We examined the hypothesis that respiratory sinus arrhythmia (RSA) is primarily a central phenomenon and thus that RSA is directly correlated with respiratory controller output. RSA was measured in nine anesthetized dogs, first during spontaneous breathing (SB) and then during constant flow ventilation (CFV), a technique whereby phasic chest wall movements and thoracic pressure swings are eliminated. Measurements of the heart rate and of the moving time averaged (MTA) phrenic neurogram during these two ventilatory modes were made during progressive hypercapnia and progressive hypoxia. RSA divided by the MTA phrenic amplitude (RSAa) showed a power-law relationship with both arterial carbon dioxide partial pressure (PaCO2) and oxygen saturation (SaO2), but with different exponents for different conditions. However, the power-law relation between RSAa and respiratory frequency had an exponent indistinguishable from -2 whether hypoxia or hypercapnia was the stimulus for increased respiratory drive, and during both CFV and spontaneous breathing (-1.9 +/- 0.4, hypoxia, SB; -1.8 +/- 0.7, hypoxia, CFV; -2.1 +/- 0.8, hypercapnia, SB; -1.9 +/- 0.7, hypercapnia, CFV). We conclude that respiratory sinus arrhythmia is centrally mediated and directly related to respiratory drive, and that changes in blood gases and phasic afferent signals affect RSA primarily by influencing respiratory drive.