Three passenger cars were subjected to relatively high speed coastdown tests, with six different types of production tires. Analysis yielded a non-linear expression for the rolling resistance coefficient of each tire, depending on road surface temperature and tire pressure. The influence of speed and load was found to be negligible. The same tires were then tested on a flat-belt tire test machine, at constant belt temperature, over the same range of load, pressure and speed as on-road. Analysis yielded a similar expression for rolling resistance on the flat-belt machine, though the power law relation for pressure differed from that determined on-road. A simple correlation was found, from which the rig measurement could be transformed to within about 13% of the on-road force. There was some correlation between tire tread temperature and the characteristic rolling resistance coefficient of each type of tire, which was independent of ambient temperature conditions.