Date of Graduation


Document Type



Punishment may decrease the rate of unpunished behavior (i.e., punishment effects may generalize). Only a few studies with animals (e.g., Honig & Slivka, 1964) have investigated the stimulus generalization of punishment effects, or how they vary systematically with physical properties of stimuli. In a recent study with humans (O'Donnell & Crosbie, in press), each response in the presence of one horizontal line was punished, and generalization of punishment was tested in the presence of nine other lines that differed only in length. Generalization gradients were not obtained, most likely because the punisher became the discriminative stimulus for punishment conditions, thus rendering the programmed stimulus dimension (line length) nonfunctional. The purpose of the present set of experiments was to eliminate the punisher as a discriminative stimulus, thereby increasing the probability of obtaining gradients. In Experiment 1, reinforcement gradients were obtained with the line-length continuum to eliminate it as a factor should punishment gradients not be obtained. In Experiments 2 and 3, the procedure used in O'Donnell and Crosbie (in press) was arranged with point loss delivered at the end of the component (Experiment 2) or on an intermittent schedule (Experiment 3). Neither procedure resulted in gradients. Traditional discrimination-training and generalization-testing procedures were used in Experiments 4 and 5, but gradients were obtained only when point loss was delivered at the end of the entire session (Experiment 5). In Experiment 5, punishment conditions were in effect without punisher delivery, which resulted in discriminative control by line length, and thus produced generalization gradients. Results of present experiments suggest that when the punisher is a discriminative stimulus for punishment, all other stimuli become irrelevant, and that punishment both increases and decreases the rate of unpunished behavior.