Rectal and axillary temperatures were measured during the daytime in 281 infants seen randomly at home and 656 at hospital under 6 months old, using mercury-in-glass thermometers. The normal temperature range derived from the babies at home was 36.7-37.9 degrees C for rectal temperature and 35.6-37.2 degrees C for axillary temperature. Rectal temperature was higher than axillary in 98% of the measurements. The mean (SD) difference between rectal and axillary temperatures was 0.7 (0.5) degrees C, with a range of 3 degrees C. When used in hospital to detect high temperature, axillary temperature had a sensitivity of 73% compared with rectal temperature. This is too insensitive for accurate detection of an infant's high temperature. Rectal temperature measurement is safer than previously suggested: perforation has occurred in less than one in two million measurements. If an infant's temperature needs to be taken, rectal temperature should be used.