A slip-predictable stochastic model is presented based on Markov renewal theory. Times between successive events are assumed to be Weibull-distributed with an increasing hazard rate. The model forecasts probabilities of earthquake occurrences conditional on the time of occurrence of the last event. The hazard along the Middle America Trench, Mexico, where data has suggested slip-predictable behavior, is obtained for illustrative purposes. Comparisons of this model with the Poisson model show that probability forecasts are underestimated with the Poisson model when there has been a seismic gap. The slip-predictable stochastic model presented in this paper represents a considerable improvement over the existing seismic hazard analysis methods. However, further testing of the model is necessary prior to wider application.