Stimulus generalization of behavioral history.

Okouchi (2003) asked 12 college students to press a key when a line looked "just right." Only one line length paid points. After 45 minutes, the team flashed new lengths to map how far the learning spread.

The set-up copied classic pigeon tests. Birds peck for food; students pressed for points. Both give clean curves of generalization.

The peak of responding sat slightly longer than the reinforced line. The curve sloped down on the short side and stretched out on the long side — the same lopsided hill seen in animals for decades.

In plain words, humans "overshoot" the trained size, just like pigeons.

Terrace (1969) first drew these hills in pigeons. Hiroto’s human curves sit on top of that data, showing the pattern is not just for birds.

Hoffman et al. (1966) found two training tones can create a twin-peaked hill. Hiroto used one trained length, so he got one peak — no contradiction, just fewer tops.

HOFFMAN et al. (1963) showed that suppression hills can last 2.5 years. Hiroto looked at minutes, not years, but both prove that once a stimulus controls behavior, the imprint is tough to erase.

Your client may call a 3-inch picture "dog" but not the 2-inch flashcard. Expect the generalization gradient: strongest near the trained size, fading as you shrink or stretch it. When you probe bigger sizes, responding can bounce back a little — plan extra trials there. Teach the exact size first, then branch out in small hops, not big leaps.