Experiments on steel samples are reported for a new type of hardness tester based on the measurement of thermal contact resistance. These experiments have been carried out using a machine-mounted probe and a newer hand-held double-probe. It is shown that, by using the double-probe, errors due to the sample not having reached ambient temperature can be avoided. Although, ideally, the thermal conductivity of the sample should be much greater than that of the probe tip, it is found that it is simple to apply a correction to the apparent hardness number if this condition is not met. The standard deviation in the Vickers hardness number H, that is yielded by the hand-held probe ranges from 10% to 14% as H changes from 150 to 850 kg mm−2, but the observations with the machine-mounted probe indicate that a much higher accuracy than this is possible.