Recent research has suggested that operant responses can be weakened when they are tested in new contexts. The present experiment was therefore designed to test whether animals can learn a context-(R-O) relation. Rats were given training sessions in context A, in which one response (R1; lever pressing or chain pulling) produced one outcome (O1) and another response (R2; chain pulling or lever pressing) produced another outcome (O2) on variable interval reinforcement schedules. These sessions were intermixed with training in context B, where R1 now produced O2 and R2 produced O1. Given the arrangement, it was possible for the animal to learn two distinct R-O associations in each specific context. To test for them, rats were then given aversion conditioning with O2 by pairing its presentation with lithium-chloride-induced illness. Following the aversion conditioning, the rats were given an extinction test with both R1 and R2 available in each context. During testing, rats showed a selective suppression in each context of the response that had been paired with the reinforcer subsequently associated with illness. Rats could not have performed this way without knowledge of the R-O associations in effect in each specific context, lending support to the hypothesis that rats learn context-(R-O) associations. However, despite a complete aversion to O2, responding was not completely suppressed, leaving the possibility open that rats form context-R associations in addition to context-(R-O) associations.