A peak shift on a line-tilt continuum.

Pigeons learned to peck when a vertical line appeared. They never got food for pecking at a 45-degree line.

After training, the birds saw lines tilted at many angles. The researchers tracked how often the birds pecked at each new tilt.

The birds pecked most at a line slightly past vertical, away from the 45-degree no-food cue. This 'peak shift' shows the pigeons acted as if the safe line was tilted even more than it really was.

Périkel et al. (1974) later showed the size of this shift grows when the 45-degree line appears less often. The same tilt setup now tells us how reinforcement rate tunes the illusion.

Griffin et al. (1977) kept the tilt task but added a light that signaled food was coming. Even without extra training, the generalization curve still slid away from the old no-food angle. The shift survives when cues change.

Grusec (1968) gave mild shock whenever the 45-degree line showed up. The birds again peaked farther from that line, proving that bad feelings alone can push the gradient over.

When you teach a child to pick the 'same' picture or the 'different' shape, watch where their true peak lands. If they learned 'not that one,' their best pick may drift away from the trained foil. Check generalization probes carefully; the real peak can hide one step past the stimulus you thought they learned.