The role of discrimination training in the generalization of punishment.

Mello (1966) worked with pigeons to see if punishment needs discrimination training. Birds pecked a key for food. Some birds first learned which key color meant "peck here." Then punishment shock arrived for pecking the wrong color.

Other birds got the same shock, but never learned the color rule. Researchers later tested many colors to map how the birds spread their fear.

Pigeons that had clear color rules showed neat hills of fear. They pecked little near the punished color and more as colors grew different.

Birds without the color lesson showed flat lines. Shock scared them the same amount across every color. No hill, no order.

SLOANE (1964) ran a similar pigeon lab two years earlier. They added extra "do not peck" colors along a flicker line. Each new stop sign pushed the birds' pecking toward the safe end. Mello (1966) used the same trick but with shock, proving the idea works for punishment too.

Fontes et al. (2018) showed the dark side: punishing one move can bring back an old problem move. Their 2018 pigeons relapsed when the new punished choice hurt. Mello (1966) tells us why—without clear color rules, the birds could not tell where the pain belonged, so fear sprayed everywhere.

Green et al. (1986) found the same need for rules in extinction. Only birds that first learned "this room means no food" later showed big recovery when the room returned. Together, the three papers shout one message: train the signal first, or your consequence floats free.

Before you add a timeout or a response cost, first teach the client exactly when the rule applies. Use clear colors, words, or pictures for "yes" and "no." Then probe with similar cases to be sure the fear or suppression follows the rule, not the room. Without that step, your punishment may scare the wrong behavior or return old problems.