Computational and behavioral mechanisms underlying selecting, stopping, and switching of actions

How actions are regulated based on environmental changes is a fundamental neurobiological question. A key component of action regulation is action inhibition, which when abnormal contributes to various neuropsychiatric disorders. We explored the mechanism of action regulation by recruiting healthy adults (n=15, ages 21-42) to perform reaches towards targets presented on a computer screen using a joystick. The experiment includes: decision-making task, in which subjects were instructed to reach towards one target or freely choose between two targets; stop-signal task, in which a stop signal appeared in some trials of the decision-making task, forcing subjects to stop; and switch task, in which subjects had to switch towards an unknown or known target location in some trials of the decision-making task. We found that the reaction time for one-target trials is always shorter than that of two-target trials, regardless of the task (two-sample t-test, p<0.001). In the stop-signal task, the probability of successfully stopping is inversely correlated with the stop signal delay. However, subjects could stop an action easier when they choose between two targets than when they are instructed to move towards one target. They also delayed their response when anticipating a stop signal, suggesting an inhibitory mechanism during planning to ensure a successful stop. Importantly, subjects did not delay their response when they anticipated a switch signal, suggesting that an inhibitory mechanism in not engaged to prepare for switching. In addition, they had slower responses to switch ongoing actions when the new target location was known (two- target trials) than when it was unknown (one-target trials) prior to movement initiation. To better understand the mechanism of action regulation, we designed a neurocomputational framework and showed that selecting and switching actions towards locations that are known prior to movement initiation could be implemented through action competition, without engaging an inhibitory mechanism.