The shape of some wavelength generalization gradients.

Blough (1961) looked back at dozens of pigeon experiments on color generalization.

The birds pecked a key for food when lights of different wavelengths appeared.

The paper asked: do the old Hull-Spence math rules predict how the birds spread their pecks?

Two rules held up. Birds showed transposition: they picked the greener green even after training.

Peak shift also appeared: after training on green vs blue, birds peaked on a still-greener hue.

Other rules failed. Gradient summation, drive level, and extinction curves did not match the math.

Harrison et al. (1975) later showed the same gradient shape holds when pigeons avoid shock instead of earning food.

Baer (1974) added that massed extinction sessions make the color hill even steeper, giving clinicians a way to tighten stimulus control.

Schmidt et al. (1969) and Brinker et al. (1975) proved peak shift needs only tiny cues—like a shorter hopper time or a 3-min break—so you can get the shift without long contrast histories.

You now know that basic color-peak rules survive across food, shock, and extinction setups.

When you want sharper stimulus control, pile on extra training or massed extinction.

When you see a learner respond harder to a novel cue just past the trained one, expect peak shift, not confusion—it is normal.