The role of physical identity of the sample and correct comparison stimulus in matching-to-sample paradigms.
Matching-to-sample works fine when sample and correct choice look different—compound stimulus history is enough.
01Research in Context
What this study did
Pigeons pecked a three-key matching-to-sample task. The sample and correct choice were line orientations that never looked the same. Researchers asked: do birds need the sample to match the correct picture?
Sessions used higher-order conditional discrimination. Birds saw a tilted line on the center key. Two side keys showed new tilts. One tilt was correct only because it had shared training history, not because it looked like the sample.
What they found
The pigeons learned the rule anyway. Their pecking followed smooth stimulus-control gradients. Accuracy did not jump when the comparison happened to resemble the sample. The birds used the whole compound cue, not simple identity.
In plain words: physical sameness is optional. Compound stimulus relations alone can drive correct matching.
How this fits with other research
Bennett et al. (1973) saw poorer accuracy when pigeons faced compound samples. That sounds like a contradiction. The difference is compound type: shared-attention pictures in 1973, line-orientation sets here. The earlier mix of overlapping pictures hurt control; the later clean line sets helped it.
Robinson et al. (1974) showed that new stimuli usually disrupt matching. Santi (1978) extends that work by proving the disruption is not about missing identity. Once compound cues are set, birds can keep performing even with fresh tilts.
Saunders et al. (1988) later tested symmetry and transitivity in these same pigeon preparations. They found none. Together the papers draw the line: pigeons can learn conditional rules without identity, but those rules still do not become equivalence classes.
Why it matters
When you build conditional-discrimination lessons for learners with autism, you do not have to make the correct choice look like the sample. You can teach arbitrary relations such as printed word to picture, or icon to object, from day one. Just ensure the stimulus compounds are clear and well separated. Check for generalization early; birds and humans alike may need extra training sets to broaden the skill.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Pick one arbitrary relation your learner needs (e.g., written name to photo). Run ten trials where the items never look alike. Record accuracy to see if identity matching is really required.
02At a glance
03Original abstract
Pigeons were trained in a higher-order conditional discrimination paradigm to assess the role of physical identity in a within-subjects design. A line orientation which was super-imposed on all response keys signalled whether a response to the matching color or a response to the nonmatching color was correct. Following training under this paradigm, stimulus control gradients were obtained by varying the angularity of the lines. Orderly gradients of stimulus control were obtained and no bias toward or away from the physically identical comparison stimulus was observed. The data were interpreted as indicating that the pigeons acquired a discrimination for each specific stimulus configuration or a set of specific stimulus-response chains based on compound stimuli in which physical identity played no special role.
Journal of the experimental analysis of behavior, 1978 · doi:10.1901/jeab.1978.29-511