We use magnetic anomaly and fracture zone crossings from 3 ¿N--24 ¿N along the East Pacific Rise to define the boundaries and relative velocities of the rigid Pacific, Rivera, North American, and Cocos plates since 0.78 Ma and to test several hypotheses regarding the nature of deformation between these plates. Crossings of the 0.78 Ma isochron from the Pacific-Rivera rise north of 22.0 ¿N are significantly better fit by the 0--0.78 Ma Pacific-North America rotation than by a Pacific-Rivera rotation, thereby supporting a recently proposed model for accretion of the northern part of the Rivera plate to North America. A significant misfit of the Pacific-Cocos rotation to crossings of anomaly 1n from the East Pacific Rise north of 16.4 ¿N suggests that seafloor north of ~16 ¿N moves relative to the rigid Pacific or Cocos plates. We also find that nine crossings of the eastern Rivera fracture zone and eight crossings of anomaly 1n flanking the Manzanillo spreading segment are fit well by a Pacific-Rivera rotation. The Rivera-Cocos angular velocity derived from the best fitting Pacific-Cocos and Pacific-Rivera angular velocities predicts 19¿3 mm yr-1 (95% confidence limit) of ~N-S convergence near the center of the Rivera-Cocos boundary. Nearby left-lateral, strike-slip earthquakes with north trending fault planes suggest that convergence is accommodated via sinistral shear along a north trending zone. The unambiguous kinematic and seismologic evidence for significant Rivera-Cocos motion indicates that the absence of a well-defined Rivera-Cocos plate boundary cannot be interpreted as evidence that Rivera-Cocos motion is slow or zero, as is postulated in several previous studies. The Rivera-Cocos angular velocity predicts oblique Rivera-Cocos convergence at the El Gordo graben and beneath the Manzanillo trough and southern Colima graben. Any extension across these features since 0.78 Ma thus cannot have resulted from divergence between the subducting Rivera and Cococs plates.¿ 1997 American Geophysical Union