OCD Learning Models: How Is OCD Maintained? Avoidance, Negative Reinforcement, Habits, and Safety Behaviors

Author: Ukrainian Psychological Hub · Published: September 15, 2026 · Editorial Policy

Learning models of obsessive-compulsive disorder explain how symptoms can become self-maintaining through repeated interactions among triggers, distress, avoidance, compulsions, short-term relief, safety behaviors, and learned action patterns. They are maintenance models: they help explain why an OCD response can become easier to repeat even when a person knows that the ritual is excessive, unwanted, or logically unnecessary.

The central learning principle is negative reinforcement. When checking, washing, reassurance seeking, mental reviewing, avoidance, or another safety response removes or prevents an aversive state — anxiety, uncertainty, disgust, guilt, incompleteness, or an anticipated surge of distress — the response can become more likely the next time a similar cue appears. A 2026 ecological momentary assessment study followed 45 people with OCD seven times per day for a week and found that stronger compulsions were associated with changes in anxiety consistent with reinforcement processes; the study also suggested that avoiding sharp increases in anxiety may itself reinforce compulsive responding. Read Swisher and Newman’s study.

That mechanism is important, but it is not a complete theory of OCD. Modern evidence also implicates threat appraisal, intolerance of uncertainty, associative learning, fear and safety learning, impaired goal-directed control, habit-like responding, and context-dependent learning. The strongest account is therefore not that OCD is “just a bad habit,” or that every obsession was acquired by conditioning. It is that several learning processes can help preserve and automate responses after symptoms have developed. A current state-of-the-art review in The BMJ similarly describes avoidance and compulsions as processes that can prevent corrective learning about feared outcomes and about a person’s ability to tolerate uncertainty and distress. See Abramowitz and colleagues’ 2026 review.

This article focuses on those learning mechanisms. For the simpler symptom-loop explanation, see OCD Cycle: What Is It?. For the separate question of how beliefs and appraisals shape obsessive meaning, see OCD Cognitive Models.

What are learning models of OCD?

Learning models describe changes in behavior and expectation that occur through experience. In OCD, they ask questions such as: What happens after a trigger? What consequence follows a compulsion? What does avoidance prevent the person from discovering? Why can a ritual become easier to initiate after repeated use? Why can a familiar cue begin to evoke an urge even when the person’s explicit belief has changed? And what new learning occurs during effective treatment?

The classic behavioral account combined associative conditioning with operant learning. A previously neutral external or internal cue could acquire aversive significance, while escape, avoidance, and ritualized responses could be strengthened because they reduce or prevent distress. Modern formulations retain the importance of reinforcement while treating the original acquisition story much more cautiously. OCD can emerge without a single identifiable conditioning event, and not every compulsion reliably lowers anxiety. Reviews of contemporary CBT therefore use behavioral learning as one layer of a broader model rather than as a universal origin story. Law and Boisseau’s review summarizes this behavioral foundation and its relationship to exposure and response prevention.

Maintenance is different from cause

A mechanism that keeps a symptom going does not have to be the mechanism that originally caused the disorder. This distinction is essential. Genetics, neurobiology, developmental factors, cognition, stress, learning history, and other influences can contribute to OCD risk; no single learning event explains the disorder as a whole. The separate OCD Causes article covers etiological evidence.

Learning models instead concentrate on what happens after obsessions, urges, sensations, or doubts are already present. If a response repeatedly changes the immediate internal state in a way that feels useful, urgent, or protective, learning can strengthen that response. This is why a maintenance formulation can be clinically useful even when nobody can identify when the first obsession began.

Negative reinforcement: the core behavioral mechanism

Negative reinforcement means that a behavior becomes more likely because it removes, reduces, or prevents something aversive. “Negative” refers to removal; it does not mean bad, unhealthy, or punitive. “Reinforcement” means an increase in the future probability or strength of a response. Negative reinforcement is therefore different from punishment, which by definition reduces behavior.

In OCD, the removed or prevented state may be fear, doubt, guilt, disgust, a sense of contamination, a feeling of incompleteness, bodily tension, or the possibility of a future spike in distress. The reinforcing consequence may be obvious relief, but it can also be subtler: “I checked, so now I have done what I could”; “I avoided the knife, so I do not have to find out what the thought means”; “I reviewed the memory again, so I have postponed the possibility of being wrong.”

This functional perspective matters because two outwardly different actions can serve the same learning function. Rechecking a lock, asking a partner for reassurance, replaying an event mentally, searching the internet, praying according to a rigid rule, or avoiding leaving home may all function to reduce uncertainty or prevent an aversive state. Conversely, the same outward behavior can be ordinary in one context and compulsive in another. OCD is not diagnosed from the shape of an action alone.

Relief does not have to be dramatic

A common oversimplification says that every compulsion produces a large drop in anxiety. Real-world data are more complicated. In the 2026 study by Swisher and Newman, anxiety reduction was one predictor of later compulsive responding, but the authors also found evidence consistent with contrast avoidance: compulsions may be reinforced when they prevent anxiety from rising sharply rather than simply making an already-high state fall. The study therefore supports the broader idea that consequences can reinforce compulsions while refining the older “ritual equals relief” formula.

This helps explain why some people report, “The ritual barely makes me feel better anymore, but I still have to do it.” A learned response can persist even when the immediate payoff has weakened, especially after extensive repetition, cueing, rule formation, or habit-like automation.

Escape and avoidance are two routes to short-term control

Escape occurs after an aversive state is already present: a person touches a surface, feels contaminated, and washes to end the feeling. Avoidance occurs before or in anticipation of the aversive state: the person stops touching public surfaces, avoids public transportation, or delegates tasks so the contamination trigger never fully occurs. Both can be negatively reinforced when they reduce contact with distress or uncertainty.

Avoidance can become especially powerful because it removes opportunities to test predictions. If someone avoids a situation and the feared outcome does not happen, OCD can interpret the result as evidence that the avoidance worked. The person receives no direct information about what would have happened without the protective response. The OCD Avoidance article examines this process in detail.

Avoidance also has clinical relevance for treatment. In a study using data from a randomized trial, Wheaton and colleagues found that greater pretreatment avoidance predicted worse outcomes in the exposure and response prevention condition and was associated with poorer adherence to between-session exposure assignments. The sample was modest and the finding requires appropriate caution, but it supports assessing avoidance rather than measuring only obvious rituals. Read the study.

Safety behaviors: protection that can block new learning

A safety behavior is an action used to prevent, minimize, monitor, or neutralize a feared outcome while entering or remaining in a situation. In OCD, safety behaviors can include carrying cleaning supplies “just in case,” positioning oneself near an exit, covertly checking bodily sensations, rehearsing what to say before a feared interaction, seeking reassurance before acting, keeping backup records, or entering an exposure while preserving a hidden ritual.

The learning problem is not simply that a safety behavior exists. The problem is what the person learns when the feared outcome fails to occur. If the conclusion is “I was safe because I performed the protective action,” the old threat model can survive. The situation itself does not become evidence that uncertainty can be tolerated without ritualized protection.

Research also warns against a rigid all-or-nothing rule. Levy and Radomsky experimentally examined safety behaviors in contamination-related fear and found that the effects of safety behavior can depend on how it is used; questions remain about whether carefully controlled safety behavior may sometimes improve treatment acceptability without preserving threat beliefs. Read the study. In clinical ERP, the key question is therefore functional: is the behavior being used as a bridge toward fuller approach and learning, or is it serving as the condition under which the person believes exposure is safe?

Compulsions as learned responses

Compulsions are repetitive behaviors or mental acts performed in response to obsessions or according to rigid rules. In a learning formulation, a compulsion can be understood partly by its consequences. If performing it repeatedly reduces distress, postpones uncertainty, creates a brief sense of completion, or seems to prevent danger, the response gains behavioral strength.

This does not make compulsions deliberate choices in the everyday moral sense. Learning changes response probability without requiring a conscious decision to “train” the behavior. People with OCD commonly recognize that the ritual is excessive or disconnected from realistic risk while simultaneously experiencing a powerful urge to perform it. For a broader clinical description, see OCD Compulsions.

Mental rituals can be reinforced too

Learning principles apply to covert responses as well as visible actions. Replaying a conversation until it feels resolved, replacing a “bad” thought with a “good” one, silently repeating phrases, mentally checking intent, testing attraction, reviewing memory, or analyzing whether one is certain can all become reinforced if they repeatedly alter an aversive internal state.

Because mental compulsions happen internally, they can be mistaken for ordinary reflection. Function and repetition matter. Reflection is usually flexible and capable of ending when enough information exists. A compulsion is driven by the need to reduce obsessional uncertainty or distress and tends to demand another round when certainty does not hold.

Why reassurance can become part of the learning loop

Reassurance can act as an interpersonal safety behavior. A feared possibility appears, another person provides certainty, distress decreases, and reassurance becomes more likely to be requested again. The short-term interpersonal success of reassurance is exactly what can make it sticky. The learner is not only the person with OCD: family members or partners can also learn that giving reassurance quickly lowers visible distress, making accommodation more likely the next time.

This does not mean supportive relationships should become cold or withholding. Effective support shifts from answering the obsession’s demand for certainty toward helping the person tolerate uncertainty and follow an agreed treatment plan. The behavioral function of reassurance matters more than whether the sentence sounds comforting.

Stimulus control and generalization: how the trigger network can expand

Learned responses are often cue-dependent. A ritual initially linked to one stimulus can become evoked by similar stimuli, contexts, thoughts, sensations, or memories. A checking response associated with the stove can spread to plugs, locks, taps, emails, forms, and conversations. A contamination response associated with one restroom can generalize to door handles, public seating, clothing, packages, and eventually the feeling of having been near a potentially contaminated object.

Generalization helps explain why OCD can become more restrictive without any new catastrophic event. The response system begins treating a wider class of cues as relevant. Avoidance then reduces opportunities to learn where the threat boundary actually lies. This is one reason effective treatment usually aims for learning that transfers across different contexts rather than mastering one perfectly controlled exposure.

Fear conditioning and extinction: useful evidence, with limits

Classical conditioning models propose that cues can acquire threat value through association and that later encounters with those cues evoke conditioned responses. This framework has influenced exposure therapy, but laboratory evidence in OCD is not simple. Cooper and Dunsmoor’s systematic review of 12 studies found moderate evidence for abnormalities in conditioned-response acquisition and comparatively stronger evidence for impaired extinction-related processes, while also noting inconsistent findings and major methodological limitations. Read the systematic review.

Steuber and McGuire’s later systematic review likewise found minimal evidence for abnormal initial fear learning across the included studies, with more support for differences in extinction and reversal learning. Read the review. Together, these findings argue against a simplistic claim that people with OCD merely “condition fear more easily.” The more plausible research question concerns how threat and safety information are updated, retained, discriminated, and retrieved across contexts.

Extinction is new learning, not erasing the old fear

Traditional explanations of exposure often emphasized habituation: remain with a feared cue long enough and distress falls. Habituation can occur, but contemporary learning theory treats it as an incomplete explanation. Extinction is better understood as new learning that competes with the old threat association. A person can learn, in effect, “This cue can be encountered without performing the ritual, and I can tolerate what follows.” The old association may remain available, which helps explain why fear can return in new contexts, after time has passed, or under stress.

Jacoby and Abramowitz reviewed inhibitory-learning approaches specifically for OCD and argued that exposure can be optimized by emphasizing expectancy violation, variability, retrieval of new learning, and reduced reliance on safety signals rather than making within-session anxiety reduction the sole target. Read their critical review. The empirical translation of inhibitory-learning strategies to OCD remains developing, so these principles are best treated as an influential framework for optimizing ERP rather than as proof that one single mechanism explains every successful treatment.

Habits and goal-directed control

Negative reinforcement explains why a response can initially be useful in the short term. Habit models ask a later question: after enough repetition, can the response become relatively insensitive to whether the original outcome still matters? In experimental psychology, goal-directed actions are sensitive to the current value of their outcomes, whereas habits are more strongly controlled by learned stimulus-response associations.

A landmark 2011 study found that people with OCD showed more “slips of action” after outcomes were devalued, consistent with reduced goal-directed control and greater reliance on habitual responding. Gillan and colleagues, 2011. A later avoidance-learning experiment found that participants with OCD continued more avoidance responses after the relevant shock outcome had been devalued, again supporting a habit-like account. Gillan and colleagues, 2014.

The habit hypothesis has become influential because it explains a familiar clinical observation: a person can know a ritual is unnecessary and still feel pulled to perform it. Yet the evidence does not justify equating OCD with excessive habit formation. Habit tasks can measure several processes at once, and impaired goal-directed control does not automatically prove that the habit system itself is overactive.

Recent studies refine the habit account

A 2024 eLife study trained action sequences over a month and found that people with OCD did not simply form objective habits faster than healthy participants. Both groups achieved automatic performance, while the OCD group reported stronger subjective habitual tendencies and showed a preference for trained sequences under some effort-based choices. The authors argued for refinements to the simple goal/habit imbalance account. Read Banca and colleagues.

Another 2024 study using a revised slips-of-action task and computational modeling found impaired flexible goal-directed control in OCD alongside differences in learning rate, perseveration, and reinforcement sensitivity. Yu and colleagues. These findings support learning-system differences while also showing that the mechanism is richer than a single “too many habits” parameter.

The strongest current formulation is therefore graded: habit-like responding and impaired goal-directed updating are relevant for at least some people and some tasks, but they are not sufficient as a stand-alone explanation of OCD. They also appear across compulsivity and other forms of psychopathology, so they cannot be used as a diagnostic marker by themselves.

Reinforcement learning and prediction error

Modern reinforcement-learning models formalize how expectations change when outcomes differ from what was predicted. A prediction error is the discrepancy between expected and obtained outcomes; learning systems use that discrepancy to update future expectations and choices. In OCD research, computational studies examine whether people update action-outcome values, reward and punishment expectations, or model-based plans differently.

These models are scientifically useful because they separate processes that can look identical at the surface. Repeating a check might reflect strong stimulus-response control, uncertainty about the action-outcome association, exaggerated sensitivity to possible negative outcomes, difficulty updating after safe outcomes, or a combination. Computational parameters are research constructs, however; a person cannot be clinically diagnosed with “model-free OCD” from everyday behavior.

How learning models and cognitive models fit together

Behavioral and cognitive accounts answer different parts of the same maintenance problem. Cognitive models emphasize why an intrusive event becomes personally significant: inflated responsibility, overestimated threat, thought-action fusion, intolerance of uncertainty, perfectionism, or the need to control thoughts can transform an ordinary intrusion into an urgent problem. Learning models explain how the responses to that urgent problem can be strengthened through their consequences and repeated cueing.

A thought can therefore acquire high significance through appraisal, trigger a ritual, produce a short-term change in distress or uncertainty, and become embedded in a reinforced response pattern. The next intrusion arrives in a system that has already learned a preferred route to relief. For the belief side of this interaction, see OCD Cognitive Models and OCD and Uncertainty.

A worked learning formulation

Consider repeated checking. A person leaves home and experiences the thought, “Maybe I did not lock the door.” The immediate consequence is doubt and perceived responsibility. Checking the lock reduces uncertainty for a few seconds. That reduction can negatively reinforce checking. Because the person leaves only after checking, the absence of burglary is easy to attribute to having checked correctly rather than to the ordinary safety of the situation. The next episode begins with less confidence in memory and a stronger learned association between departure-related doubt and returning to the door.

Over time, additional cues can enter the network: touching the handle, walking to the elevator, seeing a key, hearing a news story about theft, or remembering a previous mistake. The checking sequence can become highly practiced and partly automatic. If the person then tries to resist, the absence of the usual response can itself feel wrong or incomplete. Several mechanisms — appraisal, negative reinforcement, stimulus generalization, memory distrust, and habit-like action control — can therefore converge on one visible symptom.

Why “knowing better” may not stop the response

Explicit knowledge and learned action control are not identical systems. Someone can believe that the probability of harm is tiny and still experience an urge to check. They can understand the concept of negative reinforcement and still feel immediate relief after ritualizing. They can know that reassurance has never produced lasting certainty and still seek it under pressure. This gap between propositional knowledge and action tendency is one reason psychoeducation alone is rarely sufficient for established OCD.

Learning changes through experience. Treatment therefore asks the person to practice a different response in the presence of the trigger: approach rather than avoid, allow uncertainty rather than resolve it, and refrain from the ritual long enough for new learning to occur.

How ERP targets learning mechanisms

Exposure and response prevention (ERP) deliberately combines contact with obsessional triggers with prevention of the compulsive or safety response. NICE recommends CBT that includes ERP as a core psychological treatment for OCD, with intensity matched to impairment and preference. See the NICE guideline. A 2021 systematic review and meta-analysis of 36 randomized trials involving 2,020 participants found a large pooled advantage for CBT with ERP over all control conditions, while also showing that effect size depended strongly on the comparator and highlighting methodological limitations in the literature. Reid and colleagues.

From a learning perspective, the “response prevention” component is crucial because exposure performed while preserving the same neutralizing strategy can leave the maintenance contingency intact. The person needs opportunities to encounter the cue, experience uncertainty or discomfort, omit the ritual, and discover what happens. That discovery can include several forms of new learning: the feared event does not occur, the probability was overestimated, distress can change without ritualizing, uncertainty can be carried, a mistake can be survivable, and the urge itself can rise and fall.

For the full treatment protocol, evidence base, safety considerations, and what sessions involve, see ERP for OCD. The present article explains the learning logic rather than duplicating the treatment guide.

ERP is not a test that must produce immediate calm

If successful exposure is defined only as “anxiety went down during the exercise,” a person may begin monitoring anxiety as another safety rule. Contemporary learning approaches instead emphasize willingness to experience the trigger without ritualizing and the acquisition of flexible new expectations. Distress often decreases with practice, but within-session habituation is not a requirement for every useful exposure.

This is especially relevant to OCD presentations dominated by disgust, guilt, uncertainty, incompleteness, or taboo thoughts, where the target experience may not behave like a simple fear curve. Learning can occur even when a feeling remains uncomfortable at the end of a practice period.

Expectancy violation and behavioral experiments

One way to strengthen learning is to identify the prediction that makes a ritual seem necessary and then create a safe, ethically appropriate test in which the ritual is withheld. The outcome can update more than probability estimates. A person may learn, “I cannot get absolute certainty, and I can still continue with my day,” which is a different learning target from proving that nothing bad can ever happen.

This distinction protects ERP from becoming another certainty-seeking exercise. Real life contains ordinary risk. Treatment aims to reduce obsessive-compulsive responding to uncertainty, not to promise perfect safety.

What about medication and learning?

Learning models primarily explain behavioral maintenance and psychological treatment mechanisms. Medication can reduce OCD symptoms through biological mechanisms and may make behavioral change more feasible for some people, but it should not be described as simply “turning off reinforcement.” Psychological and pharmacological treatments operate at different explanatory levels and can be combined when clinically appropriate. For that decision-making context, see OCD Combination Treatment.

A 2025 systematic review of ERP combined with psychological add-ons identified approaches including inhibitory-learning strategies, motivational interviewing, mindfulness-based methods, cognitive interventions, and family-accommodation work, while emphasizing the need for more evidence about which additions improve outcomes beyond ERP itself. Faustino and colleagues. This is a reminder that a plausible learning mechanism is not automatically a proven treatment enhancement.

What learning models explain well

Learning models are especially strong at explaining persistence. They show why a behavior that produces an immediate protective consequence can survive despite long-term costs; why avoidance can preserve untested threat beliefs; why reassurance can become repetitive; why rituals can spread to new cues; why responses can become increasingly automatic; and why treatment requires new behavior in the presence of old triggers rather than explanation alone.

They also provide a common functional language across symptom themes. Contamination washing, checking, mental review, confession, reassurance seeking, ordering, and avoidance can look unrelated at the level of content while sharing reinforcement processes at the level of function.

What learning models do not explain by themselves

Learning models do not fully explain why a particular person develops OCD, why one theme becomes dominant, why intrusive content often targets personally important domains, why symptom severity fluctuates with broader biological and psychosocial factors, or why individuals differ in insight and comorbidity. They also do not turn every repeated or negatively reinforced behavior into OCD.

The habit literature illustrates the broader limitation. Group-level differences in goal-directed control are scientifically meaningful, yet they overlap with transdiagnostic compulsivity and other conditions. A laboratory habit effect cannot establish an OCD diagnosis, and a person with OCD does not have to show a measurable habit deficit on every task. Likewise, evidence of extinction-learning differences does not mean that every individual with OCD has a unitary “extinction defect.”

Learning models are not diagnostic tests

A negative-reinforcement loop can occur in many forms of human behavior. Avoidance appears in anxiety disorders, trauma-related disorders, chronic pain, and ordinary coping. Habits are universal. Safety behaviors can be adaptive in genuinely dangerous situations. A person may also repeat an action because of preference, routine, sensory regulation, a tic, a medical concern, or another condition.

OCD diagnosis requires a clinical assessment of obsessions and/or compulsions, distress or impairment, time consumption, context, differential diagnosis, and other relevant factors. Learning mechanisms can support a case formulation after assessment; they cannot substitute for diagnosis.

How to recognize a maintenance contingency

A useful functional question is not “Does this look strange?” but “What changes immediately after the response, and what does that consequence teach the system?” If an action repeatedly produces a brief reduction in doubt, blocks a feared sensation, postpones uncertainty, secures reassurance, or creates a sense of completion, it may be part of a maintenance contingency. The next question is whether repeated use narrows behavior, increases dependence on the response, or prevents disconfirming experience.

This kind of analysis is more informative than debating whether a behavior is “really a habit” in everyday language. Clinical formulation focuses on the relationship among trigger, prediction, response, immediate consequence, and longer-term learning.

Frequently asked questions

What is the main learning model of OCD?

The classic behavioral model proposes that obsession-related cues evoke distress and that compulsions or avoidance are maintained because they reduce or prevent that distress through negative reinforcement. Modern learning accounts add safety learning, generalization, habit and goal-directed control, reinforcement learning, and context-dependent extinction.

What is negative reinforcement in OCD?

Negative reinforcement occurs when an OCD response becomes more likely because it removes, reduces, or prevents something aversive. The aversive state may be anxiety, doubt, disgust, guilt, incompleteness, uncertainty, or an anticipated rise in distress. Negative reinforcement is not punishment.

Do compulsions always reduce anxiety?

No. Relief can be partial, delayed, brief, or absent, and some compulsions can increase distress. Recent ecological evidence suggests that reinforcement may include both decreases in anxiety and prevention of sharp anxiety increases, which is one reason the simple “compulsion always lowers anxiety” model is too narrow. Swisher and Newman, 2026.

Is avoidance a compulsion?

Avoidance can function as a compulsive or safety response when it is used to prevent obsessional distress or feared consequences, but avoidance is not automatically OCD. Its clinical meaning depends on function, context, rigidity, impairment, and the wider symptom pattern. See OCD Avoidance.

Are safety behaviors always harmful?

No universal rule fits every situation. Real safety behavior is necessary in real risk, and some therapy research has explored carefully controlled temporary safety behaviors. In OCD treatment, the concern is whether a behavior teaches that the situation was manageable only because the safety behavior was present, thereby preserving dependence on it.

Is OCD a habit disorder?

Habit-like responding is an important research model of compulsivity, and several studies have found impaired goal-directed control or excessive persistence after outcome devaluation in OCD. Newer work complicates a simple “excess habit” explanation. Habit mechanisms are best understood as one component of OCD rather than a complete definition of the disorder.

Why do I still feel an urge when I know the ritual makes no sense?

Explicit beliefs and learned action tendencies can diverge. A highly practiced response can be strongly cued even when conscious knowledge says it is unnecessary. Negative reinforcement, stimulus-response learning, uncertainty, and habit-like control can all contribute to that gap.

Does resisting a compulsion erase the fear association?

Usually it is more accurate to think in terms of new learning rather than erasure. Exposure with response prevention creates alternative associations and expectations that can compete with older threat learning. Because old learning can remain available, practice across contexts and relapse-prevention planning can matter.

Is habituation necessary for ERP to work?

No. Anxiety reduction during exposure can occur, but contemporary inhibitory-learning approaches emphasize expectancy violation, flexibility, and retrieval of new learning rather than requiring distress to fall during every exercise. Jacoby and Abramowitz, 2016.

Can reassurance reinforce OCD?

Yes, when reassurance repeatedly resolves obsessional uncertainty or distress, it can become negatively reinforcing and increasingly sought. Support can remain warm and responsive while declining to provide repeated certainty to the obsession.

Can mental reviewing be negatively reinforced?

Yes. A covert behavior can be reinforced if it changes an aversive state. Replaying a memory, checking intent, neutralizing a thought, or testing feelings may therefore function like a visible ritual when it is repeatedly used to obtain certainty or relief.

Do learning models explain the cause of OCD?

They contribute to understanding both acquisition and persistence, but their strongest clinical role is explaining maintenance. OCD is multifactorial. Conditioning, reinforcement, and habit processes interact with cognitive, biological, developmental, and contextual influences rather than replacing them.

What treatment directly targets these learning processes?

ERP directly targets avoidance, rituals, and safety responses by combining planned exposure to obsessional triggers with prevention of compulsive responding. This creates opportunities for corrective and inhibitory learning. ERP for OCD explains the treatment itself in detail.

The central point

OCD can persist because the responses used to obtain immediate safety, certainty, relief, or completion can teach the system to use those responses again. Avoidance blocks information. Compulsions can be negatively reinforced. Safety behaviors can preserve the belief that protection was necessary. Repetition can strengthen cue-response patterns, while impaired goal-directed updating may make those patterns harder to revise. Treatment changes the learning environment by allowing contact with triggers and uncertainty without completing the old protective sequence.

The scientific picture is plural rather than singular. Negative reinforcement has strong theoretical and growing ecological support; avoidance is clinically consequential; extinction and inhibitory learning help explain exposure; and habit research identifies meaningful but nonexclusive abnormalities in goal-directed control. Together, these mechanisms explain why OCD can feel simultaneously irrational, urgent, repetitive, and resistant to reassurance — and why new learning requires new behavior.

Related Articles

• OCD Cognitive Models: How Is OCD Explained? Responsibility, Threat, Thought-Action Fusion, and Beliefs

• ERP for OCD: What Is Exposure and Response Prevention? How It Works, Evidence, and What Treatment Involves

• OCD Avoidance: What Is It? How Avoiding Triggers Maintains Obsessive-Compulsive Symptoms

References

Abramowitz, J. S., Abramovitch, A., McKay, D., & Draffin, A. (2026). Management of obsessive-compulsive disorder in adults. BMJ, 392, e083443. https://doi.org/10.1136/bmj-2024-083443

Banca, P., Herrojo Ruiz, M., Gonzalez-Zalba, M. F., Biria, M., Marzuki, A. A., Piercy, T., Sule, A., Fineberg, N. A., & Robbins, T. W. (2024). Action sequence learning, habits, and automaticity in obsessive-compulsive disorder. eLife, 12, RP87346. https://doi.org/10.7554/eLife.87346

Cooper, S. E., & Dunsmoor, J. E. (2021). Fear conditioning and extinction in obsessive-compulsive disorder: A systematic review. Neuroscience & Biobehavioral Reviews, 129, 75–94. https://doi.org/10.1016/j.neubiorev.2021.07.026

Faustino, D., Braga, R., Faria, M. J., Gonçalves, M. M., & Oliveira, J. T. (2025). A systematic review on how to combine exposure and response prevention with add-ons for the treatment of obsessive-compulsive disorder. Psychotherapy, 62(2), 132–143. https://doi.org/10.1037/pst0000560

Gillan, C. M., Papmeyer, M., Morein-Zamir, S., Sahakian, B. J., Fineberg, N. A., Robbins, T. W., & de Wit, S. (2011). Disruption in the balance between goal-directed behavior and habit learning in obsessive-compulsive disorder. American Journal of Psychiatry, 168(7), 718–726. https://doi.org/10.1176/appi.ajp.2011.10071062

Gillan, C. M., Morein-Zamir, S., Urcelay, G. P., Sule, A., Voon, V., Apergis-Schoute, A. M., Fineberg, N. A., Sahakian, B. J., & Robbins, T. W. (2014). Enhanced avoidance habits in obsessive-compulsive disorder. Biological Psychiatry, 75(8), 631–638. https://doi.org/10.1016/j.biopsych.2013.02.002

Jacoby, R. J., & Abramowitz, J. S. (2016). Inhibitory learning approaches to exposure therapy: A critical review and translation to obsessive-compulsive disorder. Clinical Psychology Review, 49, 28–40. https://doi.org/10.1016/j.cpr.2016.07.001

Law, C., & Boisseau, C. L. (2019). Exposure and response prevention in the treatment of obsessive-compulsive disorder: Current perspectives. Psychology Research and Behavior Management, 12, 1167–1174. https://doi.org/10.2147/PRBM.S211117

Levy, H. C., & Radomsky, A. S. (2016). Are all safety behaviours created equal? A comparison of novel and routinely used safety behaviours in obsessive-compulsive disorder. Cognitive Behaviour Therapy, 45(5), 367–379. https://doi.org/10.1080/16506073.2016.1184712

National Institute for Health and Care Excellence. (2005, current online guidance). Obsessive-compulsive disorder and body dysmorphic disorder: Treatment (CG31), Recommendations. https://www.nice.org.uk/guidance/cg31/chapter/Recommendations

Reid, J. E., Laws, K. R., Drummond, L., Vismara, M., Grancini, B., Mpavaenda, D., & Fineberg, N. A. (2021). Cognitive behavioural therapy with exposure and response prevention in the treatment of obsessive-compulsive disorder: A systematic review and meta-analysis of randomised controlled trials. Comprehensive Psychiatry, 106, 152223. https://doi.org/10.1016/j.comppsych.2021.152223

Steuber, E. R., & McGuire, J. F. (2022). A systematic review of fear learning, extinction learning, and reversal learning in obsessive-compulsive disorder: Implications for treatment. Journal of Clinical Psychiatry, 83(6), 21r14432. https://doi.org/10.4088/JCP.21r14432

Swisher, V. S., & Newman, M. G. (2026). Why compulsions persist: An ecological momentary assessment study of the reinforcement of compulsions. Journal of Affective Disorders, 394, 120530. https://doi.org/10.1016/j.jad.2025.120530

Wheaton, M. G., Gershkovich, M., Gallagher, T., Foa, E. B., & Simpson, H. B. (2018). Behavioral avoidance predicts treatment outcome with exposure and response prevention for obsessive-compulsive disorder. Depression and Anxiety, 35(3), 256–263. https://doi.org/10.1002/da.22720

Yu, Q., Gao, F., Li, C., Xia, J., Cao, Y., Wang, X., Xiao, C., Lu, J., Liu, Q., Fan, J., & Zhu, X. (2024). Compulsion is associated with impaired goal-directed and habitual learning and responding in obsessive-compulsive disorder. International Journal of Clinical and Health Psychology, 24(4), 100531. https://doi.org/10.1016/j.ijchp.2024.100531