Predicted task success and outcome history mediate eye movements toward locations with high informational or motivational value.

Abstract

Previous research revealed that the action outcomes of self-generated movements can be predicted internally before outcome feedback becomes available. These sensorimotor predictions have been shown to guide gaze toward locations in the environment with high informational value for feedback processing. To test whether they can further be used to assign specific values to different parts of the visible space in the context of motor learning, we assessed systematic differences between gaze toward targets with high informational value (showing graded movement outcome feedback) and targets with high motivational value (showing the monetary reward in a current trial) as a function of (predicted) task success (hit vs. miss) in a goal-directed throwing task. We observed large interindividual differences in the proportion of fixations on each target, highlighting that stable individual tendencies strongly influence whether gaze is directed toward informational or motivational content. However, individual baseline proportions were fine-tuned by the throwing outcome of previous trials (e.g., a larger proportion of fixations on the informational target in trials that followed a miss in the previous trial) and the predicted outcome of a current trial (a larger proportion of fixations on the motivational target in predicted hits compared to misses). Thus, we conclude that, in the context of motor learning, humans combine outcome history with sensorimotor predictions, including the predictive valuation of outcome feedback, to assign value to different parts of the visible space and to guide gaze at highly valued locations, albeit modestly and subordinate to more stable individual preferences.NEW & NOTEWORTHY We elucidate the potential benefits of internally generated action outcome predictions with respect to the optimization of information uptake in the context of motor learning. Here, we show that outcome predictions are used in combination with outcome history to weigh the value of different parts of the environment, albeit being subordinate to stable individual gaze preferences.