The two-factor theory of avoidance behavior predicts that a warning signal that precedes shock will have an aversive function, and if this is true then presenting a warning signal contingent upon a response should punish that response. In the present study, rats responded in a two-component multiple schedule of free-operant shock avoidance. Both components used a response-shock (RS) interval of 30 s, and a shock-shock (SS) interval of 5 s. A distinct warning signal was presented in each component, and signals differed primarily in terms of their respective durations. In one component, the signal appeared 29 s before RS shocks, and in the other component the signal appeared 5 s before RS shocks. Following the establishment of stable baseline responding under this arrangement, a conditioned punishment phase was conducted. The 5-s signal was presented briefly (0.5 s) following responses in the component with the 29-s signal. This had little effect on response rates and IRT distributions in the 29-s-Signal component for four of five rats. Relative to baseline, the fifth rat had lower overall response rates and fewer short IRTs in the 29-s-Signal component during conditioned punishment. When this fifth rat was introduced to a sham punishment condition, in which an arbitrary stimulus was presented for 0.5 s following responses in the 29-s-Signal component, response rates and IRT distributions in that component approximated baseline results. However, subsequent attempts to replicate baseline and sham punishment conditions with this rat were not successful. Taken together, presenting warning signals contingent upon a response did not suppress responding, as the two-factor theory would predict.