Seasonality of sudden infant death syndrome (SIDS) is a well-established epidemiological finding. The purpose of the study was to determine whether this feature varied significantly with age at death. In total, 13990 cases of SIDS in Scotland, England and Wales during 1982-1992 were studied by age group at death. Seasonality was established by fitting a sinusoidal curve and for each set of monthly data the peak position in the year and its magnitude were determined. Weighted regression revealed significant differences in peak position and amplitude of seasonal variation between those dying at < or = 4 months and those aged > or = 5 months at death. Those infants in the younger age group were more likely to die earlier in the winter months and had a smaller variation in seasonality. The peak (acrophase) months were January for < or = 4 months and February for > or = 5 months at death. Weighted regressions of peak position and amplitude on age at death had p-values of <0.001 and <0.01, respectively. A log linear model relating SIDS incidence to month of birth, month of death and age was able to explain some of these findings. The findings support the hypothesis that in SIDS there may be more than one infant cohort, each of which passes through a vulnerable developmental window at different ages.