Paleomagnetic data have been obtained to address a problem at the Pallett Creek paleoseismological site: the 9 mm/yr slip rate determined from three-dimensional mapping of late Holocene offsets across discrete faults is only a quarter of the expected value. We suspected that nonbrittle deformation adjacent to the faults might account for the 26 mm/yr discrepancy. In our search for the missing slip we collected and analyzed 264 paleomagnetic samples from a 53-m-wide transect across the fault zone. Half the samples came from a unit deposited immediately after a large earthquake of about A.D. 1480; these samples were affected by two large earthquakes that involved rupture at the site in 1812 and 1857. We collected the other half of the samples from a slightly older bed, one that was deposited before the earthquake of about A.D. 1480. Relative to “control” groups composed of 10 samples and collected 50 m from the fault, samples closer to the fault display clockwise rotations of 30° or less. If interpreted as block rotations, the data from the older unit imply that it has sustained a total of 14.0 ± 2.9 m of dextral warp during the past three major earthquakes and that the younger unit has experienced a total of 8.5 ± 1.0 m of warp during the most recent two. Combining these values with the amounts of dextral slip across the mapped fault planes yields dextral offsets of 5.5, 6.25, and 6.25 m for the events of A.D. 1480, 1812, and 1857 and a slip rate of 35.6 ± 6.7 mm/yr. This slip rate, averaged over the past three complete seismic cycles, is consistent with published rates from other sites. Offsets associated with the past three events are remarkably similar. These amounts, however, appear independent of the length of interseismic cycles. These observations suggest (1) that this part of the San Andreas fault has a characteristic strength and (2) that conventional concepts of strain accumulation and relief (for example, time- and slip-predictable models of earthquake occurrence) are unrealistic.