This study used functional magnetic resonance imaging (fMRI) to clarify the sites of brain activity associated with the antidepressant effects of sleep deprivation (SD). We hypothesized: 1) depressed responders' baseline ventral anterior cingulate (AC) perfusion will be greater than that of nonresponders and controls; 2) following partial sleep deprivation (PSD), ventral AC perfusion will significantly decrease in responders only. Seventeen unmedicated outpatients with current major depression and eight controls received perfusion-weighted fMRI and structural MRI at baseline and following 1 night of late-night PSD. Talairach-transformed gray matter masks were merged with Talairach Daemon-based region of interest (ROI) templates. Baseline left ventral AC (LVAC) perfusion was greater in responders than nonresponders. There was no difference involving the medial frontal cortex. Responders' LVAC perfusion dropped from baseline to PSD scans compared with nonresponders and controls, as did perfusion in the right dorsal AC. In the patient group as a whole, decrease in LVAC perfusion from baseline to PSD scans correlated directly with the decrease in the modified 17-item Hamilton Depression Rating Scale (HDRS17) between baseline and PSD conditions. These data--the first using fMRI--show greater anatomic specificity than previous findings of SD and depression in linking decreased brain activity in this area with clinical improvement.