Effects of spatial rearrangement of object components on picture recognition in pigeons.
Pigeons decide ‘same object’ by where the parts sit, so shuffle stimuli during training to lock control to the parts, not the layout.
01Research in Context
What this study did
The team trained pigeons to peck at pictures of whole objects. Then they tested the birds with the same pictures scrambled—parts moved around.
They wanted to know if the pigeons still said “same thing” when the layout changed.
What they found
Recognition dropped in a tidy order. Small scrambles hurt a little. Big scrambles hurt a lot.
The birds were judging by where the pieces sat, not just that the pieces were present.
How this fits with other research
Stevenson (1966) showed that spacing test items flattens generalization curves. Kirkpatrick-Steger et al. (1996) adds that rearranging the parts inside one picture does the same.
Sainsbury (1971) found that pushing elements closer helps pigeons learn “this pattern means no food.” The new study flips it: pulling parts apart wrecks recognition. Same factor—proximity—explains both boosts and drops.
Mueller et al. (2000) later proved pigeons can ignore location if you train them in many spots. Together the papers warn: early training with fixed layouts locks control to position, not just shape.
Why it matters
When you teach a child to match “dog,” keep the photo in different places and angles from day one. If you always flash the card dead-center, the learner may latch onto screen location instead of the dog features. Rotate, shift, and slightly scramble early examples so the useful parts—not their old seat—gain control.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Move your picture cards to a new corner of the table each trial for the next week.
02At a glance
03Original abstract
Five pigeons were first trained to discriminate among line drawings of four objects: a watering can, an iron, a desk lamp, and a sailboat. The birds were then tested with eight versions of each object, in which the object's components were vertically and horizontally rearranged. The pigeons displayed different degrees of generalization decrement to the different scrambled versions of the objects. Two analyses helped to clarify the nature of the varied accuracy scores. First, cluster analyses disclosed subsets of components that were related to test performance. Second, although the clusters varied somewhat across birds for a given object, there was reliable concordance among the subjects in their rankings of the individual scramblings, suggesting that the pigeons may have attended to common aspects of the drawings.
Journal of the experimental analysis of behavior, 1996 · doi:10.1901/jeab.1996.65-465