The relative aversiveness of signalled vs unsignalled escapable and inescapable shock.

Badia et al. (1972) worked with lab rats in a two-sided box. Each side gave the same mild electric shocks, but only one side turned on a light before every shock.

The rats could press a lever to stay on the signaled side or switch to the unsignaled side. The team tracked which side the animals picked over many sessions.

Every rat chose to keep the light-and-shock side. They pressed the lever to stay there and avoided the side where shocks arrived without warning.

The light did not remove or reduce the shock; it only announced it. Still, the animals acted as if the warned shocks hurt less.

Macdonald (1973) ran almost the same test and got the same result, a clean direct replication. The finding held when the setup changed from escape to punishment schedules.

Liberman et al. (1973) pushed further: rats still picked the signaled side even when those shocks were nine times longer or three times stronger, proving the warning effect is robust.

Hymowitz (1981) looks like a contradiction at first. That study found no clear advantage for signaled shock, but it measured how much rats pressed a lever for food, not which side they chose. Different task, different story—choice tasks reveal the preference, suppression tasks do not.

If you must use mild punishment or unavoidable corrections, give a clear, consistent warning stimulus first. A two-second red card, a beep, or the word “reset” can act like the rat’s light. Clients will then treat the consequence as predictable, not arbitrary, and problem behavior is less likely to spike or spread. Keep the signal dependable; the later studies show that spotty warnings lose their power.