In a model study, the behaviour of two sets of muscles in controlling multi-joint arm movements is compared. Both the sensory and the motor accuracy of the set containing bi-articular muscles were in general better than those of the set containing only mono-articular muscles. Accuracy considerations can explain differences in strategies for the control of redundant muscle sets between situations which do not differ biomechanically from each other. Furthermore, the role of bi-articular muscles for the robustness of motor programming is discussed.