An objective method is presented to group discrete self-injurious behavior (SIB) events into bouts. Survival analysis was used to determine how long after an SIB the probability of observing a subsequent SIB remained elevated. This estimated bout length criterion time point was determined individually for 19 subjects with developmental disorders and severe SIB. Bout length criterion estimates ranged from 0.6 to 15.5 seconds. When participants' SIBs were grouped into bouts using the bout length criterion estimate, the rate of bouts and the average number of SIBs in each bout were found to be more stable than the rate of individual SIBs. This approach may facilitate research on factors that initiate and maintain SIB.