What Is an Extinction Burst? Evidence and Examples

An extinction burst is a temporary increase in the frequency, duration or intensity of a previously reinforced response after its maintaining reinforcement is withdrawn. It can occur during extinction, but it is not inevitable. An observed rise does not prove either the function of a response or that extinction is under way.

There is no universal timetable. The evidence does not support a fixed three-to-five-day rule for schools.

Fisher et al. (2023) describe bursting as a possible event whose reported rate depends on how it is defined and measured. The often quoted figures come from Lerman, Iwata and Wallace (1999). They analysed 41 clinical data sets from people with moderate to profound intellectual disability who had been referred for treatment of self-injurious behaviour.

Of 21 data sets in which extinction was used alone, 13 met that study's burst rule, or 62%. These numbers are not a rate for classrooms, and the first three sessions in the rule are not three school days.

‌

What Is an Extinction Burst?

Operant extinction occurs when a response no longer produces the result that had reinforced it. The definition rests on the link between that response and that result. It is not defined by an adult becoming silent or seeming to ignore someone. Researchers must first show that the result had maintained the response (Fisher et al., 2023; Lloveras et al., 2023).

An extinction burst is one possible change near the start of extinction. A study may look for a rise in rate, duration, intensity or another measured feature. Some studies also report new forms of response or aggression. These are related risks, but they do not form one fixed pattern.

The idea belongs to operant conditioning, but this article has a narrower task. It explains what the burst label means and what it cannot prove. The word burst can make the event sound clear-cut. In fact, a response can vary before and after any change in its results.

Researchers choose a baseline, a feature to measure and a period to study. Those choices affect whether a change meets their burst rule. Fisher et al. (2023), Nist and Shahan (2021), and Katz and Lattal (2021) warn against calling the event universal or fully settled.

Extinction is a procedure or learning process. A burst is one possible early pattern. Escalation on its own is only an observation.

‌

How Long Does an Extinction Burst Last?

No fixed duration applies across people, responses and study designs. Research may use trials, sessions or other time windows. These units are not the same as school days. A session may last minutes or much longer, and several sessions may occur in one day.

Nist and Shahan (2021) used rat experiments and a new analysis of the data. The measured rate of bursting changed when they changed the time window. It also changed when time spent consuming a reinforcer formed part of the baseline. Katz and Lattal (2021) showed why the baseline, transition point and response measure also matter.

These studies help to explain measurement. They do not give a classroom countdown. Past reinforcement and its schedule can affect how long a response lasts when its result stops.

Even so, there is no sound rule that one uneven adult response creates a permanent effect. The schedule, history and setting all matter.

A useful report must state what response was measured and what had reinforced it. It must also state what changed, how often observations took place and which rule was used. Without those facts, a duration claim cannot be judged.

In a school, a calendar cannot show that staff have found the right cause. Concern, change or rising risk calls for assessment and review, not a generic timetable.

‌

What the Evidence Shows

The evidence comes from several kinds of work. Laboratory studies test behaviour under tightly controlled schedules. Applied single-case studies test a plan with one person at a time, often in a clinical service.

Reviews compare terms and reported rates. Guidance sets standards for safe assessment and care. None of these is a large trial of ordinary classroom behaviour.

Lerman, Iwata and Wallace (1999) supply the best-known applied result. Their 41 clinical data sets concerned self-injurious behaviour in people with moderate to profound intellectual disability. In 21 cases, extinction was used on its own.

Thirteen met the study's burst rule. That rule compared one of the first three treatment sessions with the final five baseline sessions.

The result is often given as 62 per cent. Its base is the 21 extinction-alone data sets, not all 41 cases. It is not a rate for all people, learners or classroom conduct.

The three-session window is also not evidence of a peak on day three. Population, target response, setting and study rule limit what the result can mean.

The same review found fewer bursts and less aggression when extinction formed part of a treatment package (Lerman et al., 1999). This was a link in past clinical records, not proof of one safer package. It does not permit a school recipe.

Any alternative response must meet the assessed need, be easy to use and produce a useful result. Risk, skill, consent, resources and close review also matter.

Later reviews explain why estimates differ. They point to the burst rule, unit of study, baseline method, schedule, response measure and treatment design (Fisher et al., 2023). Basic work shows that small choices in measurement can alter the result (Nist & Shahan, 2021).

The evidence supports possible bursting in some settings. It does not support a stable rate for schools.

Read the study note transcript

‌

Extinction Bursts: Evidence and Boundaries

Core idea

A possible temporary rise near the start of operant extinction. Not inevitable. Not the same as ignoring a person.

One clinical study

Population: moderate to profound intellectual disability; treatment referral for self-injurious behaviour.

Evidence: 41 clinical data sets; 13 of 21 extinction-alone data sets met the burst criterion.

Boundary: not a classroom prevalence estimate.

No fixed timetable

Detection depends on definitions, measures, baseline and time window. Sessions are not days.

Distinct, not a sequence

• Extinction: contingency changes.
• Bursting: possible early increase.
• Relapse effects: possible recurrence.

In schools

Escalation proves neither extinction nor function. Safety, safeguarding, SEND support and reasonable adjustments come first. Complex or risky cases need assessment by appropriately qualified professionals.

Memory line: Possible, not proof.

Sources: Lerman, Iwata and Wallace (1999); Fisher et al. (2023); NICE NG11 (2015).

On a smaller screen, swipe across to read all columns.

‌

Extinction, Bursting and Related Effects

In Pavlovian studies, extinction does not appear to erase the first learning. New learning forms when conditions change, and the old response can return. Bouton (2004) reviews Pavlovian relapse effects such as spontaneous recovery and renewal. These are not stages in one fixed sequence.

An extinction burst is a possible rise near the start of extinction. Spontaneous recovery is the return of a response after time has passed. Renewal is a return after the context changes.

Resurgence is the return of an older response when a newer option stops producing a useful result. Each term marks a different link between past learning and current conditions.

This distinction avoids a common error. A response seen after a weekend or holiday cannot be called spontaneous recovery from timing alone. Researchers would need to know the learning history and present conditions. Laboratory evidence for relapse does not prove a set school pattern or a fixed length.

Classical conditioning also uses the word extinction, but its design differs. Readers can compare extinction in Pavlovian conditioning. This article deals with results that follow operant responses.

Its history connects with B. F. Skinner's operant account, Thorndike's law of effect and John B. Watson's early behaviourism.

Direct results are not a full account of learning. Bandura's social learning theory studies models and observation. Wider social cognitive theories include two-way links among thought, action and setting.

These theories do not refute operant findings. They show why one account of results should not explain all classroom behaviour.

‌

A Controlled Example

Consider a laboratory device with a button and a light. Earlier tests show that the light keeps button pressing going. During the baseline phase, each press turns on the light.

The press is the measured response, and the light is the result that maintains it. During extinction, pressing no longer turns on the light.

Suppose pressing briefly rises above the stated baseline measure after this change. It then falls. Under a clear study rule, the early rise may count as a burst.

The example is controlled because the response, result, schedule, baseline and time window are known. It needs no story about what the subject thinks.

A similar graph would not prove extinction if pressing rose for another cause. The button may have changed. An instruction may be unclear.

Another result may now follow a press. A sound label needs knowledge of the full design, not just the shape of a line.

This example should not be copied onto a learner. Movement, refusal, silence, distress and requests occur within social, health, sensory and learning settings. An adult cannot control those conditions by changing one response. A school observation should lead to careful inquiry, not an attempt to copy the laboratory procedure.

‌

Why Escalation Is Not Diagnostic

Escalation can occur during extinction, but it has many possible causes. Pain, illness, anxiety, trauma, sensory overload, unclear speech, hard work, peer events and changes in the setting can look alike. A new result may also follow the response. What adults see cannot identify which account is right.

A functional assessment is more than an adult's view of what a learner gains or avoids. NICE guidance asks teams to study triggers, settings and results. It also asks them to consider speech, physical and mental health, trauma, sensory needs and the wider setting. Direct observation should support the work, and the depth of study should match the risk (National Institute for Health and Care Excellence, 2013, 2015).

The learner's view matters, as do facts from family members and relevant staff. Information from more than one setting can help. Teams should gather it together and should not assume that a parent, peer or colleague caused a problem.

A sound idea about function is still a hypothesis, not a fact. It needs careful testing and review within ethical and safety limits. A change in behaviour shows that some condition changed, but not which one. Routes for health, SEND, mental health and safeguarding remain important in their own right.

The word ignore adds risk. Lloveras et al. (2023) list six concerns: danger, lack of feasibility, the wrong function, low social acceptance, vague wording and unwanted spread to other conduct.

An instruction to ignore fails to say which result will change. It can also hide the care, speech, trust and safety responses that must remain.

‌

Interpretation in Schools

Extinction bursts have a valid place in the study of behaviourism as a learning theory. They are not a general way to manage a class. The theory helps scholars ask exact questions about links and measures. It cannot turn a visible rise into a diagnosis or justify the loss of support.

School review should start with access to learning and taking part. Is the task clear and within reach? Can the learner ask for help, a pause or an explanation?

Could pain, hunger, fatigue, medicine or another health issue play a part? Teams should also consider sensory load, peers, a change in routine and past events.

An existing break, movement option or lower demand may be an agreed adjustment. It should not be removed because someone assumes that it rewards avoidance. A personal plan can change how support is given, but only after sound assessment. It also needs a clear reason, the learner's and family's views where suitable, risk limits and a review date.

Department for Education (2024) guidance tells schools to maintain a safe and supportive setting. Schools should plan for SEND-related triggers where they can, offer support and make reasonable adjustments for disabled learners. Keeping Children Safe in Education also requires staff to act when conduct or other signs may point to harm (Department for Education, 2025). A theory label cannot pause these duties.

Leaders can test whether a claim has crossed its evidence limit. Does it treat a clinical self-injury study as a study of ordinary lessons? Does it turn sessions into days or infer function from looks?

Does a plan state what support remains and when it must stop? These are checks on quality and safety, not steps for teachers to run an experiment.

The concept has a modest place among major learning theories. It shows how an exact idea can become unsafe when moved between settings without care. Its value for education lies in seeing both the idea and the point where the evidence ends.

‌

Professional and Safeguarding Boundaries

Complex or risky conduct needs suitable professional support. The right route may include the SENCO, designated safeguarding lead, educational psychologist, health professional or qualified behavioural practitioner. Each role has limits. No job title removes the need for joint, person-centred work.

NICE guidance stresses prevention and quality of life for the groups it covers. It calls for useful ways to communicate, clear written duties, training, data and frequent review (National Institute for Health and Care Excellence, 2013, 2015). The BACB ethics code also requires skill, assessment, low risk and ongoing review for its certified analysts (Behavior Analyst Certification Board, 2020). That code does not qualify teachers or turn an article into a plan.

Self-harm, harm to others, absconding, severe distress, a sign of abuse or a medical need requires the agreed safety or health response. Staff should not relabel such events as an expected burst. They should protect the learner and others, record relevant facts and start the right review.

The term planned ignoring must not mean loss of care, trust, speech, communication access or attention to safety. A specialist plan that changes a certain result must name that result. It must state what support stays in place and how an easy alternative response will be taught and met. It also needs named staff, useful data and clear events that make the plan stop or change.

Prevention and the least restrictive support offer a sounder guide than rigid use of a procedure. Fidelity matters only while a plan remains lawful, ethical, practical and safe. Evidence of harm, a wrong hypothesis or an unmet need calls for review. It is not a reason to defend consistency.

‌

Limitations and Unresolved Questions

The first limit is the definition. Researchers differ over the baseline, response measure and time window. A short rise can meet one rule and fail another. Reported rates cannot be compared well until those choices are clear.

The second limit is transfer. Laboratory studies offer control, but many use non-human subjects and simple designs. Applied studies often use a small set of cases and people in specialist services.

These methods can answer narrow questions. They cannot give a population rate for schools.

The third limit concerns cause. Several accounts of bursting are still being tested (Fisher et al., 2023). A story about a person trying harder may make sense, but it is not proof. Scholars should keep the observed change apart from claims about thought or motive.

The fourth limit is the move into practice. Past clinical records link treatment packages with fewer bursts, but they do not reveal one best package. The assessed need, ease of access, consent or assent, staff skill, resources and risk all affect a plan. Whole-class use would be a further leap.

The fifth limit is ethical. Some studies concern severe self-injury or aggression. Their results cannot be cut away from the needs and rights of the people involved.

They must not be used to make danger seem normal. Research still needs clearer terms, better tests of cause and outcomes that matter to people. It must also meet strict ethical rules.

‌

Frequently Asked Questions

Extinction-burst questions often seek a firm timetable or a sign that a plan works. The research offers neither. The answers below separate the narrow learning concept from claims about cause, classroom use and safety. Each answer keeps the study setting and limits in view.

‌

Are extinction bursts inevitable?

No. Studies observe them under some extinction designs, but reported rates vary widely. Study rules, schedules, baseline methods, response measures and time windows affect detection. It is sounder to say that a burst is possible than to call it a required stage.

‌

How long does an extinction burst last?

There is no universal length. A study may report trials or sessions under one procedure, not days that apply in school. Any claim must name the response, history, setting, method and time window. The approved evidence gives no fixed classroom range.

‌

Does worsening behaviour prove that extinction is working?

No. A rise does not prove that extinction is taking place, that the cause was found or that a plan works. It can mark a wrong idea, an unmet health or speech need, a hard setting or harm. Safety and review still decide what happens next.

‌

Is an extinction burst the same as spontaneous recovery?

No. A burst is a possible rise near the start of extinction. Spontaneous recovery is the return of a response after time has passed. Renewal follows a change in context.

Resurgence is the return of an older response when a newer option stops producing a useful result.

‌

Can teachers identify the function of behaviour by observation?

Observation gives useful data, but looks alone do not show function. Assessment may need to cover speech, health, trauma, sensory needs, curriculum, context, events before a response and its results. Complex or risky cases need the right skills and input from more than one profession.

‌

Is planned ignoring an extinction procedure?

The phrase is too vague for a safe answer without details. Extinction concerns a named response and the result that maintains it. An order to ignore names neither one and can spread to care, speech or safety.

Lloveras et al. (2023) advise exact descriptions instead.

‌

Further Reading

• Basic and applied research on extinction bursts
  Fisher, Greer, Shahan and Norris (2023) review terms, reported rates, possible causes and limits on transfer.
• Side effects of extinction
  Lerman, Iwata and Wallace (1999) report the clinical self-injury data sets often cited in work on bursting and aggression.
• What is an extinction burst?
  Katz and Lattal (2021) study the choices involved in defining a change in behaviour.
• Recommendations regarding use of the term ignore
  Lloveras et al. (2023) explain why exact descriptions are safer than generic language about ignoring.
• NICE NG11 recommendations
  This guidance covers person-centred assessment, prevention, risk, support and review for its named population.

‌

References