An infrared tympanic membrane thermometer (FIRST temp) said to approximate core temperature accurately is being marketed as a noninvasive, quick, and easy-to-use clinical instrument. The determination of tympanic membrane temperatures by this device was compared with the determination of oral, rectal, or axillary temperatures by a conventional glass thermometer. Subjects were patients of a pediatric group practice in Houston, Texas. FIRST temp and conventional temperature determinations on individual patients were completed within 5 minutes of each other. The presence or absence of otitis media was noted by the examining physician. Agreement between the two methods of temperature determination was assessed by calculating limits of agreement within which 95% (±2 standard deviations) of individual differences would fall. The location of conventional thermometer (oral, rectal, axillary), time interval between the two separate measures, and the presence or absence of otitis media were entered into a multiple regression analysis to determine whether these factors influenced the observed differences between the two methods. A total of 144 patients were enrolled in the study; oral comparisons were obtained in 92 (57%) patients, rectal in 35 (24%), and axillary in 29 (19%). The upper and lower limits of agreement between temperature methods were 1.12°C and 0.89°C and the mean difference was –0.12°C. Regression analysis revealed that only the site of conventional thermometer location (oral, rectal, axillary) was a significant predictor of FIRST temp/conventional differences. Each site had a range of agreement greater than 1.65°C; axillary temperatures had the greatest range (–0.94°C to +1.30°C). Although ease and time reduction with use are attractive properties of the FIRST temp device, these data show that it is unreliable compared with conventional methods of temperature determination.