How differential reinforcement works
Every differential reinforcement procedure has two parts. Reinforcement is delivered for the target behavior or condition, and it is withheld for the problem behavior, which is extinction: the problem behavior stops producing the consequence that was maintaining it.
The procedure works best when it is function-based, meaning the reinforcer delivered for the alternative is the same one the problem behavior was producing. A learner who hits to escape tasks will not give up hitting for praise; they might give it up for a break they can ask for. That is why differential reinforcement usually follows a functional behavior assessment, and why it is a core part of most behavior intervention plans.
The four main types
Each type answers a different version of the same question: what should earn the reinforcer now that the problem behavior does not?
| Type | What is reinforced | Example |
|---|---|---|
| DRA: differential reinforcement of alternative behavior | A specific appropriate behavior that serves the same function as the problem behavior. | A learner who grabs toys is taught to say "my turn" and gets the toy for asking, never for grabbing. |
| DRI: differential reinforcement of incompatible behavior | A behavior that cannot physically occur at the same time as the problem behavior. | A learner who flaps their hands during work earns reinforcement for hands on the materials, which cannot happen while flapping. |
| DRO: differential reinforcement of other behavior | The absence of the problem behavior for a set interval, or at a set moment. | Reinforcement after every 2 minutes with no hitting; a hit restarts the 2-minute timer. |
| DRL: differential reinforcement of low rates | The problem behavior occurring at or below a set rate, for behaviors that are acceptable at low rates. | Reinforcement when a learner asks no more than 4 questions during a 20-minute lesson. |
DRA and functional communication training
DRA is usually the first choice when the function of the problem behavior is known and an appropriate alternative can be taught. Its most common form is functional communication training, in which the alternative is a communicative response - a word, sign, picture card, or device message - that asks for the reinforcer the problem behavior used to produce.
The alternative has to be easier and more reliable than the problem behavior, especially at first. Early in training, every request is typically honored; once the request is established, it can be thinned to realistic schedules, for example by introducing short waits or by honoring requests only at certain times.
DRO variations
DRO delivers reinforcement for not doing something. In interval DRO, reinforcement follows an interval in which the behavior did not occur at all; in momentary DRO, it follows the absence of the behavior at the moment the interval ends. Intervals can be fixed or variable, and resetting or non-resetting: in a resetting DRO, an occurrence restarts the interval timer.
Set the starting interval short enough that the learner can earn reinforcement - typically shorter than the usual time between occurrences in baseline - and lengthen it gradually as the behavior decreases. DRO does not teach a replacement behavior, and it can accidentally reinforce whatever else the learner happened to be doing, including a different problem behavior, so it is often combined with teaching an alternative.
DRL variations
DRL suits behaviors that are fine in moderation and a problem only in excess: asking questions, requesting help, raising a hand. In full-session DRL, reinforcement follows a session in which the behavior occurred no more than a set number of times. In interval DRL, the session is divided into intervals and each interval at or below the limit earns reinforcement. In spaced-responding DRL, a response is reinforced only if a minimum time has passed since the previous one.
The counterpart, differential reinforcement of high rates (DRH), reinforces a behavior when it occurs at or above a set rate, and is used to increase behaviors that occur too slowly or too rarely.
Choosing a procedure
Start with function and topography. When the function is known and a replacement can be taught, DRA addresses the cause directly. When a behavior has an obvious physical opposite, DRI can work well, ideally with an incompatible behavior that is useful in itself. When there is no single replacement and the behavior should stop entirely, or occurs at a very high rate, DRO is often used. When the behavior is acceptable at a lower rate, DRL keeps it while reducing it.
In practice these are often combined, and combined with antecedent strategies that make the problem behavior less likely in the first place. Those choices, and how everyone involved will carry them out, belong in a written behavior intervention plan.
Measuring whether it is working
Measure both sides. The problem behavior is measured with whichever system fits it - frequency, duration, or interval recording - against a baseline taken before the procedure starts. The replacement behavior is a skill being taught, measured as an acquisition target. A procedure that reduces the problem behavior without the replacement growing is a warning sign that something else is maintaining the change.
Expect the possibility of an extinction burst: when a behavior stops producing its reinforcer, it often increases in frequency or intensity for a while before it decreases. A plan should say how staff respond if that happens, so the burst is not reinforced. Check implementation, too: differential reinforcement delivered inconsistently - a reinforcer occasionally given for the problem behavior - teaches the learner that persistence pays.
How Cadence measures differential reinforcement
In Cadence, a differential reinforcement program is usually two targets. The problem behavior is measured with frequency, duration, or interval recording, evaluated against an at-or-below mastery rule, and the replacement behavior is an acquisition target with its own criterion. Both are graphed automatically from committed session data with phase-change lines, so the session a procedure changed is visible on the graph.
On the Clinical and Practice plans, trend, variability, and level change are computed for each target and narrated in plain language, which helps a BCBA tell a real reduction from a temporary dip or an extinction burst.