Shivering (electromyographic activity of the pectoral muscle), oxygen consumption, and body temperature were measured from undisturbed pigeons for periods of several weeks, and segments from the midparts of each phase of the light-dark cycle were compared at various ambient temperatures and feeding regimes. Behavior was recorded with a video camera. None of the observed types of behavior (e.g., walking, preening, feeding, drinking, pecking, defecation) induced spurious electrical activity in the pectoral muscle. On the other hand, none of these behaviors directly inhibited ongoing shivering. There was no difference in the mean level of shivering between the light (L) and dark (D) phases of the day in any of the conditions, although body temperature was 2 degreesC higher during L. Measurements of integrated electromyogram (EMG) with high temporal resolution (28 samples/s) showed that, at 1 degreesC, shivering in the pectoral muscle was present for more than 98% of the time. Plots of oxygen consumption against root mean square EMG were obtained in each condition by a filtering procedure that excludes data points in which oxygen consumption is affected by motor activity. These plots showed that the increase in heat production induced by a unit increase in pectoral EMG was lower in D than in L and that it was further lowered by fasting. The amplitude spectra of raw EMG signals were similar in all conditions. Spectra of demodulated (rectified, low-pass filtered) EMG showed a distinct rhythmicity around 8 Hz at 21 degreesC that was further enhanced by fasting but absent at 1 degreesC. This suggests that the degree of synchronization and pattern of recruitment of motor units are specific for various temperatures and feeding regimes, and may partly explain the variable relation between heat production and muscle electrical activity. The results emphasize the advantages of long-term measurements for understanding the control of thermogenesis in birds.