Limits on reinterpreting instrumental conditioning in terms of classical conditioning