Shaping across dimensions of behavior

Some behavior simply isn't in the learner's repertoire yet, so there is nothing to reinforce. Shaping solves that problem: you reinforce whatever the learner can already do that is closest to the goal, then gradually ask for a little more.

Why it matters

Shaping is how you build first words, longer attending, louder voices, faster responding, and smoother motor skills when prompting alone isn't enough. It is also how problem behavior gets worse by accident. A caregiver who gives in only when a tantrum gets loud enough is shaping louder tantrums without meaning to.

Using it well means picking a good first approximation, naming the dimension being shaped, deciding what to do when progress stalls, and telling shaping apart from chaining and stimulus fading.

Key ideas

• Shaping is differential reinforcement of successive approximations. Responses that meet the current criterion are reinforced; responses that don't meet it are placed on extinction.
• Extinction does useful work. When an earlier approximation stops producing reinforcement, responding becomes more variable. Some of those new variations come closer to the goal, and those are the ones you now reinforce.
• Start with something the learner already does. The first approximation should occur at least sometimes and resemble the terminal behavior on the dimension you care about.
• Shaping can change form or another dimension.
  • Across topographies: each step is a different form of the response, such as "b" → "ba" → "ball", or walking with support → a few independent steps → walking across the room.
  • Within a topography: the form stays the same while another dimension changes, such as rate, duration, latency, or magnitude (for example, force or volume).
• Step size is a judgment call. Steps that are too large produce extinction-like breakdown: less responding, more errors, sometimes emotional responding. Steps that are too small, or lingering too long at one step, can make an intermediate approximation hard to move past.
• Back up when needed. If performance falls apart after a criterion change, return to the last approximation the learner met reliably, then move forward in smaller steps.
• Reinforce immediately. Deliver the reinforcer right after a response that meets the criterion. A conditioned reinforcer, such as a click or a quick "Yes!", can mark the exact response.
• Shaping can be combined with other tools. Prompts or models can speed things up, as long as you plan to fade them.

Definitions

• Shaping: building a new behavior, or a new level of an existing behavior, by differentially reinforcing responses that come progressively closer to a terminal behavior.
• Successive approximations: the sequence of intermediate responses, each closer to the terminal behavior than the one before, that are reinforced in turn.
• Terminal behavior: the final form or level of the behavior that shaping is meant to produce, stated with a clear mastery criterion.
• Differential reinforcement: reinforcing responses that meet a criterion while withholding reinforcement for responses that do not.
• Response differentiation: the result of differential reinforcement, a new set of responses that share the features that were reinforced.

Examples from a session

In each case, define the terminal behavior first, write the planned approximations down, and keep data so that each criterion change is based on performance, not on the calendar.

Non-example: Diego's teacher reinforces every 1-minute stretch of sitting for a month and never raises the criterion. That is reinforcement of an existing level, not shaping, because the criterion never moves toward the goal.

Common mistakes / exam traps

• Confusing shaping with chaining. Shaping changes the requirement for a single response, either its form or a dimension such as duration. Chaining links separate responses into a sequence, like the steps of handwashing.
• Choosing a first step the learner never emits. A response that doesn't occur cannot be reinforced. Pick something already in the repertoire.
• Jumping too far. Moving from 2 minutes to 10 minutes of sitting in one step invites breakdown. Use smaller steps and data-based criteria.
• Forgetting that problem behavior can be shaped. Reinforcing only intense or long episodes can make them more intense or longer.

Quick check

Related topics

• Differential reinforcement (DRA, DRO, DRL, DRH) with and without extinction (G.2)
• Fading prompts: prompt delay and stimulus fading (G.8)
• Selecting and running chaining procedures (G.12)
• Operant vs. respondent extinction, as procedure and as process (B.11)
• Measuring time-based dimensions: duration, latency, and IRT (C.4)