Schedules of conditioned reinforcement during experimental extinction.

KELLEHER (1961) wrote a think-piece, not an experiment.

He asked: if a stimulus has become a conditioned reinforcer, what happens when food stops?

He used Skinner’s 1938 rat data to argue that these ‘secondary’ reinforcers can keep responses alive longer during extinction.

The paper claims that conditioned reinforcers act like glue.

Even after the main reinforcer is gone, the animal keeps pressing if lights or clicks had previously predicted food.

In short, extinction takes longer when ‘good-news’ signals are still in the box.

Nevin et al. (2005) later tested the idea with pigeons. They showed that faster blinking lights raised peck rate, but the birds still quit quickly when food vanished. The signal boosted performance, not staying power.

Kohlenberg (1973) found the same glue effect only when the light stayed on for the whole wait and sometimes signaled the bigger payoff. Brief flashes or equal-payoff cues lost their power.

Buskist et al. (1988) added a timer rule: a half-second signal can bridge up to 9 s of delay, but 27 s needs the light to stay on. Together these studies draw the borders of the glue: signals help only if they are long, reliable, and richer at times.

Before you start extinction, list every sound, tablet ding, or teacher nod that might be a conditioned reinforcer. Remove or neutralize those stimuli if you want the target response to drop fast. If you must keep the stimulus, make it brief and equal across conditions so it loses value.