We use geodetic measurements from very long baseline interferometry (VLBI) to test whether the Pacific-North American plate velocity averaged over several years (1984--1987) of direct observation equals that averaged over millions of years as inferred from other data. We furthermore test whether the Pacific-North America velocity estimated from VLBI parallels the San Andreas fault, transform faults and earthquake slip vectors in the Gulf of California, and earthquake slip vectors along the Queen Charlotte Fault, along the Alaskan peninsula, and along the Kamchatkan peninsula. We assume that VLBI sites at Fairbanks (Alaska), Fort Davis (Texas), Richmond (Florida), Westford (Massachusetts), and Platteville (Colorado) lie on the North American plate and that VLBI sites on the Hawaiian island of Kauai, on the Marshall island of Kwajalein, and at Vandenberg Air Force Base (California) lie on the Pacific plate. The Euler vector (0.806¿/m.y. about 47.5¿N, 79.9¿W) that we estimate from VLBI is well constrained; its three-dimensional 95% confidence ellipsoid axes are only 5--15% as long as the Euler vector itself. The VLBI Euler vector is nearly identical to the Pacific-North America Euler vector of plate motion model NUVEL-1, which is determined from transform fault azimuths, earthquake slip vectors, and spreading rates from marine magnetic anomalies that integrate motions since 3 Ma.

The VLBI Euler vector predicts that Pacific-North America motions is on average 10¿¿4¿ clockwise of earthquake slip vectors along the Queen Charlotte fault, suggesting that these slip vectors are biased measures of the direction of relative plate motion. The VLBI Euler vector predicts that Pacific-North America motion is 6¿¿4¿ clockwise of the San Andreas fault in central California, reinforcing the conclusion from plate motion models that the San Andreas fault does not parallel Pacific-North American plate motion. The VLBI Euler vector predicts that Pacific-North America motion is parallel to most transform faults and earthquake slip vectors in the Gulf of California and to most earthquake slip vectors along the Alaskan and Kamchatkan peninsulas. The Pacific-North America velocity averaged over several years equals its velocity averaged over millions of years with the difference along the mutual boundary nowhere exceedng 4¿7 mm/yr. Therefore the motions between the Pacific and North American plates appear to be steady. Moreover, our results imply that one can combine rates averaged over intervals as long as millions of years with rates averaged over intervals as short as a few years when constructing models of lithosphere kinematics. ¿ American Geophysical Union 1990