Plate motions relative to the hotspots over the past 4 to 7 Myr are investigated with a goal of determining the shortest time interval over which reliable volcanic propagation rates and segment trends can be estimated. The rate and trend uncertainties are objectively determined from the dispersion of volcano age and of volcano location and are used to test the mutual consistency of the trends and rates. Ten hotspot data sets are constructed from overlapping time intervals with various durations and starting times. Our preferred hotspot data set, HS3, consists of two volcanic propagation rates and eleven segment trends from four plates. It averages plate motion over the past ≈5.8 Myr, which is almost twice the length of time (3.2 Myr) over which the NUVEL-1A global set of relative plate angular velocities is estimated. HS3-NUVEL1A, our preferred set of angular velocities of 15 plates relative to the hotspots, was constructed from the HS3 data set while constraining the relative plate angular velocities to consistency with NUVEL-1A. No hotspots are in significant relative motion, but the 95 per cent confidence limit on motion is typically ±20 to ±40 km Myr−1 and ranges up to ±145 km Myr−1. The uncertainties of the new angular velocities of plates relative to the hotspots are smaller than those of previously published HS2-NUVEL1 (Gripp & Gordon 1990), while being averaged over a shorter and much more uniform time interval. Nine of the fourteen HS2-NUVEL1 angular velocities lie outside the 95 per cent confidence region of the corresponding HS3-NUVEL1A angular velocity, while all fourteen of the HS3-NUVEL1A angular velocities lie inside the 95 per cent confidence region of the corresponding HS2-NUVEL1 angular velocity. The HS2-NUVEL1 Pacific Plate angular velocity lies inside the 95 per cent confidence region of the HS3-NUVEL1A Pacific Plate angular velocity, but the 0 to 3 Ma Pacific Plate angular velocity of Wessel & Kroenke (1997) lies far outside the confidence region. We show that the change in trend of the Hawaiian hotspot over the past 2 to 3 Myr has no counterpart on other chains and therefore provides no basis for inferring a change in Pacific Plate motion relative to global hotspots. The current angular velocity of the Pacific Plate can be shown to differ from the average over the past 47 Myr in rate but not in orientation, with the current rotation being about 50 per cent faster (1.06 ± 0.10 deg Myr−1) than the average (0.70 deg Myr−1) since the 47-Myr-old bend in the Hawaiian–Emperor chain.