Chaining, Shaping and Token Systems

Some skills are too big to learn all at once. Others don't exist yet in any finished form. And some behavior needs a way to earn rewards that doesn't stop the activity. This lesson covers three procedures for these problems:

• Chaining teaches a task made of several separate actions.
• Shaping gradually changes one response until it reaches a goal.
• Token systems let a learner earn small items now and trade them later.

As an RBT, you carry out these procedures as your supervisor writes them. You don't redesign them.

By the end, you should be able to:

• Write or improve a task analysis so each step is one action an observer can see.
• Explain how forward, backward and total-task chaining differ, and read their data correctly.
• Tell shaping apart from chaining, and record which approximation actually occurred.
• Run a token system as written: deliver tokens, track the balance, run exchanges and report problems.

Before you start: A reinforcer is something delivered after a behavior that makes that behavior more likely in the future. It is defined by its effect. Something the learner likes is not automatically a reinforcer.

Break a task into steps you can see

A task analysis is a list of the small steps that make up a task. In plain terms, you write the task the way you would watch it: one action at a time.

Two rules make a step usable:

• One discrete action. The step has a clear start and end, and an observer can see whether it happened.
• A manageable size for this learner. A step that is easy for one learner may hide several actions for another.

Suppose a first draft for filling a watering can says: "Fill the can at the sink up to the marked line." That sounds like one step. But an observer would actually see several actions: putting the can under the tap, turning the tap on, and turning it off at the right moment. If the learner got stuck, you couldn't tell which part caused the problem.

Here is the same task broken into discrete steps:

1. Place the can in the sink under the tap.
2. Turn the tap on.
3. Turn the tap off when the water reaches the marked line.
4. Lift the filled can out of the sink.

Each step is one action you can see. Step 3 includes its end point, so an observer can judge it: was the water at the line when the tap went off? "Understands how to fill the can" would not be a step, because you can't see understanding. You can only see actions.

Change one fact. For a learner who already uses taps easily, steps 2 and 3 might not need to be separate. Step size depends on the learner, and your supervisor decides the final version. Your part is to follow the steps as written and to report a step that seems to contain more than one action or is too hard for the learner.

Three ways to teach the same chain

A chain is a task where separate actions happen in a set order. The three chaining methods never change that order. The learner always fills the can from step 1 to step 4. What changes is which steps the learner is responsible for doing on their own. You assist with the rest.

• Forward chaining: The learner starts by practicing step 1. You help with steps 2–4. Once step 1 meets your supervisor's mastery criterion, step 2 is added. The learner now does steps 1 and 2 on their own, and you help with steps 3–4.
• Backward chaining: The learner starts by practicing step 4, so they finish the task themselves. You help with steps 1–3. Once step 4 is mastered, step 3 is added. The learner now does steps 3 and 4, and you help with steps 1–2. "Taught in reverse order" describes the order in which steps are added. It does not mean the task is performed backward.
• Total task: The learner attempts every step each time. You assist only where needed.

In forward and backward chaining, mastered steps don't drop out. The learner keeps doing them independently, so the independent part of the task grows with each added step.

Here is one session under each method, at the first teaching stage. I means independent and A means assisted.

Read the data carefully

In both forward and backward chaining, the learner was independent on 1 of 4 steps (1 ÷ 4 = 0.25, or 25%). But the three assisted steps were planned, not failures. "Independent on the step being taught" and "independent on 25% of the whole task" answer different questions. When you report a number, name its denominator: what you counted out of what.

Write what you saw, too. "Turned the tap off at the line without help" is an observation. "Seemed bored" is an interpretation.

If the session ended before step 4, record step 4 as not observed, using whatever code your data sheet provides. Don't record it as assisted or as zero.

Try this: under backward chaining, a coworker says, "So the learner lifts the can out first, then turns the tap on." What's wrong?

Answer: Backward chaining keeps the real order

The task still runs from step 1 to step 4. You assist with steps 1–3 in order, and the learner then does step 4, lifting the can out. "Backward" refers only to which step is taught first: the last one.

Shaping: one response, closer and closer

Shaping means reinforcing successive approximations. These are versions of a response that come closer and closer to a goal response, called the terminal behavior.

Suppose the goal is for a learner to say "ball" to ask for a ball during play. The learner currently says "b." A possible sequence is "b," then "ba," then "baw," then "ball." Your supervisor decides three things:

• which approximation currently earns reinforcement,
• when to move to the next one, and
• what mastery means.

There is no universal count of correct responses that applies to every program.

How shaping differs from chaining. Shaping changes one response. Chaining links separate actions. Change one fact in the example: if the goal became "pick up the ball, walk to the bin, drop it in," that is three separate actions. That calls for chaining, not shaping.

Your role:

• Reinforce the approximation the plan names.
• Record exactly what you heard. Write "said 'ba,'" not "tried to say ball."
• Report if progress stalls. The step size may be too large, or the reinforcer may not be working. Your supervisor decides what to change.

Token systems: items that buy something

A token economy has five parts:

1. Target behavior: An observable action that earns a token.
2. Tokens: Stickers, chips or checkmarks. A token is a conditioned reinforcer: in a token system, its value comes from what it can be exchanged for. A learner might already like an item used as a token, such as a sticker, but the system works because tokens are linked to things the learner can buy with them. A token counts as a reinforcer only if the target behavior actually increases.
3. Earned-token record: Where the balance is tracked, sometimes called the "bank."
4. Backup reinforcers: The items or activities tokens buy.
5. Exchange rate and exchange opportunity: How many tokens each item costs, and when and where trading happens.

Example plan:

• The learner earns 1 token for each finished worksheet placed in the tray.
• 4 tokens buys 5 minutes of a tablet game.
• Exchanges happen at 10:30 and 12:00.
• Leftover tokens carry over to the next exchange.

When you deliver a token, say why: "You put your worksheet in the tray, here's a token."

The learner starts with 0 tokens and earns 6 by 10:30. They buy one game: 6 − 4 = 2 tokens left, which carry over to 12:00. Whether leftovers carry over is a plan decision. If the plan doesn't say, ask your supervisor.

Keep time anchors straight. "Earned 6 tokens by 10:30" and "earns 6 tokens per hour" describe different things. Don't swap one for the other.

Follow the plan exactly:

• Don't change prices or the menu on your own.
• Don't put basic needs, such as water, meals and bathroom access, on the menu or make a learner earn them. Treat this as a safety boundary: a token plan should never be the reason a learner goes without them. If a plan seems to do this, raise it with your supervisor right away.
• Don't take tokens away unless the documented plan includes this and your supervisor has verified your competence. Taking away earned tokens as a consequence for a behavior, in order to reduce that behavior, is a punishment procedure called response cost. Spending tokens at an exchange is not response cost. Neither is fixing a counting mistake by following your workplace's data-correction procedure.

Report to your supervisor:

• The behavior isn't increasing.
• The learner ignores the tokens.
• The plan is unclear, for example about what counts as "finished."
• You missed giving a token. Record it honestly: "token not delivered" is not the same as "not earned."

Practice on your own

1. A coworker drafts this task analysis for washing hands:

1. Wash hands with soap.
2. Dry hands properly.

Rewrite it so each step is one action an observer can see. Then explain what was wrong with each original step.

Answer: Practice 1

"Wash hands with soap" hides several actions, so you couldn't tell which part a learner missed. "Dry hands properly" can't be judged, because "properly" isn't something you can see. One possible rewrite:

1. Turn the tap on.
2. Put both hands under the water.
3. Press the soap pump once.
4. Rub hands together until both are covered in soap.
5. Hold hands under the water until no soap is visible.
6. Turn the tap off.
7. Pull one paper towel from the dispenser.
8. Wipe both hands with the towel until they no longer drip.
9. Drop the towel in the bin.

Your list may be split differently. Check each step against the two rules: is it one action with a visible start and end, and is it a manageable size for the learner? Your supervisor approves the final version.

2. The RBT helps with steps 1–3 of a four-step task, and the learner completes step 4 alone. Once step 4 is mastered, step 3 is added. Which chaining method is this?

Answer: Practice 2

This is backward chaining. The learner performs the final step, and earlier steps are added one at a time. The task is still done in its normal order.

3. In a total-task session of a four-step task, the data read: step 1 I, step 2 A, step 3 I. The session ended before step 4. A coworker writes "2 of 4 independent, 50%." What's wrong with this?

Answer: Practice 3

Step 4 was not observed. Counting it in the denominator treats it as a failure. The accurate summary is: 2 of 3 observed steps were independent, and step 4 was not observed.

4. The goal is a clear "cup." For three sessions, the learner has stayed at "kuh," the current approximation, and has not produced a closer version. What should you do?

Answer: Practice 4

Record exactly what you heard and report the stall to your supervisor. Don't skip ahead or set a new criterion yourself. The step may be too large or the reinforcer may not be working, and those are supervisor decisions.

5. Classify each goal as shaping or chaining, and say why.

• (a) The learner currently says "m." The goal is a clear "more."
• (b) The goal is putting on a coat: put one arm in a sleeve, put the other arm in, then pull the zipper up.

Answer: Practice 5

(a) is shaping. There is one response, and each version comes closer to "more." (b) is chaining. It links three separate actions that happen in a set order. The tempting mistake is calling (a) chaining because it moves through stages. Stages of one response that gets closer to a goal are shaping. A series of different actions is a chain.

6. Here is a token plan:

• 1 token per page read.
• The menu: a sticker costs 3 tokens, 5 minutes of drawing costs 5 tokens.
• Leftover tokens carry over to the next session.

The learner starts with 2 tokens and earns 9. At the exchange, they buy drawing and one sticker. What is their ending balance?

Answer: Practice 6

First add what they had to what they earned: 2 + 9 = 11 tokens. Then subtract what they spent: 11 − 5 − 3 = 3 tokens, which carry over to the next session.

7. A learner shouts during work. A coworker says, "Take two tokens back." The plan says nothing about removing tokens. What do you do?

Answer: Practice 7

Don't remove any tokens. Taking tokens away because of the shouting would be response cost, a punishment procedure. It requires a documented plan and supervisor-verified competence. Follow the plan as written and raise the concern with your supervisor.

What to remember

• Task analysis: Each step is one observable action with a clear start and end, sized for the learner. Your supervisor sets the final steps; report a step that hides several actions.
• Chaining: The real task order never changes. Forward, backward and total-task chaining differ in which steps the learner is responsible for independently. In forward and backward chaining, mastered steps stay independent as new steps are added.
• Data: Name your denominator, separate observations from interpretations, and record missing steps as not observed.
• Shaping: One response is reinforced as it gets closer to a goal. A series of separate actions is a chain instead. The supervisor sets the criteria, and you record and report.
• Token systems: In a token system, tokens get their value from backup reinforcers. Follow the exchange rate and timing in the plan. Don't remove tokens as a consequence unless the documented plan includes response cost and your competence is verified. Never make basic needs something to earn. Report anything unclear or ineffective.

Sources

• Sam, A., & AFIRM Team. (2024). Task Analysis, updated. The University of North Carolina at Chapel Hill, Frank Porter Graham Child Development Institute, Autism Focused Intervention Resources and Modules. https://afirm.fpg.unc.edu/wp-content/uploads/Task-Analysis-Brief-Packet-Sam-AFIRM-Team-Updated-2024.pdf
• Sam, A., & AFIRM Team. (2024). Reinforcement, updated. The University of North Carolina at Chapel Hill, Frank Porter Graham Child Development Institute, Autism Focused Intervention Resources and Modules. https://afirm.fpg.unc.edu/wp-content/uploads/Reinforcement-Brief-Packet-Sam-AFIRM-Team-Updated-2024.pdf
• Silva, F. J. (1999). Beyond the "hot-and-cold" game: A demonstration of computer-controlled shaping. PubMed record 10495835. https://pubmed.ncbi.nlm.nih.gov/10495835/
• University of North Carolina at Charlotte. Guidelines for the Effective Application of Shaping (faculty course webpage, "Application Exercises"). Accessed 2026-11-04. https://webpages.charlotte.edu/gcampbe1/5270/APPLICATION%20EXERCISES.htm
• Behavior Analyst Certification Board. (2023). RBT test content outline (3rd ed.). https://www.bacb.com/wp-content/rbt-outline-3rdEd/
• Behavior Analyst Certification Board. (2021). RBT ethics code (2.0). https://www.bacb.com/wp-content/rbt-ethics-code

Independent study resource

This study resource is not affiliated with, sponsored by, or endorsed by the Behavior Analyst Certification Board (BACB). BACB, Registered Behavior Technician and RBT are registered trademarks of the Behavior Analyst Certification Board. Source citations do not imply endorsement by their authors or publishers.