The gradual transitions that are found between unsuccessful and successful shock strengths in percent success or dose-response curves suggest that defibrillation is a probabilistic phenomenon. This concept appears to be reinforced by the fact that a frequency distribution is observed in defibrillation threshold data and that a dose-response relationship is also obtained by integration of the frequency distribution. The purpose of this study was to investigate whether a deterministic threshold model (based on experimental results) could produce 1) gradual transitions in dose-response curves, and 2) a threshold frequency distribution for individual subjects. In the experimental phase of the study, a linear deterministic relationship was found between transthoracic threshold current and defibrillation episode number (other variables held constant) in pentobarbital-anesthetized dogs. The correlation coefficient for each dog was between 0.77 and 0.98 (P less than 0.01), and both positive and negative slopes were found. Based on these results, threshold current was modeled for computer simulation as a linear function of episode number. The model was thus purely deterministic with no random variability. For each simulated experiment, several parameters were varied: order of shocks (increment, decrement, random order), slope of threshold function, and percent error of the initial threshold. Several hundred computer simulations were performed to determine the effect of varying these parameters. In all cases, threshold-frequency distributions and sigmoidal dose-response curves with gradual transitions were produced. The results of this investigation demonstrate that the apparent probabilistic behavior of defibrillation can be produced by a deterministic relationship.