Extinction vs spontaneous recovery: why an extinguished response comes back

Extinction is the weakening of a learned response when the event that it predicted, or the reinforcer that it earned, stops coming. Spontaneous recovery is the return of that extinguished response after time passes, with no new training. Extinction comes first and lowers the response. Spontaneous recovery comes later and shows that extinction did not erase what was learned. Both happen after classical conditioning and after operant conditioning.

• Ivan Pavlov showed both in one lecture. A dog’s salivation to a metronome fell from 10 drops to 3 in seven trials without food. Twenty-three minutes later, the dog gave 6 drops.
• At the first test, recovery is often partial. Pavlov wrote that established reflexes return “sooner or later to their full strength”, but most of his own first tests show a smaller return.
• Spontaneous recovery is one of four well-studied ways in which an extinguished response comes back. The other three are renewal, reinstatement and rapid reacquisition. Together, they are the main evidence that extinction is new learning, not erasure.
• An extinction burst is a different event: a short rise in responding at the start of operant extinction. An analysis of 113 sets of applied data found a burst in only 24% of them.
• The cause of spontaneous recovery is not settled. Researchers propose several mechanisms, and the psychologist Robert Rescorla concluded that it probably has more than one source.

Extinction vs spontaneous recovery: the difference in one table

What extinction means

The APA Dictionary of Psychology gives extinction two senses in each type of conditioning. In classical conditioning, extinction is a procedure in which the pairing of stimuli stops, “either by presenting the conditioned stimulus alone or by presenting the conditioned stimulus and the unconditioned stimulus independently of one another”. It is also the result of that procedure: “a gradual decline in the probability and magnitude of the conditioned response”. In operant conditioning, extinction is a procedure in which “reinforcement is discontinued”. Its result is “a decline in the rate of the formerly reinforced response”.

So an exam answer to “what is extinction?” usually needs two parts: the pairing or the reinforcement stops, and the response becomes weaker. Our comparison of classical and operant conditioning explains the two kinds of learning that extinction applies to.

Extinction is not punishment: nothing is added or taken away after the behaviour, and the consequence simply stops. Our article on negative reinforcement and punishment explains why extinction is a third thing next to them. In operant conditioning, the history of reinforcement also matters. A behaviour that was reinforced only some of the time resists extinction more than a behaviour that was reinforced every time. The APA Dictionary calls this the partial reinforcement effect.

Pavlov chose the word with care. He wrote that the term “does not imply any hypothesis as to the exact mechanism by which the phenomenon is brought about”. Extinction describes a result. It does not say that the learning is gone.

Pavlov’s numbers: extinction and recovery in one lecture

Pavlov described spontaneous recovery in his lectures on conditioned reflexes, published in English in 1927. In Lecture IV, he extinguished a reflex in front of his audience. The dog had learned to salivate to the sound of a metronome. The metronome sounded for 30 seconds every 2 minutes, and no food followed.

The decline was not smooth, because trials 3 and 5 went up again. But the trend was down, and the dog also started to salivate later. After 23 minutes without the metronome, Pavlov tested the dog again. “Considerable recovery has therefore taken place spontaneously during the lecture,” he said.

In two more examples from his laboratory, a dog saw meat powder at a short distance but did not get it.

• In an experiment by Babkin, the meat powder appeared every 3 minutes. Salivation fell from 1.0 cubic centimetres to 0.6, 0.3 and 0.1, and then to 0.0 twice. After 2 hours, it was 0.15.
• In an experiment by Eliason, the meat powder appeared every 10 minutes. Salivation fell from 8 drops to 3 and then to 0. After 20 minutes, it was 7 drops.

Pavlov wrote that fully established reflexes “invariably and spontaneously return sooner or later to their full strength”. His own first tests show a less simple result. Babkin’s dog came back to 15% of its first value after 2 hours, the demonstration dog to 6 of 10 drops, and Eliason’s dog to 7 of 8 drops. The recovery is real, but at the first test it is often partial. According to Pavlov, the time to full recovery can be anything “from a few minutes to a number of hours”.

Two more observations from the same lecture are still part of modern research:

• Recovery can be made faster. Pavlov wrote that it is enough to give the unconditioned stimulus again, alone or with the conditioned stimulus. Today, a return after the unconditioned stimulus alone is called reinstatement.
• Pavlov concluded that extinction “must be regarded as a special form of inhibition”. It is not “an irreparable destruction of the conditioned reflex”, and his proof was that “the extinguished reflexes spontaneously regenerate in course of time”.

Why the response comes back: extinction is new learning

If extinction erased the original learning, the response would have nothing to return from. Modern research agrees with Pavlov here. In a 2004 review, the psychologist Mark Bouton concluded that extinction “does not destroy the original learning, but instead generates new learning that is especially context-dependent”.

In Bouton’s account, the conditioned stimulus has two learned meanings after extinction: the original link with food or shock, and the newer link with no food or no shock. In his words: “As is true for an ambiguous word, the context is crucial in selecting between them.” The place, the time and the state of the learner decide which meaning controls the response. Bouton called extinction “just one example of a retroactive inhibition phenomenon in which new learning inhibits old”. Memory research has a related term, retroactive interference, for new learning that blocks the recall of older learning.

For spontaneous recovery, Bouton and his colleagues argued that time acts like a context. The learner learns extinction at one time, and a later test happens at a different time. On this view, spontaneous recovery “can be seen as the renewal effect that occurs when the CS is tested outside its temporal context”. CS is the usual abbreviation for the conditioned stimulus. One finding supports the idea. In a 1993 study by Brooks and Bouton, a cue that was presented during extinction reduced spontaneous recovery when it was presented again just before the test. A 1994 study found the same result for renewal. The cue seems to remind the learner of extinction.

This is one account, not the settled answer. In a review in the same 2004 journal issue, Rescorla called spontaneous recovery “one of the most basic phenomena of Pavlovian conditioning”. He also wrote that the research on what controls it is “relatively sparse and quite mixed”, and that “its very ubiquity suggests that spontaneous recovery has multiple sources”. Some points are firm. In the “standard description”, responding is “partially restored in a test session that is administered after a delay”. Recovery increases with time. And the inference that extinction does not remove all of the original learning “seems secure”, although spontaneous recovery alone does not prove that all of it remains.

Two experiments show why one explanation is not enough. In a second 2004 paper, Rescorla reported four experiments with rats and pigeons. Spontaneous recovery was greater when extinction came soon after conditioning than when it came later. According to the abstract, “many accounts of spontaneous recovery that attribute it to the fading of learning that occurred during extinction” do not predict this result. In 2009, Heather Sissons and Ralph Miller gave rats two phases of training with opposite meanings for one stimulus, then tested them 4 or 27 days later. After the long delay, the first-learned association came back. They concluded that second-learned associations fade faster than first-learned ones.

Four ways an extinguished response comes back

Spontaneous recovery is the best-known return effect, but it is not the only one. Bouton’s review describes four effects that bring back an extinguished response. Each effect has a different trigger.

Renewal has three versions. In ABA renewal, conditioning happens in context A, extinction in context B, and the test in context A. In ABC renewal, the test happens in a third context. In AAB renewal, conditioning and extinction both happen in context A, and the test happens in context B. Bouton wrote that renewal had been seen “in virtually every conditioning preparation in which it has been investigated”. In fear conditioning with rats, it appeared when 84 extinction trials followed 8 conditioning trials. Other studies found it after as many as 160 extinction trials, but a 2003 report suggested that it might not survive 800.

The other effects also have limits. Reacquisition is not always rapid. In fear conditioning and in flavour aversion learning, reacquisition can be slower than learning with a new stimulus. In fear conditioning, this slow reacquisition needs extensive extinction training. For reinstatement tests held 24 hours or more after the unconditioned stimulus, Bouton wrote that the evidence “strongly suggests that the effect is due to conditioning of the context”. The animal links the setting with the event, and the setting then brings back the response.

These effects are not only classical. A 2014 review by Travis Todd, Drina Vurbic and Mark Bouton reported that “all three forms of renewal (ABA, AAB, and ABC) have now been demonstrated after the extinction of instrumental behavior”, which is operant behaviour. Its general conclusion is direct: “Extinction is not the same as erasure.”

Pavlov also described a fifth trigger, called disinhibition. The APA Dictionary defines disinhibition in conditioning as “the reappearance of responding, which has stopped occurring as a result of exposure to extinction, when a new stimulus is presented”.

Extinction burst vs spontaneous recovery

These two terms are often confused because both describe an increase in a behaviour that no longer earns a reinforcer. The difference is the timing. The APA Dictionary defines a burst as a series of responses at a relatively high rate, for example “at the onset of extinction (an extinction burst) when conditioned responses are no longer rewarded”.

The burst is less common than people often assume. In 1995, Dorothea Lerman and Brian Iwata examined 113 sets of extinction data from applied research. They found bursting in 24% of the cases. When extinction was used together with other procedures, the rate was 12%. When extinction was the only intervention, the rate was 36%. A later study by Lerman, Iwata and Wallace looked at 41 data sets from the treatment of self-injurious behaviour. Bursts or increases in aggression occurred in nearly half of the cases, and less often when extinction was part of a treatment package.

A Reddit discussion among MCAT students shows two common errors. One answer says that a burst happens “right before the behavior is eliminated for good”, but the return effects show that extinction often does not remove a behaviour for good. Another says that spontaneous recovery is “mostly associated with classical conditioning”. The APA Dictionary defines spontaneous recovery as a return “following either operant or classical conditioning”, and Rescorla’s 2004 experiments included instrumental training.

Resurgence vs spontaneous recovery

Resurgence is the return of an older behaviour when a newer behaviour stops working. The APA Dictionary’s example of resurgence uses a rat and two levers. First, presses on lever A earn food. Next, presses on A no longer earn food, and presses on B do. The rat stops pressing A and presses B. Then B also stops earning food. “As responding on B declines, pressing on A will increase temporarily.”

Spontaneous recovery needs only one behaviour and a pause. Resurgence needs two behaviours, and the trigger is the extinction of the newer one. Todd, Vurbic and Bouton argue that “resurgence can be understood as another form of the renewal effect”, conceptually similar to ABC renewal. On their view, the extinction of the newer behaviour changes the context, and the old behaviour returns in that new context.

Spontaneous recovery outside the laboratory

Extinction is also the basis of some treatments. The APA Dictionary describes exposure therapy for anxiety disorders as “systematic and repeated confrontation with a feared stimulus”, and it lists extinction among the processes behind it. In a 2013 review, Bram Vervliet, Michelle Craske and Dirk Hermans wrote that exposure-based treatments “generally are very effective, but relapse is not uncommon”. Their summary of the laboratory work is that fear extinction is “relatively easy to ‘learn’ but difficult to ‘remember’”. They concluded that treatments will benefit from more focus on the long-term retrieval of extinction. If a fear or a habit that improved with treatment comes back, tell the therapist or the clinician who treated it. A return of fear is a known pattern, and treatment plans can address it.

The same term appears in research on habituation, the decrease in response to a stimulus that repeats. A 2009 consensus paper by Catharine Rankin and colleagues lists spontaneous recovery as a characteristic of habituation: “If the stimulus is withheld after response decrement, the response recovers at least partially over the observation time.” This is not dishabituation. In dishabituation, a different stimulus makes the decreased response stronger again.

Classifying exam scenarios

Exam questions describe a situation and ask which term fits. Ask two questions. Did the response decrease, increase or come back? And what happened just before: the start of extinction, a pause, a change of place, the event alone, new pairings, or the extinction of a newer behaviour?

Three traps are common. First, a burst and spontaneous recovery are both increases, but a burst comes at the start of extinction and spontaneous recovery comes after a pause. Second, if the place changes as well as the time, the answer is usually renewal. Third, reinstatement needs the event alone, and rapid reacquisition needs the signal and the event together.

Keeping the terms apart

Extinction, spontaneous recovery, renewal, reinstatement, rapid reacquisition, extinction burst and resurgence all describe what happens after the event or the reinforcer stops, so they blur easily. In Psychology World you can save terms like these to one list. Then you can drill the list as flashcards, as multiple-choice questions once it has four or more terms, or by typing the answer. What each study mode does is set out on the home page. Our guide to memorising psychology terminology explains which study methods have the strongest evidence.