To determine whether structural brain abnormalities in patients with depression are related to cortisol state, we examined the relationship between the dexamethasone suppression test (DST) and brain magnetic resonance imaging (MRI) in 40 inpatients with severe depression referred for electroconvulsive therapy (ECT). Prior to ECT, 27 (68%) of the patients exhibited nonsuppression on the DST. Frontal lobe volume was negatively correlated with peak post-dexamethasone cortisone (r = -0.37) and was 13% smaller in DST nonsuppressors than suppressors; these findings were no longer significant after adjustments for age, gender, and cranial size. Lateral and third ventricular volumes were also correlated with peak postdexamethasone cortisol (r = 0.34 and 0.33, respectively), but not after adjustments for age, gender, and cranial size. Subcortical hyperintensity was associated with peak postdexamethasone cortisol and was more common in DST nonsuppressors than suppressors. Again these findings were no longer significant after adjustments for age. Finally, longitudinal DST and brain MRI studies in 11 of these patients revealed no changes in regional brain volumes nor in postdexamethasone cortisol up to six months after ECT. However, within individual patients, postdexamethasone cortisol was positively (and significantly) correlated with frontal lobe volume.