Fault segments identified from geologic and historic evidence have sometimes been adopted as features limiting the likely extends of earthquake ruptures. There is no doubt that individual segments can sometimes join together to produce larger earthquakes. This work is a trial of an objective method to determine the probability of multisegment ruptures. The frequency of occurrence of events on all conjectured combinations of adjacent segments in northern California is found by fitting to both geologic slip rates and to an assumed distribution of event sizes for the region as a whole. Uncertainty in the shape of the distribution near the maximum magnitude has a large effect on the solution. Frequencies of individual events cannot be determined, but it is possible to find a set of frequencies to fit a model closely. A robust conclusion for the San Francisco Bay region is that large multisegment events occur on the San Andreas and San Gregorio faults, but single-segment events predominate on the extended Hayward and Calaveras strands of segments.