In three observing‐response experiments relevant to the information hypothesis of conditioned reinforcement, the basic procedure was one in which an observing response produced one stimulus on trials that terminated in non‐contingent reinforcement and another stimulus on trials that terminated in a brief timeout. In Experiment I, the observing response consisted of a single peck or a short fixed‐ratio schedule (FR 3 or FR 6), depending on the type of trial. If the single peck produced the negative stimulus and the fixed ratio produced the positive stimulus, observing responses were maintained. If the single peck produced the positive stimulus and the fixed‐ratio produced the negative stimulus, observing responses were not maintained on negative trials. In the second experiment, the response key was either white or dark at the beginning of a trial, indicating whether it was a positive or negative trial. Observing responses continued to be maintained on positive trials but not on negative trials. In Experiment III, only positive or negative trials were scheduled for several sessions. Observing responses extinguished regardless of whether positive or negative trials were scheduled. The results do not support the hypothesis that making the stimuli produced by observing responses redundant will reduce observing re