When people are asked to estimate the chances of logically related events, answers often exhibit probabilistic incoherence. We describe a method for finding a coherent approximation to incoherent estimates of probability, and present some preliminary data on its use.