Records of two hospitals and of an association of mothers of twins were analysed for the years 1924-1980, finding a total of 656,109 and 1185 twin births, respectively. Twinning rates were higher in summer or fall, in general, but these seasonal variations occasionally changed within a short period of years, as in 1973-75 in the Association group. The twinning rate was also affected by the month of mother's birth. It was low among the mothers born around May-July in maternities up to 1960, and these mothers did not show seasonal variation of twin births. The unlike-sexed twinning rate of the mothers born in May-July was also lower than that of the other mothers, and constantly low even at higher maternal ages. A possible explanation for these phenomena is discussed.

Birth dates of 1,536 twin-pairs in 1971-1984 were collected from the members of an association of twins' mothers. The seasonal variation of twinning changed every 2-4 years. Years when twinning rate was higher in the summer-fall season (1971-72, 1976-77, 1982-84) and those when a peak of the rate was not observed (1973-75, 1978-81) appeared alternately. In years with a summer-fall peak, the elevation of twinning in the summer-fall season was detected consistently in both like- and unlike-sexed and in both MZ and DZ twin groups. The twinning seasonality in these years, however, was not evident in twin births of mothers who were born in May-July. These results suggest the possibility that seasonal factors which influence the twinning rate be not multiple-ovulation-inducing but probably abortion-inducing factors and most likely seasonally epidemic microbes.