To test the hypothesis that effective refractory period (ERP) testing is an accurate and reproducible measure of the cardiac refractory period, 4 pigs and 12 dogs were studied. We define the conventional ERP as the longest S1-S2 interval associated with noncapture and the triplicate ERP as the average of three ERP values determined by the up-down algorithm. We also determined the probability curve for noncapture after testing multiple S1-S2 intervals. The results showed that the difference between the coupling intervals associated with a 0 and a 100% probability of noncapture was 13 +/- 1 (range 12-16) ms for pigs and 9 +/- 2 (range 4-12) ms for dogs. The conventional and the triplicate ERP were associated with a 71 +/- 20% (range 26-100%) and a 60 +/- 21% (range 12-100%) probability of noncapture, respectively. We conclude that the ERP is a probability function. Therefore a single ERP determination cannot be used as an accurate and reproducible measure of cardiac refractory period unless it is used to detect potential differences in ERP that exceed 16 ms in pigs or 12 ms in dogs. A probability function curve determination will be needed if smaller differences in the ERP are to be detected.