A newly developed, very long-term ( approximately 7 days) ambulatory monitoring system for assessing beat-to-beat heart rate variability (HRV) and body movements (BM) was used to study the mechanism(s) responsible for the long-period oscillation in human HRV. Data continuously collected from five healthy subjects were analyzed by 1) standard auto- and cross-spectral techniques, 2) a cross-Wigner distribution (WD; a time-frequency analysis) between BM and HRV for 10-s averaged data, and 3) coarse-graining spectral analysis for 600 successive cardiac cycles. The results showed 1) a clear circadian rhythm in HRV and BM, 2) a 1/f (beta)-type spectrum in HRV and BM at ultradian frequencies, and 3) coherent relationships between BM and HRV only at specific ultradian as well as circadian frequencies, indicated by significant (P < 0.05) levels of the squared coherence and temporal localizations of the covariance between BM and HRV in the cross-WD. In a single subject, an instance in which the behavioral (mean BM) and autonomic [HRV power >0.15 Hz and mean heart rate (HR)] rhythmicities were dissociated occurred when the individual had an irregular daily life. It was concluded that the long-term HRV in normal humans contained persistent oscillations synchronized with those of BM at ultradian frequencies but could not be explained exclusively by activity levels of the subjects.