Comparing Alternative Computational Models of the Stroop Task Using Effective Connectivity Analysis of fMRI Data

Methodological advances have made it possible to generate fMRI predictions for cognitive architectures, such as ACT-R, thus expanding the range of model predictions and making it possible to distinguish between alternative models that produce otherwise identical behavioral patterns. However, for tasks associated with relatively brief response times, fMRI predictions are often not sufficient to compare alternative models. In this paper, we outline a method based on effective connectivity, which significantly augments the amount of information that can be extracted from fMRI data to distinguish between models. We show the application of this method in the case of two competing ACT-R models of the Stroop task. Although the models make, predictably, identical behavioral and BOLD time-course predictions, patterns of functional connectivity favor one model over the other. Finally, we show that the same data suggests directions in which both models should be revised.