To test hypotheses relating the absolute velocities of the plates to driving models for the present plate system, we have determined the absolute plate motions in the early Tertiary for a number of force models using the assumption that these forces exert no net torque on the lithosphere. All absolute motion models are based on a self-consistent set of relative plate velocities obtained from published finite plate rotations and on the Jurdy and Van der Voo map of plate boundaries at 55 m.y. B.P. The derived pattern of plate velocities in the early Tertiary is consistent with paleomagnetic and paleosedimentation data and differs in important respects from the present pattern: (1) The clear separation of the plate population at present into fast (7--10 cm/yr) oceaic plates and slow (generally 0--2 cm/yr) continental plates does not occur 55 m.y. ago. The rms absolute velocities of subducted oceanic plates at 55 m.y. are neither uniformly fast (4--10 cm/yr) nor consistently greater than those for dominantly continental plates (0--5 cm/yr.) (2) There is a tendency at 55 m.y., as at present, for lithosphere in the equatorial half of the earth to move faster than in the polar half, though at 55 m.y. the coordinate pole giving the highest equatorial rms velocity is far from the geographic pole. (3) The predicted motions do not agree with the hypothesis of fixed hot spots and the observed trends of seamount chains and aseismic ridges, the poorest agreement being in the Pacific. From point 1 we infer that viscous drag beneath continents is not significantly greater than that beneath oceanic lithosphere and that forces at subduction zones are not the sole regulators of the velocities of subducted oceanic plates. Observation 2 weakens the case for a possible connection between plate speeds and earth rotation. From point 3 we concur with others that either substantial relative motion among the asthenospheric roots of hot spots is required or an origin unrelated to hot spots for at least some of the seamount chains and aseismic ridges must be postulated. The bends in the acific seamount chains may be associated with a reorientation of stress in the Pacific plate as its northern boundary switched from a ridge to a subduction zone.