The effect of overtraining on behavioral contrast and the peak-shift.

Researchers gave pigeons extra discrimination practice after they had already learned the task. They wanted to see if this 'overtraining' would make a peak shift more or less likely when the birds were later tested with new colors.

The team varied how long the extra training lasted and how the color stimuli were arranged. Then they checked whether any of these factors could predict if a peak shift would show up.

No clear pattern emerged. Some birds with lots of extra training showed a peak shift; others did not. The amount of responding to the trained color just before the test did not forecast what would happen.

In short, overtraining duration and stimulus setup were not reliable predictors of peak shift during generalization tests.

Reynolds (1966) had already shown that very long discrimination training can wipe out both behavioral contrast and peak shift. Dukhayyil et al. (1973) now add that, even when the shift survives, you cannot guess its presence from the bird's final response rate.

Halpern et al. (1966) found that equal response rates abolish peak shift. The new data agree: response rate at the end of training tells you nothing about whether a shift will appear later.

Schmidt et al. (1969) got peak shift by changing reinforcement duration cues. Their positive result highlights that only certain training tweaks, not simply 'more of the same,' control the shift.

If you are teaching a client to tell two colors, sounds, or shapes apart, do not assume that extra drill after mastery will create (or prevent) a peak shift. The shift may come or go, but the learner's final performance level will not tip you off. Focus on the discrimination itself, not on reaching a magic number of 'overtraining' trials.