With administration of 2 mg of dexamethasone 9 h before, serum cortisol levels were determined for the times just before and 30 min after the first and last electroconvulsive therapies (ECTs) of a series of at least six sessions. Twelve drug-free patients with melancholic or psychotic depression completed the study. Response to treatment was taken as a fall in the value of the Carroll rating scale. The average cortisol elevation over baseline was by 575% with the first ECT, 10 times larger than other studies in the literature that were done without dexamethasone. With the final ECT the elevation was lower (p < 0.01), averaging 181%. Post-ECT cortisol levels (and elevations over baseline) were lower after the final ECT than after the initial ECT, with 10 of the 11 patients who responded to this treatment (p < 0.005 vs. 50%). Consistent with this, the lone nonresponder showed a large increase in serum cortisol level and elevation over baseline from the first to the last treatment. The one responder who went against this trend had an unusually high baseline cortisol level, 24.1 μg/dl, which can be considered to disqualify his inclusion. The near-unanimous fall of post-ECT cortisol levels with remission suggests that defects in cortisol regulation are much more prevalent (p < 0.01) than those detected by the dexamethasone suppression test (36% in this sample).