An experimental study has been carried out to investigate the effect of sanding on the electrical isolation of a wheel/rail contact. Sand is applied to the wheel/rail interface to increase adhesion in both braking and traction. Train detection, for signalling purposes, can be by means of track circuits. Signalling block occupancy is triggered by the wheelset of the train ‘shorting out’ the track circuit. Sand in the wheel/rail interface means that contact between the wheelsets and the track may be compromised, inhibiting train identification. Tests were performed on a twin-disc machine where rail and wheel steel discs are loaded together and driven under controlled conditions of rolling and slip. Sand was fed into the disc contact through a standard compressed air sanding valve. The electrical circuit used was a simplified simulation of the TI21 track circuit. The application of sand with and without water to the discs was carried out under a range of mild and severe test conditions. The results indicated that a transition exists in the sand flowrate below which there is a measurable, but not severe, change in voltage, but above which the contact conductance decreases by an order of magnitude and the voltage tends towards its open-circuit value. The total isolation time also showed a similar transition. Contact resistance was modelled assuming full disc separation by a sand layer and partial disc contact with some sand present. Traction was monitored during the tests. A wet contact showed approximately half the traction of a dry contact. The addition of sand increased the traction to levels observed in a dry contact. Idealizations inherent in the test method mean that it represents a severe case. The disc geometry is smaller than a wheel/rail contact and both are in rotational motion. The sand nozzle was placed closer to the interface, leading to greater sand entrainment and low inductance. A fast data acquisition rate made the test circuit more sensitive to small fluctuations in isolation than an actual track circuit. Given these limitations, it is likely that the test method, at its present stage of development, should be used as a means to assess qualitatively the relative effects on electrical isolation of different contaminants.