Differential reinforcement is one of the most flexible tools in ABA. But DRO, DRA, DRI, and DRL each serve a different function — and choosing the wrong one for a target behavior can slow progress or create new problems. Here is how to match the procedure to the goal.
What Is Differential Reinforcement?
Differential reinforcement means reinforcing some behaviors (or the absence of a behavior) while withholding reinforcement for the target problem behavior. It is paired with extinction — the target behavior no longer produces reinforcement — to reduce challenging behavior while simultaneously building or maintaining appropriate ones.
There are four main procedures, each defined by what gets reinforced:
| Procedure | Full Name | What Is Reinforced |
|---|---|---|
| DRO | Differential Reinforcement of Other Behavior | Absence of the target behavior during an interval |
| DRA | Differential Reinforcement of Alternative Behavior | A functionally appropriate alternative behavior |
| DRI | Differential Reinforcement of Incompatible Behavior | A behavior physically incompatible with the problem behavior |
| DRL | Differential Reinforcement of Low Rates | A reduced frequency (not elimination) of the behavior |
The wrong choice wastes sessions and can frustrate both learner and practitioner. The right choice depends on three things: whether the function is known, whether safety is a concern, and whether the goal is elimination or reduction.
DRO: Reinforcing the Absence
When to use it: You want to decrease the frequency of a behavior, and you are less concerned about what the learner does instead — as long as the target behavior does not occur.
How it works: Set an interval. If the target behavior does not occur during that interval, deliver reinforcement. Reset the interval if it does occur.
Fixed-interval DRO (reinforce if no aggression for 2 minutes) is easier to implement consistently. Variable-interval DRO is more resistant to pattern learning — the learner cannot briefly suppress the behavior just before a predictable reinforcement moment and then return to baseline immediately after.
Common mistakes to avoid:
- Intervals set too long for the current baseline rate — the learner cannot access reinforcement, and motivation drops quickly
- Not pairing DRO with extinction — the problem behavior may still produce reinforcement from other sources, undermining the procedure
- Failing to systematically increase the interval as the behavior decreases
DRO is often the most straightforward differential reinforcement procedure to implement, which makes it appealing when session complexity is already high. But because it reinforces the absence of a behavior rather than a specific replacement, it does not teach a new skill on its own. If skill acquisition is a goal, layer DRO under a DRA program.
DRA: Building the Better Alternative
When to use it: The problem behavior has a clear function, and you want to teach a more appropriate behavior that serves the same function.
DRA is Functional Communication Training’s closest relative. When FBA data confirm the function, DRA teaches a behavior that meets the same need — a request for a break instead of elopement, a tap on the shoulder instead of grabbing, a picture exchange instead of property destruction.
Why it is often the right choice: DRA directly addresses the why behind the behavior. If the function is attention, you teach an appropriate attention-seeking behavior and reinforce it richly. The problem behavior undergoes extinction — it no longer works. Over time the learner has a reliable, socially appropriate way to meet the need.
Common mistakes to avoid:
- Using DRA without extinction — the problem behavior still works, so the learner has no reason to switch to the more effortful alternative
- Choosing an alternative that requires more effort than the problem behavior — behavior obeys an effort economy, and the path of least resistance wins
- Poor FBA data leading to a mismatched alternative that does not actually serve the same function
DRA produces the most durable behavior change of the four procedures, but it requires solid FBA data underneath it. If the function is unknown or multi-maintained, DRO while the FBA is being completed is the safer interim choice.
DRI: The Physical Incompatibility Approach
When to use it: You have a target behavior that physically cannot occur at the same time as the replacement behavior.
Examples: reinforcing hands flat on lap to reduce self-injurious hand movements; reinforcing walking to reduce running in high-risk settings; reinforcing sitting to address persistent out-of-seat behavior.
Why it is valuable: Incompatibility is a strong mechanism because the problem behavior literally cannot occur when the incompatible behavior is occurring. For behaviors carrying significant safety risk, DRI provides reliable suppression during the skill acquisition phase.
Common mistakes to avoid:
- Choosing an incompatible behavior the learner cannot yet perform fluently — you have added two challenges at once, and both suffer
- Failing to program generalization — an RBT may reinforce hands-in-lap in session reliably, but what is the incompatible behavior in the community, in the car, at school?
DRI works best as part of a larger package rather than in isolation. Pair it with antecedent modification and a DRA program when possible, so the learner acquires a skill that travels beyond the therapy room.
DRL: Reducing Rate Without Eliminating
When to use it: The behavior is acceptable at low rates but problematic at high rates. The goal is reduction, not elimination.
Classic examples: frequent verbal questioning, repetitive physical affection, topic perseveration. Some of these behaviors are socially appropriate in moderation. DRL targets the rate, not the behavior itself, which preserves appropriate instances while bringing the frequency into a functional range.
Two DRL variants:
- Full-session DRL: If the total behavior count stays at or below criterion for the session, deliver reinforcement at session end. Best for shorter sessions or behaviors easily measured in aggregate.
- Spaced-responding DRL (IRT-based): Reinforce only when a minimum inter-response time has elapsed since the last occurrence. Better for behaviors where temporal spacing matters — for example, re-asking the same question before the previous answer has been used.
Common mistakes to avoid:
- Criterion set too close to baseline — the learner does not experience early success, and motivation drops before any thinning occurs
- Not communicating the criterion clearly when appropriate — some learners benefit from an explicit behavioral contract paired with DRL, especially when verbal explanation is a strength
- Forgetting to continue thinning the criterion as rate decreases
Choosing the Right Procedure: A Quick Decision Framework
| Situation | Recommended Procedure |
|---|---|
| Function is known; an appropriate alternative exists | DRA |
| Behavior is a safety risk; an incompatible replacement is feasible | DRI |
| Rapid suppression is needed while FBA is still in progress | DRO |
| Behavior is socially appropriate at low rates but problematic when high | DRL |
| Multiple competing behaviors need addressing simultaneously | DRO + DRA combination |
These procedures are frequently combined. A DRA program usually includes a DRO interval to cover instances where the alternative behavior is not immediately selected. DRI often pairs with DRL when a behavior cannot be fully eliminated but needs both behavioral shaping and pacing. The goal is always a comprehensive package — not a single procedure applied in isolation.
How Kipr Helps
Selecting the right differential reinforcement procedure is one challenge. Implementing it with fidelity across settings, sessions, and team members is another. Most implementation errors come not from knowledge gaps but from a lack of reps — practitioners who have not had enough practice making the moment-to-moment decisions that DR procedures demand in real time.
Kipr lets BCBAs and RBTs practice differential reinforcement scenarios in AI-powered simulations before running them with real learners. That means the interval decisions, the extinction pairing, and the alternative behavior selection all get tested in a low-stakes environment where a mistake does not cost a session — and feedback is immediate.
Join the Kipr waitlist for early access.