Autoshaping: further study of "negative automaintenance".

The team worked with pigeons in a small lab space. Each bird saw a lit key followed by grain.

They tested four plans. Reward: peck earned food. Omission: peck canceled food. Extinction: no food. Blank: no light, no food.

Birds pecked most when pecking paid off. They still pecked when pecking blocked food, just less.

Extinction and blank trials produced almost no pecks. The light-food link alone kept the response alive.

Neuringer (1973) showed pigeons start pecking faster with shorter lights. T et al. used the same light timing to test what happens after the response starts.

Kirby et al. (1981) later repeated the omission plan with rats. Rats stopped pawing but switched to nosing the lever, showing omission reshapes rather than erases behavior.

Herrnstein et al. (1979) pushed further. They delayed both food and light change after each peck. Responding still held, proving the signal itself can keep the behavior going even when every consequence is bad.

The light-food pairing can keep a response running even when the response loses food. When you fade reinforcement, check if the stimulus itself is now driving the behavior. If it is, change the stimulus or add a new contingency, not just more extinction.