Sharply angular boulders as large as 320 kg sit on the Racetrack Playa, Death Valley, California; trails leading to them indicate that the rocks have moved large distances. The process has never been witnessed. Although high winds and a wetted surface seem necessary, controversy persists about the need for other conditions, especially ice sheets. On the basis of experiments with a wetted Racetrack surface (soft mud ∼3 cm deep), we find the effective coefficient of friction to be surprisingly high, ∼0.8. Movement by wind alone of moderate-sized (20 kg) rocks with cubic shape requires sustained winds close to the ground of ∼80 m/s (∼180 mph). Larger flat-lying rocks require much higher winds. To assess the ice-sheet hypothesis, we mapped a large number of tracks into a precise coordinate system with an electronic theodolite. Certain tracks, separated by <1 to ∼830 m, have nearly identical curving patterns near their starts. The distance between a distinctive bend on one such track and its mate on another matches the distance for another mated pair of bends on these same tracks within several centimetres, even for tracks 830 m apart. As proposed by Stanley (1955), it seems that the rocks, resting on mud, were locked into a single floating ice sheet, in this case at least 850 X 500 m. Final resting places of these rocks are much more widely scattered than their starting points, suggesting that the sheet broke into smaller plates. Large ice sheets can move rocks even with light winds and may explain the gentle curvature of tracks hundreds of metres long, a pattern very unlikely with gusty high winds and no ice.