This study documents three localities in the Franciscan accretionary complex of northern California, now adjacent to the San Andreas fault, that were overprinted thermally between 13.9 and 12.2 Ma: Point Delgada-Shelter Cove (King Range terrane); Bolinas Ridge (San Bruno Mountain terrane); and Mount San Bruno (San Bruno Mountain terrane). Vein assemblages of quartz, carbonate, sulfide minerals, and adularia were precipitated locally in highly fractured wall rock. Vitrinite reflectance (Rm) values and illite crystallinity decrease away from the zones of metalliferous veins, where peak wall-rock temperatures, as determined from Rm, were as high as 315??C. The ??18O values of quartz and calcite indicate that two separate types of fluid contributed to vein precipitation. Higher ??18O fluids produced widespread quartz and calcite veins that are typical of the regional paleothermal regime. The widespread veins are by-products of heat conduction and diffuse fluid flow during zeolite and prehnite-pumpellyite-grade metamorphism, and we interpret their paleofluids to have evolved through dehydration reactions and/or extensive isotopic exchange with accreted Franciscan rocks. Lower ??18O fluids, in contrast, evolved from relatively high temperature exchange between seawater (or meteoric water) and basaltic and/or sedimentary host rocks; focused flow of those fluids resulted in local deposition of the metalliferous veins. Heat sources for the three paleothermal anomalies remain uncertain and may have been unrelated to one another. Higher temperature metalliferous fluids in the King Range terrane could have advected either from a site of ridge-trench interaction north of the Mendocino fracture zone or from a "slabless window" in the wake of the northward migrating Mendocino triple junction. A separate paradox involves the amount of Quaternary offset of Franciscan basement rocks near Shelter Cove by on-land faults that some regard as the main active trace of the San Andreas plate boundary. Contouring of vitrinite reflectance values to the north of an area affected by A.D. 1906 surface rupture indicates that the maximum dextral offset within the interior of the King Range terrane is only 2.5 km. If this fault extends inland, and if it has been accommodating most of the strike-slip component of San Andreas offset at a rate of 3-4 cm/yr, then its activity began only 83-62 ka. This interpretation would also mean that a longer term trace of the San Andreas fault must be nearby, either offshore or along the northeast boundary of the King Range terrane. An offshore fault trace would be consistent with peak heating of King Range strata north of the Mendocino triple junction. Conversely, shifting the fault to the east would be compatible with a slabless window heat source and long-distance northward translation of the King Range terrane after peak heating.