Nine healthy females were studied about the time of the spring equinox, while living in student accommodation and aware of the passage of solar time. After 7 control days, during which a conventional lifestyle was lived, subjects underwent a 24-h ‘constant routine’, followed by 17 ‘days’ on a 27-h ‘day’ (9 h sleep and 18 h wake). Throughout the experiment, regular recordings of (non-dominant) wrist activity (every 30 s) and rectal temperature (every 6 min) were made. Only the control and 27-h (experimental) ‘days’ have been analysed in the present report. From each subject, 24-h profiles of raw temperature (consisting of 240 points) were obtained: one (control days), by averaging the control days; the other (experimental days), by conflating 16 consecutive 27-h ‘days’. Activity data were first collected into 240 points by summing them over 6-min intervals; they were then converted into three data sets (each of 240 points) for control and, separately, for experimental days. These data sets were summed activities in the previous 18 min (A18), summed activity over the previous 18–30 min (A30), and summed activity over the previous 30–42 min (A42). The raw temperature data sets for control and experimental days were sepa-rately analysed by ANCOVA using two time-of-day factors: ‘hours’ (24 levels) and ‘six-minute-intervals’ (10 levels). The covariate was the three activity data sets; in order to make the analysis more versatile, a cubic polynomial model was used, with a linear, quadratic and cubic term for each of these activity data sets. Moreover, the effects of activity upon core temperature were separately assessed for four 6-h sections of the 24-h profile, centred on its low, rising, high and falling phases. The main results were as follows: 1. All three activity data sets made significant contributions to the model, but that by the A30 data set was the most powerful of the three. This supports the use of activity files covering the previous 30 min in other ‘purification’ methods. 2. Although the linear term was the one that was significant most frequently, quadratic and (negative) cubic terms were also present on several occasions. This result indicates that the effect of activity upon core temperature can be approximated by a linear function (as has been done in other ‘purification’ models), but that, with wider ranges of activity, a sigmoid curve would be more accurate, indicative of the process of thermoregulation. 3. During the experimental days, the effect of activity upon temperature was greater in the rising than the falling temperature phase, and greater in the low than in the high phase. These results are predicted from current understanding of the circadian rhythm of thermoregulation. 4. During control days, the effects of activity were more complex, probably due to the factors that were present at some, but not all, phases — factors such as sleep, meals, changes in posture, lighting, and so on. 5. The ANCOVA also enabled the temperature profiles, corrected for the effects of activity (and, therefore, to be considered as ‘purified’), to be displayed. We conclude that the use of ANCOVA to tackle the problem of ‘purifying’ raw temperature data is a promising one. So far, it has produced results that accord with those from other ‘purification’ methods and with predictions based on our current understanding of the circadian rhythm of thermoregulation.
Read full abstract