Recovery of responses during mild punishment.

Rachlin (1966) worked with pigeons that pecked a key for food.

The birds faced two schedules. In one, mild shocks sometimes followed pecks. In the other, no shocks happened.

After punishment, the team stopped all food and watched how the birds reacted during extinction.

The pecking fully bounced back while shocks still occurred.

When shocks stayed on during extinction, that component died out faster.

When shocks were turned off, the once-punished component slowed extinction, showing a lingering after-effect.

Davison et al. (1968) ran a near-copy design and also saw equal suppression across rich and lean food schedules. Their birds kept low error rates after shocks ended, matching the recovery theme.

Sailor (1971) pushed the idea further. He dialed shock up and down and found an upside-down U: mid-level shock briefly spiked unpunished extinction responding. This parametric twist extends H’s basic recovery curve.

SIDMAELLIOTT et al. (1962) came first. They showed punishing one response can make alternate responses rise. H’s later data fit that contrast picture, turning the 1962 finding into a recovery rule.

For you, the lesson is timing. If you pause punishment before you start extinction, the old problem behavior may stick around longer. Keep the mild punisher in place while you withhold reinforcement and the behavior drops faster. Test this in your next multiple-schedule program or DRA setup.