The reliability of daily routine oral temperature measured by the nursing staff with an electronic thermometer was tested by comparing the measurements with rectal temperature measured with a mercury thermometer. In each patient, rectal temperature was measured by the author just after the oral temperature measurement. The nursing staff was not informed about the investigation. In a small group of 12 patients with clinical suspicion of fever, six patients had normal oral temperatures (less than or equal to 37 degrees C). Three of these had rectal temperatures greater than or equal to 38 degrees C. In another group of 11 patients selected because of a normal oral morning temperature (less than or equal to 36.5 degrees C) seven had slightly elevated rectal temperatures (37.1-37.3 degrees C). Thus, a significant number of patients in the two groups had false normal oral temperatures. In five of the 12 and four of the 11 patients, the rectal temperature was 1 degree C or more above the oral temperature. The reason for measuring too low oral temperatures seems to be difficulties in avoiding external influence. The use of different measuring points each with its own lower limit of fever might lead to misinterpretation of the patients' temperature curve. Abolition of the mercury thermometer represents environmental progress. Unfortunately the sensor of the electronic thermometer, which is discarded after use, contains considerable amounts of different metals, mainly copper but also heavy metals. The rectum should remain the site of measurement of body temperature and the digital reuse thermometer the measuring device.