A signal-estimation theory approach to the problem of optimizing defibrillation thresholding techniques leading to an optimum stepping algorithm is presented. The results guarantee maximum average performance by achieving the minimum average cost. It is shown that Bayesian statistics, which can be adapted to any preferred stepping algorithm, give estimates for the defibrillation threshold of minimum average cost. An example of mean squared error (MSE) estimation for the two-step up-down algorithm, which minimizes the variance of the estimates, is analyzed for a variety of conditions. It is concluded that a large variety of defibrillation thresholding techniques can be analyzed by minimizing the cost of a specific experiment and the total average cost. The resulting stepping algorithms and estimators will be optimum for the selected cost functions.