Line-orientation generalization following signalled-reinforcer training.

Griffin et al. (1977) worked with five pigeons in a lab chamber. The birds pecked a key that showed tilted black lines.

A small blank spot on the key lit up only when a food reinforcer was ready. This signal let the bird know "food is coming now."

The team never taught the birds to discriminate line angles. They simply tracked how the birds generalized across many tilts after training.

Every pigeon showed a clear peak shift. The strongest pecking moved away from the training angle, even without explicit S- training.

The blank-key signal had segmented the schedule into "food moments" and "non-food moments." That segmentation alone was enough to reshape the generalization gradient.

Bloomfield (1967) first showed peak shift with line-tilt stimuli, but only after true discrimination training with an S-. Griffin et al. (1977) got the same shift without ever punishing the wrong angle.

Périkel et al. (1974) also moved the peak by changing relative reinforcement rates. Both studies prove the gradient is plastic, yet P et al. did it with a simple signal instead of rate juggling.

Smith et al. (2010) later extended the idea to noncontingent reinforcement. Their signal deepened response suppression, showing the same cue that helps in NCR can also reorganize generalization.

You can reshape stimulus control without extra trials or punishment. Just add a brief cue that marks when reinforcement is available. Next time a learner over-generalizes, try pairing a small visual or auditory signal with each reinforcer and then probe the gradient. You may see the peak drift naturally, saving you from lengthy discrimination sets.