We present NNR-NUVEL1, a model of plate velocities relative to the unique reference frame defined by requiring no-net-rotation of the lithosphere while constraining relative plate velocities to equal those in global plate motion model NUVEL-1 [DeMets et al., 1990>. Differences between NNR-NUVEL 1 and no-net-rotation plate motion model AM0-2 [Minster and Jordan, 1978> are as large as 15 mm/yr. In NNR-NUVEL1, the Pacific plate rotates in a right-handed sense relative to the no-net-rotation reference frame at 0.67¿/m.y. about 63¿S, 107¿E. This rotation nearly parallels, but is 32% slower than, the Pacific-hotspot Euler vector of HS2-NUVEL1, which is a global model of plate-hotspot velocities constrained to consistency with NUVEL-1 [Gripp and Gordon, 1990>. At Hawaii the Pacific plate moves relative to the no-net-rotation reference frame at 70 mm/yr, which is 25 mm/yr slower than the Pacific plate moves relative to the hotspots. Differences between NNR-NUVEL1 and HS2-NUVEL1 are described completely by a right-handed rotation of 0.33¿/m.y. about 49¿S, 65¿E. The 99% confidence ellipsoids for plate-hotspot motion in model HS2-NUVEL1 exclude the corresponding Euler vectors from model NNR-NUVEL1. Thus the no-net-rotation reference frame differs significantly from the hotspot reference frame.

If the difference between reference frames is caused by motion of the hotspots relative to a mean mantle reference frame, then hostpots beneath the Pacific plate move with coherent motion towards the east-southeast. Alternatively, the difference between reference frames may show that the uniform drag, no-net-torque reference frame, which is kinematically equivalent to the no-net-rotation reference frame, is based on a dynamically incorrect premise. Possible exceptions to the assumption of uniform drag include a net torque on the lithosphere due to attached subducting slabs and greater resistance to plate motion beneath continents than beneath oceans. ¿American Geophysical Union 1991