This article summarizes the 1988, 1990, 1995, and 2002 Working Groups on California Earthquake Probabilities (wgceps). Each of these studies used the best available science to make a time-dependent earthquake forecast. All involved applying elastic-rebound-theory–motivated recurrence models to estimate the probability of rupture on discrete fault segments. The focus was necessarily limited to those faults that had sufficient information to make a time-dependent analysis, although the later studies included probabilities for other events as well. This manuscript does not give a point-by-point critique or review of the previous wgceps, both because it would take considerably more text and because such discussions will necessarily be documented by future working groups as they justify changes. One opinion that is emphasized, however, is that future wgceps should resist overcomplexity in certain aspects of the model given basic assumptions that have been made and/or limitations in our understanding of the earthquake system. For example, one might argue that the level of sophistication applied by previous wgceps with respect to recurrence-interval variability and uncertainty was overkill given their basic assumptions that large earthquakes obey segment boundaries, ruptures never jump from one fault to another, and earthquake-clustering effects can be ignored.