Selective punishment of interresponse times.

Galbicka et al. (1981) worked with rats that pressed a lever for food.

The rats earned food on a VI schedule, meaning pellets came at random times if they kept pressing.

The team gave a quick shock whenever a rat waited too long between two presses. They only shocked the long waits, not the short ones.

Shock for long waits cut those waits down.

The rats started pressing faster, but only the long pauses were hit with shock.

Total pressing stayed high, showing punishment can trim one part of a response pattern without killing the whole behavior.

DARDANO et al. (1964) did an early lab test of selective punishment. They shocked one of two levers and saw rats switch to the safe lever. Their work set the stage for picking which part of behavior to punish.

Jarrold et al. (1994) moved the idea to a classroom. They punished both mild and severe hits, not just the worst ones. When they shocked only the severe hits, the aggression stayed. Punishing the full range worked. This child case echoes the rat lab: you must punish the whole response class you want to change.

Fontes et al. (2025) later showed limits. In fast-changing schedules, the direct-suppression rule broke. Their 2025 lab study warns that real-world settings with quick shifts may not follow the clean IRT pattern seen here.

You can use punishment like a scalpel, not a hammer. If a client stalls too long between tasks, deliver a mild punisher only for the long pauses. Keep reinforcement flowing for quick responses. Check that you are punishing every instance of the target delay, not just the longest ones, or the shape may not hold.