Abstract

Working memory is essential for many of our distinctly human abilities, including reasoning, problem solving, and planning. Research spanning many decades has helped to refine our understanding of this high-level function as comprising several hierarchically organized components, some which maintain information in the conscious mind, and others which manipulate and reorganize this information in useful ways. In the neocortex, these processes are likely implemented by a distributed frontoparietal network, with more posterior regions serving to maintain volatile information, and more anterior regions subserving the manipulation of this information. Recent meta-analytic findings have identified the anterior lateral prefrontal cortex, in particular, as being generally engaged by working memory tasks, while the posterior lateral prefrontal cortex was more strongly associated with the cognitive load required by these tasks. These findings suggest specific roles for these regions in the cognitive control processes underlying working memory. To further characterize these regions, we applied three distinct seed-based methods for determining cortical connectivity. Specifically, we employed meta-analytic connectivity mapping across task-based fMRI experiments, resting-state BOLD correlations, and VBM-based structural covariance. We found a frontoparietal pattern of convergence which strongly resembled the working memory networks identified in previous research. A contrast between anterior and posterior parts of the lateral prefrontal cortex revealed distinct connectivity patterns consistent with the idea of a hierarchical organization of frontoparietal networks. Moreover, we found a distributed network that was anticorrelated with the anterior seed region, which included most of the default mode network and a subcomponent related to social and emotional processing. These findings fit well with the internal attention model of working memory, in which representation of information is processed according to an anteroposterior gradient of abstract-to-concrete representations.