A systematic replication investigating the efficiency and effectiveness of restricted‐ and free‐operant programming

Bulla and colleagues asked a simple question. If you keep reinforcement the same, does it matter whether the learner can move freely between two buttons or must stay on one?

They taught neurotypical adults to tell new numerals apart. One group used a restricted setup: press the left key until correct, then the right key until correct. The free group could hop between keys at any time. A control group got no special rules.

All three groups earned points on the same schedule. The only difference was how they were allowed to respond.

Restricted-operant teaching won. People in that group learned the new numeral pairs faster than the free-choice or control groups.

Fluency at the end was the same for everyone. The gain was in speed of learning, not final skill level.

McCormack et al. (2019) pooled 43 experiments and found that pairing one unique reward with each correct response speeds learning. Bulla’s setup does exactly that: one key equals one reinforcer, so the result lines up.

McGonigle et al. (1982) showed that limiting rich-schedule responses makes lean-schedule responses rise. Bulla flips this idea into teaching: by restricting where the learner can respond, they get more correct trials per minute and learn faster.

Fabbretti et al. (1997) saw pigeons prefer two keys over one even when payoff was equal. Bulla’s humans had no preference phase, yet they learned faster when the choice was taken away. Together the papers hint that preference and learning efficiency can pull in opposite directions.

If you run discrete trials, try a quick restricted phase. Lock the student into one response option until a few correct hits occur, then switch. You may shave sessions off acquisition without changing reinforcers or task difficulty. Keep the option open for maintenance once the skill is solid.