When do errors in reinforcer delivery affect learning? A parametric analysis of treatment integrity

Bergmann et al. (2021) asked a simple question. How many reinforcers can you miss before learning breaks down?

They ran a computer lesson with neurotypical adults. The program taught new picture matches. Sometimes it forgot to deliver the reinforcer. The team varied the miss rate from 0% to 30%.

Learning stayed on track when up to 15% of reinforcers were skipped. Beyond that, mastery slowed or stopped.

In short, you can flub one in every six or seven rewards and still get acquisition.

Argueta et al. (2024) reviewed 31 studies that tried to create new conditioned reinforcers. Their paper shows that half of those attempts failed. Bergmann’s work explains part of the reason: if the new reinforcer is also delivered late or missed too often, it can’t do its job.

Grosch et al. (1981) taught adults with severe disabilities to play darts. Every correct step earned praise or tokens—no misses. All learners mastered the game. Their perfect reinforcement schedule lines up with Bergmann’s ceiling: stay under 15% errors and you’re safe.

Sievers et al. (2020) ran an eight-session safety class for adults with ID. They kept reinforcement tight and saw clear gains. Again, the outcome matches Bergmann’s rule: consistent delivery keeps acquisition smooth.

You don’t need robotic perfection. If a tablet glitches, a therapist looks away, or a token drops on the floor, relax—one missed reinforcer in six won’t ruin the program. Do keep a quick tally during sessions. When misses creep past 15%, pause and fix the delivery system before continuing.