A cross-over study of 24 Ss with self-injurious behavior (SIB) was conducted over a continuous 10-week period. Treatment with naltrexone (NTX) was provided for 3 weeks in a randomized, reversal design with different doses or placebo each week. Videotaped observations (20 hr/subject), neurological examinations, and ratings of adaptive and maladaptive behavior were collected. Treatment with 2 mg/kg NTX produced at least a 50% reduction in SIB in a significant (p <.01) number of Ss. The 1.0 mg/kg was less effective (p <.02), and no significant change was observed at 0.5 mg/kg. Eighteen of 21 Ss achieved at least a 25% reduction in SIB after treatment of at least 1 dose of NTX (p <.0001). More than half of the Ss (52%) had a < 50% reduction (p <.001), and a significant number of Ss (33%) decreased SIB by more than 75% after at least 1 dose of NTX. Significant improvement was measured after NTX on measures of learning and attention.