You cannot reinforce a behavior that has never happened. That sounds obvious and it rules out most of what we want to teach: a child who has never rolled a ball toward someone cannot be rewarded for rolling it, and waiting for a full roll to appear on its own can take a very long time.
Shaping is the way out. You take something the child already does, reward it, and then progressively reward only the versions that are closer to where you are going.
Figure — From touching a ball to rolling it
- Touch
- Push
- Roll toward another person
The four decisions
Where you are going. Write the target down, concretely enough that two people would agree it happened. “Plays better with others” is not a target. “Rolls the ball at least a foot toward a partner” is.
Where you are starting. Watch and find the closest thing the child does today. Touching the ball at all. A hand raised halfway. Sitting for four seconds. If you cannot find a starting behavior, shaping is the wrong tool right now and the job is to arrange a situation where something happens at all.
The steps in between. Sketch them, but hold the sketch loosely — the child’s actual behavior will suggest better steps than your plan did.
When to move. This is the one that decides everything, and it has no formula. The usable rule: move up when the current version is happening easily and repeatedly, and not before. If you have to wait for it, you are not there yet.
The part that feels wrong
At step four, the version you were happily accepting at step three stops earning anything.
This feels unfair and it is the mechanism. If a light touch keeps getting the ball handed back for another turn, a real push has no reason to appear — the child already has a working option and it is cheaper. So the old approximation goes quiet, the child tries a bit harder, and that slightly better version is what you catch.
Two things follow from that. Expect a wobble when you raise the bar: the child will try the old version a few times, more insistently, before producing something new. That is the normal shape of the change. And do not raise two steps at once, because if the child cannot reach the new bar, nothing gets reinforced at all, the behavior falls apart, and you end up further back than you started.
If that happens, the fix is not to push harder. Drop back to the last step that was working, get it solid again, and make the next step smaller.
Shaping is not prompt fading
These get confused constantly, and the confusion is expensive.
Prompting and fading keep the response the same and change the help: you show the child, then show less, then point, then nothing. The behavior you want is there from the first trial, propped up.
Shaping, on its own, gives no help at all — what changes is the behavior itself, which starts as something else and becomes what you wanted. In practice the two are often combined rather than kept apart (next paragraph), so this is the definition to reach for, not a rule about how a program has to be built.
The practical consequence: if the child can already produce the response with help, you want fading, not shaping. If the response does not exist in any form, help cannot summon it and shaping is the tool. Plenty of programs use both — shape a rough version into existence, then use prompts to sharpen it — but knowing which one you are running tells you what to do when it stalls.
Where it goes wrong
Never raising the bar. A common failure, and the most comfortable one: the child is succeeding, everyone is pleased, and nothing has changed since February. Writing down the step and the date is what makes this visible.
Raising it too fast. The other direction. The behavior stops being reinforced, the child stops producing it, and the work of several sessions goes.
A target nobody can score. If two adults disagree about whether it happened, the criterion moves without anyone deciding to move it, and the child gets an inconsistent rule.
Shaping something the child does not want to do anything about. Shaping works with what is already happening. When nothing is happening, the question is not which step comes first — it is what would make the situation worth engaging with at all.
Further reading
- The mand — the request, which is usually the first thing worth shaping.
- When progress stalls — how to tell a flat line from a step that was never raised.