Conditional probability, calculated using sequential analysis techniques in four time conditions (2, 10, 30, and 60 seconds), provided evidence that successive episodes of self-injury were sequentially dependent. This unique distribution of sequential association for self-injurious behavior (SIB) was not related to frequency or rate of occurrence. Compared with other environmental and behavioral events, the best predictor of SIB was an earlier SIB episode, consistent with a contagious distribution. This study is one of the few in which sequential analysis techniques were applied to data derived from a large group of individuals with severe behavior disorders. It may be the first in which this analytic tool was used to investigate systematically successive occurrences of SIB as it takes place in vivo.