To test the hypothesis that ventricular fibrillation (VF) threshold testing is a probability function, 12 open-chest dogs were studied. The VF thresholds were tested by scanning the T wave with either the single premature stimulus method or with the train stimulus method. The dose-response curve method was used to determine the probability of inducing VF with different strengths of premature stimulation. Conventional methods and up-down methods were also used to test the VF threshold. The results showed that the VF threshold is a probability function. The conventional method VF threshold for the single premature stimulus and for the train stimulus methods corresponded to the current strength associated with a 23.4 +/- 13.2 and a 33.2 +/- 19.5% probability of VF (P < 0.05), respectively. In comparison, the triplicate VF threshold determined by the up-down algorithm method resulted in values that were not significantly different from a 50% probability of inducing VF, with a good correlation (r = 0.90, P < 0.001 for single and r = 0.89, P = 0.003 for the train stimulus method). We conclude that: 1) The VF threshold is a probability function, and 2) the triplicate VF threshold determined by the up-down algorithm method is the best alternative to the dose-response curve method in estimating 50% probability of inducing VF.