Minimal impact of proprioceptive loss on implicit sensorimotor adaptation and perceived movement outcome

Abstract

Our ability to produce successful goal-directed actions involves multiple learning processes. Among these, implicit adaptation is of utmost importance, keeping our sensorimotor system well-calibrated in response to changes in the body and environment. Implicit adaptation is assumed to be driven by a sensory prediction error, the difference between the predicted and actual sensory consequences of a movement. Whereas most models of implicit adaptation have focused on how visual information defines the sensory prediction error, we have recently proposed that this error signal is kinesthetic, the difference between the desired and perceived hand position, with adaptation serving to align these two signals and restore optimal motor performance (Tsay et al., 2022). Here, we examined implicit adaptation and kinesthetic perception in rare individuals who lack proprioceptive signals from the upper limbs. We used a visuomotor rotation task designed to isolate implicit adaptation while simultaneously probing the participants' perceived hand position. Consistent with prior work, control participants exhibited robust implicit adaptation and the signature of kinesthetic re-alignment, an initial bias in perceived hand position towards the visual cursor and a gradual shift back to the movement goal. Strikingly, the time course of both implicit adaptation and kinesthetic re-alignment was preserved in the deafferented group, suggesting that proprioceptive afferents are not necessary for implicit adaptation and kinesthetic re-alignment. We propose that a kinesthetic prediction error derived from efferent motor signals is sufficient to drive implicit adaptation and to re-algin a biased percept of hand position with the movement goal.

Evidence for an efferent-based prediction contributing to implicit motor adaptation

“…Thus, we predicted that participants would exhibit implicit adaptation during the isometric aiming task by slowly drifting the direction of their force exertion away from the target when exposed to clamped visual feedback. Further, we also predicted that adaptation would reach similar asymptotic levels compared to those found in previous clamped visual feedback reaching studies (e.g., Kim et al, 2018;Morehead et al, 2017;Tsay et al, 2023), based on the notion that the efferent sensory component would compensate for the missing afferent contributions.…”

“…However, in a recent study examining implicit adaptation during target-directed reaching movements in rare individuals with deafferentation (severe, or complete, proprioceptive loss), Tsay et al (2023) found similar adaptation between deafferented individuals and healthy controls. Previous studies utilizing visuomotor rotations or forcefield perturbations have also shown comparable levels of adaptation between individuals with deafferentation and healthy control participants (Bernier et al, 2006;Sarlegna et al, 2010).…”

Perceptual error based on Bayesian cue combination drives implicit motor adaptation

“…The importance of perceptual error was first implied in the PReMo model ( Tsay et al, 2022b ), which also proposed that a misperceived hand location is the driving signal for implicit adaptation. However, this previous work improperly called this perceptual error a proprioceptive error, leading to the impression that it is a type of sensory error (but see their new unpublished Tsay et al, 2024 ). The two models differ fundamentally in their conceptualization of how different cues contribute to the error signal.…”

Perceptual error based on Bayesian cue combination drives implicit motor adaptation

“…The importance of perceptual error was first implied in the PReMo model (Tsay, Kim, Haith, et al, 2022), which also proposed that a misperceived hand location is the driving signal for implicit adaptation. However, this previous work improperly called this perceptual error a proprioceptive error, leading to the impression that it is a type of sensory error (but see their new unpublished Tsay et al, 2024). The two models differ fundamentally in their conceptualization of how different cues contribute to the error signal.…”