Extinction Burst: Why Behavior Spikes Before It Drops
An extinction burst is a short spike in behavior when reinforcement stops. Learn why it happens, when it is risky, and how BCBAs plan for it.
Key takeaway
An extinction burst is a short spike in a behavior right after the reward for that behavior stops. For a little while, the behavior gets more frequent or more intense before it fades.

Prediction and Probabilities: Three foundational equations to successful behavior reduction
Free learning path on this topic
Master PFA and SBT
Matt Harrington's 12-article walk through practical functional assessment and skill-based treatment, from Hanley's 2014 study to the enhanced choice model. Then assent, since SBT depends on it, and the ethical guardrails that keep behavior reduction defensible. Five CEUs, one of them ethics.
Experts who cover this
5 recordings on openceu cover extinction burst.
- Featured talk2 Learning CEUs
Prediction and Probabilities: Three foundational equations to successful behavior reduction
Matt HarringtonThe Behaviorist Bookclub
Extinction bursts are framed as the main risk of extinction-based procedures, including DRO, and are the reason Matt reaches for NCR instead when a client's…
- 0.5 Learning CEUs
Rad N Bad: The Forbidden E-Word — Extinction, Its Side Effects, and Compassionate Implementation
Sean Yocum, LBA & Michael CarreroHickory Learning Group
The speakers explain that behavior typically worsens before it improves once reinforcement is withheld, describing this temporary escalation as the extinction…
- 1 Learning CEU
Collaboration and Building Rapport: Starting a Strong Foundation With Caregivers and Stakeholders in the School or Community
Eric SoltysikThe Behaviorist Bookclub
Used to explain how a teacher's attempt at planned ignoring/extinction can backfire if problem behavior is reinforced intermittently once the teacher 'hits her…
- 0.5 Learning CEUs
Rad N Bad: Your Error Correction Sucks — Error-Correction Assessments, Trials to Mastery, and Learner-Informed Procedures
Sean Yocum, LBA & Michael CarreroHickory Learning Group
An increase in effort or problem behavior under a poorly-matched error correction procedure is compared to an extinction burst — you may eventually get there…
- 2.5 Learning CEUs
5 Days of Manding Mastery
Matt HarringtonThe Behaviorist Bookclub
Cited from the Landa prompting-timing study: delaying the prompt after problem behavior (a changeover delay) was more likely to produce an extinction burst…
On this page · 8 sections▾
An extinction burst is a short spike in a behavior right after the reward for that behavior stops. For a little while, the behavior gets more frequent or more intense before it fades.
That spike is not just an odd side effect. It shapes which tools a clinician picks for behavior reduction. Matt Harrington and Dr. Emily Ice both treat the burst as a real risk to plan for. Their talks show how the setting, the family, and the type of behavior all change the plan.
What an extinction burst looks like#
Picture a child who used to get picked up every time they cried at bedtime. Then the crying stops getting a response. The crying does not fade right away. For a short time, it may get louder or happen more often. That jump is the burst.
Matt Harrington describes it simply. When a burst happens, the behavior "spikes in intensity and frequency." Intensity means how strong the behavior is. Frequency means how often it happens. A burst can push up one or both.
The key word is short. A burst is a temporary rise, not a new normal. Researchers describe it as a brief increase in response rate early in the switch to extinction. Extinction here just means the old reward no longer follows the behavior.
Not every person shows a burst, either. Some clients go straight to a slow decline. Others show a sharp spike first. That is part of what makes the burst hard to predict. A clinician can't assume it will happen, but they also can't assume it won't. So the safest plan treats the burst as a real possibility from day one.
Why bursts hide inside other procedures#
Many people link bursts only to plain extinction. Matt Harrington points out that the risk can hide inside other plans too. His example is DRO. DRO stands for differential reinforcement of other behavior. In plain terms, the client earns a reward when the target behavior does not happen for a set stretch of time.
That sounds gentle. But Harrington notes there is a catch. If the behavior does happen, the client usually misses the reward for that interval. That missed reward works like extinction. So a DRO can trigger the same spike.
one worry that I always have about DRO is that there's a lot of extinction within DRO, because if a behavior occurs, then oftentimes that client's not going to access reinforcement. And thus, they might hit an extinction contingency that then might lead to an extinction burst.
From the talk — Matt Harrington
An extinction contingency is the rule that a behavior no longer earns its reward. Harrington's point is that this rule is baked into DRO. The procedure is not "extinction free" just because it adds rewards for other behavior.
He also shares a rule of thumb for setting a DRO interval. Start at about half the inter-response time. That is the average gap between one behavior and the next. If self-injury happens about every two minutes, the first interval might be one minute. A shorter interval gives the client more chances to earn reinforcement.
When the setting can't handle a burst#
The real question is not only "will a burst happen?" It is also "what happens if it does?" Harrington frames this as a question about the environment around the client.
For some behaviors, a short spike is stressful but safe. For others, it is dangerous. Harrington's example is head banging, or self-injury in general. Self-injury means behavior that causes harm to the person's own body.
If an extinction burst happens, and that behavior spikes in intensity and frequency, we could be looking at a trip to the hospital. And thus, we don't want to risk an extinction burst, so then we use NCR instead.
From the talk — Matt Harrington
NCR stands for noncontingent reinforcement. In plain terms, the client gets the reward on a time schedule, no matter what they are doing. Because the reward keeps coming, there is no sudden loss to push a spike. Harrington says NCR gives a rich reinforcement environment. DRO only does that if things go well.
He sums up the choice in one line. NCR is great "when the environment around you can't sustain an extinction burst." That framing is useful. It turns the burst from a vague worry into a concrete check. Can this child, this staff team, and this room safely ride out a spike? If not, pick a tool that avoids one.
Harrington also notes the decision depends on the client. There are times he would choose DRO over NCR. The burst risk is one factor, weighed with the rest.
Bursts and the timing of prompts#
Bursts also show up in teaching, not just in behavior plans. In his manding series, Harrington links bursts to prompt timing. A mand is a request, like a child asking for juice. A prompt is the help a teacher gives so the child can respond the right way.
He reviews a study that compared when to prompt after problem behavior. Prompting right after problem behavior did not raise problem behavior for eight of eleven participants. In fact, it was tied to a decrease. Prompting after the reward was removed also looked like a solid option.
The delayed approach was different. Waiting to prompt after problem behavior sometimes led to a burst.
We want, wherever we prompt, whether it be following, before, after, wherever, we want it to cut out the challenging behavior and not result in an extinction burst, which the delaying method sometimes did result in an extinction burst.
From the talk — Matt Harrington
His practical takeaway is about how long to wait before prompting. With more severe behavior, he suggests starting with more immediate prompts. Then fade the help slowly over time. This is the idea behind a progressive prompt delay. The wait grows a little at a time instead of all at once. The full walkthrough lives in 5 Days of Manding Mastery.
Bursts at bedtime#
Dr. Emily Ice brings the burst into a setting many families know well: the night. Her talk on sleep shows why a burst can be too much for a household to carry.
Sleep problems are complex, she explains. Some causes are medical, like sleep apnea or restless leg syndrome. Behavior analysts can't fix those. So she urges a medical consult to look at the biological side.
Then she turns to treatment. Her message is to prescribe with empathy. A clinician who recommends extinction for sleep won't be in the home at 2 a.m. The parents will be.
don't push for extinction because you're not going to be there to walk through the extinction burst at nighttime when everybody's already tired and it's hard enough.
From the talk — Dr. Emily Ice
Dr. Ice adds that extinction is probably not the "end all be all answer" for sleep. In her view, a clinician often doesn't need it at all. She stresses having the knowledge, experience, and mentorship to build a protocol that fits the family. She also asks clinicians to talk with caregivers about any history of trauma. Sleep comes when a person feels safe, she says, and plans should respect that.
Her point lines up with Harrington's. Both ask the same question from different angles. Who has to live through the burst, and can they do it safely?
Planning ahead for a burst#
Put the two talks together, and a clear planning habit appears. Before using any procedure with extinction inside it, ask a few questions.
First, how dangerous is the behavior if it spikes? Harrington's head banging example shows the stakes. If a spike could mean a hospital visit, a burst-free option like NCR may be wiser.
Second, who will be there when it happens? Dr. Ice's bedtime example shows that tired parents carry the burst alone at night. A plan they can't follow is not a good plan.
Third, does the procedure hide extinction? DRO and delayed prompting both can. Harrington's examples show that the label on a procedure doesn't tell the whole story.
Fourth, is there a richer path? Harrington favors NCR's rich reinforcement for high-risk cases. For teaching, he favors immediate prompts that fade slowly. Dr. Ice favors gentler sleep protocols built with the family.
None of this means bursts are always a reason to avoid a plan. It means the burst should be part of the plan. Caregivers should know it might happen, what it may look like, and what to do.
What the research says#
Bursts are common enough to plan for. One clinical record review looked at 108 clients in treatment (Muething et al., 2024). About 24% showed an extinction burst. That rate matched earlier studies. Burst size shrank across sessions once extinction was in place. The authors urge clinicians to prepare for short bursts during extinction-based treatment.
Why do bursts happen at all? One theory says reinforcers give access to a valued activity (Shahan, 2022). That activity competes with the target behavior for time. When the reward stops, that competition goes away. The target behavior can rise for a moment. Then it drops as extinction lowers its value.
This theory also explains a key clinical finding. Bursts may shrink when another source of reward is available. A rat study tested this directly (Shahan & Avellaneda, 2025). With no alternative reward, a strong burst appeared. With an alternative reward, bursts were smaller and less common. A larger alternative reward cut them even more.
A wider review pulled basic and applied research together (Fisher et al., 2023). It notes that past estimates of how often bursts occur vary a lot. It then uses a model called the temporally weighted matching law. That model offers ideas clinicians could use to lessen or prevent bursts.
FAQ#
Is an extinction burst a sign that treatment is failing? Not by itself. A burst is a short spike that often comes before the behavior drops. Still, plan for it, because the spike can be risky for some behaviors.
Can a burst happen without plain extinction? Yes. Matt Harrington notes that DRO contains extinction, since a missed interval means no reward. Delayed prompting during teaching can also produce a burst.
How can clinicians reduce the chance of a burst? Research suggests offering another source of reward can make bursts smaller and less common. For high-risk behavior, Harrington often reaches for NCR, which keeps rewards coming on a schedule.
Turn this topic into a CEU
You just studied this. Now get credit for it.
Watch Prediction and Probabilities: Three foundational equations to successful behavior reduction with Matt Harrington and earn 2 free BCBA CEUs. Detailed certificate, delivered the moment you finish.
Keep going with a free learning path
Master PFA and SBT
Matt Harrington's 12-article walk through practical functional assessment and skill-based treatment, from Hanley's 2014 study to the enhanced choice model. Then assent, since SBT depends on it, and the ethical guardrails that keep behavior reduction defensible. Five CEUs, one of them ethics.
Want the primary literature? Read the Extinction Burst research roundup on our sister site, Behaviorist Book Club.

