Inference by exclusion can be exhibited by deductively responding to new stimuli that are presented in the context of familiar stimuli. We investigated exclusion-based responding in pigeons using a 2-alternative forced-choice discrimination task. In Phase 1, pigeons learned to associate 2 stimuli (A and B) with Response 1 and 2 stimuli (C and D) with Response 2. Following successful acquisition of these stimulus–response pairings, pigeons advanced to Phase 2, in which stimuli A and B were now reassigned to Response 2. Based on their Phase 1 training, pigeons should initially choose Response 1 when presented with A and B in Phase 2 (this response is now incorrect, but the birds would not yet have had the opportunity to learn the new stimulus–response associations). Also, in Phase 2, stimuli E and F—new stimuli replacing stimuli C and D—were concurrently presented and assigned to Response 1. Without prior training, pigeons’ initial responding to E and F in Phase 2 should be at chance. However, if the pigeons were to apply an exclusion rule (stimuli E and F stand in opposition to stimuli A and B), then they might initially choose Response 2 for new stimuli E and F because they are concurrently choosing Response 1 for stimuli A and B. If that is the case, then choice accuracy for stimuli E and F should also be below chance. Indeed, our pigeons responded at reliably below chance levels to stimuli E and F, consistent with their exhibiting an exclusion rule–based strategy, which could actually arise from a more mechanical underlying process such as acquired equivalence formation.