In a study of A-V and V-A propagation 61 dogs were used, anesthetized with pentobarbital, with the heart denervated, the adrenals ligated, and the S.A. node crushed. In 15 hearts, unusually long A-V or V-A latencies were recorded. Changes in the stimulus intervals often caused abrupt shortenings to the usual values, or resulted in alternations between the short and long latencies. It is inferred that there exist normally two processes or paths for A-V and V-A propagation. If the rapid process is open, it will be used. But if it becomes blocked, the slow process is still available and allows propagation.