The emission of a fixed number of responses by rats was followed by food reinforcement. This fixed number could be accumulated in any way from two continuously available but mutually incompatible response classes, bar pressing, and not bar pressing for a fixed time period. A preference for one response class was arranged by specifying different maximum reinforcement rates for the two classes. Under selective punishment conditions, the preferred response occasionally led to both food and electric shock, while the non-preferred response led only to food. Selective punishment effects were measured through changes in the preference to the two responses in the sequence. The actions of shock intensity, deprivation, the specification of the non-preferred response, and three drugs were investigated. The results were broadly similar to the work reported by Dardano and Sauerbrunn (1964), who found localized increases in interresponse times before punished responses in fixed-ratio schedules. Performance under this procedure was found to be stable and sensitive to each of the experimental variables examined.In a fixed-ratio (FR) schedule, the emission of a fixed number of responses is followed by reinforcement. A number of experiments have investigated the effect of punishing one ordinal response in the FR sequence (Dardano and Sauerbrunn, 1964;DeArmond-Edwards, 1966;Donahoe and Schulte, 1967;Richardson and Donahoe, 1967;Davison, 1968). These workers reported that when punishment (electric shock) was contingent upon the terminal response of the ratio (the reinforced response), increases in interresponse times occurred just before the punished response.Selective punishment techniques are not generally satisfactory for parametric investigations of the variables controlling punishment effects. If animals are trained at a number of shock intensities and deprivation levels (without causing responding to cease), difficulties occur in replicating quantitative measures of responding within each animal (Davison, 1968). If the shock intensity is raised so that responding ceases, food reinforcement is no longer obtained and the effects of punishment must be measured in extinction. Many problems arise in analyzing data from such rapidly changing behavioral baselines (Sidman, 1960).