Beyond conflict monitoring: Cognitive control and the neural basis of thinking before you act

Abstract

Cognitive control refers to the processes by which individual cognitive functions are coordinated in the service of higher level goals. The anterior cingulate cortex (ACC) in the middle front of the brain monitors performance, and it is activated when the need for control is greater, as in difficult situations or when errors occur. Since the late 1990s, the ACC has been thought to signal when there is internal conflict between competing action plans, so that the conflict can be resolved. More recently, an alternative model has reconceptualized the computational role of ACC as predicting and evaluating the likely outcomes of a planned action before actions are made. This new predicted response outcome (PRO) model accounts for a broader range of findings and suggests that the ACC might support the cognitive operations by which individuals can "think before you act" in order to avoid risky or otherwise poor choices.

Keywords: anterior cingulate cortex; cognitive control; conflict monitoring; performance monitoring; risk.

Figure 1

The Predicted Response Outcome (PRO) model of anterior cingulate cortex. (a) The Prediction units generate a timed prediction of what outcomes are expected, with what probability, and when. (b) Timecourse of Prediction units. Greater probabilities are associated with greater prediction activity, and the activity peaks at the time when the outcome is expected. Prediction signals of aversive outcomes might influence decisions away from risk. (c) The Evaluation units compute negative surprise, i.e. they detect when an expected outcome fails to occur. This is computed simply by subtracting the actual outcome signal off (via inhibition) from the Prediction unit signal. (d) Timecourse of Evaluation unit activity. Events that are predicted with a high probability yield greater surprise signals when they fail to occur (solid line), relative to when they do occur as expected (dashed line). This mechanism accounts for error signals within ACC and might drive corrective actions when errors occur. (e) Outcome units signal when an actual outcome occurs. (f) An Outcome unit is transiently activated at the moment when a corresponding outcome actually occurs (dashed line), or if no outcome occurs, the Outcome unit remains inactive (solid line).

Figure 2

Conflict effects without conflict computation as predicted by the PRO model. (A) In the standard flanker task instructed here by the “MIDDLE” cue, subjects must press a button corresponding to the direction of the arrow in the middle and ignore the flankers. Here, the middle arrow points to the left, and so the subject presses only the left button. (B) In the modified flanker task instructed here by the “ALL” cue, subjects must press one or both buttons as cued by both the middle and the flanker arrows. Here, the middle and flanker arrows point in both directions, but the subject need not suppress the response to the incongruent flankers. So the subject presses both buttons simultaneously. (C) The dorsal ACC shows greater activity for the incongruent relative to the congruent conditions, in both the “middle” and “all” conditions, suggesting that apparent response conflict effects in ACC may not depend on conflict. Adapted from Brown, 2009.