Abstract
Predicting the temporal consequences of our motor commands is a necessary skill for many daily activities. We optimize these predictions by using sensory error signals. Errors can stem from inaccurate predictions, but also from noise in the sensorimotor system or other external influences. The brain therefore needs to assess which errors need to be corrected for and which do not. Many studies have used perturbation paradigms to study our ability to predict the consequences of our actions, but adaptation to the perturbation is often not complete. State-space models of motor learning have attributed this lack of adaptation to a trial-to-trial forgetting factor. To explain the lack of adaptation, we propose a more parsimonious alternative, which does not involve forgetting. A temporal perturbation that is introduced during a motor task increases uncertainty in the measurement. To investigate if this added uncertainty could provide a more simple explanation for the lack of adaptation, we modeled participants' adaptation to a gradually increasing delay in a simple motor task. A target on a screen moved horizontally towards a vertical line (at different speeds/distances). Participants were instructed to press a button when the target reached the line. In separate conditions either visual or auditory (unisensory), or combined (mulitsensory) feedback was provided when the subject pressed the button. Each condition consisted of 135 trials and the sensory feedback of the button-press was gradually delayed with 1 ms/trial. Participants started accounting for these delays by pressing the button earlier. We modeled the behavior of the participants with a non-stationary Kalman filter, in which the measurement error increased gradually with the delay. Our model could not only explain the lack of adaptation to the imposed delay, but also the higher adaptation to multisensory versus unisensory feedback, which could not be explained by a forgetting factor.
Meeting abstract presented at VSS 2018