Action costs rapidly and automatically interfere with reward-based decision-making in a reaching task

Abstract

It is widely assumed that we select actions we value the most. While the influence of rewards on decision-making has been extensively studied, evidence regarding the influence of motor costs is scarce. Specifically, how and when motor costs are integrated in the decision process is unclear. Twenty-two right-handed human participants performed a reward-based target selection task by reaching with their right arm toward one of two visual targets. Targets were positioned in different directions according to biomechanical preference, such that one target was systematically associated with a lower motor cost than the other. Only one of the two targets was rewarded, either in a congruent or incongruent manner with respect to the associated motor cost. A timed-response paradigm was used to manipulate participants’ reaction times (RT). Results showed that when the rewarded target carried the highest motor cost, movements produced at short RT (<350ms) were deviated toward the other (i.e., non-rewarded, low-cost) target. In this context participants needed an additional 150ms delay to reach the same percentage of rewarded trials as when the low-cost target was rewarded. Crucially, motor costs affected the total earnings of participants. These results demonstrate a robust interference of motor costs in a simple reward-based decision-making task. They point to the rapid and automatic integration of motor costs at an early stage of processing, potentially through the direct modulation of competing action representations in parieto-frontal regions. The progressive overcoming of this bias with increasing RT is likely achieved through top-down signaling pertaining to expected rewards.

SIGNIFICANCE STATEMENT

Rapid evaluation of expected action costs for action selection possesses an adaptive value in ecological settings. The present work shows that these motor costs quickly and automatically bias decisions supposedly based on reward information, leading to lesser earnings when rewards and motor costs are incongruent. This bias is progressively overcome with increasing reaction times, consistent with the perspective of a hierarchical influence of different decisional variables on action representations based on their level of abstraction. Overall, these findings highlight the need to consider motor costs when using dynamic motor tasks for studying decision-making, especially under temporal pressure.