Correspondence between saying and doing, typically studied in young children and individuals with developmental disabilities, was examined as an instance of conditional stimulus control. In Experiment 1, 3 pigeons were exposed to a two-component repeated-trials procedure. In the first-sample or say-component, two response keys transilluminated by different colored lights were presented and the pigeon pecked one of the keys. After 1 s of darkness in the chamber, the second-choice or do-component was presented, in which the two keys again were transilluminated, one by the color selected in the first component and the second by another color. Selecting the color that matched that selected in the say component resulted in access to food. Selecting the other color produced a blackout of the chamber. After an intertrial interval (ITI), the next say component was programmed, and the procedure was repeated. Correspondence remained at chance levels through several manipulations of ITI duration and sample response requirement. When a correction procedure was added such that only the originally selected sample stimulus was re-presented until a correct choice response occurred, reliable correspondence developed in 2 pigeons. This correspondence was eliminated by making reinforcement independent of correspondence and subsequently was reestablished when reinforcement again depended on correspondence. In Experiment 2, 3 other pigeons rapidly acquired correspondence under the final procedure used in Experiment 1. Increasing the time interval between the say and do components diminished correspondence. The results of the two experiments suggest how correspondence may be considered an instance of conditional stimulus control and that it is possible to construct a homologue of human say-do correspondence with pigeons.