Self-injuring behavior (SIB) is a life-threatening behavior exhibited by many species, including humans, and has no known cause and no agreed upon treatment. The role of the stress axis in the maintenance of this mysterious behavior was examined in subjects with life-long SIB. Over a 6-year period, 40 hr of direct observations of behavior and the environment were recorded on palmtop computers while 36 residential subjects (28 target and 8 control subjects) conducted their daily activities. Blood samples were collected in morning and evening for all subjects and within minutes after a self-injuring act in 28 target subjects who exhibited SIB to determine levels of ACTH and B-endorphin (BE). Self-injuring events in the patient group were significantly sequentially dependent (i.e., the only predictor of a self-injuring act was an antecedent self-injuring act). Higher morning levels of BE relative to ACTH predicted [r(df=27) = .57, p < .001] the sequentially dependent pattern of SIB. This effect was validated in a subgroup retested several months later [r(df=22) = .60, p < .001]. A subgroup of seven subjects exhibiting sequentially dependent patterns were administered an opiate blocker (naltrexone) in a double-blind, crossover design with an additional 14 hr/week of observation for 7 weeks. Naltrexone challenge interrupted the sequential pattern (improved behavior) in subjects with elevated BE immediately following SIB (r = .85, p < .01). The pattern of results supported the conclusion that the stress axis played a significant role in the maintenance of complex episodes of self-injury.