Four young stutterers were observed during 10 weekly sessions. Each session was divided into pretreatment, treatment, and posttreatment segments. Redeemable tokens were administered contingent upon stuttering behaviors in the treatment segment of the Experimental condition. In a Parallel Control condition, no tokens were administered during the entire session. The differences between the pretreatment and treatment segments were compared for the two conditions. Three subjects had dramatically fewer stuttering behaviors when tokens were being administered, while the fourth had more stuttering under the same condition. The subject whose stuttering increased had a history of therapy in which voluntary, “faked” stuttering had been called for, and the behaviors that increased were judged to be of this type. The decreases were interpreted as suggesting that the contingent tokens acted to countercondition the aversiveness of the stuttering experience, which reduced the anticipation of stuttering and hence the stuttering itself. The increase was felt to be simple reinforcement. The counterconditioning interpretation was borne out in two clinical applications in which money was presented contingent on stuttering behaviors judged to be aversive to the stutterer, and in which dramatically sudden, but long-lasting, improvement was seen.