A critical issue in testing theories of observing is whether the stimulus associated with extinction (the S—) reinforces observing responses. In previous experiments, subjects have been trained to make observing responses that produce both the S— and the stimulus correlated with reinforcement (the S+). Then, either the S+ or the S— has been withheld. Conflicting results have been attributed to differences among species. In the present experiments, pecking one key by master pigeons was reinforced with grain on a variable‐ratio extinction schedule. Yoked pigeons received the grain on a variable‐interval, extinction schedule controlled by the variable‐ratio performances of the master birds. For both groups, concurrent pecking on a second key was reinforced on a variable‐interval schedule with displays of discriminative stimuli. Subsequently, either the S+ or the S— was eliminated from the procedure. Omission of S+ produced a large decrease, as predicted by traditional conditioned reinforcement accounts of observing. By itself, S— did not maintain observing. A smaller and less reliable decrease, comparable to that obtained by Lieberman (1972) with rhesus monkeys, occurred when S— was eliminated. This replication with pigeons of Lieberman's results indicates that they are not species‐specific, and the fact that observing was not maintained by S— alone suggests that the decrease obtained when S— was omitted is not attributable to th