Eight healthy males were studied for a total of 13 subject-days to assess if gut (from an ingested pill) and axilla (from a thermally insulated skin probe) temperatures would act as a substitute for rectal temperature in field studies of the circadian rhythm of core temperature. Subjects slept and went about their activities, indoors and outdoors, normally. Regular recordings (at 6min intervals) were made of temperatures from the three sites. In addition, activity was measured (by a sensor on the nondominant wrist) so that the raw temperature data could be "purified," that is, corrected for the direct effects of sleep and activity. Inspection of the raw data indicated that there was a close parallelism between rectal and gut temperatures, but that the parallelism between rectal and insulated axilla temperatures was less reliable. This parallelism was supported by initial calculations of the correlations between rectal and gut temperatures (high and positive) and between rectal and insulated axilla (lower, though still positive) temperatures. Calculation of the limits of agreement between the parameters of the cosine curves fitted to the raw data confirmed that the rectal and gut temperatures were far closer with regard to acrophase and amplitude than were rectal and insulated axilla temperatures (-0.31 ± 0.89 vs. +0.75 ± 6.03 h and +0.002 ± 0.116 vs. +0.083 ± 0.625°C, respectively). After purification of the temperature data, the limits of agreement for the cosine parameters acrophase and amplitude still indicated that there was a closer agreement between rectal and gut temperatures than between rectal and insulated axilla temperatures (-0.30 ± 1.12 vs. +0.5 8 ± 6.69 h, and +0.007 ± 0.116 vs. +0.104 ± 0.620°C, respectively). Part of the explanation of this difference was the unreliable relationships between temperature changes in insulated axilla temperature and bursts of activity and going to bed. It is concluded that, whereas gut temperature is a viable alternative to rectal temperature (from the viewpoints of both user acceptability and the reliability of data obtained), insulated axilla temperature, though acceptable to subjects, is unreliable from an experimental viewpoint.