Abstract
To measure the extent to which the motor system compensates for motor error, studies of motor learning and adaptation typically perturb visual or proprioceptive feedback and measure the compensatory response. These perturbations are often large compared to trial-to-trial error to increase the signal-to-noise ratio of compensatory responses. However, large, detectable perturbations can be unnatural and allow subjects to use conscious mechanisms for compensation. Conscious, explicit motor compensation processes may have different parameters and recruit different neural circuitry compared to low-level, automatic mechanisms. To isolate low-level mechanisms, alternative experimental designs provide increased power per trial while using small perturbations so that subjects are unaware of the perturbation. However, to maximize the signal-to-noise ratio of compensation measurements, perturbations should still be as large as possible while remaining unnoticed. We measured the upper limit of undetectable perturbations. We also investigated how subjects combined perturbed visual feedback with the proprioceptive signal to estimate self-generated motor error. Participants made fast center-out reaches on a tabletop while viewing a frontoparallel monitor. On half of these reaches, feedback was perturbed in either gain or direction. No visual feedback of the finger position was displayed on the monitor except for the final, possibly perturbed, reach endpoint. Subjects indicated whether they believed the trajectory had been perturbed, at which point the true reach endpoint was displayed. Participants reliably detected perturbations larger than 1.5 times the trial-to-trial endpoint SD. Detection was primarily a function of displayed visual error feedback, indicating a nearly complete disregard of proprioceptive information. Subjects may be unable to combine visual and proprioceptive feedback to detect perturbation when there is spatial separation between the hand and feedback and/or may downweight proprioception in this task because it is too unreliable.
Meeting abstract presented at VSS 2016