INTRODUCTION: Long-term heart rate variability (HRV) measures, including the standard deviation of means of successive 5-minute epochs of R-R interval intervals (SDANN) and the power law slope (beta), are important prognostic measures, yet their physiologic basis is unknown. We tested the hypothesis that long-term HRV arises from physical activity in a randomized cross-over study in patients with rate-responsive pacemakers. METHODS AND RESULTS: Ten patients with complete heart block and dual-chamber pacemakers underwent 24-hour periods of ambulatory ECG in each of three pacing modes: atrially tracked, fixed-rate, and rate-responsive pacing. SDANN, ultra low frequency (ULF; frequencies <0.0033 Hz), and beta slope were calculated; and high-frequency power and root mean square of consecutive normal R-R intervals (rMSSD) were calculated as measures of short-term HRV, which have autonomic origins. Long-term HRV measures were similar with atrially tracked and rate-responsive pacing and were much greater than in fixed-rate pacing (SDANN P = 0.0001; ULF P = 0.0001; beta slope P = 0.0002). Short-term HRV measures were similarly low in fixed-rate and rate-responsive pacing (P = NS) and were significantly lower than with atrially tracked pacing (P = 0.0034). CONCLUSION: Rate-responsive pacing reproduces long-term, but not short-term, measures of HRV, suggesting that they may be markers of heart rate responses to patient activity.