Temperatures were measured using an electronic thermometer in an emergency department to determine the relationship between oral or rectal and axillary measurements. A total of 164 data pairs were obtained--95 in afebrile children, and 69 in febrile children. The correlation coefficient was .74 for oral-axillary pairs, and .70 for rectal-axillary pairs. The mean difference between oral and axillary temperatures was 1.17 degrees C +/- 0.72 degrees C, and between rectal and axillary temperatures was 1.81 degrees C +/- 0.97 degrees C. Using 37.4 degrees C (greater than or equal to 2 SDs) axillary as the upper limit of normal, the sensitivity, specificity, and positive and negative predictive values were calculated for detecting a fever. The sensitivity was 46%; specificity, 99%; positive predictive value, 97%; and negative predictive value, 72% for combined oral-axillary and rectal-axillary data. It was concluded that axillary temperatures are not sensitive enough to determine a fever when measured with an electronic thermometer. Electronic thermometers should be used to determine oral or rectal temperatures; axillary temperatures may be misleading and should be abandoned in the outpatient setting.