Reexamination of geodetic data has shown that significant crustal deformation both preceded and followed the great 1906 earthquake (M=8.3). In the ~50 years prior to 1906, tensor shear strain rates averaged over an ~100-km-wide region spanning the San Andreas system ranged from 0.4 to 0.9¿10-6 yr-1, rates being higher east of the San Andreas fault than west of it. Part of this accumulated strain was released abruptly by seismic slip on the fault in 1906, but aseismic strain release continued at a rate of ~1.2¿10-6 yr-1 near the fault for about 30 years following the earthquake. Since that time the rate has been near 0.3¿10-6 yr-1 both close to the fault and in a broad region surrounding it. The mechanism suggested to account for these observations is the progressive failure of most of a 450-km-long segment of the San Andreas plate boundary during an ~80-year interval. The seismic slip, averaging near 4 m, is relatively shallow, being constrained by the data to depths no greater than about 10 km. The seismic moment determined from long-period surface wave amplitudes is 4.0¿1027 dyn cm and agrees well with the geodetic estimates. Continued aseismic slip of 3--4 m below the seismic zone to depths of ~30 km explains all of the postseismic data well. The small number of preearthquake observations are consistent with rapid slip on the fault at greater depths. Finally, the observation of a high rate of strain accumulation prior to 1906 casts considerable doubt on the often quoted 100-year periodicity of great earthquakes on the San Andreas fault, which appears to have originated with H. F. Reid's inferences based on the pre-1906 deformation.