Computational Modeling of Cognitive Control in a Flanker Task

AbstractCognitive control refers to the ability to adjust our thoughts and behaviors in order to achieve internalized goals. In the past, researchers have proposed several models of cognitive control to account for the characteristic patterns of response times observed in the tasks (e.g., Botvinick, Braver, Barch, Carter, & Cohen, 2001). The goal of this study is to evaluate empirical validity of such models in an experiment. To that end, we compared two models of cognitive control, the conflict monitoring model and the expectancy-based model. Each model was implemented in two different modeling frameworks, neural networks and simple linear models. The relative fits of the four models were then evaluated and compared based on observed data from a flanker task experiment. The model comparison results showed that performance of the simple linear models was entirely comparable to that of the neural network models. We also constructed and fitted hierarchical Bayesian latent mixture versions of the linear models to investigate individual differences. The result suggests that no single model of cognitive control, whether conflict monitoring or expectancy-based, would be able to account for individual performance on the task.


Return to previous page