Pigeons' discrimination of Michotte's launching effect.

The team showed pigeons short cartoons. One ball moved, hit a second ball, and the second ball rolled away. This is called the "launching effect."

Birds got food if they pecked only when the second ball moved right after contact. They had to ignore clips where the second ball stayed still. Training ran for many daily sessions.

The pigeons finally learned to pick the "launch" clips, but accuracy stayed low. Probe tests showed the birds were not watching the true cause. They pecked when the balls touched at the same spot on the screen, even if no launch happened.

THOMAS et al. (1963) also worked with pigeons. They proved birds must first master side-by-side matching before they can handle delayed matching. Lancioni et al. (2006) shows the same rule: simple cues must be firm before abstract ones can take over.

Brown et al. (1994) taught pigeons to form equivalence classes. Those birds built brand-new relations after training. The 2006 birds never reached that stage; they stayed stuck on surface details. Together the studies warn that strong stimulus control is rare without careful shaping.

Mello (1966) showed that punishment produces flat generalization unless the bird first learns a clear S+ and S-. The 2006 study echoes this: when the intended cue is too subtle, animals grab any handy difference, just like the flat gradients K saw.

Your learner may look like he "gets it" yet be cueing on a tiny accident of the task. Run probe trials that keep the target relation but move or remove the accidental cue. Only when accuracy holds do you claim true concept learning.