Extinction in Dog Behavior: Why Learned Behavior Disappears – and Returns

2. What Is Extinction? A Definition

In behavioral psychology, extinction is the procedure of no longer delivering reinforcement for a previously reinforced behavior. As a result, the frequency, intensity, or duration of that behavior decreases over time.

Extinction is not the same as forgetting. Forgetting implies that the memory of the behavior fades passively with time. Extinction is an active learning process: the dog learns that the old behavior no longer works, that the contingency between behavior and reward has changed.

For example:

• A dog that jumps on the counter and finds food is reinforced. If the owner removes all food from the counter, the dog may continue jumping for a while (extinction burst), then gradually stop. The dog has learned that jumping no longer produces food.

• A dog that barks at the door and receives attention (even negative attention like “quiet!”) may continue barking. If the owner consistently ignores the barking (extinction), the barking will eventually decrease.

Critically, extinction does not erase the original learning. The underlying memory remains intact, which is why the behavior can return later.

Extinction represents not “unlearning” but the formation of a second, context‑dependent inhibitory memory trace that suppresses the original association in a situation‑specific manner. This is the central insight of modern learning theory (Bouton, 2002; Craske et al., 2014).

3. The Neurobiology of Extinction – Not Erasure, But Inhibitory Learning

Early researchers believed extinction erased the original conditioned response. Modern neuroscience has shown otherwise. Extinction is not unlearning; it is inhibitory learning – the process of forming a new inhibitory memory trace. The brain forms a new memory that competes with the original memory.

3.1 The Role of the Prefrontal Cortex

The prefrontal cortex (PFC) plays a central role in extinction learning. When a dog learns that a previously rewarded behavior no longer produces reinforcement, the PFC inhibits the original response by activating the infralimbic cortex, which projects to the amygdala and suppresses fear or reward responses.

In extinction, the ventromedial prefrontal cortex (vmPFC) – a region involved in safety learning and inhibition – becomes active. Dogs with compromised PFC function (e.g., due to chronic stress, age‑related cognitive decline, or poor early development) show impaired extinction learning.

3.2 The Amygdala and Extinction

The amygdala encodes the original conditioned emotional response (fear or excitement). Extinction does not remove that amygdala memory. Instead, the PFC forms a new, inhibitory connection that “overrides” the amygdala’s response. The original memory remains, which is why spontaneous recovery is possible.

3.3 NMDA Receptors and Extinction Consolidation

Extinction learning, like initial learning, requires the activation of NMDA receptors in the amygdala and PFC. Research in rodents shows that blocking NMDA receptors prevents extinction. Enhancing NMDA function (e.g., with D‑cycloserine) can accelerate extinction. While pharmacological enhancement is not practical in dog training, the principle underscores that extinction is an active, biologically demanding process.

Note on species extrapolation: While most mechanistic insights stem from rodent models, converging evidence across species suggests that similar neural circuits are involved in mammals, including canines.

For more on prefrontal cortex function in impulse control, see prefrontal cortex and self‑control in dogs.

For the role of dopamine in reward learning and extinction, see dopamine and learning in canine neurochemistry.

4. A Central Feature: The Context‑Dependence of Extinction

A fundamental characteristic of extinction learning is its pronounced context‑dependence. This context‑dependence is not a secondary feature but a defining property of extinction learning. While original conditioning (e.g., a dog learning that a clicker predicts food) generalizes relatively robustly across different environments, extinction remains strongly bound to the specific context in which it was learned (Bouton, 2002).

In practical terms: a dog that learns to ignore the doorbell through extinction training in the living room may still react to the doorbell when the family visits a friend’s house. The inhibitory memory trace does not automatically transfer to a new setting. This is why renewal (see Section 7) is so common and why trainers must actively work to generalize extinction across contexts.

5. Extinction in Operant vs. Classical Conditioning

Extinction occurs in both operant and classical conditioning, but the mechanisms and applications differ.

5.1 Extinction in Operant Conditioning (Behavior – Consequence)

In operant conditioning, extinction means withholding the reinforcer that previously followed the behavior.

Original learning: Dog sits → gets treat

Extinction procedure: Owner stops giving treat for sitting

Outcome: Sitting decreases over time

---

Original learning: Dog jumps on guests → gets attention (even “off!”)

Extinction procedure: Owner ignores jumping completely

Outcome: Jumping decreases

Key features of operant extinction:

• Extinction burst – An initial increase in the behavior’s frequency, intensity, or duration. Extinction bursts are not a separate learning mechanism but a short‑term behavioral escalation resulting from the discrepancy between expectation and the absence of reinforcement. Response output temporarily increases as the organism samples previously reinforced behaviors.

• Extinction‑induced aggression – Frustration from withheld reinforcement can, in some cases, lead to aggressive responses (e.g., nipping, growling). This is not inevitable but is more likely when extinction is applied suddenly to behaviors that were previously reinforced on a continuous schedule, especially in dogs with low frustration tolerance.

• Extinction‑induced variability – The dog may try new, untrained behaviors in an attempt to obtain reinforcement. This can be useful if the owner captures and reinforces an alternative behavior.

5.2 Extinction in Classical Conditioning (Stimulus – Emotion)

In classical conditioning, extinction means presenting the conditioned stimulus (CS) without the unconditioned stimulus (US) repeatedly, until the CS no longer elicits the conditioned response (CR).

Original learning: Leash click → food (salivation)

Extinction procedure: Leash click presented alone, no food

Outcome: Salivation decreases

---

Original learning: Stranger (CS) → painful leash pop (US) → fear (CR)

Extinction procedure: Stranger presented without pain, but with safety

Outcome: Fear decreases gradually

Important difference: In classical conditioning, the conditioned response is not a voluntary behavior but an emotional/physiological reaction. Extinction of fear is slower and more context‑dependent than extinction of an operant response. It also carries a high risk of renewal.

Spontaneous recovery (see Section 6) is often interpreted as a temporary shift in retrieval probability rather than a change in underlying associative strength. The inhibitory memory trace is still present but becomes temporarily less accessible.

5.3 Interaction Between Operant and Classical Extinction – A Key Insight

In real‑world training situations, operant and classical processes rarely act in isolation. Operant extinction (e.g., ignoring a jumping behavior) can simultaneously influence classical processes, for example by gradually weakening the dog’s emotional response to the presence of visitors (the jumping previously predicted attention; when attention no longer follows, the predictive value of the visitors changes). Conversely, classical extinction of fear can reduce the intensity of operant avoidance behaviors driven by that fear.

Understanding this interplay is essential for trainers: extinguishing an overt behavior without addressing the underlying emotional conditioning may leave the dog “compliant but still anxious.” Addressing both systems is the path to lasting change.

For a deeper look at how classical conditioning underlies emotional responses, see reactivity in dogs – a neurological perspective.

6. Spontaneous Recovery – Why Extinguished Behaviors Return

The most frustrating and misunderstood phenomenon in dog training is spontaneous recovery: the reappearance of a previously extinguished behavior after a period of rest, without any new reinforcement.

6.1 What Spontaneous Recovery Looks Like

• You train your dog to stop jumping on visitors by ignoring the behavior. After several weeks, the jumping stops. Then, after a weekend with no visitors, the first guest arrives – and your dog jumps again, just as enthusiastically as before.

• You counter‑condition your dog’s fear of the vacuum cleaner by pairing it with high‑value treats. The dog becomes relaxed around the vacuum. After two weeks of not using the vacuum, you bring it out again – and the dog hides.

Spontaneous recovery is not a relapse. It is a natural property of extinction learning. The original memory is still present, and after a period of rest, it can briefly resurface before extinction re‑asserts itself.

6.2 Why Does Spontaneous Recovery Occur?

The dominant explanation is that extinction creates a second, inhibitory memory trace that competes with the original, excitatory memory. After rest, the inhibitory memory trace is less accessible, allowing the original memory to momentarily dominate. With further extinction trials, the inhibitory memory strengthens again.

Spontaneous recovery is typically a function of the time elapsed since the last extinction session: the longer the interval, the more likely the behavior is to reappear. Each successive recovery is usually less intense than the previous one and extinguishes more rapidly.

6.3 Implications for Training

• Expect spontaneous recovery. It is not a sign of failure.

• When the behavior returns, do not punish. Simply repeat the extinction procedure (withhold reinforcement).

• Use maintenance training – periodic reinforcement of the alternative behavior – to strengthen the new habit and reduce the likelihood of recovery.

7. Other Forms of Relapse – Renewal, Resurgence, and Reinstatement

Spontaneous recovery is one of several ways extinguished behaviors can return. Trainers should also be aware of:

7.1 Renewal

Renewal occurs when the behavior returns because the context changes. If extinction takes place in one environment (e.g., the living room), the extinguished behavior may reappear when the dog is in a different environment (e.g., the park).

There are three types of renewal:

• ABA renewal – Original learning in context A, extinction in context B, test in context A → behavior returns.

• ABC renewal – Original learning in A, extinction in B, test in a novel context C → behavior returns.

• AAB renewal – Original learning and extinction both in A, test in A → no renewal. This is the most stable.

Practical implication: To minimize renewal, conduct extinction (and retraining) in the same contexts where the behavior originally occurred, or in multiple contexts (context‑fading). Train the alternative behavior in as many relevant settings as possible.

7.2 Resurgence

Resurgence occurs when an extinguished behavior returns after a newer, reinforced behavior is itself extinguished.

Example:

• Original behavior: dog jumps on counter to get food (extinguished).

• New behavior: dog sits on mat to get food (reinforced).

• Then, the owner stops reinforcing the “sit on mat” behavior (extinction of the new behavior). The dog may revert to jumping on the counter – the original behavior.

Resurgence is important in training because it shows that extinguishing a desirable replacement behavior can inadvertently bring back the problem behavior. Always ensure the replacement behavior remains reinforced (even intermittently) until the original behavior is truly dormant.

7.3 Reinstatement

Reinstatement occurs when a single unpaired presentation of the original unconditioned stimulus (US) or reinforcer causes the extinguished response to return.

Example: A dog that has stopped reacting to the mail carrier (extinction) sees the mail carrier drop a treat by accident (unpaired US). The dog may suddenly bark again.

Reinstatement shows how sensitive extinction is to unexpected rewards. In training, it means that even one accidental reinforcement (e.g., a guest giving attention after the owner asked them to ignore the dog) can temporarily reinstate the old behavior.

7.4 A Unifying Theoretical Framework

Spontaneous recovery, renewal, resurgence, and reinstatement can be understood theoretically as different expressions of retrieval competition between excitatory and inhibitory memory traces. From this perspective, relapse phenomena do not require separate mechanisms but can be understood as variations in retrieval conditions (Bouton, 2002). The original association is never deleted; it remains available and can be expressed when conditions favor its retrieval (e.g., after time, after a context change, after extinction of a competing behavior, or after an unexpected reward). This framework highlights why extinction is inherently fragile unless actively maintained and generalized.

For strategies to manage frustration during extinction bursts, see neurobiology of frustration in dogs.

8. The Inhibitory Learning Model (Craske et al., 2014) – An Evidence‑Based Upgrade

Modern clinical and experimental approaches (e.g., Craske et al., 2014) conceptualize extinction primarily as optimizing inhibitory learning, not as simply reducing the original response. The goal is not merely to make the undesirable behavior disappear, but to maximize the retrievability of the inhibitory memory trace across contexts and over time.

A key driver of inhibitory learning is expectancy violation: extinction trials are most effective when the expected outcome (e.g., reward or threat) fails to occur in a salient and noticeable way. The dog must actively perceive that the old contingency has changed; otherwise, it may learn only that reinforcement is occasionally absent, not that it has permanently stopped.

According to this model, effective extinction training should:

• Expect and normalize spontaneous recovery – It is a sign that inhibition is competing with excitation, not a failure.

• Vary the extinction context – Conduct extinction trials in multiple environments to reduce renewal.

• Occasionally “retrieve” the extinction memory – Brief, spaced reminders (e.g., a single extinction trial) can strengthen the long‑term accessibility of inhibition.

• Combine extinction with alternative reinforcement – Differential reinforcement of incompatible behaviors builds a new, positive habit that supports inhibition.

For dog trainers, this means: rather than aiming to “eliminate” a problem behavior through repeated non‑reinforcement alone, actively structure the learning environment to make the inhibitory memory trace as strong and context‑independent as possible, and ensure that the dog experiences clear expectancy violations.

9. Factors That Influence Extinction in Dogs

Not all extinctions are equally effective. Several factors determine how quickly a behavior extinguishes and how likely it is to return.

9.1 Reinforcement Schedule Before Extinction

Behavior maintained on a continuous reinforcement schedule (every response is reinforced) extinguishes quickly – but also recovers more strongly. Behavior maintained on a partial (intermittent) reinforcement schedule (e.g., variable ratio) is more resistant to extinction (the partial reinforcement extinction effect, PREE) but once extinguished, it is less likely to show spontaneous recovery because the dog is already used to periods without reward.

Practical takeaway: To build durable, relapse‑resistant behaviors, train on variable reinforcement schedules early. Do not rely solely on continuous reinforcement.

9.2 Extinction Burst Management

Extinction bursts can be intense and may include frustration‑based aggression. If the burst is too severe, the owner may accidentally reinforce the behavior by giving in (e.g., “The dog barked louder, so I opened the door.”). This teaches the dog that persistence works – making extinction harder the next time.

Solution: Anticipate the burst. Have a management plan (e.g., use a barrier, leave the room, redirect to an incompatible behavior). Do not reinforce the burst.

9.3 Conditioned Reinforcers (Markers)

A clicker or other conditioned reinforcer can signal to the dog which behavior is being reinforced. In extinction, the owner stops using the marker for the undesired behavior. However, if the dog hears the marker (even accidentally) while performing the old behavior, it can serve as a reinforcer and delay extinction.

Practical tip: Keep markers precise and only use them for targeted behaviors. Avoid marking the very behavior you are trying to extinguish.

9.4 Emotional State During Extinction

Extinction is frustrating. The dog experiences negative prediction error – dopamine drops below baseline, creating a state of disappointment and, in some cases, anger. If the dog is already stressed or anxious, extinction can worsen emotional states.

Dogs with chronic stress (high cortisol) show impaired extinction learning because cortisol damages prefrontal function. Address underlying stress before using extinction.

9.5 Alternative Reinforcement (Differential Reinforcement)

Extinction is most effective and least frustrating when combined with differential reinforcement of an alternative behavior (DRA). Instead of simply letting the old behavior fade, teach and reinforce a new, incompatible behavior.

Example:

• Old behavior: jumping on guests (extinction – no attention).

• New behavior: sitting on a mat (reinforced with treats and attention).

DRA reduces frustration, speeds up extinction, and builds a positive habit that replaces the old one.

For more on chronic stress and its effects on learning, see neurobiology of chronic stress in dogs – cortisol impact.

10. Extinction in Dog Training – Practical Applications

Understanding extinction allows trainers to use it strategically, without causing emotional harm or unintended relapse.

10.1 Common Extinction Procedures in Training

• Ignoring attention‑seeking behavior – Barking, pawing, nudging. Extinction works when the owner consistently withholds all attention (eye contact, verbal response, touch) until the dog offers a different behavior.

• Stopping reinforcement for pulling on leash – The dog pulls but the handler stops walking; only forward movement (the reinforcer) is contingent on loose leash. This is extinction (pulling no longer produces movement) combined with negative reinforcement (loose leash ends the stop).

• Reducing jumping on counters – The owner removes all food from counters so that counter‑surfing never yields food. This is extinction (no reinforcement).

10.2 When NOT to Use Extinction Alone

Extinction should not be used as a standalone procedure for:

• Fear‑based aggression – Withholding reinforcement for aggression does not address the underlying fear. In fact, extinction may increase aggression (extinction burst) and worsen the dog’s emotional state.

• Compulsive behaviors – Compulsive disorders (e.g., tail chasing, flank sucking) are not maintained by external reinforcement in the same way. Extinction is ineffective; medication and behavioral therapy are required.

• Behaviors that are self‑reinforcing – Some behaviors (e.g., chasing leaves, barking at echoes) are inherently reinforcing. Extinction (e.g., removing the leaf) is often impossible.

For fear‑based aggression, focus on counter‑conditioning and desensitization, not extinction. See aversive training methods – neurological effects in dogs for why punishment and extinction alone can backfire.

10.3 Using Extinction in Separation Anxiety

Separation‑related behaviors (howling, destruction) are often driven by panic, not by external reinforcement. Extinction (e.g., ignoring the dog upon return) may worsen anxiety. Instead, treat the emotional state with systematic desensitization and counter‑conditioning.

For more, see neurobiology of separation anxiety in dogs and separation anxiety in dogs – neurobiology and panic.

11. Preventing Relapse – Building Extinction‑Resistant Behaviors

To minimize spontaneous recovery, renewal, resurgence, and reinstatement, follow these evidence‑based guidelines:

1. Train on variable reinforcement before considering a behavior “learned.” Intermittent rewards make the behavior more resistant to extinction and reduce relapse.

2. Practice extinctions in multiple contexts (different rooms, locations, with different people). This reduces renewal.

3. Use DRA (teach an incompatible behavior) and keep reinforcing it even after the old behavior has stopped.

4. Apply “extinction reminders” – periodically practice the extinction procedure (e.g., ignore a jump) even after the behavior is gone, to strengthen the inhibitory memory trace.

5. Do not allow accidental reinforcement – Enlist family members and visitors to consistently withhold reinforcement for the old behavior.

6. Monitor emotional state – If the dog shows signs of frustration, anxiety, or aggression during extinction, reduce intensity, add alternative reinforcement, or consult a professional.

12. Summary Table: Extinction and Relapse Phenomena 

Extinction (Operant)

• Definition: Withholding reinforcement for a previously reinforced behavior

• What happens: Behavior decreases over time; extinction burst + possible frustration

• Typical outcome: Not erasure; new inhibitory memory trace competes with original

Extinction (Classical)

• Definition: Presenting CS without US repeatedly

• What happens: Conditioned emotional response (fear, excitement) fades gradually

• Typical outcome: Original fear memory remains; renewal is strong; strong context‑dependence

Spontaneous Recovery

• Definition: Reappearance of extinguished behavior after rest, without new reinforcement

• Why it occurs: Time‑dependent loss of accessibility of inhibitory memory trace (shift in retrieval probability)

• Training implication: Expect it; do not punish; repeat extinction; each recovery is weaker

Renewal

• Definition: Return of extinguished behavior after context change

• Why it occurs: Extinction is context‑specific; new context restores original memory

• Training implication: Train in multiple contexts; use context‑fading

Resurgence

• Definition: Return of original extinguished behavior after a newer, reinforced behavior is extinguished

• Why it occurs: Extinction of the new behavior disinhibits the old memory

• Training implication: Keep reinforcing the replacement behavior

Reinstatement

• Definition: Return of extinguished behavior after a single unpaired reinforcer or US

• Why it occurs: Unexpected reward reactivates the original expectation

• Training implication: Prevent accidental reinforcement during extinction

Key Insights (Takeaways)

• Extinction does not erase learned behavior – It creates a new inhibitory memory trace that competes with the original. The original memory remains intact.

• Extinction is fundamentally context‑dependent – The inhibitory memory trace does not automatically generalize; train in multiple settings. This context‑dependence is a defining property, not a secondary feature.

• Spontaneous recovery is normal – After a period of rest, the behavior may return briefly. It is a temporary shift in retrieval probability, not a training failure.

• Expectancy violation drives inhibitory learning – The dog must clearly perceive that the expected outcome fails to occur. Salient, noticeable extinction trials are most effective.

• Partial reinforcement during training builds resistance – Dogs trained with variable reinforcement are less prone to relapse.

• Differential reinforcement of an alternative behavior (DRA) reduces frustration and accelerates extinction.

• Extinction can cause frustration bursts and, in some cases, aggression – Be prepared, do not reinforce the burst, and consider frustration tolerance.

• Do not use extinction alone for fear‑based behaviors – Address the underlying emotion with counter‑conditioning.

• Modern inhibitory learning theory (Craske et al.) emphasizes maximizing the retrievability of the inhibitory memory trace across contexts rather than simply reducing behavior.

Conclusion

Extinction is not a process of forgetting. Extinction therefore does not weaken the original learning in a permanent sense, but alters the conditions under which different memory traces are retrieved. It is a context‑dependent, competitive learning mechanism in which a new inhibitory memory trace temporarily overrides the original association without erasing it. This fundamental property explains why extinguished behaviors can return through spontaneous recovery, renewal, resurgence, or reinstatement – not as failures, but as predictable expressions of retrieval competition between excitatory and inhibitory memory traces.

From a predictive processing perspective, extinction reflects an update of expectancy models rather than deletion of prior knowledge. The dog does not “unlearn” that the bell predicted food; it learns that in this context, the bell now predicts something else (or nothing). The original prediction remains available.

For dog trainers and owners, understanding these phenomena transforms frustrating setbacks into manageable events. By using extinction strategically – in combination with variable reinforcement schedules, differential reinforcement of alternative behaviors, multiple training contexts, and careful management of frustration – trainers can build resilient behaviors that withstand the tests of time, context, and human error.

When the behavior does return (as it inevitably will, at least briefly), the informed trainer responds not with punishment, but with a calm repetition of the extinction procedure, knowing that each recovery grows weaker and shorter‑lived. Lasting behavior change comes not from hoping the memory disappears, but from systematically strengthening the inhibitory memory trace until it reliably outcompetes the original.

References

Bouton, M. E. (2002). Context, ambiguity, and unlearning: Sources of relapse after behavioral extinction. Biological Psychiatry, *52*(10), 976–986.

Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning & Memory, *11*(5), 485–494.

Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., & Vervliet, B. (2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, *58*, 10–23.

Pavlov, I. P. (1927). Conditioned reflexes. Oxford University Press.

Rescorla, R. A. (2001). Experimental extinction. In R. R. Mowrer & S. B. Klein (Eds.), Handbook of contemporary learning theories (pp. 119–154). Lawrence Erlbaum.

Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. Appleton-Century.