A stimulus given in the T-wave (a ULV stimulus) of a paced cardiac cycle can induce ventricular fibrillation (VF). However, if the ULV stimulus strength is increased, the probability of inducing VF drops in a dose-response fashion. This paper presents a hypothesis which explains this dose-response relationship. We hypothesize that the dose-response relationship arises from the variation in the location of the induction of rotating wavefronts. If the wavefronts are initiated near an anatomical obstacle, such as the A-V groove, they have a high probability of wandering into the obstacle and reducing to quiescence. If the wavefronts are initiated far from such an obstacle, they have a high probability of surviving long enough to disintegrate into VF. Since increasing the stimulus strength can reduce the distance between the A-V groove and the location of the wavefronts, the probability of inducing VF drops with increasing stimulus strength. Modeling results are presented to support a key aspect of this hypothesis. From the hypothesis, several theoretical predictions are made. It is shown that the experimental evidence supports several of these predictions.