Quantitative estimates of the rates and azimuths of Phoenix plate convergence with the Antarctic Peninsula have been derived from plate rotation calculations for two periods in the Late Cretaceous and early Tertiary (83.0–67.7 and 61.1–53.4 Ma). Published marine magnetic anomaly identifications and ‘flow lines’ interpreted from gravity anomaly maps were used in simple spherical geometry calculations to derive Phoenix–Pacific stage rotations. These were combined with published Pacific–Antarctic rotation data to determine contemporaneous Phoenix–Antarctic stage rotations. The results indicate a significant change in azimuths of relative motion between the Late Cretaceous and early Tertiary. Late Cretaceous, and perhaps earlier, oblique subduction probably caused migration of fore-arc slivers along the margin, resulting in variations in width of the accretionary prism. Comparison between synthetic magnetic profiles and a 900 km long magnetic profile across ocean floor produced at the Antarctic–Phoenix ridge during the early Tertiary establishes the time of a major decrease in spreading rate, and hence also in convergence rate, as chron C23r (52.3 Ma). The associated change in subduction dynamics may have caused the dextral transtensional deformation observed in the George VI Sound region and initiated uplift of blueschist facies rocks now exposed on Smith Island. The calculated convergence history does not provide a simple explanation for the occurrence of high-Mg# andesite lavas on Alexander Island, which were erupted more than 20 Ma before the Antarctic–Phoenix ridge arrived at the margin. The existence of these lavas implies either earlier subduction of another ridge or slab break-up beneath Alexander Island.