New paleomagnetic results from the Karmutsen Formation (Late Triassic) of Vancouver Island confirm the presence of two families of magnetizations (X and Y), both of which are inconsistent with known Mesozoic and Cenozoic geomagnetic fields of cratonic North America. The X magnetizations have coherent directions with the exception of a subset of five sites (the B subset). We argue that the deviation of the B subset is caused either by a 31 ± 13 °anticlockwise rotation of a small block relative to the main sampling areas or by a short-term excursion of the field. The X magnetization has an overall mean direction 008°, −33 °α95 = 6 °based on results from 147 oriented samples (usually 2 specimens from each) collected at 28 sites spanning about 6000 m stratigraphically. We interpret this as the original Late Triassic magnetization. The corresponding X paleopole (21°N, 44°E A95 = 6°) is strongly far-sided and right-handed with respect to the Mesozoic apparent polar wander path for cratonic North America. The paleolatitude indicated for Vancouver Island in the Late Triassic is either 18°N or 18°S, the latter being preferred on the grounds that it yields a more consistent pattern for Cordilleran magnetizations, but the ambiguity is still not settled. In either case the results show that Vancouver Island was far south of its present position relative to North America in the Late Triassic, thus confirming the previous results of Irving and Yole. The Y magnetizations, with more heterogenous properties, occur at 14 sites (66 oriented cores, usually 2 specimens each). Y magnetizations are generally softer than X and for this and other reasons we regard them as secondary and post-Triassic in age. Individual site poles for the Y magnetization are, with minor exceptions, right-handed and slightly far-sided with respect to the apparent polar wandering path for cratonic North America. The mean paleopole for Y magnetizations is situated at 70°N, 15°W A95 = 11°. Both the X and Y magnetizations are consistent with either northward motion of the westernmost Cordilleran elements accompanied by clockwise rotation, or with oblique translation from the southwest. The northward component of motion derived from X directions would be the same in both instances and amounts to 1300 or 4900 km depending on whether the northern or the southern paleolatitude option is chosen. Our preference is for the latter and we present arguments which suggest that Vancouver Island may have been originally derived from a region near to eastern Gondwana or from a block east of Gondwana that might have included Malaysia. The procedures used for the tectonic analysis of aberrant paleopoles are described in the Appendix.