Charlop, Burgio, Iwata, and Ivancic [J. Appl. Behav. Anal. 21 (1988) 89] demonstrated that varied punishment procedures produced greater or more consistent reductions of problem behavior than a constant punishment procedure. More recently, Fisher and colleagues [Res. Dev. Disabil. 15 (1994) 133; J. Appl. Behav. Anal. 27 (1994) 447] developed a systematic methodology for predicting the efficacy of various punishment procedures. Their procedure identified reinforcers and punishers (termed ''empirically derived consequences'' or EDC) that, when used in combination, reduced the destructive behavior of individuals with developmental disabilities who displayed automatically maintained destructive behavior. The current investigation combines these two lines of research by comparing the effects of constant versus varied punishers on the self-injury of two individuals with developmental disabilities. The punishing stimuli were selected via the procedures described by Fisher et al. and were predicted to be at varying levels of effectiveness. The varied presentation of punishers resulted in enhanced suppressive effects over the constant presentation of a punisher for one of two individuals, but only in comparison to a single stimulus predicted to be minimally effective. Even then, the differences were small. These results suggest that the additive effects of varied punishment are negligible if clinicians use stimuli predicted to be effective and are discussed in terms of the conditions under which stimulus variation could potentially enhance the effects of punishers.