OBJECTIVE To compare temperatures obtained with glass-mercury axillary and aural infrared thermometers with temperatures obtained with glass-mercury rectal thermometers. METHODS Blind comparison with criterion standard. METHODS Well-baby nursery at a private teaching hospital. METHODS Convenience sampling of 200 term newborn infants. METHODS Temperatures were measured simultaneously with glass-mercury rectal and axillary thermometers for a 3-minute period by one investigator. In a blinded fashion, a second investigator obtained three aural temperatures by using two tympanic membrane thermometers. One tympanic membrane reported infants' rectal-equivalent temperatures, and a second reported oral equivalent temperatures. RESULTS Temperatures obtained with glass-mercury rectal and axillary thermometers for the population were similar (37 degrees +/- 0.4 degree C vs 36.8 degrees +/- 0.3 degrees C, respectively), but 25% of measurements at these two sites differed by > 0.3 degree C. Oral-equivalent tympanic membrane temperatures were more accurate than rectal-equivalent temperatures in predicting an infant's glass-mercury axillary and rectal temperatures (75% of oral-equivalent temperatures vs fewer than 50% of rectal-equivalent temperatures were within 0.3 degree C of either glass-mercury rectal or axillary measurements). CONCLUSIONS Temperatures obtained with glass-mercury axillary and rectal thermometers are similar in most cases. However, temperatures obtained with tympanic membrane thermometers either in the rectal-equivalent mode or in the oral-equivalent mode did not accurately reflect an infant's rectal or axillary temperature. We believe that tympanic membrane temperatures should not be substituted for rectal or axillary temperatures in assessments of newborn infants.