If your learner cannot emit the target behavior at all — not even a rough version — you cannot reinforce it. That is the problem shaping solves.
TL;DR: Shaping uses differential reinforcement to build a new behavior by reinforcing successive approximations of the terminal target. You start where the learner is, reinforce steps toward the goal, and raise the criterion as performance improves. Done well, shaping builds behaviors that no prompt could produce. Done poorly, it stalls, drifts, or never terminates.
What Shaping Is (and Isn’t)
Shaping is the differential reinforcement of successive approximations toward a terminal behavior.
Three components make up every shaping program:
- A terminal behavior — the final target you are working toward (e.g., saying “help,” climbing three steps unassisted, writing the letter “A”)
- A starting behavior — the response currently in the learner’s repertoire that most closely resembles the terminal behavior
- A progression of criteria — each intermediate step, with a defined decision rule for when to advance
Shaping is not the same as prompting. Prompting introduces antecedent support to help a learner emit a behavior they could not otherwise produce. Shaping differentially reinforces responses that occur without that support and progressively shifts which responses earn reinforcement.
If you are using prompts to pull a learner through each step of your program, you are not shaping — you are running a prompt-fading procedure. The distinction matters because shaping builds independent responding from the start; chasing each step with a prompt delays independence and creates prompt dependence.
How to Design a Shaping Program
A well-designed shaping program has five components written before you begin the first session:
1. Operational definition of the terminal behavior. You cannot shape toward a vague target. “Better eye contact” is not a shaping target. “Sustained eye contact for three consecutive seconds within two seconds of an SD” is.
2. The starting approximation. Run a baseline probe session and observe what the learner already does that resembles the terminal behavior. That is your first step. If a learner makes no vocal sounds at all, the first approximation might be any vocalization in any context.
3. All intermediate steps, listed explicitly. Write them before the first session, not in the moment. How many steps depends on the complexity of the terminal behavior and the learner’s history of responding to shaping. When uncertain, use more steps — you can skip one if the learner advances quickly, but adding a step mid-program when a learner is struggling requires re-planning under pressure.
4. Mastery criteria for each step. Common formats: a response-rate criterion (e.g., five emissions in a five-minute session) or a percentage-correct criterion across a fixed opportunity set. The specific numbers matter less than having them defined before you start rather than deciding in the moment.
5. Decision rules for advancing and reverting. When does the learner move to the next approximation? When does the criterion revert to a previous step? The reversion rule is the one practitioners most often skip — and it is the one that determines whether a stalled program gets unstuck or stays stuck. Pre-specify both directions.
The Three Mistakes That Stall Shaping Programs
Criterion jumps. When a learner meets the current step, the natural impulse is to skip ahead. Two or three steps forward seems reasonable in the moment. The learner fails repeatedly. Motivation drops. Progress reverses. Stick to the pre-specified step increments; adjust them between sessions, not mid-session.
Criterion freeze. The opposite problem: a learner meets criterion, but the clinician is uncertain about the next step or reluctant to risk losing the response. Reinforcement continues for approximations the learner has already mastered. The terminal behavior is never reached. If mastery criteria are met, advance — that is what the data are for.
No reversion plan. Without a defined reversion rule, regression requires a judgment call in the moment. Judgment calls under session pressure are inconsistent across clinicians and across sessions for the same clinician. If the learner fails three consecutive trials at the current step, the criterion reverts to the prior step — or whatever rule you specified. Write it down. Follow it.
When to Use Shaping Instead of Other Procedures
Shaping is the right choice when:
- The terminal behavior does not yet exist in the learner’s repertoire and cannot be prompted into existence
- You are building a new topography — a new physical form of a response, such as a vocalization, a fine motor movement, or a motor imitation
- You are working on fluency, moving a correct but slow response toward fast and automatic
Shaping is not the right choice when:
- The terminal behavior exists in the repertoire but is not under appropriate stimulus control — use discrimination training
- The terminal behavior exists but is not being maintained — check reinforcement schedules and motivating operations before redesigning the program
- The complexity is in the sequence of steps rather than in the form of the response itself — use a task analysis with a chaining procedure instead
How Kipr Helps
Shaping requires moment-to-moment clinical judgment: when to reinforce, when to withhold, when to advance the criterion, when to revert. That judgment is a skill, and like every ABA skill, it develops through reps.
Kipr lets BCBAs and behavior technicians practice shaping scenarios with AI-simulated client personas before applying them with real learners. You can attempt a program, make the classic errors — criterion jumps, frozen criteria, missing reversion rules — receive feedback, and try again. The practice happens with an AI persona, not with a client.
Clinical judgment in the session cannot be scripted into a protocol. But it can be built through deliberate practice.
Join the Kipr waitlist for early access.