, studies limited to conditions present at the time of diagnosis cannot always be said to be concerned with the role of social and economic conditions as causative factors prior to onset. Nor can it be said, in these circumstances, that the findings either support or reject any hypothesis that any aspect of the "socio-economic environment" is a contributing factor in the onset of the disease. In fact, the populations with different rates of disease in 1940 may have been exactly alike at the time of the initial onset of the disease.Some of the difficulties in interpreting "ecological correlations" for purposes of mental disease epidemiology have been described by others (7, 8, 9, 10, 11). The chief problem, however, is believed to be related to the lack of clinical information necessary to the establishment of the time of onset of the disease. Without this information, early recognition of cases outside the hospital is not possible and the conditions prior to the development of the disease cannot be separated and studied apart from the conditions resulting from the disease.From this point of view, useful clinical data constitute the greatest need in mental disease epidemiology, regardless of the method of analysis used.