In a retrospective study, we compared the months of birth of 3,106 psychiatric inpatients to those of 1,943 surgical patients collected during the same period 1981-1991 in the same hospital, and of a sample of 10,003,572 births in France in 1977-1989. DSM III-R categories were modified so to allow a comparison with former studies, and psychiatric patients were distributed among seven categories: Bipolars (N = 294), Unipolars (N = 287), Neurotic-reactive depressions (N = 582), First Major Depressive episode (N = 214), Schizophrenia (N = 244), Schizo-Affectives (N = 52) and Other Diagnosies (N = 1,433). Months of birth were grouped in quarters and semesters, according to the usual calendar, but also to temperature and the photoperiodic cycle. The main results were: 1. A seasonnality of births in the General Population sample, with a spring maximum (p < 0.001). 2. An absence of deviation from the general population and the surgical sample among Neurotic-reactive Depressions and Other Diagnoses. 3. A deviation from the general population and from other comparison groups (surgical cases and Other Diagnoses) among Unipolars and First Major Depressive Episodes (most of those being late episodes), with a significant excess of births during the "dark" or "cold" season of the year, especially around the winter solstice. The Bipolar group followed the same tendency, though to a lesser degree and for subjects born before 1940 only. The most significant results were found among Unipolars, which differ from the general population either by quarters (p < 0.0005) or by semesters (p < 0.0005) and from surgical cases by quarters (p < 0.01) and by semesters (p < 0.001). The results were similar for First Major Depressive episodes, although this category was theoretically "anosological". As the median age was high in this category, it might group a number of late depressive episodes, near to "involutionnal melancholia". Thus, our results seem to be relevant to the traditional endogenous-psychogenetic dichotomy, with a "cold" or "dark" seasonnality of births in the first case, and no particular seasonnality in the second case. Some former studies showed the same results, but the most significant deviation was found in mania. Our results cannot be explained by differences in the sex-ratio among the categories, and only partially by an age-incidence effect or an age-cohort effect.(ABSTRACT TRUNCATED AT 400 WORDS)