Treatment with antidepressants is marked by heterogeneity of response; predicting individual response to any given agent remains problematic. Neuroimaging studies suggest that response is accompanied by physiologic changes in cerebral energy utilization, but have not provided useful markers at pretreatment baseline. Using quantitative EEG (QEEG) techniques, we investigated pretreatment neurophysiologic features to identify responders and non-responders to fluoxetine. In a double-masked study, 24 adult subjects with current major depression of the unipolar type were studied over 8 weeks while receiving fluoxetine (20 mg QD) or placebo. Neurophysiology was assessed with QEEG cordance, a measure reflecting cerebral energy utilization. Response was determined with rating scales and clinical interview. Subjects were divided into discordant and concordant groups based upon the number of electrodes exhibiting discordance. The concordant group had a more robust response than the discordant group, judged by lower final Hamilton Depression (HAM-D) mean score (8.0+/-7.5 vs. 19.6+/-4.7, P = 0.01) and final Beck Depression Inventory (BDI) mean score (14.0+/-9.4 vs. 27.8+/-3.7, P = 0.015), and by faster reduction in symptoms (HAM-D: 14.0+/-5.0 vs. 23.8+/-4.1, P = 0.004 at 1 week). Groups did not differ on pretreatment clinical or historical features. Response to placebo was not predicted by this physiologic measure. We conclude that cordance distinguishes depressed adults who will respond to treatment with fluoxetine from those who will not. This measure detects a propensity to respond to fluoxetine and may indicate a more general responsiveness to antidepressants.