Samples for paleomagnetic analysis were collected from red radiolarian ribbon chert in a Franciscan-like subduction complex (Western Baja terrane) exposed along the western margin of the Baja California Peninsula. A 40-m-thick section of chert was sampled on Cedros Island, and a 30-m-thick and two smaller sections of chert were sampled on the San Benito Islands to the west. A site in pillow basalt underlying the chert also was sampled on Cedros Island. The sections of chert on Cedros Island and the San Benito Islands contain radiolaria ranging in age from Late Triassic to Early Cretaceous. All of these rocks have similar normal-polarity direction of magnetization, indicating that they were remagnetized long after deposition. Results from a chert site, including an intrastratal fold, also indicate that the magnetization is postdepositional. A significantly improved grouping (95% confidence level) of directions after correction for the structural attitudes of the Cedros and San Benito chert sections, however, implies that the remagnetization predates breaking up and incorporation of the chert into the subduction complex. Studies of modern subduction zones imply that the attitude of the ancient subducting slab, including the chert sections, was near horizontal (dip angles <10° at the time of remagnetization. Thus, assuming paleohorizontality, a paleolatitude of 12° ± 5° N can be estimated for the Western Baja terrane in mid-Cretaceous time. The paleomagnetic directions also indicate a subsequent poleward translation of 25° ± 7° in latitude (2,800 km) and a clockwise rotation of 56° ± 7° of the Western Baja terrane with respect to North America. The paleolatitude error limits at the 95% confidence level represent minimum values, because deviations of the bedding planes from horizontal during remagnetization would contribute to the error. A 10° NE dip of the subducting slab, however, would account for only 4° of the paleolatitude anomaly, and a ≃60° NE dip would be needed to account for the entire anomaly. Modern analogs, in combination with plate-reconstruction models, imply that the northward translation of the Western Baja terrane along the ancient convergent margin of North America could have been accomplished by arc-parallel strike-slip faulting that was driven by oblique subduction.