Eliminating behavior with reinforcement.

Scientists compared two ways to make a rat stop pressing a lever. One group got plain extinction: no food ever. The other group got DRO: food popped out only when the rat did NOT press for five seconds.

The rats worked alone in small boxes. Each session tracked how fast the lever pressing faded away.

DRO beat extinction. Lever pressing dropped quicker and stayed gone. Extinction alone let the behavior bounce back later.

Only the extinction rats showed 'spontaneous recovery' the next day. The DRO rats kept quiet.

Pear et al. (1971) ran the same comparison in the same lab and saw the same win for DRO. Their yoked-control setup showed the timing of the food mattered; random food did not work as well.

Rogers-Warren et al. (1976) later topped both studies. They gave monkeys food for pressing a different lever during extinction. That trick cut the old behavior even faster than DRO, so the 1971 DRO finding was improved upon, not contradicted.

Bensemann et al. (2015) and Hangen et al. (2020) added a warning. In humans, DRO still shrinks the target behavior, but it can accidentally grow other odd behaviors because 'anything except the target' gets reinforced. Track all responses, not just the problem one.

When you need quick suppression, reinforce something specific instead of using DRO or extinction alone. If you do use DRO, watch for new odd behaviors that might pop up. Collect data on every response class, not just the one you want gone.