In the third study, Bertolino and associates used a novel approach to address these types of questions. These investigators combined spectroscopic magnetic resonance imaging of N-acetylaspartate, a surrogate marker of neuronal integrity, with PET images obtained while subjects performed tasks that tap working memory. Their results confirm that working memory tasks require the activation of a neuronal network distributed across a number of cortical and subcortical structures (7, 8). More important, their findings suggest that the function of the network can be disrupted by a disturbance in neuronal integrity (e.g., decreased N-acetylaspartate levels) that appears to be restricted to one node of the network, in this case the dorsolateral prefrontal cortex. However, the authors carefully point out that abnormal N-acetylaspartate levels could reflect focal pathological changes either of neurons intrinsic to the prefrontal cortex or of afferent inputs from the thalamus, medial temporal lobe, or other brain region. Indeed, these options are not mutually exclusive, and an either/or scenario actually seems unlikely. By virtue of the interconnections among nodes, it would be surprising if a disturbance in one region did not produce a series of alterations throughout the network, ranging from shifts in gene expression to larger-scale structural changes. In terms of the Bertolino et al. study, it may be that disturbances in multiple projection streams crossed the threshold of detectability in the prefrontal cortex, only by virtue of their confluence in that brain region.