Telling a learner "stop" rarely teaches them what to do instead. Differential reinforcement does. You reinforce one thing (a better behavior, the absence of the problem behavior, or a lower rate of it) and stop reinforcing the problem behavior, so it no longer pays off.
Why it matters
Differential reinforcement is one of the most common parts of a behavior plan, and you will often be the person who runs it. It reduces problem behavior without relying on punishment, and in most variations it builds a useful skill at the same time.
The skill is discrimination: reading a short session description, naming the variation (DRA, DRI, DRO, DRL, or FCT) from what earns the reinforcer, and spotting an RBT who reinforced the wrong behavior.
Key ideas
- Two parts, every time. Reinforce the target named in the plan, and withhold reinforcement for the problem behavior. Withholding the reinforcer that maintains a behavior is called extinction.
- Ask what earns the reinforcer. That one question tells the variations apart: a specific alternative behavior (DRA), a behavior that cannot happen at the same time (DRI), not doing the behavior for a period (DRO), or doing it less often (DRL).
- Match the function. Plans work best when the new behavior earns the same reinforcer the problem behavior used to get. If a learner hits to escape tasks, a break request that produces a short break competes directly with hitting.
- Start rich, then thin. At first the replacement behavior is reinforced every time, so it pays off better than the problem behavior. Your supervisor thins the schedule later as the new behavior gets strong.
- Watch for a burst. When problem behavior stops working, it may briefly get worse before it improves. Keep following the plan, record the data, and tell your supervisor.
Definitions
- Differential reinforcement: reinforcing one response class, or the absence of the problem behavior, while withholding reinforcement for the problem behavior.
- DRA (differential reinforcement of alternative behavior): reinforce a specific, more appropriate behavior instead of the problem behavior. The alternative does not have to rule out the problem behavior; a learner can raise a hand while also calling out.
- DRI (differential reinforcement of incompatible behavior): a kind of DRA in which the reinforced behavior cannot physically occur at the same time as the problem behavior, such as hands holding a toy instead of hands hitting.
- DRO (differential reinforcement of other behavior): deliver reinforcement when the problem behavior has not occurred for a set interval. In a resetting DRO, an occurrence restarts the timer. Also called omission training.
- DRL (differential reinforcement of low rates): reinforce the behavior only when it occurs at or below a set rate, or after enough time has passed since the last response. It suits behaviors that are fine in moderation but happen too often.
- FCT (functional communication training): a form of DRA that teaches a communication response, such as a word, sign, card, or device button, that produces the same reinforcer the problem behavior did.
Examples from a session
Five variations, five learners
- DRA (school): Leo, 10, calls out answers. His teacher and RBT call on him and praise him when he raises his hand, and calling out gets no response.
- DRI (home): Grace, 4, pulls her sister's hair while they watch TV together. Her RBT gives her a squishy toy to hold with both hands and praises her while she holds it.
- DRO (clinic): Omar, 8, spits. Every 3 minutes without spitting earns a token. If he spits, his RBT resets the timer.
- DRL (job site): Lena, 19, checks with her job coach about 30 times an hour. She earns praise and a preferred break at the end of any hour with 10 or fewer check-ins, because some checking in is useful.
- FCT (clinic): Mateo, 6, screams to get out of writing tasks. He learns to hand over a "break" card, and each time he does, his RBT gives a short break right away.
In every example the plan spells out exactly what earns the reinforcer, and the new behavior earns the same reinforcer the problem behavior used to get.
Non-example: Leo's teacher praises hand raising but still answers him whenever he calls out. Calling out keeps paying off, so this is not differential reinforcement; the second part, withholding reinforcement for the problem behavior, is missing.
Common mistakes / exam traps
- Reinforcing the problem behavior "just this once." Giving a break after screaming during FCT teaches that screaming still works, and it slows progress.
- Using DRL for dangerous behavior. DRL allows some responding, so it does not fit aggression or self-injury. Plans for those behaviors typically use DRO, DRA, or FCT.
- Confusing DRO with noncontingent reinforcement. In DRO the reinforcer depends on the problem behavior not happening; in NCR it arrives on a time schedule no matter what.
- Delivering a DRO reinforcer right after another problem behavior. DRO can accidentally strengthen whatever the learner is doing when the interval ends. Note it and tell your supervisor.
- Thinning the schedule on your own. If a learner requests a break every minute, keep following the plan and share the data. Your supervisor decides when and how to thin.
Exam trap
To name the variation, ignore the problem behavior for a moment and ask what earns the reinforcer: a different behavior (DRA), a behavior that cannot happen at the same time (DRI), time without the behavior (DRO), or a lower rate of the same behavior (DRL).
Exam trap
DRA or DRI? Both reinforce a replacement behavior. The discriminating feature is whether the two behaviors can happen at the same time: if a learner could raise a hand while still calling out, it is DRA; if hands holding a toy cannot also pull hair, it is DRI.
Quick check
Kai, 9, yells to get his teacher's attention. His plan uses DRA: when he taps his teacher's arm and says, “Excuse me,” he gets attention right away. What should happen when Kai yells?