In today's vehicles the human-machine interaction (HMI) is becoming increasingly complex. Haptic interfaces are a realistic approach to improving HMI. However, to what extent haptic interfaces can replace visual interfaces still remains an unresolved issue. The purpose of the study reported in this paper was to measure task completion time, error rate, and workload in an easy menu selection task for three different types of interface providing: haptic information, graphic information, or combined haptic and graphic information. The haptic interface used was a rotary device that could be turned and pushed. The results suggest that the combined haptic and graphic interface was preferable to the graphic interface regarding accuracy, i.e. fewer turn errors, while both interfaces were fundamentally equal concerning completion time. The workload results indicated that the combined haptic and graphic interface was less mentally demanding than either the graphic or the haptic interface. Solely haptic feedback produced, without exception, worse results than the other cases, as could be expected.