The atrioventricular (AV) node responds in a complex fashion to changes in activation rate. A variety of approaches have been used to explain these dynamic AV nodal responses, but none has been able to account fully for AV nodal behavior. Three specific rate-dependent properties of the AV node have been described: 1) time-dependent recovery after excitation, 2) an effect of short cycles to advance recovery ("facilitation"), and 3) a gradual slowing of conduction in response to sustained, high-frequency activation ("fatigue"). We hypothesized that a model incorporating quantitative descriptors of all three processes might be able to account for a wide variety of AV nodal behaviors. Quantitative descriptors of AV nodal recovery, facilitation, and fatigue were developed based on AV nodal conduction changes during selective pacing protocols in seven autonomically blocked dogs. These descriptors were incorporated into a set of mathematical equations that define AV nodal conduction of any beat based on activation history. The equations were then applied to predict pacing-induced Wenckebach periodicity in each dog. Experimental data were obtained after nine to 19 step decreases in atrial cycle length into the Wenckebach zone in each animal. Observed behaviors included complex patterns of block, a progressive increase in the level of block over 5 minutes of rapid pacing, and periods of alternating patterns of block. The model accurately predicted the onset of AV block at each cycle length, the relation between conduction ratio and cycle length as a function of time, and the changing patterns of Wenckebach periodicity during sustained atrial pacing. All three terms of the model equation (describing recovery, facilitation, and fatigue) were essential to account fully for the observed behaviors. Elimination of AV nodal fatigue from the model resulted in failure to account for time-dependent changes in Wenckebach patterns, whereas exclusion of facilitation led to consistent overestimation of the degree of AV block at each cycle length. We conclude that a mathematical model incorporating terms to describe recovery, facilitation, and fatigue accurately predicts a wide range of Wenckebach-type behavior and that complex conduction patterns of the AV node can be fully accounted for by simple functional AV nodal properties.