Human limb position sense can be measured in two ways: in a blindfolded matching task, position of one limb is indicated with the other limb. Alternatively, position of a limb, hidden from view, is indicated with a pointer, moved by pressing a lever. These experiments examined the sensory basis of position sense measured in these two ways. Position errors were measured in 14 subjects after elbow flexors or extensors had been conditioned with a half‐maximum voluntary contraction. In agreement with previous studies, in the matching trials, position errors were distributed according to a pattern consistent with the action of muscle spindles as the position sensors. In the pointing trials, all errors lay in the direction of extension of the true position of the hidden arm and their distribution was inconsistent with influences arising in muscle spindles. Vibration of elbow muscles produced an illusion of muscle lengthening during a matching task, while during the pointing task no illusion was present. Finally, the matching : pointing error difference was preserved, even when one arm was loaded with a weight or skin over the elbow was stretched. It is proposed that there are two kinds of position sense, one signalled by muscle spindles, indicating position of one part of the body relative to another. A second sense provides information about the position of the body in extrapersonal space and here we hypothesise that exteroceptors, including vision, touch and hearing, acting via a central map of the body, provide the spatial information. This article is protected by copyright. All rights reserved