Elsevier

Neurocomputing

Volume 69, Issues 10–12, June 2006, Pages 1322-1326
Neurocomputing

A model of dual control mechanisms through anterior cingulate and prefrontal cortex interactions

https://doi.org/10.1016/j.neucom.2005.12.100Get rights and content

Abstract

A computational model is presented that describes dual mechanisms of cognitive control through interactions between the prefrontal cortex (PFC) and anterior cingulate cortex (ACC). One mechanism, reactive control, consists in the transient activation of PFC, based on conflict detected in ACC over a short time-scale. The second mechanism, proactive control, consists in the sustained active maintenance of task-set information in a separate PFC module, driven by long time-scale conflict detected in a separate ACC unit. The computational function of the first mechanism is to suppress the activation of task-irrelevant information just prior to when it could interfere with responding. The role of the second mechanism is to prime task-relevant processing pathways prior to stimulus-onset, in a preparatory fashion. The model provided an excellent fit to both the behavioral and brain imaging data from a previous detailed empirical study on humans performing the color-word version of the Stroop task. The model captured changes in reaction times across conditions, accuracy, and transient and sustained activity dynamics within lateral PFC and ACC.

Introduction

A great deal of convergent research has suggested that the lateral prefrontal cortex (PFC) and anterior cingulate cortex (ACC) play a critical role in human cognitive control. The relationship between these two brain areas has been studied extensively, both in neuroimaging (for example in [3], [5]) and in neural-network models [1]. In particular, one focus has been on the role of ACC in detecting response conflict during the execution of a cognitive task, and the subsequent translation of this conflict into cognitive control. The basic hypothesis is that when high conflict occurs between different motor or behavioral responses, cognitive control mechanisms intervene to bias one response versus the others depending on the task requirement, thus overcoming the conflict.

In these previous studies PFC and ACC interactions have been characterized in terms of a single conflict–control loop mechanism: performance of certain task conditions leads to detection of response conflict, which in turn leads to the engagement or increase of cognitive control, that results in improved conflict resolution in subsequent performance. Here, motivated by previous experimental findings in humans, we develop a new neural-network model in which ACC-PFC interactions are described by two, rather than one, distinct conflict–control loops. The first mechanism, reactive control, is characterized by the transient activation of PFC based on conflict detected in ACC over a short time-scale (on the order of milliseconds). The second mechanism, proactive control, is characterized by the sustained active maintenance of task-set information in a separate PFC module, which is driven by long time-scale conflict detected in a separate ACC unit (on the order of several seconds or minutes). The computational function of the first mechanism is to suppress the activation of task-irrelevant information just prior to when it could interfere with responding. The role of the second mechanism is to prime task-relevant processing pathways prior to stimulus-onset, in a preparatory fashion.

We conducted several computational simulations with this model to examine how well it could account for detailed empirical data regarding human behavioral performance and brain activation. The task studied was the color-word version of the Stroop task, a benchmark experimental preparation for examining response conflict and cognitive control.

Section snippets

Simulation method

The model is a large-scale connectionist network [7], simulating rate code spiking activity of brain regions involved in the execution of the color-naming Stroop test. This task requires verbally responding with the name of the font-color in which a visually presented English word appears. For example, the words DOG (in green color), RED (in red color), GREEN (in blue color) would require as correct verbal responses, respectively, “green”, “red”, and “blue”. Trials can be of several different

Results

In the empirical data (Fig. 2), behavioral performance patterns (upper left panel) indicated that interference (incongruent minus neutral reaction times) and facilitation (neutral minus congruent) effects in response times were reduced in the MI condition compared to MC, with an intermediate effect in the MN condition. This is consistent with a shift from reactive to proactive control when comparing MC versus MI conditions, since proactive control should result in a tonically reduced influence

Conclusions

The model describes a non-homuncular mechanism that may explain the dramatic shift in human behavior and brain activity during different conditions of the Stroop cognitive control task. Specifically, the model suggests that local and sustained experience of conflict during performance might lead to a shift in the neural mechanisms of cognitive control engaged to perform the task. Importantly, since individuals may not have been explicitly aware of changes in task context across condition, the

Acknowledgments

This research was supported by ONR N00014-04-1-0004 and R01 (NIH) MH06607803. Christine Hoyer provided invaluable assistance in the collection and analysis of empirical data.

Nicola De Pisapia is a research associate in the Cognitive Control and Psychopathology laboratory at the Washington University in Saint Louis. He received a Laurea (equivalent to a Master) in Philosophy and Artificial Intelligence from the University of Napoli in 1995, a Master in Philosophy and Computation from Carnegie Mellon University in 2000, and a Ph.D. in Computer Science from the University of Edinburgh in 2003. His research interests include neurocomputational models of prefrontal

References (7)

  • M. Botvinick et al.

    Conflict monitoring and cognitive control

    Psychol. Rev.

    (2001)
  • T.S. Braver, C.M. Hoyer, Neural mechanisms of proactive and reactive cognitive control,...
  • C.S. Carter et al.

    Anterior cingulated cortex, error detection, and the online monitoring of performance

    Science

    (1998)
There are more references available in the full text version of this article.

Cited by (176)

View all citing articles on Scopus

Nicola De Pisapia is a research associate in the Cognitive Control and Psychopathology laboratory at the Washington University in Saint Louis. He received a Laurea (equivalent to a Master) in Philosophy and Artificial Intelligence from the University of Napoli in 1995, a Master in Philosophy and Computation from Carnegie Mellon University in 2000, and a Ph.D. in Computer Science from the University of Edinburgh in 2003. His research interests include neurocomputational models of prefrontal cortex activity, behavioral planning and cognitive control.

Todd S. Braver is an associate professor and director of the Cognitive Control and Psychopathology laboratory at the Washington University in Saint Louis. He received his Ph.D. in Cognitive Neuroscience from Carnegie Mellon University in 1997. His research interests include understanding the neural and computational mechanisms of cognitive control, their breakdown in different populations, and their interaction with individual differences and emotion.

View full text