Tactical contingencies in the experimental analysis of reinforcement and operant classes

Catania ran three tiny pilot studies. Each had one trial. A rat pressed a lever. The rat got food. The next trial came right after.

The team watched if the rat used the exact same paw position again. They wanted to know: does the food make that tiny detail more likely next time?

The answer was no. Reinforcing one exact paw spot did not raise the chance of that spot on the next trial.

Instead, the food picked a whole class of lever presses. The rat kept pressing, but the fine details drifted.

Baum (2025) looks at the same puzzle from the other end. Baum explains why ratio schedules crash when food is rare. He says the issue is molar feedback, not tiny trial links. Catania’s data back this up: moment-to-moment boosts do not drive behavior.

Russell et al. (2018) show tokens can act like food even after kids eat candy. That fits Catania’s view: reinforcers select classes, not specific moves. The token works because it joins the class, not because it copies the last response.

Regnier et al. (2022) tell us to fade token systems slowly. Their advice makes sense with Catania’s finding. If each reward does not lock in the exact move, you need extra steps to keep the whole class alive after tokens stop.

Diaz-Salvat et al. (2020) found that more response choices cut resurgence. This lines up too. When contingencies act on classes, giving extra members in the class lowers relapse risk.

Stop hunting for the perfect micro-reinforcer. Focus on building strong, wide response classes. When you teach a skill, reward any correct form at first. Then add variety and fade the reward. The class will hold even if the exact form shifts.