Summary:This article describes how fMRI can be used to examine the large-scale networks of cortical areas that subserve high-level cognition, such as sentence comprehension and visual thinking. The findings from a number of studies show that the qualitative and quantitative nature of the cognitive processes determines which cortical areas are activated (the network constituency) and the degree to which each network member is activated. For example, during sentence comprehension, activation in the left posterior temporal region and the inferior frontal gyms, as well as their right hemisphere homologs, increases as a function of the linguistic complexity of the sentence. Such findings indicate that cognition emerges from the collaboration among the multiple cortical areas that compose the large-scale networks, rather than from the aggregate of autonomously functioning modules. The patterns of activation also show systematic shifts in the activity of a network during the spontaneous recovery of function by stroke patients, demonstrating cortical plasticity in adults. Finally, the article describes some simulation models that relate the information processing activity of a computational system to its resource consumption. This construct enables a mapping from the functional properties of the cognitive systems to the biological substrate that is reflected in fMRI