When Velocity Becomes Position: Proprioceptive Contributions to State Estimation

Abstract

In this study, we investigate the relative contributions of positional- and velocity-based proprioceptive feedback to state estimation when visual feedback of the hand is unavailable. We performed a virtual reality experiment in which healthy human participants (N = 22) completed two tasks: a continuous target-tracking task and a sequential reaching task. In both cases, visual feedback of the hand position was withheld, except for a brief moment, during which an offset to the left or right was sometimes applied. By analysing the movement path following the offset visual feedback and comparing it to the non-offset condition, we can infer the extent to which the brain relies on rate-of-change feedback compared to absolute positional feedback for estimating the hand position. We compare and contrast our empirical data with simulated movements from a computational model based on Bayesian sensory integration under varied assumptions of the reliance on positional and velocity-based cues. Our human and simulated data together indicate a heavy reliance on velocity-based proprioceptive feedback, as movement offsets induced by biased visual feedback decay slowly during both target tracking and sequential reaching tasks. The eventual convergence between offset and non-offset trials demonstrates that absolute positional information is incorporated, albeit slowly. Author summaryThe brain continually estimates the current pose and movement of the body, supporting motor control and giving rise to the perception of the bodys state. When visual feedback is lacking, we rely heavily on the sense of proprioception to infer the state of the body. Key among the proprioceptive receptors are the muscle spindle afferents, which are divided into two specialised subgroups: one that signals absolute muscle length, and the other that signals the rate-of-change of muscle length. In this study, we investigate the extent to which the brain utilises rate-of-change information to supplement its inference of the current body position. While such a function would fit well within several of the most popular theories of sensorimotor control, the investigation of this process has so far received limited attention. By having a group of research participants complete two virtual reality-based experiments, we demonstrate that the performed movements remain relative to a visually induced offset. By further simulating the same tasks in a computational model based on Bayesian sensory integration, we can show that this pattern of behaviour is well explained by assuming a relatively high reliance on rate-of-change versus absolute proprioceptive feedback.