The relationship between the days on which sudden infant death syndrome (SIDS) occurred and the daily minimum temperature was examined in Auckland (1979-1984) and Christchurch (1979-1987). There was a marked winter excess of deaths in both regions. There was a significant negative correlation between the monthly mean minimum temperature and SIDS rate for both regions (r = -0.43, n = 347, P less than 0.0001). The monthly mean minimum temperature describes SIDS mortality equally as well as the three variables of daily minimum temperature, season and geographical location. There was a significant association of SIDS with minimum temperature 4 and 5 days prior to the death after adjusting for the effect of monthly mean minimum temperature. The days preceding death were on average colder than the other days, but the effect was small, especially when compared with the magnitude of the temperature differences between consecutive months.