Key points When a blindfolded subject holds his or her arm at a particular angle, its reported position shifts over time; this is known as proprioceptive drift. Here, we show that in relation to position sense at the elbow, the direction of perceived shifts is consistent with adaptation in discharge levels of sensory receptors in elbow muscles. Raising or lowering receptor discharge levels by similar amounts in opposing muscles at the elbow using muscle conditioning abolishes proprioceptive drift, but large position errors may result. The present experiments provide an explanation for proprioceptive drift and indicate that, in a forearm position‐matching task, the brain is not concerned with actual discharge levels from arm muscles, but with their difference. Abstract These experiments on the human forearm are based on the hypothesis that drift in the perceived position of a limb over time can be explained by receptor adaptation. Limb position sense was measured in 39 blindfolded subjects using a forearm‐matching task. A property of muscle, its thixotropy, a contraction history‐dependent passive stiffness, was exploited to place muscle receptors of elbow muscles in a defined state. After the arm had been held flexed and elbow flexors contracted, we observed time‐dependent changes in the perceived position of the reference arm by an average of 2.8° in the direction of elbow flexion over 30 s (Experiment 1). The direction of the drift reversed after the arm had been extended and elbow extensors contracted, with a mean shift of 3.5° over 30 s in the direction of elbow extension (Experiment 2). The time‐dependent changes could be abolished by conditioning elbow flexors and extensors in the reference arm at the test angle, although this led to large position errors during matching (±10°), depending on how the indicator arm had been conditioned (Experiments 3 and 4). When slack was introduced in the elbow muscles of both arms, by shortening muscles after the conditioning contraction, matching errors became small and there was no drift in position sense (Experiments 5 and 6). These experiments argue for a receptor‐based mechanism for proprioceptive drift and suggest that to align the two forearms, the brain monitors the difference between the afferent signals from the two arms.