We combine updated GPS velocities from the Nubian (NU), Eurasian (EU), and North American (NA) plates with 500 new 3.16-Myr-average seafloor spreading rates and nine transform fault azimuths from the northern Atlantic and Arctic basin seafloor spreading centers to estimate and test for changes in the relative motion between these plates. The numerous new seafloor spreading rates and GPS velocities improve our ability to detect recent changes in the relative motions of these plates. The angular velocity vector that best fits the EU–NA GPS velocities lies significantly north of the 3-Ma-average pole, in accord with previously published geologic evidence that the EU–NA pole has migrated northward since ∼3 Ma. Although we also find evidence for a significant post-3-Ma change in NU–NA motion, it is less compelling because the Nubian plate GPS velocity field is sparse and NU–NA seafloor spreading rates appear to have remained steady within the 1 mm yr−1 uncertainties if we systematically decrease the seafloor spreading rates to correct for outward displacement of seafloor spreading magnetic lineations. The NU–EU pole derived from GPS site velocities lies more than 30 angular degrees south of the tightly constrained 3-Ma-average estimate and predicts significantly slower and more oblique present-day NU–EU convergence in the Mediterranean. Both models for NU–EU motion pass a key test for their accuracy, namely, they correctly predict strike-slip motion along the well-mapped Gloria fault east of the Azores. The change to more oblique NU–EU motion may reflect increasing difficulty in maintaining margin-normal convergence within this continent–continent collision zone.