Abstract
AbstractMany characteristics of goal-directed movements, such as their initiation time, initial direction, and speed, are influenced both by the details of previously executed movements (i.e. action history), and by the degree to which previous movements were rewarded or punished (i.e. reward history). In reinforcement learning terms, when movements are externally cued, action and reward history jointly define the probability and magnitude of positive/negative outcomes of available options, and therefore their pre-stimulus expected value. To dissociate which of these neurocomputational variables influence sensorimotor brain processing, we studied how reach behaviour and evoked brain responses are affected by independent manipulations of action and reward history. We found that movements were initiated earlier both for more frequently repeated targets and targets associated with higher reward magnitude, but only movements to highly rewarded targets had higher movement speeds. Classical visually-evoked encephalographic (EEG) potentials (P1/N1) were not affected by either reward magnitude or target probability. There were, however, amplified midline ERP responses at centroparietal electrodes for rewarded targets and movements compared to control, but no differences between more frequently presented targets and control. Critically, the spatial precision of decoded target locations extracted from a multivariate linear decoding model of EEG data was greater for target locations associated with higher reward magnitude than for control target locations (∼150-300ms after target presentation). Again, there were no differences in the precision of decoded target direction representations between more frequent target locations and control target locations. These data suggest that the expected reward magnitude associated with an action, rather than its long-run expected value, determines the precision of early sensorimotor processing.Significance StatementWe move more quickly and more accurately toward goals that we value more highly, and this is due partly to enhanced motor preparation. However, our expectations about the value of an action depend both on the probability of its requirement and the magnitude of the reward associated with it. Here we disentangled the influence of reward magnitude and probability on early sensorimotor processing via a multivariate linear decoding approach to extract target direction from scalp encephalograms. We found that the spatial precision of decoded target direction was greater for high reward targets but not for more probable targets. Thus, early sensorimotor processing is sharpened when the magnitude of reward associated with movement to a cued target is high.HighlightsThe direction of movement can be reliably decoded from the scalp EEG from ∼80ms after target presentation.The neural representation of movement direction is more precise for targets that are associated with high reward, but not for targets that are more probable.The magnitude of reward associated with movement to a presented target, rather than the long-run expected value of the movement, sharpens the spatial precision of early sensorimotor processing.
Publisher
Cold Spring Harbor Laboratory