Analysis of response rates during stimulus generalization.

Migler et al. (1969) watched pigeons choose between two levers when new colors appeared.

The birds had learned one color meant food on the left lever, another color meant food on the right.

The team then tested many in-between colors and counted how fast the birds pecked each lever.

The color still controlled which lever the bird picked, but it did not control how fast they pecked.

Middle colors did not produce medium speeds. Instead the birds switched back and forth between the two levers.

The authors say the "medium" rate is just the average of two clear behaviors, not a weak response.

Stevenson (1966) already showed that packing test colors close together can flatten the curve. B et al. now add that the flatness may be an illusion caused by alternation, not poor stimulus control.

Schwarz et al. (1970) saw the same thing the next year: absolute peck rates looked messy while choice told a cleaner story. Both labs agree that rate alone can mislead.

Allan et al. (1991) later moved the idea into a classroom. They probed each trained picture after teaching adults to say "please." Like the pigeons, people sometimes gave the right word only when the exact training picture showed up, revealing hidden stimulus control that group averages would have missed.

When you probe generalization, look at what the learner does, not just how often they do it. A flat or medium response rate may hide sharp stimulus control that appears only when you check choice, location, or alternation. Next time you run a generalization probe, record which item the client picks or where they go, even if the speed looks the same across stimuli.