We have proposed that the dopamine (DA) system plays a modulatory role in cognitive control, by gating the access of information to PFC. This gating function arises through reward-learning mechanisms, such that the DA system can learn which environmental stimuli are most relevant for selection and maintenance as context (i.e., those stimuli which tend to be predictive of future rewards). The gating function of DA allows for flexible updating of context representations as situational demands dictate, while simultaneously providing protection from the effects of interference. The power of this theory is that, if confirmed, it will describe a system that has both the capacity to control its behavior (by gating and stabilizing memory representations within PFC) and, critically, the ability to learn how and when to do so on its own, thus avoiding the perennial problem of the homunculus (Cohen, Braver and Brown, 2002).
A series of simulation studies demonstrate both the computational viability of the theory, and its ability to account for empirical data on interference effects in the AX-CPT (Braver & Cohen, 1999; Braver and Cohen, 2000). In related work, we have shown how this model provides a new account of cognitive control deficits in schizophrenia that can account for existing data, while making novel predictions about underlying neurophysiological disturbances (Braver, Barch, and Cohen, 1999). Current studies apply the model to phenomena associated with task-switching.