A test of symmetry and transitivity in the conditional discrimination performances of pigeons.

Saunders et al. (1988) asked if pigeons can form equivalence classes. They used matching-to-sample training with colored lights. After each bird learned A-B and B-C pairs, they tested for two emergent relations.

Symmetry probes checked if the birds would reverse the trained pairs. Transitivity probes checked if the birds would link A to C without direct training.

None of the pigeons showed symmetry. None showed transitivity. The conditional relations stayed stuck to the exact trials that were trained.

In plain words, the birds acted like they had never seen the new combinations before.

The same lab had already seen trouble: Volkmar et al. (1985) got weak symmetry and zero transitivity in pigeons. The 1988 paper is a clean replication that removes the "maybe" and calls it a failure.

Later work shows the failure can be flipped. Ribes-Iñesta (1999) added class-consistent differential reinforcement and got transitivity in most pigeons. The birds did not suddenly get smarter; the procedure just gave extra help for observing the right cues.

Swisher et al. (2015) adds a spatial twist. They found symmetry in pigeons only when probe stimuli sat in the exact spots used during training. Move the lights one key over and the relation vanishes. This extends the 1988 warning: even tiny layout changes can break emergent relations.

For BCBAs, the lesson is to test, not assume. After conditional-discrimination training, run quick symmetry and transitivity probes before you call it an equivalence class. If the learner fails, add supports like differential reinforcement or keep stimulus locations identical across phases. These tweaks can turn a null result into a solid equivalence repertoire.