The Renewal Effect Experiment (Context and Extinction) – Mark Bouton and Robert Bolles

The quest to understand how organisms acquire, maintain, and extinguish learned behaviors represents one of the foundational enterprises of experimental psychology and behavioral neuroscience. For the better part of the twentieth century, associative learning theory operated under a mechanistic assumption inherited from early behaviorism: that learning reflects the continuous formation of associative bonds between stimuli and responses, and that the cessation of reinforcement leads to the progressive erosion or unlearning of those very bonds. Within this paradigm, extinction was conceptualized largely as an erasure process—a systematic uncoupling of a conditioned stimulus from its unconditioned outcome until the memory trace vanished into associative neutrality.

This tidy theoretical formulation was fundamentally disrupted in the late 1970s by a series of revolutionary investigations conducted by Mark Bouton and his mentor Robert Bolles. In their seminal 1979 paper, “Contextual control of the extinction of conditioned fear,” published in the Journal of Experimental Psychology: Animal Behavior Processes, Bouton and Bolles demonstrated that an extinguished fear response reliably and robustly re-emerges when an animal is tested outside the spatial and sensory environment in which extinction occurred. This empirical breakthrough, formally christened the Renewal Effect, decisively refuted the unlearning hypothesis. It demonstrated that extinction does not erase an underlying associative memory; rather, extinction generates a secondary, context-dependent inhibitory memory that suppresses the primary learned response only within the specific environmental context where safety was experienced.

The implications of Bouton and Bolles’ discovery reverberate across contemporary neuroscience, cognitive psychology, and clinical psychiatry. By establishing the pervasive role of background contextual cues—ranging from physical chambers and interoceptive physiological states to temporal intervals and cognitive mindsets—the renewal paradigm transformed our comprehension of memory architecture, occasion setting, and behavioral relapse. Today, the renewal effect stands as the primary theoretical framework explaining why exposure-based therapies for phobias, post-traumatic stress disorder (PTSD), and substance use disorders frequently suffer from high rates of clinical relapse once patients leave the sterile confines of the therapist’s office. This treatise provides an exhaustive, multifaceted examination of the renewal effect, tracing its historical roots, experimental architecture, neurobiological substrates, computational models, and profound translational significance.

1. Historical Foundations of Conditioning and the Problem of Extinction

1.1 Pavlovian Conditioning and Early Formulations of Extinction

The foundational principles of classical conditioning were codified in the laboratory of Russian physiologist Ivan Pavlov during his extensive investigations into digestive secretions in canines. Pavlov observed that when an emotionally neutral event—such as the sound of a metronome or the onset of a visual marker—consistently preceded an biologically significant event like the delivery of meat powder, the neutral event gradually acquired the capacity to elicit a conditional reflex (salivation). Pavlov termed the predictive signal the conditioned stimulus (CS) and the biologically imperative event the unconditioned stimulus (US), designating the newly formed reaction the conditioned response (CR). While Pavlov’s demonstration of acquisition captured the imagination of Western psychology, his conceptualization of the opposite phenomenon—extinction—proved remarkably nuanced yet historically misunderstood.

When the CS was repeatedly presented in the complete absence of the US, the amplitude of the CR progressively attenuated until it was indistinguishable from baseline behavior. Pavlov vehemently rejected the simplistic notion that extinction represented passive forgetting or the physical destruction of the underlying neural connection. Instead, he formulated the doctrine of internal inhibition, proposing that extinction represented an active, physiologically energetic process wherein the cortex established an inhibitory barrier that temporarily suppressed the excitatory reflex. To support this view, Pavlov documented the spontaneous return of the conditional reflex following a period of rest—a phenomenon he labeled spontaneous recovery—as well as the immediate release of the extinguished response when a novel, startling stimulus was introduced alongside the CS, an effect known as disinhibition.

Despite Pavlov’s prescient theoretical distinctions, the ascendancy of radical and methodological behaviorism in North America—championed by figures such as John B. Watson, Clark Hull, and B. F. Skinner—gradually flattened the understanding of extinction into a mechanical model of associative decay or unlearning. Mid-twentieth-century learning theorists largely treated experimental environments as neutral, interchangeable backdrops against which discrete, isolated stimuli interacted. Within standard stimulus-response (S-R) formulations, if an animal ceased responding during non-reinforced trials, it was assumed that the associative weight connecting the stimulus trace to the response drive had been neutralized or driven to an asymptotic baseline of zero. Background environmental cues were systematically disregarded as passive static, methodologically controlled through environmental isolation but conceptually ignored in associative equations.

This persistent unlearning consensus, however, was perpetually haunted by empirical anomalies that defied reductionist accounts. In addition to Pavlov’s spontaneous recovery, early animal researchers intermittently noted that animals trained in one apparatus, extinguished in another, and returned to the first often showed unexpected resurgences of conditioned responding. Yet, lacking a cohesive theoretical model that elevated the physical environment from passive scenery to an active cognitive variable, the field relegated these anomalies to experimental error or methodological noise. The prevailing view maintained that memory was an absolute, unitary quantity: once uncoupled from reinforcement, its behavioral expression was permanently diminished across all settings.

1.2 Robert Bolles and the Evolution of Associative Learning Theory

The intellectual groundwork necessary to challenge this mechanistic orthodoxy was laid by Robert C. Bolles in his pioneering critiques of animal learning paradigms during the late 1960s and 1970s. Bolles fundamentally disrupted the behaviorist dogma of equipotentiality—the assumption that any perceivable stimulus could be conditioned to any arbitrary motor response with equal facility. In his landmark 1970 paper on Species-Specific Defense Reactions (SSDRs), Bolles argued that animals do not enter conditioning chambers as blank slates (tabulae rasae) waiting for arbitrary S-R habits to be stamped in by reinforcement. Instead, evolutionary selection pressures pre-configure organisms with an innate repertoire of defensive behaviors—such as freezing, fleeing, or fighting—that are automatically triggered when danger is anticipated.

Bolles’ theoretical pivot was explicitly cognitive and ethological. He contended that animals do not merely acquire mechanical motor habits; rather, they learn expectancies about the causal structure of their environment. In a defensive conditioning scenario, an animal learns an S-S* expectancy—namely, that a specific environmental cue predicts the occurrence of an aversive unconditioned stimulus (such as an electric footshock). By re-conceptualizing learning as the acquisition of informational expectancies, Bolles opened the door to investigating the organism’s broader cognitive representation of its surroundings. He argued that animals continuously process background environmental variables, forming expectations not just about isolated cues, but about the spatial and sensory context in which survival threats materialize.

It was within this intellectual crucible at the University of Washington that Bolles mentored a young Mark E. Bouton. Bolles recognized that traditional associative paradigms suffered from a severe methodological and conceptual limitation: they failed to separate the direct associative value of focal stimuli from the modulatory influence exerted by the contextual background. Bolles urged Bouton to interrogate how background stimuli interact with discrete target signals during both the acquisition and extinction of fear. Bolles perceived that if extinction were merely unlearning, an animal’s learned safety should be unconditional and ubiquitous. However, if learning involved structural cognitive representations of the environment, then shifts in the physical background could expose fundamental asymmetries in how acquired fear and newly learned safety are cataloged, stored, and retrieved by the nervous system.

The collaboration between Bolles and Bouton marked a turning point in the philosophy of experimental psychology. Together, they designed paradigms that treated the conditioning chamber not as an empty vacuum designed to showcase isolated reflexes, but as an explicit sensory matrix composed of unique tactile, olfactory, visual, and auditory elements. By methodically manipulating these background configurations independent of the discrete CS, they sought to isolate the contextual boundaries of memory retention, laying the architectural foundation for modern context-dependent retrieval theory.

1.3 The Theoretical Impasse: Erasure Versus Retrieval Failure

By the early 1970s, associative learning theory had reached an apex of mathematical formalization with the publication of the Rescorla-Wagner Model (1972). The model presented an elegant mathematical account of Pavlovian conditioning based on the discrepancy between the expected outcome and the actual outcome of an experience—a concept formalized as the prediction error ($\Delta V = \alpha \beta (\lambda – \sum V)$). When a novel CS is paired with a US, the associative strength ($V$) increases because the presentation of the US exceeds expectations ($\lambda > \sum V$). Crucially, during extinction, the model dictated that the associative strength of the CS must decrease:

$$\Delta V = \alpha \beta (0 – V) = -\alpha \beta V$$

Because the expected US does not occur ($lambda = 0$), the discrepancy between expectation and reality produces a negative prediction error. This explicitly subtracts associative weight from the CS, trial by trial, until $V$ returns to zero. In the pristine mathematics of the Rescorla-Wagner formulation, extinction was unequivocally operationalized as unlearning—the literal, algebraic erasure of the previously acquired associative weight. If two stimuli competed for associative strength, or if a stimulus was extinguished, its internal associative link was systematically depleted. The model possessed no native mechanism to represent an intact CS-US memory trace existing simultaneously beneath an extinguished behavioral surface, nor did it assign an independent retrieval or occasion-setting role to background contextual cues unless those cues were modeled simply as additional discrete stimuli summing linearly with the CS.

Yet, the persistence of spontaneous recovery stood as a continuous empirical rebuke to the Rescorla-Wagner model’s core premise. If an associative connection had been degraded to zero associative weight, the passage of an arbitrary time interval should theoretically possess no power to restore it; zero multiplied by time remains zero. To rescue erasure models, theorists frequently posited ad-hoc mechanisms, such as incomplete extinction, temporary habituation of neural pathways, or statistical fluctuations in baseline responding. None of these accounts, however, provided a unified, falsifiable cognitive mechanism that explained how an apparently extinguished memory could instantly reconstitute itself without further training trials.

Bouton and Bolles entered this theoretical fray by proposing a paradigm shift: they hypothesized that extinction might represent a profound retrieval failure rather than an associative unlearning event. Drawing inspiration from human memory research—specifically the encoding specificity principle of Endel Tulving and retrieval cue paradigms—they postulated that the physical context acts as an interpretive filter or retrieval cue. If an organism acquires fear in one context and undergoes extinction in another, the extinguished performance might simply reflect the animal’s reliance on the local contextual cues of the extinction setting to retrieve an inhibitory “safety” memory. If the animal were removed from that extinction setting, the inhibitory retrieval cue would be lost, potentially unmasking the intact, dormant CS-US excitatory association. This theoretical formulation framed the 1979 investigation: an experimental challenge designed to determine whether extinction permanently alters the target associative trace or merely leaves it behaviorally latent, awaiting the appropriate contextual catalyst to reignite.

2. The Seminal 1979 Bouton and Bolles Experiment

2.1 Experimental Architecture: The ABA Paradigm

To provide empirical verification for their retrieval hypothesis, Bouton and Bolles developed the classic ABA renewal paradigm, an experimental design of exceptional elegance and methodological rigor. Published in their 1979 article, the experiment utilized male Sprague-Dawley rats as subjects and was divided into three distinct chronological phases: Fear Acquisition (Phase 1), Extinction (Phase 2), and Testing (Phase 3). The core operational hypothesis was straightforward: if extinction erases the underlying fear memory, the response to the extinguished CS should remain completely extinguished regardless of where the testing occurs. Conversely, if extinction is context-dependent, returning the animal to the original acquisition context (Context A) following extinction in a secondary context (Context B) should trigger a dramatic rebound of the conditioned fear response.

In Phase 1, subjects were placed in Context A and subjected to classical fear conditioning. The conditioned stimulus (CS) was an audiovisual cue (such as an intermittent tone paired with a visual change), and the unconditioned stimulus (US) was a mild, brief electric shock delivered through the grid floor of the chamber. Over multiple pairings, the animals formed a robust Pavlovian association between the CS and the shock. To quantitatively measure fear with high resolution, Bouton and Bolles utilized the conditioned suppression paradigm (originally developed by Estes and Skinner in 1941). The animals had previously been trained to press a lever for an appetitive food reward on a variable-interval schedule. When an animal experiences fear, its innate freezing behavior competes with and suppresses this ongoing, appetitive operant lever-pressing behavior. Thus, the degree of suppression directly indexes the intensity of the internal fear state.

In Phase 2, the subjects were transferred to an entirely distinct physical environment, Context B. Over a series of days, the animals were exposed to systematic extinction training: the CS was presented repeatedly across dozens of trials without any shock delivery. Over time, the rate of lever pressing during CS presentations rebounded to baseline levels, demonstrating that the behavioral fear response had been fully extinguished within the confines of Context B. At the conclusion of this phase, the animals exhibited no freezing to the CS; by all traditional operational metrics, the fear had been eliminated.

The decisive test occurred in Phase 3. The rats were divided into distinct experimental and control cohorts. The experimental group was returned to the original acquisition environment, Context A, and presented with the non-reinforced CS (the classic ABA design). A primary control group remained in the extinction environment, Context B, and received identical non-reinforced presentations of the CS (an ABB design). Crucially, no shocks were administered during this testing phase to either group. The experiment was designed to isolate a single independent variable: the match or mismatch between the testing environment and the extinction environment.

2.2 Physical and Sensory Manipulation of Contextual Environments

A critical challenge in experimental psychology is ensuring that the manipulation of “context” constitutes a genuinely holistic, multidimensional sensory environment, rather than a single confounding discrete cue. Bouton and Bolles instituted meticulous environmental controls to construct two profoundly differentiated contextual ecosystems, Context A and Context B, ensuring that animals could clearly distinguish between the two spaces across all major sensory modalities.

Tactile features of the chambers were starkly contrasted to ensure distinct somatic feedback during locomotion and lever pressing. One context featured standard, cold, parallel stainless-steel grid rods across the entire floor, while the alternative context featured a solid plexiglass floor covered with a thick, textured layer of aromatic cedar or pine wood chips. Furthermore, the structural walls of the chambers were modified: one setting utilized smooth, transparent acrylic walls, while the other incorporated rough, opaque, diagonally striped aluminum panels that completely transformed the tactile and visual texture of the enclosure.

Olfactory and visual differentiations were systematically maintained to engage the highly sensitive sensory apparatus of the rat. Context A was scented with a distinct volatile compound, such as a 10% solution of acetic acid (vinegar) wiped down across the catch pans prior to each session, creating a sharp, pungent olfactory backdrop. In contrast, Context B was infused with an entirely different essence, such as an anise (licorice) extract or a mild peppermint solution. Visually, ambient illumination was sharply differentiated: Context A operated under standard overhead fluorescent room lighting, whereas Context B was shrouded in near-darkness, illuminated only by a dim, localized 28-volt red light bulb positioned directly above the lever mechanism.

Auditory backgrounds were meticulously engineered to prevent external environmental acoustic bleed from serving as a shared contextual cue. One context incorporated a continuous, high-amplitude white noise generator operating at 70 decibels, providing a flat acoustic masking background. The alternative chamber lacked this white noise, utilizing instead the distinct, rhythmic mechanical hum of an exhaust ventilation fan operating at a substantially lower pitch and acoustic profile. Finally, strict hygiene and cleaning protocols were enforced to eliminate contextual cross-contamination: the experimenters used separate cleaning agents, separate transport cages, and even different holding rooms between sessions to ensure that lingering odors or handling artifacts could not bridge the psychological barrier between Context A and Context B.

2.3 Empirical Findings and the Demonstration of Renewal

The empirical results published by Bouton and Bolles in 1979 delivered a decisive verdict that permanently destabilized classical unlearning theory. During the Phase 3 testing sessions, animals tested in the extinction context (the ABB control group) continued to demonstrate complete extinction: their lever-pressing rates remained entirely uninterrupted during the presentation of the non-reinforced CS, indicating a persistent absence of behavioral fear. However, when the experimental subjects were presented with the non-reinforced CS within the original acquisition context (the ABA group), they exhibited a massive, statistically robust return of conditioned suppression. The animals immediately ceased pressing the food lever, oriented away from the feeder, and engaged in profound behavioral freezing throughout the duration of the CS presentation.

This resurgence occurred despite the fact that the CS had been systematically extinguished to absolute zero responding only 24 hours prior in Context B. Because this testing occurred in the complete absence of any shock presentations, the sudden reappearance of the fear response could not be attributed to new learning or re-acquisition. Bouton and Bolles termed this rebound the Renewal Effect. The experiment empirically verified that the conditioned stimulus had never lost its intrinsic associative capacity to evoke a mental representation of the aversive unconditioned stimulus; rather, the expression of that association had merely been suppressed by the specific environmental backdrop of Context B.

Critically, Bouton and Bolles conducted exhaustive statistical verifications to prove that the renewed fear was not simply an artifact of baseline fear conditioned to Context A itself. Because the shock had been delivered within Context A during Phase 1, it was mathematically plausible that Context A had acquired its own direct excitatory associative strength (context conditioning), which might merely summate with a partially extinguished CS to produce the observed suppression. By measuring the pre-CS rates of lever pressing—the animal’s operant baseline immediately before the CS onset—they demonstrated that animals were pressing the lever actively and consistently in Context A prior to the CS presentation. The suppression of lever pressing was triggered specifically, exclusively, and instantaneously by the onset of the CS itself. The renewal effect was an associative phenomenon of the target conditioned stimulus, modulated entirely by the cognitive shift in environmental context.

3. Typology of Renewal Paradigms: ABA, ABC, and AAB

3.1 The Classical ABA Renewal Design

Following the seminal 1979 breakthrough, Mark Bouton and his contemporaries sought to map the boundary conditions and operational variations of contextual resurgence. The classical ABA renewal paradigm remains the most widely recognized, structurally robust, and historically impactful design within this literature. In this architecture, associative acquisition occurs in Context A, extinction is conducted systematically within Context B, and final non-reinforced behavioral testing takes place back within Context A. Methodologically, the design serves as an exact experimental analogue for human clinical experiences, wherein an individual acquires an aversive fear or addiction within their everyday environment, undergoes successful clinical intervention in a therapeutic setting, and subsequently returns to their natural environment only to experience a catastrophic return of symptoms.

Mechanistically, ABA renewal generates the highest effect size and most pronounced magnitude of behavioral recovery among all known contextual variations. In ABA testing, two distinct psychological forces converge to maximize the expression of the conditioned response. First, the animal experiences a departure from the extinction context (Context B), which strips the CS of the local inhibitory retrieval cues necessary to maintain the second-order safety memory. Second, the animal is returned to the exact physical space where the original excitatory CS-US engram was encoded. This physical homecoming provides a high-fidelity environmental match that actively facilitates the retrieval of the primary excitatory memory trace.

Because Context A has a historical pairing with the US, theoretical debates persisted for several years regarding the exact contribution of direct context-US associations. While Bouton systematically ruled out simple summation through pre-CS baseline controls, the unique status of Context A as an excitatory environment undeniably contributes a powerful motivational valence that primes the nervous system. The ABA design proves unequivocally that extinction is not the structural undoing of the acquisition engram, but rather an unstable, environmentally locked inhibitory veneer that shatters the moment the organism returns to the scene of its original trauma.

3.2 ABC Renewal: Departure from the Extinction Context

While the ABA paradigm demonstrated that returning to the training context revitalized the extinguished response, it left open a fundamental associative question: Was the re-emergence of the conditioned response dependent upon returning specifically to the *original* acquisition environment, or was it driven simply by *leaving* the extinction environment? To resolve this theoretical ambiguity, Bouton and his colleagues developed the ABC renewal paradigm. In this design, acquisition is conducted in Context A, extinction is conducted in Context B, but the critical non-reinforced testing phase is executed within an entirely novel, third setting: Context C.

Context C is an environment in which the animal has never received a shock, nor has it ever experienced an extinction trial with the CS. It is a completely neutral, orthogonal sensory ecosystem featuring distinct tactile, olfactory, visual, and acoustic properties that share no structural overlap with Context A or Context B. When animals are tested with the CS in Context C, researchers observe a profound and reliable renewal of the conditioned response. Although the magnitude of ABC renewal is typically slightly lower than that observed in classical ABA renewal, the suppression of behavior or expression of freezing is substantial, statistically unmistakable, and profoundly disruptive to classical unlearning frameworks.

The demonstration of ABC renewal was monumental for learning theory because it systematically dismantled any remaining summation hypotheses. In Context C, there are zero direct context-US associations; Context C has never been paired with shock, meaning its direct associative value ($V_C$) is absolute zero. Therefore, the return of the conditioned response cannot be explained as the summation of background excitation with the CS. Instead, the ABC design proves that simply departing from the extinction context is entirely sufficient to unleash the conditioned response. It establishes that extinction learning does not generalize across space; rather, extinction is intrinsically local, tethered exclusively to the specific environmental coordinates where the organism learned that the danger was absent.

3.3 AAB Renewal: Contextual Transitions After Co-located Learning

To further probe the architectural asymmetry between acquisition and extinction, researchers formulated the AAB renewal paradigm. In this variant, both the initial fear acquisition (Phase 1) and the subsequent extinction training (Phase 2) are conducted within the exact same physical space: Context A. The testing phase (Phase 3) is then executed within an entirely novel environment, Context B. This design represents a profound theoretical inversion: the animal has only ever experienced the CS within Context A. In that single environment, the animal learned first that the CS predicted shock, and then learned that the CS no longer predicted shock.

When the animal is introduced to Context B and presented with the CS, a striking phenomenon occurs: the conditioned response re-emerges. Even though Context B was never associated with the unconditioned stimulus, the mere act of transitioning the animal away from the shared acquisition-and-extinction environment causes a failure of extinction retrieval, allowing the original conditioned response to surface. The magnitude of AAB renewal is generally the most subtle of the three paradigms—producing a smaller effect size than ABA or ABC renewal—because Context A houses both the excitatory and inhibitory associations, creating an ambiguous training history within a single space.

The theoretical significance of AAB renewal cannot be overstated. It demonstrates unequivocally that the context-dependence of extinction is not merely an artifact of shifting chambers between Phase 1 and Phase 2. Even when an animal receives hundreds of extinction trials in the very same room where it was originally conditioned, the newly acquired safety memory remains hyper-localized to that specific chamber. Initial excitatory learning possesses an inherent biological prerogative to generalize broadly across space, whereas extinction learning remains chronically context-bound, frail, and vulnerable to catastrophic retrieval failure the moment the organism crosses an environmental threshold.

4. Mechanisms of Extinction: Erasure Versus New Inhibitory Learning

4.1 Extinction as Second-Order Inhibitory Learning

The comprehensive empirical validation of the ABA, ABC, and AAB renewal paradigms required a complete reformulation of the cognitive and neural mechanics of extinction. The dominant paradigm, crystallized by Mark Bouton, conceptualizes extinction not as the structural erasure or modification of the primary CS-US engram, but as an active process of second-order inhibitory learning. When an organism undergoes extinction, the brain forms a secondary, competing memory trace: an inhibitory association often conceptualized as a CS-noUS link. This secondary trace does not destroy the original excitatory CS-US pathway; rather, it develops an active, parallel downstream circuit designed to suppress the behavioral expression of the primary trace.

Consequently, the post-extinction nervous system contains two distinct, coexisting, and diametrically opposed memory traces regarding the same conditioned stimulus:

• The Primary Excitatory Memory Trace (CS $\rightarrow$ US): Encodes the predictive relationship between the stimulus and the unconditioned aversive event; formed rapidly, resistant to temporal degradation, and inherently generalized across spatial environments.
• The Secondary Inhibitory Memory Trace (CS $\rightarrow$ noUS): Encodes the omission of the expected outcome; formed slowly over repetitive non-reinforced exposures, highly fragile, and fundamentally dependent upon local contextual retrieval cues for behavioral activation.

This coexistence creates a profound biological asymmetry. While initial fear conditioning occurs with remarkable evolutionary efficiency—often requiring only a single CS-US pairing to establish life-long behavioral suppression—extinction demands dozens or hundreds of non-reinforced presentations to take hold. More crucially, the primary excitatory trace possesses an intrinsic capacity to transfer seamlessly across diverse environmental contexts, a feature that provides obvious evolutionary survival advantages: a predator encountered at a waterhole remains dangerous when encountered in a dense thicket. In contrast, the secondary inhibitory memory trace behaves as a conditional, highly context-dependent exception to the general rule. The organism does not learn that the world is universally safe; it learns only that the CS does not predict danger within this specific, localized setting.

4.2 The Occasion-Setting Properties of Context

To mathematically and conceptually formalize how context controls which of these competing traces is executed, Bouton adapted the concept of occasion setting from associative learning theorists like Peter Holland and Robert Rescorla. In traditional conditioning, a stimulus acts as a simple conditioned excitor or inhibitor, directly eliciting or suppressing a response. In contrast, an occasion setter does not directly evoke a response on its own; instead, it acts as a hierarchical, modulatory switch that dictates whether another stimulus is permitted to activate its own downstream associative linkages.

Within Bouton’s framework, context functions precisely as a hierarchical occasion setter or conditional gating mechanism. Context does not need to possess direct excitatory or inhibitory associative strength to exert profound behavioral control. Rather, the background environmental context sets the occasion for the activation of either the CS-US trace or the CS-noUS trace. When an animal is situated within the extinction context (Context B), the physical, sensory, and interoceptive cues of that environment actively unlock the inhibitory CS-noUS node, permitting the suppression of the conditioned response. When the animal is transported outside of Context B (into Context A or Context C), this contextual gating mechanism is disengaged. The inhibitory node is rendered silent, and the default excitatory CS-US association immediately commands the motor circuitry, producing the behavioral resurgence known as renewal.

This occasion-setting model explains why simple linear summation equations—such as those found in early iterations of the Rescorla-Wagner model—fail to capture the dynamics of renewal. Context does not simply add or subtract discrete mathematical units of fear; it modulates the cognitive access pathways to competing memory representations. The context acts like a linguistic syntactic frame that disambiguates a homonym: just as the word “bank” changes its operational meaning based on whether the context is a financial institution or a river, a conditioned stimulus changes its behavioral meaning based on the contextual architecture in which it is perceived.

4.3 Temporal and Retrospective Dynamics of Ambiguity

Central to Bouton’s theoretical framework is the principle of associative ambiguity. A stimulus that has undergone both acquisition and extinction becomes inherently ambiguous to the organism’s cognitive architecture. Prior to extinction, the CS possesses a clear, unambiguous informational status: it consistently predicts the delivery of an unconditioned stimulus. Because there is no conflicting information, the organism does not require contextual cues to interpret the stimulus; the fear response is deployed universally, irrespective of background environmental shifts.

The introduction of non-reinforced extinction trials fractures this clarity. The organism is confronted with a profound informational contradiction: the CS, which previously heralded danger, now heralds nothing. This contradiction creates high prediction errors and forces the cognitive apparatus to resolve the associative conflict. According to Bouton’s retrieval model, the nervous system resolves this ambiguity by making the *second-learned meaning* fundamentally context-dependent. The rule of memory access dictates that while the first-learned meaning serves as the organism’s universal default assumption, all subsequent, conflicting modifications to that meaning require specific environmental qualifiers to be retrieved.

This temporal asymmetry explains why acquisition generalizes while extinction remains tethered to its originating context. The chronological sequence of experience determines cognitive priority. The organism establishes the primary rule: “CS predicts shock.” It then establishes the subordinate, conditional rule: “CS predicts safety, but only in Context B.” This formulation holds significant implications for the scheduling of extinction sessions. Experiments investigating massed versus distributed extinction show that while highly massed extinction trials can rapidly suppress responding within a single session, distributed extinction trials across multiple days and varying temporal contexts promote superior long-term retention of safety, precisely because they begin to decouple the inhibitory trace from a single, narrow contextual frame.

5. Taxonomy of Context: Beyond Physical Chambers

5.1 Interoceptive and Pharmacological Contexts

While experimental paradigms typically manipulate external physical chambers featuring distinct walls, floors, and smells, Bouton’s information-processing model defines “context” far more expansively. Context encompasses the entire continuous background matrix against which a stimulus is processed, including the internal milieu of the organism. Consequently, interoceptive and pharmacological states function as potent contextual retrieval cues capable of gating learned fear and behavioral extinction.

Decades of research into state-dependent learning have shown that if an animal acquires a response under the influence of a centrally acting pharmacological agent, the behavioral response often fails to express when the drug is cleared from the bloodstream. Translating this to renewal paradigms, researchers have demonstrated profound interoceptive renewal. If an animal undergoes fear extinction while under the influence of a therapeutic compound—such as a benzodiazepine (e.g., chlordiazepoxide), an anxiolytic, an alcohol state, or a stimulant—the newly acquired extinction memory becomes bound to the interoceptive neurochemical context produced by the drug.

When the pharmaceutical agent washes out of the animal’s central nervous system, testing the CS in a drug-free, sober state triggers immediate renewal of the conditioned fear response. Conversely, if fear is extinguished in a sober state and the animal is subsequently exposed to the CS while experiencing acute physiological intoxication, stress-induced hormonal spikes (such as elevated corticosterone or adrenaline), or altered metabolic states (such as severe hunger or satiety), the shift in internal somatic feedback acts identical to an environmental chamber shift, precipitating a resurgence of the original fear. This reveals that the physiological state of the body is an indelible component of the contextual tapestry that governs memory retrieval.

5.2 Temporal Context: Time as an Environmental Feature

One of Bouton’s most conceptually radical and intellectually satisfying theoretical achievements was the unification of Spontaneous Recovery and the Renewal Effect under a single, overarching temporal context framework. For nearly a century, Pavlov’s spontaneous recovery—the reappearance of an extinguished response following the mere passage of time—was treated as an isolated, enigmatic chronological anomaly. Bouton proposed that time itself functions as a dynamic, continuously shifting environmental context.

In any experimental setting, an organism is immersed in a continuous stream of internal and external stimuli that change systematically across time. The temporal framework of an extinction session creates a distinct “temporal context” composed of the immediate physiological, circadian, and environmental states present during that specific window of time. Immediately following extinction, the animal remains within the immediate temporal shadow of the extinction session; testing the animal immediately (e.g., within minutes) yields robust suppression of fear, because the temporal context matches the extinction environment perfectly.

However, as hours, days, or weeks elapse, the organism’s internal and external baseline inexorably shifts. The passage of time inexorably moves the animal into a novel temporal context. Bouton demonstrated that spontaneous recovery can be accurately conceptualized as an instance of temporal ABC renewal: the animal acquires fear at Time 1 (Context A), extinguishes fear at Time 2 (Context B), and is tested at Time 3 (Context C). Because Time 3 represents an entirely new temporal environment distinct from the extinction window at Time 2, the fragile, context-bound inhibitory CS-noUS memory trace fails to retrieve, and the durable, default excitatory memory re-asserts itself. By formalizing time as a continuously drifting contextual vector, Bouton resolved one of the oldest mysteries in learning theory, subsuming temporal decay and spatial renewal under a single computational architecture.

5.3 Cognitive and Social Contexts

As associative learning paradigms scaled from rodent models into human cognitive psychology, the definition of context broadened to encompass higher-order mental frameworks, including explicit instructional sets, rule-based contexts, and complex social dynamics. In human conditioning experiments, researchers can generate robust renewal effects without altering the physical room or sensory apparatus, simply by shifting the participant’s cognitive framing or explicit instructional set.

For example, if human participants acquire an expectancy of an aversive shock when viewing an image of a geometric shape under a specific cognitive rule (e.g., “The system is operating in Evaluation Mode”), undergo extinction under a secondary rule (“The system is operating in Practice Mode”), and are subsequently tested under the original evaluation rule, conditioned autonomic skin conductance responses (SCR) and subjective shock expectancies instantly renew. The participant’s mental construction of the task’s rules acts as an internal, occasion-setting cognitive context that selectively gates access to the competing memory traces.

Furthermore, social presence and affective mood states function as exceptionally powerful contextual determinants of retrieval in humans. The identity, demographic characteristics, and emotional expression of an experimenter or therapist can serve as the primary contextual anchor for extinction learning. An individual who extinguishes a social anxiety response in the presence of an empathetic, non-threatening therapist may experience complete renewal of panic when encountering an identical social evaluative stressor in the presence of peers or authority figures. Similarly, internal emotional landscapes—such as transient depressive states or acute bouts of generalized anxiety—act as affective contexts. An extinction memory consolidated during a calm, low-stress period fails to retrieve when the individual is plunged into a high-stress affective context, unleashing suppressed maladaptive responses through internal contextual shifts.

6. Neurobiological Architecture of Renewal and Contextual Gating

6.1 The Hippocampus as the Contextual Processor

The translation of Bouton and Bolles’ psychological model into structural neurobiology has pinpointed a sophisticated, highly integrated neural network responsible for contextual encoding, memory storage, and inhibitory gating. At the epicenter of this network is the hippocampus, universally recognized as the master computational engine for processing complex spatial, temporal, and multimodal environmental configurations.

Unlike simple sensory cortices that process unimodal discrete stimuli (such as an isolated auditory tone), the hippocampus synthesizes convergent inputs from the perirhinal, parahippocampal, and entorhinal cortices into a coherent, high-dimensional representation of the background environment. Decades of lesion and pharmacological inactivation studies have established that the dorsal and ventral hippocampus are indispensable for the execution of renewal. If the dorsal hippocampus is surgically lesioned or temporarily silenced via microinfusions of the GABA receptor agonist muscimol prior to testing, an animal’s ability to demonstrate ABA and ABC renewal is completely abolished. The animal behaves as though it cannot recognize the contextual mismatch between Context A and Context B, causing the extinguished safety behavior to inappropriately generalize across environments.

Pyramidal neurons within the CA1 subfield and the subiculum generate unique population codes that fire selectively in response to specific environmental configurations. These hippocampal contextual representations serve as the neural instantiation of Bouton’s occasion setters. When an animal transitions from Context B to Context A, distinct hippocampal engram ensembles are activated. These ensembles project directly and indirectly via monosynaptic and polysynaptic pathways to downstream associative nodes in the amygdala and prefrontal cortex, effectively sending a neurobiological command signal that indicates which environmental matrix the organism currently occupies.

6.2 The Amygdaloid Circuitry: Locus of the CS-US Engram

While the hippocampus maps the environmental context, the physical locus of the primary conditioned fear memory resides within the nuclear complexes of the amygdala, specifically the lateral amygdala (LA), the basolateral amygdala (BLA), and the central amygdala (CeA). The lateral nucleus of the amygdala serves as the principal sensory interface, receiving direct sensory projections regarding the CS from the auditory thalamus and cortex, as well as nociceptive somatosensory inputs regarding the unconditioned shock from the spinal and thalamic pathways.

During Phase 1 acquisition, the simultaneous convergence of the CS and US inputs within the LA drives robust long-term potentiation (LTP), physically altering synaptic weights and establishing an enduring CS-US engram. Modern optogenetic and immediate-early gene (IEG) tagging techniques have definitively proven what Bouton deduced behaviorally: *this original synaptic engram within the amygdala survives extinction completely intact*. Extinction training does not depotentiate or erase these potentiated synapses within the LA.

The basolateral nucleus of the amygdala (BLA) serves as a critical integration hub where contextual information from the hippocampus intersects with the discrete CS representation. The BLA houses distinct, functionally segregated populations of projection neurons: “fear neurons” that fire vigorously during acquisition and renewal, driving downstream outputs via the central nucleus (CeA) to trigger autonomic freeze responses, and “extinction neurons” that fire selectively when an extinguished CS is encountered within a safe environment. The central amygdala (CeA) operates as the primary motor and autonomic gateway, projecting directly to the periaqueductal gray (PAG) to execute behavioral freezing, the lateral hypothalamus to accelerate heart rate, and the paraventricular nucleus of the hypothalamus to trigger endocrine stress cascades. Renewal occurs precisely because hippocampal contextual signals bias the intra-amygdalar microcircuitry, silencing extinction neurons and re-activating fear projection neurons when the CS is detected outside the safety context.

6.3 The Medial Prefrontal Cortex (mPFC) and Inhibitory Control

The definitive top-down orchestration of contextual extinction retrieval is mediated by the medial prefrontal cortex (mPFC), specifically via the distinct functional subdivisions of the infralimbic (IL) and prelimbic (PL) cortices. Research spearheaded by Gregory Quirk, Mohammed Milad, and their collaborators has elucidated a breathtakingly elegant tripartite circuit operating between the hippocampus, mPFC, and amygdala that mechanistically executes the Renewal Effect.

The infralimbic (IL) cortex is the master anatomical site for the encoding and retrieval of the extinction memory. During Phase 2 extinction training, IL neurons undergo synaptic modifications that allow them to fire selectively to the non-reinforced CS. When the IL is activated, its long-range glutamatergic projections target a specialized cluster of GABAergic interneurons nestled within the amygdala, known as the intercalated (ITC) cell masses. When excited by the IL, these inhibitory ITC cells release GABA directly onto the central nucleus (CeA), acting as an active electrical brake that blocks the transmission of fear signals from the LA to the brainstem. Thus, extinction is driven by IL-mediated top-down inhibition of fear output.

Crucially, this IL-ITC inhibitory brake is strictly dependent upon direct inputs from the hippocampus. When the animal is in Context B, the hippocampus provides an enabling, permissive excitatory drive to the IL cortex, allowing it to suppress the amygdala. However, when the animal is moved to Context A or Context C, hippocampal projections shift their target. Instead of supporting the IL, hippocampal contextual signals excite the prelimbic (PL) cortex. The PL cortex directly projects to the basolateral amygdala, exciting the amygdaloid “fear neurons” and overriding prefrontal inhibition. Simultaneously, the lack of appropriate hippocampal input to the IL causes the IL-ITC inhibitory brake to collapse. Free from prefrontal suppression, the lateral amygdala’s intact CS-US engram fires without impediment, driving CeA motor pathways and producing full-scale behavioral renewal.

7. Renewal Versus Other Forms of Behavioral Relapse

7.1 Spontaneous Recovery: The Drift of Temporal Context

In the landscape of behavioral relapse, the Renewal Effect does not operate in total isolation; it belongs to a family of recovery phenomena that demonstrate the permanence of associative learning. The most famous of these is spontaneous recovery, originally cataloged by Pavlov and systematically integrated into retrieval theory by Bouton. Spontaneous recovery is operationally defined as the reappearance of an extinguished conditioned response following a significant chronological delay between extinction training and non-reinforced testing, with no intervening physical environmental changes.

The fundamental distinction between physical renewal and spontaneous recovery lies in the nature of the contextual manipulation. Classical renewal explicitly alters the spatial, visual, or tactile dimensions of the external chamber (e.g., ABA or ABC). Spontaneous recovery, by contrast, relies entirely upon the internal, continuous drift of temporal context. As established in Section 5.2, time serves as an invisible contextual coordinate. Immediately following extinction, the subject’s neurological state is dominated by the recent, highly salient memory of non-reinforcement, reinforced by residual neurochemical markers of the extinction session.

As the retention interval extends from hours to weeks, this temporal context inevitably drifts away from the distinct temporal state associated with the extinction trials. When the CS is subsequently presented, the organism experiences an internal contextual mismatch identical to moving from Context B to Context C. Remarkably, empirical experiments demonstrate that physical renewal and spontaneous recovery exhibit powerful additive properties. If an animal is tested after a long temporal delay (spontaneous recovery) *and* within a novel physical environment (ABC renewal), the magnitude of the returned fear response is vastly greater than that produced by either manipulation alone. This additivity provides profound empirical support for Bouton’s unified retrieval model, demonstrating that temporal and physical cues operate via identical underlying cognitive mechanisms.

7.2 Reinstatement: Unconditioned Stimulus Exposure

A second major form of behavioral relapse is reinstatement. Reinstatement is operationally defined as the re-emergence of an extinguished conditioned response produced by the presentation of the unconditioned stimulus (US) alone, completely unsignaled and isolated from the conditioned stimulus, prior to the final testing phase. For instance, an animal conditioned to fear a tone via shock pairings and subsequently extinguished to the tone alone is suddenly subjected to two or three unheralded footshocks in the chamber without the tone playing. When the tone is subsequently presented, behavioral fear surges back with violent intensity.

For decades, reinstatement was viewed as a general motivational priming effect, with theorists hypothesizing that exposure to the traumatic US simply hyper-sensitized the organism’s emotional nervous system. However, rigorous investigations by Mark Bouton, Robert Bolles, and their contemporaries revealed that reinstatement is, at its core, a fundamentally context-dependent phenomenon. Reinstatement only occurs if the unsignaled US presentations are delivered within the *exact same context* where the final CS testing occurs.

The associative architecture of reinstatement is distinct from renewal. During the delivery of the unsignaled shocks, the experimental context itself forms a direct, excitatory context-US association. When the extinguished CS is subsequently presented within that shock-conditioned context, the animal uses the excitatory context as an occasion setter that biases retrieval toward the excitatory CS-US trace. If the unsignaled shocks are delivered in an entirely different room, and the animal is tested with the CS in a neutral environment, reinstatement fails to materialize. Thus, reinstatement is not a generic sensitization artifact; it is an associative phenomenon governed entirely by contextual conditioning, cementing the principle that context dictates whether extinguished memories remain dormant or re-awaken.

7.3 Resurgence: Extinction of Alternative Operant Responses

While renewal, spontaneous recovery, and reinstatement have their primary historical origins within classical Pavlovian paradigms, behavioral relapse is equally prevalent and destructive within operant (instrumental) conditioning frameworks. The operant analog most closely aligned with these mechanisms is resurgence. Resurgence is operationally defined as the reappearance of a previously extinguished target operant response when an alternative, newly reinforced response is subsequently subjected to extinction.

In a typical resurgence paradigm, a primary behavior (Response 1, such as pressing Lever A for a drug or food reward) is systematically trained and subsequently extinguished. Following this extinction, the experimenters introduce and reinforce an alternative, constructive behavior (Response 2, such as pressing Lever B). The animal rapidly learns to engage in Response 2 to obtain the reinforcer, while Response 1 remains completely dormant. However, the moment Response 2 is also subjected to extinction (no longer producing reinforcement), the animal instantaneously abandons Lever B and experiences a massive, spontaneous return to pressing Lever A, despite Lever A continuing to yield no reward whatsoever.

Resurgence shares profound theoretical and practical parallels with classical renewal. In both cases, the organism relies on an underlying hierarchy of behavioral strategies. When a newly acquired behavioral strategy (whether it is an inhibitory safety memory in renewal, or an alternative operant response in resurgence) fails to produce expected outcomes or loses its local environmental reinforcement, the cognitive control system immediately retreats down its evolutionary and learning history, reactivating the foundational, primary learned response. Resurgence and renewal frequently interact in clinical landscapes: if an individual attempts to extinguish an addictive habit using a replacement coping behavior, the failure of that replacement behavior—combined with an environmental context shift—creates a catastrophic compound relapse scenario.

8. Methodological Rigor and Controls in Bouton and Bolles’ Research

8.1 Ruling Out Non-Associative Explanations

To fundamentally overthrow the entrenched unlearning consensus of twentieth-century behaviorism, Mark Bouton and Robert Bolles had to establish experimental controls of unprecedented empirical rigor. Skeptics of early retrieval concepts frequently attempted to explain away the re-emergence of behavior using non-associative psychological mechanisms, arguing that the return of responding was merely an artifact of pseudoconditioning, generalized hyperreactivity, or sensory sensitization induced by novel environmental transitions.

To decisively dismantle these counter-arguments, Bouton and Bolles integrated explicitly unpaired control groups into their paradigms. In these control cohorts, subjects received the exact same number of CS presentations and unconditioned shocks during Phase 1, but the presentations were completely random and explicitly unpaired in time, preventing the formation of an associative CS-US predictive bond. When these unpaired control animals were subsequently transitioned through Phase 2 extinction and Phase 3 testing across different contexts (ABA and ABC), they demonstrated zero conditioned suppression upon presentation of the CS. This confirmed that the renewal effect was an exclusively associative phenomenon that required the prior encoding of a predictive relationship; it could not be generated by generalized shock exposure or non-specific trauma.

Furthermore, Bouton and Bolles instituted rigorous habituation controls to account for the intrinsic novelty of environmental shifts. Introducing a rat to an unfamiliar chamber naturally provokes exploratory behavior, sensory orienting reflexes, and heightened vigilance, which can temporarily disrupt baseline operant lever pressing. By systematically pre-exposing all subjects to Contexts A, B, and C for extensive periods prior to the initiation of experimental trials, the researchers habituated these innate orienting responses to absolute baseline. They demonstrated that the contextual transition itself caused no suppression of behavior; the suppression emerged exclusively when the target CS was introduced within the mismatched environment, ruling out sensitization and order effects entirely.

8.2 Controlling for Context-US Direct Associations

The most dangerous methodological confound threatening the validity of the classical ABA renewal effect was the potential presence of direct context-US excitatory associations. In the ABA paradigm, Phase 1 acquisition takes place within Context A. During these conditioning trials, the animal not only associates the discrete CS with the shock, but inevitably associates the physical chamber itself (Context A) with the shock. If Context A becomes a conditioned excitor in its own right, critics argued that testing the CS in Context A might simply reflect the additive summation of baseline contextual fear with a partially extinguished CS, rather than a genuine failure of extinction retrieval.

Bouton and Bolles conquered this confound through brilliant experimental design: the implementation of context extinction protocols prior to testing. Following the completion of Phase 2 extinction of the CS in Context B, the researchers placed the animals back into Context A for hours across multiple days in the complete absence of any CS presentations or shocks. During these prolonged sessions, the animals fully extinguished their direct associative fear of Context A. Their baseline operant lever pressing recovered entirely, demonstrating that the physical chamber had lost all its direct excitatory capacity to elicit freezing or emotional suppression.

Remarkably, when the extinguished CS was subsequently presented within the thoroughly extinguished Context A, the renewal effect emerged with unmitigated, full-scale strength. Even though Context A had been rendered completely benign and incapable of driving suppression on its own, it retained its full potency as a hierarchical occasion setter. This empirical masterstroke disproved the summation hypothesis once and for all. It proved that an environment’s capacity to serve as a contextual retrieval cue is mathematically and neurobiologically independent of its own direct associative value.

8.3 Quantification of Behavior: Conditioned Suppression Ratios

The quantitative precision of Bouton and Bolles’ findings was rooted in their rigorous mathematical operationalization of behavioral fear, utilizing the Annau-Kamin suppression ratio. Developed by Z. Annau and Leon Kamin in 1961, this ratio provides a standardized, mathematically bounded metric that transforms raw operant lever-pressing counts into an exact index of conditioned emotional response, isolating internal fear states from idiosyncratic differences in an animal’s baseline motor activity.

The suppression ratio is calculated according to the formula:

$$\text{Suppression Ratio} = \frac{\text{CS Responding}}{\text{CS Responding} + \text{Pre-CS Responding}}$$

The mathematical boundaries and behavioral interpretations of this formulation are profound:

• A ratio of 0.00: Indicates *total suppression* of responding. The animal pressed the lever zero times during the presentation of the CS, reflecting maximum, profound conditioned fear and complete behavioral freezing.
• A ratio of 0.50: Indicates *zero suppression* of responding. The rate of lever pressing during the CS is identical to the baseline rate immediately preceding the CS, signifying that the stimulus elicits no emotional response whatsoever (complete extinction or absence of fear).
• Ratios between 0.00 and 0.50: Provide an exceptionally sensitive, linear continuum measuring graded intensities of fear, allowing for subtle statistical discriminations across experimental cohorts.

To ensure that the denominator of this ratio remained stable across hundreds of experimental subjects, Bouton and Bolles trained rats on robust variable-interval (VI) food reinforcement schedules (such as VI 60-second or VI 90-second schedules). These schedules produce remarkably steady, highly resilient baselines of operant lever pressing that resist spontaneous drift. By subjecting these suppression ratios to multi-factor analyses of variance (ANOVA) and tracking them across individual trials, Bouton and Bolles produced datasets of extraordinary statistical power, leaving no empirical ambiguity regarding the mathematical reality of the Renewal Effect.

9. Bouton’s Information Processing and Retrieval Theory

9.1 The Ambiguity-Retrieval Model

Synthesizing decades of experimental data, Mark Bouton formulated the Ambiguity-Retrieval Model, which remains the preeminent theoretical framework explaining context-dependent memory dynamics. At its core, the model is an information-processing theory that views conditioning not as the mechanical etching of neural reflexes, but as the active construction of an internal cognitive database. The central tenet of the model is that *ambiguity is the direct trigger for contextual gating*.

Bouton draws a compelling cognitive parallel between associative learning and psycholinguistics, specifically the phenomenon of lexical ambiguity and polysemy. When a person encounters a word with a single, unambiguous definition, the cognitive system accesses its meaning automatically, without needing to analyze the surrounding sentence. However, when an individual encounters a polysemous word—such as “trunk” (which can signify an elephant’s proboscis, a storage chest, or an automobile compartment)—the linguistic system immediately and unconsciously interrogates the surrounding linguistic context to disambiguate the word’s operational meaning.

Within associative learning, a conditioned stimulus that has been paired with shock and subsequently extinguished is the behavioral equivalent of a polysemous word. It possesses two contradictory operational definitions:

• Meaning 1: $\text{CS} \rightarrow \text{US}$ (Danger / Reinforcement)
• Meaning 2: $\text{CS} \rightarrow \text{noUS}$ (Safety / Non-reinforcement)

Because the CS is now fundamentally ambiguous, the brain cannot execute a response safely without additional informational input. It is precisely at this moment of associative conflict that the contextual processing system is recruited. The context acts as the semantic framework that resolves the ambiguity. If the context contains the features of the extinction environment, Meaning 2 is selected. If the context deviates from the extinction environment, the system relies on its hard-wired, evolutionary default rule: assume the original meaning (Meaning 1) remains in effect. The organism prioritizes survival over safety, treating danger as universal and safety as an exception requiring explicit environmental verification.

9.2 Memory Representation and Node Activation

To translate the Ambiguity-Retrieval Model into a structural architecture, Bouton utilized network node representations akin to spreading activation models in cognitive science. Within this network, mental representations are conceptualized as interconnected nodes: the Conditioned Stimulus (CS) node, the Unconditioned Stimulus (US) node, the Contextual node, and a specialized Inhibitory Intermediary node.

During Phase 1 conditioning, a direct, highly stable excitatory pathway is established between the CS node and the US node. The activation threshold of this primary excitatory link is exceptionally low; once established, activation spreading from the CS node travels directly to the US node, triggering the conditioned response. During Phase 2 extinction, the nervous system does not sever this connection. Instead, it constructs an entirely new Inhibitory Intermediary node that lies in parallel across the pathway. When this inhibitory node is fully activated, it actively intercepts and quenches the electrical or cognitive transmission traveling between the CS and the US nodes.

Crucially, the Inhibitory Intermediary node has an exceptionally high intrinsic activation threshold. It cannot be excited by the CS alone; it requires a simultaneous, convergent input from the Context node representing the extinction environment. When the animal is in Context B, the Context B node fires continuously, providing the baseline sub-threshold excitation necessary for the CS to fully activate the inhibitory link. The circuit closes, and fear is suppressed. However, the moment the animal shifts into Context A or Context C, the Context B node ceases firing. Deprived of contextual support, the Inhibitory Intermediary node falls dormant. When the CS is presented, its activation bypasses the silent inhibitory node, surging down the unhindered primary excitatory pathway to light up the US node and trigger full behavioral renewal.

9.3 Computational Models of Contextual Gating

The behavioral dynamics discovered by Bouton and Bolles exposed the mathematical limits of the Rescorla-Wagner model and prompted a revolution in computational neuroscience. Contemporary learning theorists have successfully formalized Bouton’s insights by constructing sophisticated connectionist networks, modified attentional frameworks like the Pearce-Hall model, and advanced Bayesian latent cause inference models.

In modern Bayesian formulations of extinction—pioneered by researchers like Samuel Gershman and Yael Niv—the brain is modeled as an inference engine that continuously calculates the probability distribution of underlying latent causes in the environment. An animal does not simply form direct associations between raw sensory inputs; it attempts to infer the hidden, unobservable states of the world that give rise to those sensory experiences. During acquisition in Context A, the brain infers the existence of a single, highly dangerous latent cause (Latent Cause 1: “Aversive State”).

When the animal is moved to Context B and reinforcement ceases, the discrepancy in environmental cues combined with the sudden omission of the shock leads the Bayesian inference engine to conclude that a *new* latent cause has emerged (Latent Cause 2: “Benign Safety State”). The animal does not update or overwrite the parameters of Latent Cause 1; it segregates its experiences into two entirely separate cognitive files. When the animal is subsequently moved back to Context A, the sensory features of the room provide overwhelming Bayesian evidence that the organism has returned to the domain of Latent Cause 1. The animal assigns a near-unity probability to the active state of Latent Cause 1, instantly producing ABA renewal. Even in ABC renewal, the departure from Context B lowers the probability that Latent Cause 2 is active, forcing the system to distribute probability mass back toward the original, highly salient Latent Cause 1. These computational frameworks demonstrate that renewal is the mathematically optimal response of a rational inference engine navigating a dynamic, uncertain world.

10. Clinical Implications: Anxiety Disorders and Exposure Therapy

10.1 The Clinic as Context B: The Exposure Therapy Dilemma

The discovery of the Renewal Effect delivered a monumental, paradigm-shifting shock to clinical psychology, specifically to the practice of exposure therapy and cognitive behavioral therapy (CBT) for anxiety disorders, specific phobias, panic disorder, and post-traumatic stress disorder (PTSD). Exposure therapy was historically conceptualized under the same unlearning assumptions that dominated early behaviorism: clinicians believed that by exposing a patient to a feared stimulus (such as a spider, a social setting, or a traumatic memory) in the absence of catastrophic harm, the underlying fear association would be permanently extinguished or unlearned.

However, clinicians were perpetually baffled by a tragic and widespread phenomenon: high rates of clinical relapse. Patients would achieve complete, transformative reductions in fear within the clinical setting, demonstrating absolute calm and mastery in the presence of their phobic trigger. Yet, upon returning to their natural environments—their homes, workplaces, or the physical site of their original trauma—their paralyzing terror, autonomic hyperarousal, and panic would suddenly, inexplicably return at full strength. This post-therapeutic relapse had long been blamed on patient non-compliance, therapeutic resistance, or insufficient exposure dosage.

Mark Bouton’s work shattered these misconceptions by proving that exposure therapy is simply the clinical execution of experimental extinction, and the therapist’s office is none other than Context B. In the safety of the clinical suite, in the presence of a calm, supportive therapist, surrounded by distinct sensory, physical, and interoceptive cues, the patient constructs a localized, context-dependent inhibitory safety memory (CS-noUS). The patient has not unlearned their fear; they have merely learned that the stimulus is safe *in the presence of the therapist and within the walls of the clinic*. The moment the patient steps outside into the real world (Context A or Context C), the contextual retrieval cues supporting the inhibitory memory vanish. The tripartite neurocircuitry shifts, the prefrontal inhibitory brake collapses, the amygdala’s original engram re-asserts control, and the patient experiences catastrophic ABA or ABC renewal of their pathological anxiety.

10.2 Clinical Strategies to Circumvent the Renewal Effect

Recognizing exposure therapy through the lens of contextual retrieval allowed clinical scientists to engineer revolutionary, evidence-based modifications designed to structurally insulate patients against the Renewal Effect. Rather than relying on simple habituation within the clinic, modern exposure protocols deploy targeted strategies to broaden the generalizability of extinction learning:

• Multiple-Context Exposure Training: Rather than conducting all exposure sessions within the same therapeutic office, clinicians conduct systematic exposures across a wide, highly diverse array of physical environments. By extinguishing fear to the stimulus in the clinic, in the patient’s home, in an outdoor park, in a bustling commercial setting, and across diverse sensory backdrops, the inhibitory safety memory is decoupled from any single contextual coordinate. In computational terms, the patient learns that safety is not a localized exception, but a generalized rule that applies across diverse latent causes.
• Extinction Retrieval Cues: Clinicians provide patients with concrete, portable sensory objects—such as a specific piece of jewelry, a distinctive scented oil, a unique card, or a customized digital smartphone widget—that are present continuously during successful extinction sessions in Context B. When the patient enters high-risk real-world environments (Context A or Context C), they actively carry and interact with this physical object. The object serves as a portable, physical piece of Context B, functioning as a bridge that retrieves the prefrontal inhibitory memory and suppresses amygdala-driven fear in real time.
• Deepened Extinction via Compound Stimuli: Borrowing directly from associative learning principles, clinicians combine multiple independently extinguished fear cues and present them together during subsequent exposure sessions. The unexpected non-occurrence of harm following a compound presentation generates massive negative prediction errors, vastly accelerating and deepening the inhibitory associative strength of the safety trace and rendering it far more resilient against contextual renewal.
• Maximizing Expectancy Violations: Pioneered by Michelle Craske and colleagues, modern exposure therapy shifts its primary goal away from subjective fear reduction (habituation) toward maximizing the violation of explicit cognitive expectancies. By focusing purely on disconfirming the patient’s catastrophic predictions, clinicians engage deeper prefrontal cortical plasticity, consolidating safety memories that transfer across environmental shifts with vastly greater efficacy.

10.3 Interoceptive Generalization in Cognitive Behavioral Therapy (CBT)

A profound breakthrough emerging from Bouton’s taxonomy of context was the clinical realization that a patient’s internal physiological state constitutes an inescapable contextual landscape. In conditions such as panic disorder, agoraphobia, and health anxiety, the conditioned stimuli that provoke terror are often internal somatic sensations: an accelerated heart rate, shortness of breath, dizziness, or peripheral paresthesia. When individuals experience these sensations, their nervous system misinterprets them as catastrophic harbingers of imminent death, myocardial infarction, or mental collapse.

Traditional exposure therapy often fails to resolve these conditions because extinction is conducted while the patient is physically sedentary, calm, and physiologically stable. In essence, the safety memory is consolidated within an interoceptive context of profound somatic equilibrium. The moment the patient experiences sudden physiological arousal in everyday life—triggered by intense physical exercise, sexual intercourse, sudden temperature shifts, caffeine consumption, or brief environmental fright—this surge in autonomic tone serves as an interoceptive context shift that immediately unleashes massive renewal of panic.

To inoculate patients against interoceptive renewal, modern cognitive behavioral protocols mandate the aggressive integration of interoceptive exposure exercises directly into therapy. Clinicians deliberately induce acute physiological distress: having the patient hyperventilate through a narrow straw to simulate suffocation, spin in an office chair to trigger vestibular disorientation, run vigorously in place to provoke tachycardia, or breathe carbon dioxide-enriched air to stimulate autonomic panic cascades. By extinguishing the patient’s catastrophic fear across these hyper-aroused somatic states, clinicians systematically bind the inhibitory safety memory to high-arousal interoceptive contexts. Safety is successfully learned not merely when the body is calm, but precisely when the body is in an acute state of physiological activation, effectively neutralizing interoceptive renewal in the real world.

11. Translational Applications: Addiction, Cue-Reactivity, and Relapse

11.1 Substance Use Disorders as Conditioned Pathologies

While the historical foundations of the Renewal Effect were forged within aversive fear conditioning frameworks, the identical neurobiological and associative principles govern appetitive conditioning, particularly the devastating pathology of substance use disorders. Addiction is fundamentally a chronic, relapsing disorder of associative learning and neuroplasticity. Through repetitive drug consumption, environmental stimuli that consistently precede drug intake—such as specific rooms, drug paraphernalia, social acquaintances, sensory aromas, and neighborhoods—become extraordinarily potent Pavlovian conditioned stimuli.

These drug-associated cues acquire the capacity to evoke powerful conditioned responses: intense subjective craving, anticipatory dopaminergic surges within the nucleus accumbens, and profound compensatory physiological adaptations designed to counter the incoming chemical insult. Cue-exposure therapy was enthusiastically adopted in addiction medicine as an empirical strategy to extinguish this cue-reactivity. Patients in specialized, residential rehabilitation clinics were systematically exposed to drug paraphernalia (e.g., syringes, pipes, alcohol bottles) across repetitive non-reinforced sessions until their physiological craving and autonomic reactivity dropped to baseline.

Yet, the long-term clinical success of cue-exposure therapy within addiction medicine has been notoriously abysmal. Mark Bouton’s renewal framework exposes the exact mathematical and architectural reason for this systemic therapeutic failure: the residential rehabilitation clinic is an absolute, archetypal Context B. In the tranquil, highly regimented, and completely drug-free environment of the clinic, the patient successfully establishes an inhibitory CS-noDrug memory. However, the moment the patient completes their program and returns to their home neighborhood, their social circles, or the physical environments where they previously used drugs (Context A), a violent, uncontrollable ABA renewal of cue-reactivity occurs. The prefrontal control networks fail to retrieve the clinic-bound safety memory, dopamine surges violently in response to local cues, craving overwhelms behavioral resistance, and the individual suffers an immediate, devastating chemical relapse.

11.2 Behavioral Addictions and Compulsive Overeating

The reach of the renewal paradigm extends seamlessly into modern behavioral addictions and metabolic health crises, particularly compulsive overeating, obesity, and digital screen dependencies. In our modern obesogenic environment, humans are continuously bombarded by Pavlovian conditioned cues associated with highly palatable, ultra-processed, calorie-dense foods. The golden arches of a fast-food franchise, the distinct sensory chime of an advertisement, the visual architecture of a bakery storefront, and the sensory smell of caramelized sugars act as conditioned stimuli that trigger cephalic-phase digestive reflexes, insulin release, and profound appetitive drive.

When individuals engage in lifestyle interventions, medical weight loss programs, or structured diets, they essentially attempt to extinguish their behavioral and physiological reactivity to these food cues within a highly artificial, localized context: the structured diet program (Context B). They achieve success within the controlled environment of their home kitchen or specialized wellness retreat. However, when they step back into their standard occupational environment, attend social celebrations, or encounter environmental stress (Context A or C), the contextual shift triggers profound renewal of food craving and compulsive consumption. Compulsive overeating is fundamentally an associative renewal event, wherein ancient evolutionary feeding circuitry overrides fragile, context-bound cognitive restraint.

Similarly, behavioral addictions involving digital interfaces—such as pathological online gambling, compulsive social media scrolling, and problematic gaming—are anchored within complex digital contexts. The sensory interfaces of modern smartphones, with their vibrant push notifications, variable-ratio reward schedules, and tactile touchscreen interactions, act as continuous, pervasive contextual hubs. Attempting to extinguish these compulsions through temporary digital detoxes or isolated behavioral interventions routinely fails due to AAB and ABC renewal dynamics. The moment the user re-enters the digital sensory context, the underlying associative engrams reignite, demonstrating that modern technological interfaces exploit the exact same contextual vulnerabilities documented by Bouton and Bolles in 1979.

11.3 Novel Pharmacological and Neuromodulatory Augmentations

To overcome the intrinsic fragility of extinction learning and conquer the Renewal Effect across addiction and psychiatric disorders, neuroscientists are pioneering revolutionary pharmacological and neuromodulatory interventions designed to chemically alter the physical consolidation and generalizability of extinction memories:

• D-Cycloserine (DCS): A partial agonist at the glycine-binding site of the NMDA receptor, D-cycloserine does not act as a direct anxiolytic; rather, it directly enhances long-term potentiation during learning. When administered immediately before or after an extinction session, DCS dramatically accelerates the consolidation of the newly formed inhibitory CS-noUS memory trace within the infralimbic cortex and amygdala. Crucially, preclinical research demonstrates that DCS-augmented extinction memories possess superior structural stability, exhibiting significant resistance to both ABA and ABC renewal by enabling the safety memory to generalize across contextual borders.
• Epigenetic Modulation via HDAC Inhibitors: Long-term memory storage requires chromatin remodeling and gene transcription. Compounds that inhibit histone deacetylases (HDACs), such as sodium butyrate or vorinostat, maintain DNA in an open, transcriptionally permissive state. Administering HDAC inhibitors during extinction training transforms transient, context-bound safety memories into exceptionally robust, permanent traces that permanently resist contextual recovery, essentially forcing the brain to treat extinction as a permanent rewriting of behavioral priorities.
• Reconsolidation Disruption Protocols: Perhaps the most conceptually radical departure from Bouton’s dilemma is the attempt to bypass extinction entirely by targeting memory reconsolidation. When a consolidated fear memory is briefly retrieved via a single CS presentation, the underlying molecular engram temporarily destabilizes into a labile, plastic state before being re-stabilized (reconsolidated). By administering the beta-adrenergic receptor antagonist propranolol during this critical reconsolidation window, researchers can biochemically block protein synthesis and erase the emotional component of the primary CS-US engram within the amygdala. Because reconsolidation disruption structurally dismantles the original memory rather than constructing a competing inhibitory trace, the treated fear cannot undergo renewal, spontaneous recovery, or reinstatement.
• Non-Invasive Brain Stimulation (TMS/tDCS): Modern clinical neuroscience employs repetitive Transcranial Magnetic Stimulation (rTMS) and transcranial Direct Current Stimulation (tDCS) to selectively amplify cortical excitability over the ventromedial and medial prefrontal cortices during exposure sessions. By artificially driving prefrontal pyramidal output, these neuromodulatory tools strengthen top-down inhibitory gating over the amygdala, helping patients maintain extinction retrieval even when navigating novel real-world environments.

12. Critical Perspectives, Unresolved Questions, and Future Horizons

12.1 Boundary Conditions and Failures of Renewal

Despite the extraordinary robustness of the Renewal Effect across species, sensory modalities, and motivational states, contemporary research has identified distinct boundary conditions where renewal fails to materialize, revealing critical limitations in the universality of context-dependent gating. The most significant parameter dictating the presence or absence of renewal is the phenomenon of extinction overtraining. If an organism is subjected to massive, exhaustive overtraining during Phase 2—receiving hundreds or thousands of non-reinforced CS exposures far beyond the point where behavioral responding reaches zero—the inhibitory memory trace gradually escapes its hippocampal contextual constraints.

Through mechanisms of progressive cortical consolidation, massively overtrained extinction memories eventually establish direct, autonomous inhibitory control over downstream motor pathways that no longer require permissive gating signals from the hippocampus. In such states, testing the CS in Context A or Context C yields complete behavioral suppression, indicating that the organism has finally consolidated a truly generalized, context-independent safety rule. However, achieving this level of overtraining in clinical human populations is extraordinarily difficult, demanding vast therapeutic resources and significant patient distress.

Furthermore, structural variables such as the physical intensity of the conditioned stimulus, the biological severity of the unconditioned stimulus, and the original schedule of reinforcement exert profound influences on renewal susceptibility. Fears conditioned using intensely traumatic, high-voltage shocks demonstrate near-invulnerable persistence; their primary engrams are so deeply etched within the basolateral amygdala that even extensive extinction fails to prevent rapid ABA renewal. Conversely, fears established via partial reinforcement schedules—where the CS is paired with the US on only a fraction of trials—prove exceptionally resistant to the initial acquisition of extinction, yet paradoxically show complex, unpredictable patterns of contextual recovery. Finally, developmental trajectory serves as a definitive biological boundary: pre-weanling, juvenile rodents (under postnatal day 18) fail to exhibit the renewal effect entirely. In juvenile brains, extinction operates as genuine, physical *unlearning* or memory erasure, because the complex hippocampal-prefrontal inhibitory networks required for occasion setting and contextual gating have not yet anatomically matured.

12.2 Theoretical Challenges to Bouton’s Model

While Mark Bouton’s Ambiguity-Retrieval Model remains the dominant cognitive paradigm in associative learning, it has faced sustained, sophisticated theoretical challenges from alternative schools of thought. The most prominent alternative is the Configural Learning Theory formulated by John M. Pearce. Pearce fundamentally rejects the elementistic, occasion-setting premise that a CS and a context are processed as separate, hierarchical entities. Instead, Pearce posits that organisms process their entire sensory environment as a single, indivisible, unitary configural whole.

Within Pearce’s configural framework, the animal in Context A does not perceive a discrete “Tone” against a background “Room A”; rather, it perceives a singular, holistic compound stimulus: “[Tone-Room-A]”. When the animal is conditioned, it is this unique configural compound that acquires associative strength. When the animal is transferred to Context B and presented with the tone, it perceives an entirely different configural stimulus: “[Tone-Room-B]”. Extinction systematically neutralizes the associative weight of this second configural unit. When the animal is returned to Context A during Phase 3, the configural stimulus “[Tone-Room-A]” is re-introduced. Responding occurs not because of contextual retrieval gating or occasion setting, but simply because “[Tone-Room-A]” still possesses its original, intact associative weight, while generalization between the two distinct configural compounds is incomplete. While Pearce’s model struggles to explain certain nuanced nuances of ABC renewal and context extinction controls as elegantly as Bouton’s retrieval theory, it provides a fierce, mathematically rigorous competitor that challenges the necessity of hierarchical occasion setters.

Furthermore, fierce debates persist regarding the absolute necessity of the hippocampus for all forms of context-dependent behavior. Emerging research using cell-specific optogenetic silencing has shown that animals can, under specific training parameters, retain rudimentary context-dependent discrimination even when the dorsal hippocampus is completely deactivated, suggesting that alternative structures—such as the retrosplenial cortex, the perirhinal cortex, or direct sensory-cortical microcircuits—may possess sufficient computational capacity to encode primitive contextual associations. Finally, the semantic, abstract nature of human social and cognitive contexts presents continuous ecological challenges to laboratory models, forcing scientists to interrogate whether the spatial chamber models developed by Bouton and Bolles truly capture the vast multidimensional complexity of the human experiential landscape.

12.3 Synthesizing Molecular Neuroscience and Behavioral Ecology

The contemporary frontier of learning theory lies at the breathtaking intersection of molecular engram biology and behavioral ecology. Using revolutionary viral-genetic technologies—such as activity-dependent immediate-early gene tagging (e.g., c-Fos-tTA/TRE systems), fiber photometry, and deep-brain two-photon calcium imaging—modern neuroscientists can literally visualize, manipulate, and track living memory engrams in real time as animals traverse renewal paradigms.

Groundbreaking work by researchers such as Susumu Tonegawa has confirmed the physical coexistence of the competing engrams predicted by Bouton. Scientists can now tag the specific ensemble of lateral amygdala neurons that encode the initial fear memory with light-sensitive channelrhodopsin proteins, and subsequently tag the separate ensemble of infralimbic and intercalated neurons that encode the extinction memory. During ABA renewal transitions, deep-brain imaging reveals the exact millisecond when the hippocampal input switches its electrical alignment, silencing the extinction ensemble and triggering coordinated, synchronous firing within the primary fear engram. This represents the ultimate empirical vindication of Bouton and Bolles’ theoretical vision: the physical, visible proof that extinction is a conflict between competing physical networks, mediated by the master switch of contextual processing.

Simultaneously, machine learning algorithms and computational computer vision platforms are transforming our understanding of the fine-grained micro-behavioral architecture of renewal. Rather than relying solely on coarse, binary metrics like lever-press suppression ratios or total freezing duration, artificial intelligence systems can track hundreds of subtle, sub-second kinematic variables: postural micromovements, whisker deflections, head-orienting saccades, and respiratory rhythm variations. These computational analyses demonstrate that an animal begins processing the contextual shift and anticipating the return of danger long before the explicit conditioned stimulus is even introduced. As molecular neurobiology, artificial intelligence, and classical behavioral ecology converge, we move steadily closer to a unified, multi-scale theory of memory persistence—a comprehensive biological model that traces a continuous, causal line from synaptic phosphorylation events inside single dendritic spines all the way to the complex, tragic, and magnificent behavioral adaptations that govern mammalian survival in an ever-shifting world.

Conclusion

The experimental and theoretical breakthroughs initiated by Mark Bouton and Robert Bolles in 1979 permanently transformed the landscape of psychological science and behavioral neuroscience. By demonstrating that the simple act of shifting an organism’s sensory and spatial environment reliably revives an extinguished conditioned response, the Renewal Effect effectively demolished the mid-twentieth-century dogma of associative unlearning. Bouton and Bolles proved that extinction is not the physical erasure of a memory trace, but an active, sophisticated, and remarkably dynamic form of second-order inhibitory learning—a cognitive process wherein a fragile safety memory is continuously modulated by the occasion-setting power of environmental context.

From the initial demonstrations of the classic ABA paradigm to the intricate taxonomic dissections of ABC and AAB designs, their work illuminated the fundamental biological asymmetry of memory: acquisition is universally biased to generalize across space and time to safeguard survival, whereas extinction remains fundamentally locked to the environmental coordinates where safety was discovered. The neurobiological elucidation of this phenomenon—unveiling the intricate tripartite dance between the hippocampus as the contextual processor, the amygdala as the permanent home of the emotional engram, and the medial prefrontal cortex as the context-dependent inhibitory executioner—stands as one of the crowning triumphs of modern cognitive and behavioral neuroscience.

Ultimately, the Renewal Effect represents far more than an elegant laboratory curiosity; it is a profound clinical and philosophical window into the human condition. It provides the foundational theoretical framework for understanding why individuals suffering from debilitating anxiety disorders, post-traumatic stress, and destructive chemical addictions frequently experience catastrophic relapse precisely when they venture outside the protective confines of the clinic. By deciphering the invisible contextual threads that tie our memories to our environments, Bouton and Bolles not only revolutionized our understanding of the architecture of the mind, but also bestowed upon modern medicine the vital conceptual blueprints required to engineer enduring, context-resilient interventions, ultimately helping humans reclaim mastery over their fears and behaviors across all the diverse landscapes of life.

References

• Annau, Z., & Kamin, L. J. (1961). The conditioned emotional response as a function of intensity of the US. Journal of Comparative and Physiological Psychology, 54(4), 428–432. https://doi.org/10.1037/h0042199
• Bolles, R. C. (1970). Species-specific defense reactions and avoidance learning. Psychological Review, 77(1), 32–48. https://doi.org/10.1037/h0028589
• Bouton, M. E. (1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning. Psychological Bulletin, 114(1), 80–99. https://doi.org/10.1037/0033-2909.114.1.80
• Bouton, M. E. (2002). Context, ambiguity, and unlearning: Sources of relapse after behavioral extinction. Biological Psychiatry, 52(10), 976–986. https://doi.org/10.1016/S0006-3223(02)01546-9
• Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning & Memory, 11(5), 485–494. https://doi.org/10.1101/lm.78804
• Bouton, M. E., & Bolles, R. C. (1979). Contextual control of the extinction of conditioned fear. Journal of Experimental Psychology: Animal Behavior Processes, 5(4), 368–378. https://doi.org/10.1037/0097-7403.5.4.368
• Bouton, M. E., & King, D. A. (1983). Contextual control of conditioned fear: Tests for associative context-US effects. Journal of Experimental Psychology: Animal Behavior Processes, 9(3), 248–265. https://doi.org/10.1037/0097-7403.9.3.248
• 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. https://doi.org/10.1016/j.brat.2014.04.006
• Estes, W. K., & Skinner, B. F. (1941). Some quantitative properties of anxiety. Journal of Experimental Psychology, 29(5), 390–400. https://doi.org/10.1037/h0062283
• Gershman, S. J., Blei, D. M., & Niv, Y. (2010). Context, learning, and extinction: An inductive inference approach. Psychological Review, 117(1), 197–209. https://doi.org/10.1037/a0017808
• Milad, M. R., & Quirk, G. J. (2002). Neurons in medial prefrontal cortex signal memory for fear extinction. Nature, 420(6911), 70–74. https://doi.org/10.1038/nature01138
• Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex (G. V. Anrep, Trans.). Oxford University Press.
• Pearce, J. M. (1987). A model for stimulus generalization in Pavlovian conditioning. Psychological Review, 94(1), 61–73. https://doi.org/10.1037/0033-295X.94.1.61
• Quirk, G. J., & Mueller, D. (2008). Neural mechanisms of extinction learning and retrieval. Neuropsychopharmacology, 33(1), 56–72. https://doi.org/10.1038/sj.npp.1301555
• Rescorla, R. A. (1988). Pavlovian conditioning: It’s not what you think it is. American Psychologist, 43(3), 151–160. https://doi.org/10.1037/0003-066X.43.3.151
• Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical Conditioning II: Current Research and Theory (pp. 64–99). Appleton-Century-Crofts.
• Tonegawa, S., Liu, X., Ramirez, S., & Redondo, R. (2015). Memory engram cells have come of age. Neuron, 87(5), 918–931. https://doi.org/10.1016/j.neuron.2015.08.002