Body temperatures of 99 term and 44 preterm infants were measured at four sites: core (5 cm beyond the anus, with an electronic telethermometer), rectum (2 cm, with a mercury-in-glass thermometer), axilla, and between the skin and mattress. Temperatures measured at the four sites agreed closely in this group of largely normothermic infants. However, five of seven term infants with abnormal core temperature (greater than 1.5 SD below or above the mean) would have been judged to be normothermic by each of the three other measurements. The temperatures in preterm infants were lower and varied less with the site of measurement, indicating a smaller core-surface temperature gradient because of their relative lack of thermal insulation by body fat. Axillary temperature was as reliable as rectal temperature measured in the usual way with a mercury-in-glass thermometer. Measurement of the temperature between the skin and mattress was nearly as accurate as the other more frequently used methods. Ninety percent of temperatures were within 0.1 degree C of their final stabilization readings by 5 minutes for each type of thermometer and measurement site.