The San Andreas fault in California has long been thought to be far weaker than given by the Anderson-Byerlee state because of the lack of a heat-flow anomaly adjacent to the fault as predicted by a steady-state conductive model of frictional heating. Here this issue is reexamined by testing whether stress data in California are consistent with the weak-fault hypothesis for the San Andreas fault, as implied by the heat-flow model, or with the San Andreas fault obeying Anderson-Byerlee mechanics.