Numerous studies have demonstrated that disruptive classroom behavior can be decreased by delivering tokens contingent on periods of time during which children do not engage in it or by removing tokens contingent on its occurrence. To date, the best controlled of these studies have consistently reported the two procedures to be equally effective. However, the effects of the token contingencies have been confounded with those of instructions. The present study compared these two procedures when no instructions were given regarding the token contingencies. Token delivery was not effective in decreasing disruptive behavior in any of the children, while a combination of token delivery and removal was effective for three of four children. The results suggest that the combined procedure may be effective with certain populations that are not readily controlled by instruction. (Author)