When deciding how to allocate cognitive control to a given task, people must consider both positive outcomes (e.g., praise) and negative outcomes (e.g., admonishment). However, it is unclear how these two forms of incentives differentially influence the amount and type of cognitive control a person chooses to allocate to achieve task goals. To address this question, we had participants perform a self-paced incentivized cognitive control task, varying the magnitude of reward for a correct response and punishment for an incorrect response. Formalizing control allocation as a process of adjusting parameters of a drift diffusion model (DDM), we show that participants engaged in different strategies in response to reward (primarily adjusting drift rate) versus punishment (primarily adjusting response threshold). We demonstrate that this divergent set of strategies is optimal for maximizing reward rate while minimizing effort costs. Finally, we show that these dissociable patterns of behavior enable us to infer the motivational salience of positive versus negative incentives for a given individual. These results provide a normative and mechanistic account for how reward and punishment differentially influence the adaptive allocation of cognitive control.
bioRxiv Subject Collection: Neuroscience