BACKGROUND Atrioventricular (AV) conduction time varies on a beat-by-beat basis in response to the influences of cardiac efferent autonomic activity and rate-dependent electrical recovery processes. The goals of this study were to distinguish these effects on AV conduction time and to compare the variability in sinoatrial and AV nodal function. RESULTS The PR interval on the surface ECG served as an index of AV conduction time in this study of 14 adult human subjects undergoing a random interval breathing protocol. P and R waves were located by a template-matching algorithm. Spectral analysis allowed frequency-domain comparisons between PR and RR interval variability. Spectra of PR and RR intervals had similar power distributions, although the power of the RR interval spectra was much greater. Autonomic blockade with atropine plus propranolol reduced the power of both spectra. Standing significantly decreased the spectral power from 0.15 to 0.5 Hz for PR and RR spectra, and introduced a peak near 0.1 Hz in the mean PR and RR spectra, although the latter finding was significant only for the RR interval spectra. Propranolol had no significant effects on the PR and RR interval spectra. Linear regression analysis allowed quantification of the autonomic and recovery effects on AV conduction and showed which effect predominated. Simple linear regression confirmed in adults a previous finding in children that conduction time may be either positively or negatively correlated with cycle length. By multiple regression and transfer function analysis, the inverse relation seen in some subjects was attributed to the effect of recovery from the preceding cycle. With the preceding recovery period accounted for, the conduction time and cycle length of the current beat were positively correlated, presumably due to the parallel autonomic effects on the sinoatrial and AV nodes. The magnitude of the recovery effect predicted by the regression analysis was similar to published values. CONCLUSIONS A noninvasive evaluation of the surface ECG can be used to compare variability in AV conduction time and cycle length and characterize the effects of autonomic efferent activity and rate-related recovery on AV nodal function.