The effects of twelve sessions of biofeedback-assisted relaxation on blood glucose were tested in a controlled trial of 16 patients with insulin-dependent diabetes mellitus. Treatment consisted of electromyograph biofeedback, thermal biofeedback, relaxation therapy, and diabetes education. The controls received only diabetes education. All patients monitored blood glucose daily. The treated group improved on average blood glucose, percent of values above 200 mg/dl, and number of values at target. Two biologic indicators of blood glucose were used as correlates of self-reported blood glucose. Fructosamine and glycosylated hemoglobin were significantly correlated with values determined by self-monitoring of blood glucose at pretest. Fructosamine was a better index of short-term change than glycosylated hemoglobin. These results, though limited by size of population and reliance on self-reported blood glucose, support our earlier study and are promising for biofeedback-assisted relaxation as an adjunct to conventional therapy of insulin-dependent diabetes.