The angular displacements necessary for 70% correct detection were determined in normal subjects at the shoulder and elbow joints and at the terminal joint of the middle finger. Angular velocities of displacement between 0.125 degrees and 160 degrees/s were tested. Each joint was tested in the mid-range of its normal excursion. The joints were carefully supported for testing and the muscles acting at the joints were relaxed. When assessed in terms of angular displacements and angular velocities, proprioceptive performance at the shoulder and elbow joints was superior to that at the finger joint. Optimal performance at the finger joint was attained over the range of angular velocities from 10 degrees to 80 degrees/s. Optimal performance at both more proximal joints was optimal over a wider range (2 degrees-80 degrees/s). Active pointing movements made without vision of the moving part were performed at each joint at velocities within the range of optimal proprioceptive performance. However, when detection levels and displacement velocities were expressed in terms of linear displacements and velocities at the finger tip for all three joints, the finger joint gave the best performance and the shoulder the worst. In practical terms, therefore, displacements of a given linear extent are best detected if they move distal joints and worst if they move proximal joints. For the elbow and finger joints the detection level and velocity data were expressed also in terms of proportional changes in the lengths of muscles operating at these joints, and as proportional changes in the distance between the points of attachment of the joint capsules. Analysis in terms of proportional changes of muscle length showed remarkably similar performance levels at both joints. This suggests that intramuscular receptors are important determinants of proprioceptive performance. Analysis in terms of joint capsular displacement did not unify the data: on this form of analysis proprioceptive performance at the elbow joint was superior.