SOLUTION OF THE INTERMEDIATE SIZE PROBLEM BY PIGEONS.

The researcher asked pigeons to pick the middle-sized square from three choices. First the birds learned the task with food rewards. Then the team turned the food off and showed new squares the birds had never seen.

During these no-food tests, the birds still had to pick the middle. The goal was to see if they used the exact size they had learned or the relationship 'bigger than A, smaller than C.'

In extinction, the pigeons pecked the square that was closest to the trained middle size. They did not pick the true middle of each new set. This shows they were under absolute, not relational, stimulus control.

THOMAS et al. (1963) had already shown pigeons can learn delayed matching. KIEFFETHOMAS (1965) used the same birds and lab but asked a harder question: do they learn a rule or just the picture? The answer was 'just the picture.'

Neuringer (1973) later confirmed that extinction tests expose what the bird really learned. When training contrasted two click rates, extinction gave a sharp peak. When training only said 'present vs absent,' the peak vanished. Both papers show extinction is the clearest window into stimulus control.

Brown et al. (1994) moved beyond simple control and built stimulus equivalence in pigeons. Their transitive relations proved birds can learn rules, but only if the procedure explicitly teaches them. KIEFFETHOMAS (1965) did not teach a rule, so none appeared.

If you want a learner to follow a rule (bigger than, softer than, same as), reinforce comparisons across many examples. If you only reward one exact picture, the learner will lock onto that picture. Run brief extinction probes early; they reveal the real controlling cues before bad habits set in.