Major depression is a heterogeneous condition, and the search for neural correlates specific to clinically defined subtypes has been inconclusive. Theoretical considerations implicate frontostriatal, particularly subgenual prefrontal cortex (PFC), dysfunction in the pathophysiology of melancholia-a subtype of depression characterized by anhedonia-but no empirical evidence has been found yet for such a link. To test the hypothesis that melancholic, but not nonmelancholic depression, is associated with the subgenual PFC impairment, concurrent measurement of brain electrical (electroencephalogram, EEG) and metabolic (positron emission tomography, PET) activity were obtained in 38 unmedicated subjects with DSM-IV major depressive disorder (20 melancholic, 18 nonmelancholic subjects), and 18 comparison subjects. EEG data were analyzed with a tomographic source localization method that computed the cortical three-dimensional distribution of current density for standard frequency bands, allowing voxelwise correlations between the EEG and PET data. Voxel-based morphometry analyses of structural magnetic resonance imaging (MRI) data were performed to assess potential structural abnormalities in melancholia. Melancholia was associated with reduced activity in the subgenual PFC (Brodmann area 25), manifested by increased inhibitory delta activity (1.5-6.0 Hz) and decreased glucose metabolism, which themselves were inversely correlated. Following antidepressant treatment, depressed subjects with the largest reductions in depression severity showed the lowest post-treatment subgenual PFC delta activity. Analyses of structural MRI revealed no group differences in the subgenual PFC, but in melancholic subjects, a negative correlation between gray matter density and age emerged. Based on preclinical evidence, we suggest that subgenual PFC dysfunction in melancholia may be associated with blunted hedonic response and exaggerated stress responsiveness.