In the classic split-bar determination of thermal contact resistance the temperature drop across the interface is estimated by extrapolating a temperature gradient measured remotely. It is shown that this can give rise to substantial errors which cannot greatly be reduced by increasing the number of measurements. It is suggested that due to extrapolation errors few interface temperature drops have ever been determined to better than 1/2 °K, and that this may account for some of the discrepancies between published contact resistances, particularly those measured at high loads.