A study is described in which a modified glucose tolerance test was performed in each of five bulimic female patients on two separate occasions. Two of the patients were poorly controlled outpatients and three were inpatients in a strict hospital program of nutritional rehabilitation and psychotherapy. Plasma glucose, insulin, and cortisol levels, in response to a glucose load and following 3 days of normal nutrition, were compared with those documented following engagement in various bulimic behaviors during the preceding 24 hours (i.e., fasting, binging, and vomiting.). In all five patients, following bulimic behaviors, peak insulin levels in response to a glucose load and the area under the curve of insulin secretion was increased whereas cortisol levels were lower, even in the one patient who was hypercortisolemic. However, the insulin peaks were less in evidence in the treated patients than in those who continued to engage in bulimic behavior, and there was also less evidence of cortisol suppression in the treated patients. Furthermore, insulin secretion as measured by T , , , occurred more rapidly after engagement in bulimic behavior and was delayed in treated patients following normal nutrition. lt is postulated that insulin potentiation and suppression of stress hormones as reflected by serum cortisol are important self-perpetuating mechanisms in bulimia and that these are related to the occurrence of intermittent starvation and its relief through binge eating. Treatment reduces this effect by reestablishing regular nutrition.Self-induced starvation is an important component of both anorexia nervosa and bulimia (Pirke et al., 1985) where binge eating has been said to represent