Extinction in Operant Conditioning

How extinction works

Every reinforced behavior exists because of a contingency: do this, get that. Extinction breaks the contingency. The rat that has learned to press a lever for food presses, and nothing happens. It presses again. Over the next hour its rate climbs, wobbles, and slides toward zero, tracing the extinction curve Skinner recorded in 1938.^[1]

• A rat presses a lever → no pellet, ever again → pressing declines over the session.
• A child whines for a cookie → whining is never again followed by a cookie → whining declines over days.
• You text a friend who has stopped replying → no replies → you text less and eventually stop.

Two things distinguish extinction from other ways of reducing behavior. It works only on the reinforcer that has been maintaining the behavior; withholding something else is not extinction. And it is slow and bumpy, because a history of the behavior paying off does not vanish when the payoff stops — it is tested, protested, and eventually overwritten. How long that takes depends on the reinforcement history, as the partial reinforcement extinction effect explains.

Watch it happen

The simulator below starts in extinction: a subject with a history of reinforcement, and no more reinforcers. Run it and watch the burst, the decline, and the flattening record. Switch the schedule to compare how quickly behavior built on continuous versus intermittent reinforcement gives up.

Operant vs. respondent extinction

"Extinction" is used in both branches of conditioning. In respondent (classical) extinction, a conditioned stimulus is presented repeatedly without the unconditioned stimulus — the bell rings, no food follows — and the conditioned response fades. In operant extinction, a behavior occurs repeatedly without its reinforcer and the behavior fades. Both show extinction curves, spontaneous recovery, and renewal, and both are now understood as new inhibitory learning rather than unlearning.^[2] The difference is what is disconnected: a stimulus from a stimulus, or a behavior from its consequence. Operant vs. classical conditioning in full ›

What is an extinction burst?

An extinction burst is a temporary increase in the frequency, intensity, or variability of a behavior immediately after its reinforcer is withheld, before the behavior declines. Press the elevator button; nothing happens; you press again, harder, three times, then hold it, then try the other button. More, harder, different — that is the burst. The organism's history says the behavior should work, and vigorous, varied responding is the evolved answer to a missing payoff.

The burst is also the engine of shaping: when a trainer stops reinforcing the current approximation, the burst of variability produces the slightly better response to reinforce next. But in everyday life the burst is where almost everyone gives up. The child whose whining has stopped producing candy whines louder, then cries, then throws himself on the floor — and a parent who gives in at that point has reinforced the floor-throwing.

How common is it? A review of 113 published data sets in which extinction was applied to problem behavior found a burst in about a quarter of cases, more often when extinction was used alone than when combined with procedures such as differential reinforcement.^[3] A later analysis of 41 cases of self-injury treated with extinction found bursts in 39% and extinction-induced aggression in 22%, again less often when extinction was combined with other treatments.^[4] The burst is worth planning for but not inevitable, and the way to make it less likely is also the way to make extinction work better: reinforce something else at the same time.

Extinction-induced aggression and emotional responding

Withholding an expected reinforcer also elicits emotional behavior: agitation, frustration, and aggression, often aimed at whatever is nearby. In a landmark experiment, pigeons pecking a key for grain were placed on extinction, and when the grain stopped they attacked a restrained pigeon at the other end of the chamber — a target that had nothing to do with the reinforcer.^[5] The human version is familiar: the vending machine gets kicked, the customer-service line gets yelled at, a child on extinction hits a sibling. Emotional responding is the second main side effect of extinction, after the burst, and a central reason it is rarely used alone for serious behavior.^[4][6]

Spontaneous recovery, resurgence, and renewal

Extinguished behavior comes back in at least three distinct ways, each with its own trigger. Knowing which is which tells you what to do about it.

Spontaneous recovery

Spontaneous recovery is the reappearance of an extinguished behavior after time away from the extinction setting. The rat that stopped pressing by the end of Monday's session starts again on Tuesday, at a lower rate, and stops sooner. First described by Pavlov for conditioned reflexes, it occurs just as reliably for operant behavior, and each recovery is smaller than the last if reinforcement stays withheld.^[7] A behavior that seems gone on Friday may be back on Monday; that is not a sign extinction failed.

Resurgence

Resurgence is the reappearance of a previously reinforced behavior when a more recently reinforced behavior is placed on extinction. Teach a pigeon to peck key A for food, extinguish it, teach it to peck key B, then extinguish B — and the pigeon starts pecking A again.^[8][9] This explains a common form of relapse: a child learns to ask politely instead of screaming, then in a busy classroom polite requests go unanswered, and the screaming returns.^[10] The remedy is to keep the replacement behavior reliably reinforced, especially where the old behavior once worked.

Renewal

Renewal is the return of an extinguished behavior when the context changes. Mark Bouton's research showed that extinction learning is unusually tied to the place and cues in which it happened, whereas the original learning transfers broadly — which is why a fear extinguished in a therapist's office returns in the parking garage, and a behavior extinguished at the clinic returns at home.^[2][11] The remedy is to conduct extinction in every context that matters.

The partial reinforcement extinction effect

The most important variable in how long extinction takes is the schedule on which the behavior was reinforced. Behavior reinforced intermittently is far more resistant to extinction than behavior reinforced every time. This is the partial reinforcement extinction effect, first demonstrated systematically by Lloyd Humphreys in 1939: responses reinforced only half the time persisted much longer without reinforcement than responses reinforced every time.^[12]

The effect is paradoxical — less reinforcement produces more persistence — and two theories survived decades of testing. Abram Amsel's frustration theory holds that organisms on intermittent schedules learn to keep responding through the frustration of nonreward, so nonreward during extinction is nothing new.^[13] E. J. Capaldi's sequential theory holds that the memory of unreinforced trials becomes a cue for responding.^[14] In plain language: an organism on a continuous schedule can tell instantly that the rules have changed; one on a variable schedule cannot, because unrewarded stretches were always part of the deal.

The consequences are everywhere. Slot-machine players keep pulling through long droughts. A parent who gives in to whining "only sometimes" has built whining on a variable schedule and faces a long extinction. And a habit you want to keep should be moved from continuous to intermittent reinforcement once established, so it survives the days the reinforcer does not show up.

Resistance to extinction and behavioral momentum

Resistance to extinction is one case of what John Nevin called behavioral momentum. In physics, momentum depends on velocity and mass; in Nevin's analogy, a behavior's velocity is its rate and its mass is its resistance to disruption — by extinction, satiation, distraction, or free food. His central finding is that mass depends on the rate of reinforcement obtained in the presence of a stimulus, largely independent of response rate: behavior from a richly reinforced context is harder to disrupt than behavior from a lean one, even at the same rate.^[15][16]

This has a counterintuitive implication for treatment. Reinforcing an alternative behavior makes the alternative more likely, but it also enriches the context, which can make the problem behavior more resistant to extinction when its reinforcement is later withheld.^[16] Practitioners deliver the alternative reinforcement in clearly distinct situations and expect the persistence. For the rest of us, momentum is why a well-established habit, good or bad, shrugs off disruption.

Extinction is not ignoring

"Just ignore it" is the folk version of extinction, and it is wrong often enough to be dangerous. Ignoring is extinction only when the reinforcer maintaining the behavior is your attention. Extinction means withholding whichever reinforcer is doing the maintaining; a functional analysis identifies it, and the procedure must match it.^[17][18]

The escape row is the one that catches parents and teachers. A child who tantrums to get out of putting on shoes is not seeking attention, and ignoring the tantrum while the shoes stay off is not extinction; it is the reinforcer, delivered. Escape extinction means the shoes still go on — planned carefully, with calm, safety, and often an easier demand to begin with. More on escape-maintained behavior ›

Examples of extinction

The bedtime example is a real case, published in 1959. A 21-month-old boy screamed when his parents left the room, keeping one of them there for up to two hours a night. The parents put him to bed, left, and did not return. He cried for about 45 minutes the first night, less on subsequent nights, and by the tenth night not at all. A week later an aunt put him to bed, returned when he cried, and the tantrums came back at full strength — requiring a second extinction, which again succeeded.^[19] The case contains the whole story: burst, curve, accidental intermittent reinforcement, recovery, and the persistence needed to see it through.

How to use extinction

1. Identify the maintaining reinforcer. Attention? Escape? An item? A sensation? If you cannot tell, the procedure cannot be designed.
2. Make sure you control that reinforcer. If a classroom behavior is maintained by peers' laughter, a teacher's ignoring changes nothing.
3. Decide whether you can withstand the burst. If the behavior is dangerous at its worst — self-injury, aggression, bolting — extinction alone is not appropriate, and a professional should design the plan.
4. Reinforce a replacement at the same time. Extinction says what no longer works; differential reinforcement says what does. The combination reduces bursts and aggression.
5. Get everyone on the same page. One family member or one weekend with the grandparents can reinforce the behavior intermittently and undo weeks of progress.
6. Expect recovery, resurgence, and renewal. Returns after a break, under stress, and in new settings are normal, and each is weaker if the plan holds.
7. Track it. Extinction curves are bumpy; without data a bad week looks like failure.

When not to use extinction

• When the behavior is dangerous and the burst could cause harm.
• When you cannot control the maintaining reinforcer — peer attention, sensory consequences you cannot block, reinforcers delivered by others.
• When you cannot be consistent; intermittent extinction is intermittent reinforcement.
• When an antecedent change would remove the need — moving the cookie jar is easier than extinguishing cookie-seeking. Antecedent strategies in the ABC model ›

Combining extinction with differential reinforcement

Extinction is rarely used alone in modern practice: by itself it produces the burst and the emotional side effects and leaves a hole where the behavior was. The standard package pairs it with reinforcement of something else.^[6][20]

• DRA (alternative behavior): reinforce an appropriate behavior that gets the same result. The child who screamed for help is taught to tap the adult's arm; tapping is answered every time, screaming never.
• DRI (incompatible behavior): reinforce a behavior that cannot happen at the same time. Sitting is reinforced; jumping up cannot coexist with it.
• DRO (other behavior): reinforce the passage of time without the behavior. Every five minutes without shouting earns a point.
• Noncontingent reinforcement: deliver the maintaining reinforcer freely on a schedule, so the behavior has nothing to earn.

With a replacement in place, extinction becomes a redirection rather than a standoff: the reinforcer the learner wants is still available, through the door you chose. The same applies when the learner is you — the reliable way to extinguish a habit you do not want is to make its payoff available from one you do. Building habits with operant conditioning › · Positive reinforcement in depth ›

Key takeaways

• Extinction breaks the contingency: the behavior occurs and its reinforcer no longer arrives. Nothing is added or removed, so it is not one of the four quadrants, and it is not forgetting but new learning laid over the old.
• The extinction burst is a temporary rise in the frequency, intensity, or variability of the behavior. Giving in during the burst reinforces a more intense version on an intermittent schedule, so do not start extinction unless you can outlast it.
• Extinguished behavior returns in three ways: spontaneous recovery after time away, resurgence when a newer behavior stops being reinforced, and renewal when the context changes. Each is weaker if the plan holds, and none means extinction failed.
• Behavior reinforced intermittently is far more resistant to extinction than behavior reinforced every time: the partial reinforcement extinction effect. Giving in "only sometimes" builds whining on a variable schedule and guarantees a long extinction.
• Ignoring is extinction only when attention is the maintaining reinforcer; escape-maintained behavior requires escape extinction, in which the demand stays in place. In practice, pair extinction with reinforcement of a replacement, which reduces bursts and aggression.

Check yourself

A child tantrums when told to put on shoes. The parent ignores the tantrum, and the shoes stay off. Is this extinction?

No. The tantrum is maintained by escape from the demand, not by attention, so ignoring it while the shoes stay off delivers the reinforcer. Escape extinction means the shoes still go on; the behavior no longer ends the demand.

A child whines for cookies. Parent A never gives a cookie for whining again. Parent B takes away screen time for each whine. Which one is extinction?

Parent A. Extinction stops delivering the reinforcer the behavior used to produce; nothing is added or removed. Parent B removes a reinforcer the child already has, which is negative punishment, not extinction.

Bedtime crying that stopped last week returns after a weekend at the grandparents' house. Has extinction failed?

Not necessarily. A return after time away from the extinction setting is spontaneous recovery, which is normal and weaker each time if reinforcement stays withheld. The real risk is that someone responded to the crying over the weekend: one intermittent reinforcement can bring the behavior back at full strength.

Two children whined for candy. One was always given candy for whining; the other got it only sometimes. Both families now stop entirely. Whose whining fades faster, and why?

The child who was always given candy. Behavior reinforced every time is far easier to extinguish than behavior reinforced intermittently, which is the partial reinforcement extinction effect. The child on the continuous schedule can tell instantly that the rules have changed; the other cannot, because unrewarded stretches were always part of the deal.

Explain it to a friend. Explain why a behavior can get worse before it fades once its payoff stops, using an example that has nothing to do with children or pets.