Limitations of Proprioceptive Working Memory

arXiv:2510.21996v1 Announce Type: new
Abstract: Recalling previously experienced movements is essential for a range of activities, including sports, music, and rehabilitation, yet little is known about the accuracy and decay of proprioceptive working memory. We examined how introducing a short-term memory component affected movement reproduction accuracy by comparing movement reproduction under two conditions: simultaneous reproduction (SimRep) and memorized reproduction (MemRep). In Experiment 1 (N = 191), participants felt a 5-s haptic trajectory with one hand and reproduced it with the other hand simultaneously or immediately after the template ended. Errors were greater in MemRep than SimRep (31.1 deg vs. 21.5 deg, p < 0.001). MemRep trajectories showed systematic temporal distortions: participants lagged fast movements and led slow ones (R = -0.32, p = 0.01), unlike the ~279 ms lag in SimRep. In Experiment 2 (N = 33), we varied template durations (2-8 s). Longer durations increased error for MemRep but not SimRep (p < 0.001). During MemRep, accuracy declined steadily, with replay-template correlations dropping from ~0.4 to ~0.1 over ~3 s, while SimRep correlations rose from ~0.25 to ~0.6. In ~10% of MemRep templates, participants moved in the wrong direction initially, especially for low-amplitude movements (p < 0.001). Templates with more than four movements showed element omission; after four movements had been reproduced participants ceased movement prematurely, affecting up to 40% of 8-s templates. These findings show that transferring proprioceptive experiences into working memory introduces systematic temporal and structural distortions. Accuracy decays within seconds, and memory span for movement trajectories was limited to four movements.