Accounting for Stroop Task Neuroimaging Data: Lateral Interactions & Frontal Representations

This work integrates new neuroimaging results with existing behavioral and computational studies to produce a model of attentional control within the Stroop task. The Stroop task is the paradigmatic example of how attentional control can facilitate processing of weak, less wellpracticed behaviors, such as the ability to name the ink color of a word in the face of the more dominant response of reading the word itself. Attentional control is implemented in the model as an excitatory bias from a maintained frontal task set interacting with competitive dynamics between posterior cortical pathways that process relatively domainspecific information (e.g., ink color vs. words). Our model advances previous models by including lateral excitatory and inhibitory connectivity (via interneurons) based on biological principles. We show that the complex dynamics supported by these lateral connections, together with assumptions about the nature of frontal task representations, can produce counter-intuitive patterns of neural activation that have been observed in fMRI studies of the Stroop task, which appeared to contradict existing models. In sum, we show that the neuroimaging data support a model of attentional control as a top-down excitatory bias coupled with competition for activation within and between cortical areas.