Probabilistic seismic hazard (PSH) maps for southern California produced from the models of Ward (1994), the Working Group on California Earthquake Probabilities (1995), and the U.S. Geological Survey and California Division of Mines and Geology (Frankel et al., 1996; Petersen et al., 1996) show the peak ground accelerations predicted with each model to occur at 10% probability in 50 years, and the probability that 0.2 g will occur in 30 years, for “rock” site conditions. Differences among the maps range up to 0.4 g and 50%, respectively. We examine the locations and magnitudes of the differences as a basis to define the issues and avenues of research that may lead to more confident estimates of PSH in the future. Our analysis shows that three major factors contribute to the observed differences between the maps. They are the size of maximum magnitude assigned to a given fault, the proportion of predicted earthquakes that are distributed off the major faults, and the use of geodetic strain data to predict earthquake rates.