Glossary
Reward approximation
What is reward approximation?
Reward approximation, also called shaping, means rewarding successive approximations, or steps toward a desired behavior, and gradually asking for a closer match before giving the reward.
How it works
A trainer might first reward a rough version of a new movement, then reward increasingly precise versions. What earns the reward changes as the learner improves.
A shaping example
Imagine teaching a dog to touch a wall target with its nose. At first, a small food reward follows turning toward the target. Once that response is established, turning alone no longer earns the reward: the dog must approach it. The next criterion is touching the target with its nose. Each rewarded response more closely approximates the final action.
This hypothetical sequence illustrates changing reward criteria, not a guarantee of smooth progress. If a new criterion is too large a step, the learner may stop producing a response that earns reinforcement. The trainer must observe performance rather than advance solely because a fixed amount of time has passed.
Why it matters
Shaping turns a large jump into a staircase of smaller steps. The learner can work toward a behavior they cannot yet perform in full.
The BCT Taxonomy defines this as technique 14.4. The changing standard for earning a reward distinguishes shaping from simply making a task harder.