OBJECTIVE Sleep deprivation is a rapid, nonpharmacologic antidepressant intervention that is effective for a subset of depressed patients. The objective of this study was to identify which brain structures' activity differentiates responders from nonresponders and to study how metabolism in these brain regions changes with mood. METHODS Regional cerebral glucose metabolism was assessed by positron emission tomography (PET) with [18F]deoxyglucose (FDG) before and after total sleep deprivation in 15 unmedicated awake patients with unipolar major depression and 15 normal control subjects, who did the continuous performance test during FDG uptake. RESULTS After sleep deprivation, four patients showed a 40% or more improvement on the Hamilton Rating Scale for Depression. Before sleep deprivation the depressed responders had a significantly higher cingulate cortex metabolic rate than the depressed nonresponders, and this normalized after sleep deprivation. The normal control subjects and nonresponding depressed patients showed no change in cingulate metabolic rate after sleep deprivation. CONCLUSIONS Overactivation of the limbic system as assessed by PET scans may characterize a subset of depressed patients. Normalization of activity with sleep deprivation is associated with a decrease in depression.