Past thermal comfort research has shown differences in the thermal sensation votes given in field and laboratory settings. However, such research tends to compare the votes of different groups of people in different environments rather than comparing the same people in each environment. Therefore, a two-phase study was conducted of the thermal comfort of 30 BRE employees in their home, in their office, and in a climate chamber. In the first phase each subject spent two 3 h sessions in each environment and the temperature was adjusted between sessions within the range 18-26 degrees C. Data loggers were used to record the air and mean radiant temperature, air velocity and relative humidity; subjective ratings of thermal sensation were obtained using questionnaires. The subjects wore the same clothing in each session and were allowed to conduct only sedentary activities. The reported thermal sensation votes were compared with those predicted using ISO 7730. The observed neutral temperatures for each of the three environments differed by up to 2 degrees C and were up to 1 degrees C different to those predicted. This finding has implications for energy use. In the second phase, the subjects were studied in their home and office only. No restrictions were imposed on clothing, previous or current activities, or environmental conditions. The observed thermal sensation votes were very poorly correlated with those predicted and with operative temperature.