The term procedural fidelity refers to the degree to which a procedure is implemented as designed. Reduced procedural fidelity negatively affects differential reinforcement of alternative behavior (DRA) and response cost, but little is known about its effects on differential reinforcement of other behavior (DRO). We attempted to identify a potential “critical level” of fidelity during DRO implementation by parametrically manipulating fidelity across five levels (100%, 80%, 60%, 40%, and 20% fidelity) with college students who engaged in an arbitrary response (clicking a mouse on moving circles). Reduced fidelity involved delivery of reinforcers following unwanted behavior. Experimental control was demonstrated through a multielement design, in which 100% fidelity and reduced fidelity rapidly alternated, embedded in a reversal design, in which the level of reduced fidelity varied by phase. Two patterns of responding emerged: the DRO was effective at all levels of fidelity (for three participants) or lost its efficacy at 60% or lower fidelity (for the other three participants). Results suggest that delivering reinforcers following unwanted behavior may yield unpredictable outcomes during DRO. Several avenues for additional research are suggested.