Differential Reinforcement in ABA: How It Works

Differential reinforcement is a method that rewards one behavior more than others. The behavior you want gets the best outcome available. Other behaviors that serve the same purpose get much less.

That simple setup does two jobs at the same time. It builds up the skill you want to see more often. It also brings down behavior you do not want, without using punishment. This page covers the general principle. The named versions, like DRO and DRA, are specific procedures with their own rules.

What the procedure actually says#

Anlara Hoskins built a short course around one clean definition of this term. She borrows it from behavior analyst Tim Vollmer, and it is worth reading slowly.

differential reinforcement is providing the greater reinforcement along at least one dimension, contingent on the occurrence of one form or type of behavior while minimizing reinforcement for another form or type of behavior From the talk — Anlara Hoskins

Every piece of that sentence is doing work. "Form or type of behavior" is topography, which just means what the behavior looks like. "Contingent on" means the consequence only shows up after that behavior. A "dimension" is the part of the reward you can turn up or down. You can change how much a person gets, how good it is, how long it lasts, or how fast it arrives.

Hoskins is also strict about the word reinforcement itself. It only counts as reinforcement if the target behavior actually goes up later. If the behavior stays flat, you did not reinforce anything. You just handed something out. That check keeps teams honest about whether a plan is working.

Why the other behaviors fade on their own#

The drop in problem behavior is not the direct target. It is a side effect of the better option paying more.

Hoskins frames it as a choice between options that all lead to the same goal. A student who wants to leave a hard task might yell, might flop on the floor, or might ask for a break. Those behaviors belong to the same response class, which means a group of behaviors that get the same result. When only one of them earns the best outcome, the rest lose value.

Behavior then can decrease through differential reinforcement because the other options that are not receiving that winning consequence, that ideal outcome. From the talk — Anlara Hoskins

You can only pick one option at a time. So the option with the best payoff wins more often, and the others slowly drop off. Nothing bad has to happen to the person for that decline to occur. The gap between outcomes is the whole engine.

It is not punishment wearing a new name#

This is the point both Hoskins and Dr. Kaci Ellis push hardest, and for different audiences.

Hoskins draws a clean line between the two ideas. Punishment picks one behavior and adds something unpleasant to shrink it. Differential reinforcement picks a behavior and grows it.

So, remember, punishment is targeting a behavior for decrease by providing an aversive consequence. Differential reinforcement is going to be targeting a behavior for increase From the talk — Anlara Hoskins

Ellis hears the same confusion from school staff. In her school-based work, she lists differential reinforcement under responses that teach rather than punish. She says adults often arrive wanting a penalty, not a plan.

So we're not just punishing because a lot of people have that in their head. They just want kids to be punished. It's not a punishment. We need to be teaching students what it is that we want them to be doing. From the talk — Dr. Kaci Ellis

Ellis still leaves room for consequences, because behavior always has them. Her point is about the goal. The plan should end with the student knowing what to do instead.

Saying it in words a teacher can use#

A plan only works if the adult in the room can run it. Ellis spends real time on translation, because the term itself means nothing to most staff.

Differential reinforcement, I'm sure we all are familiar, but explaining to teachers in terms in language that they can understand is really important when we're doing differential reinforcement, because we need them to understand what it is exactly that we are asking them to do in the classroom. From the talk — Dr. Kaci Ellis

Her broader approach is to tie every strategy back to what the behavior gets the student, its function. That just means the reason the behavior happens, like getting attention or getting out of a task. When a teacher sees how a behavior meets a function, the plan stops feeling random. Ellis also ranks classroom responses from least intensive to most intensive, starting with simple error correction. Differential reinforcement sits in that same teaching family, not in the discipline family. Her full session, Practical Takeaways for School-Based Behavior Analysts, walks through how she runs these conversations with teachers and admin.

What it looks like at the dinner table#

Dr. Holly Gover uses differential reinforcement inside feeding work, where every step is small and stressful for the child.

Her setup started with three reinforcement levels, often described as red, yellow, and green. Each level lines up with a step in the shaping plan. Shaping means rewarding closer and closer tries until you reach the full skill. A child who comes to the table earns something. A child who touches the food earns more. A child who takes the bite earns the best outcome available.

It's shaping, it's differential reinforcement, and we are letting kids opt out at any time. From the talk — Dr. Holly Gover

The opt-out piece matters. Gover always lets the child leave the table. That keeps the plan from turning into forced compliance. The child's best move is to stay and try, not to escape. Gover also matches the presentation style to the child's language level, so the steps make sense to that specific kid.

The gap has to be big enough to notice#

A plan can be technically correct and still fail. If both options pay about the same, the person has no reason to switch.

Gover describes adjusting the size of the reward rather than the rules around it. With some preteens she worked with, full reinforcement meant two minutes of access to their preferred items. Partial reinforcement meant one minute. The contingency never changed. Only the duration and quality moved.

That is the practical version of Hoskins's phrase "along at least one dimension." You do not need a different reward for each level. You need the same reward at clearly different sizes. Gover also says the steps themselves should be easy for the child to name. If the child can tell the difference between the steps, and can tell the difference in the payoff, the plan will teach. If the child cannot tell, the plan is just noise.

Where experts disagree: how exact the plan must be#

The sources agree on the principle. They pull apart on how much the specific setup matters.

Gover treats the structure as loose on purpose. She notes that people have taken her three levels and run wild with them in creative ways. She is fine with that.

So long as you're using meaningful differential reinforcement, that should drive the bus. Reinforcement is driving the bus and it needs to be differential. From the talk — Dr. Holly Gover

Ellis leans the other way when the plan is headed to a classroom. She is handing the plan to a teacher with twenty other kids in the room, so she wants the ask spelled out in plain language, not left to clinical judgment on the fly.

Both instincts are defensible, and the difference is mostly about setting. A clinician running her own session can adjust on the fly. A teacher running a plan between math and lunch cannot. The safe read is to be flexible about the parameters and strict about how clearly you explain them.

Using the same principle on adults#

Brian Middleton applies differential reinforcement to a target most people never consider. He uses it on the field's own culture.

His concern is how professionals talk about people. Calling someone difficult or dramatic is a mentalistic explanation, meaning it blames a made-up inner trait instead of looking at the situation. Middleton says arguing that habit down does not work.

how do we stop that? Well, we don't stop it. We create differential reinforcement. We show people that, hey, this is better. This is a much more effective approach. From the talk — Brian Middleton

The move is the same one you would use with a client. Make the better option pay off more clearly, and the old option fades. In his case, the payoff is that a function-based view actually solves the problem, while a label never does. He ties this directly to how people get treated.

And, of course, this creates greater acceptance for individuals and shaping culture to be inclusive. From the talk — Brian Middleton

What the research says#

Research on this principle keeps circling one honest problem. Gains can slip when the setting changes.

One study looked at whether fading the background context helps. Behavior came back for almost every participant without fading. Only half came back among participants who got fading (Jackson & Kestner, 2026, Journal of the Experimental Analysis of Behavior). Teaching across a gradual change in setting seems to hold the gains better.

A second team tested a related fix by training in several contexts instead of one. Rotating through three settings cut the return of old behavior in both experiments. The trade-off was real, though. The problem behavior dropped more slowly at first (Podlesnik, Martinez-Perez, Montague, Ritchey, Lamperski, & Kuroda, 2026, Journal of the Experimental Analysis of Behavior).

Applied work shows the principle handles tough, real tasks. In one study, differential reinforcement plus a slow fade-in of the materials taught five autistic adults to cooperate with nasal swab tests. All five still cooperated at follow-up (Briere, Janetzke, Fleck, & Bourret, 2025, Journal of Applied Behavior Analysis).

Sometimes reinforcement alone is not enough. On a playground, a differential reinforcement plan stopped unsafe behavior for one of four children. The other three needed a brief time-out added. Peer interaction and self-reported enjoyment of recess did not suffer (Linton, Gomes, & Donaldson, 2025, Journal of Applied Behavior Analysis).

FAQ#

What is differential reinforcement in ABA? It is a procedure that gives the strongest reinforcement to one behavior while giving little or none to similar behaviors. The preferred behavior goes up. The others go down as a side effect.

Is differential reinforcement the same as punishment? No. Punishment adds something unpleasant to shrink a behavior. Differential reinforcement targets a behavior for increase, and the unwanted behaviors fade because they stop paying off.

How do I explain differential reinforcement to a teacher? Skip the term and describe the action. Tell the teacher exactly which behavior earns the good outcome. Explain what that outcome is, and what happens when the other behavior shows up instead.