Avoidance Learning

Abstract

Avoidance learning is a form of operant conditioning in which an organism learns a response that prevents an aversive event, so that the successful response is reinforced by the non-occurrence of harm. Its central puzzle is theoretical: if the feared event never happens, what reinforces the response that keeps it away? Mowrer's two-factor theory answered that a warning signal first acquires fear through classical conditioning and the response terminating the signal is reinforced by fear reduction, an account later challenged by avoidance's extraordinary persistence, by Sidman's signal-free procedure, and by Bolles's species-specific defense reactions. Modern work recasts avoidance as reinforcement learning in amygdala-based threat circuits and identifies maladaptive avoidance as the mechanism that sustains the anxiety disorders. This article develops the two-factor account, its challenges, its computational and neural reformulation, and its clinical significance, with interactive demonstrations.

Keywords: avoidance learning, two-factor theory, negative reinforcement

Avoidance learning is the process by which an organism acquires a response that forestalls an aversive event, and it is distinguished from escape learning by its timing: in escape the response terminates an aversive stimulus that is already present, whereas in avoidance the response prevents the aversive stimulus from arriving at all (Mowrer & Lamoreaux, 1946). This apparently small difference creates the field's defining problem. A response reinforced by escaping shock is easy to explain, because the shock is there to be escaped; a response reinforced by avoiding shock is not, because a successful avoidance response is followed by nothing—no shock, no obvious reinforcer, only the continued absence of harm. Explaining how the absence of an event can strengthen a response has organized the study of avoidance for eighty years, and the successive answers—Mowrer's mediating fear, Sidman's shock-frequency reduction, Bolles's innate defensive repertoire, and the reinforcement-learning and neural accounts of the present day—trace the same movement from a reflexive to an inferential picture that runs through the whole of conditioning. The stakes are not only theoretical: pathological avoidance is the behavior that keeps anxiety disorders alive, because a person who escapes a feared situation never learns that it was safe, and so the treatment of anxiety is in large part the undoing of learned avoidance (Pittig et al., 2018). The sections below develop the two-factor theory and its mechanism, the persistence that made avoidance famous, Sidman's signal-free procedure, Bolles's species-specific defense reactions, the computational and neural reformulation of the paradigm, and the human paradigms that connect avoidance to clinical anxiety.

What Avoidance Learning Is

Avoidance learning is defined by a contingency between a response and the prevention of an aversive outcome: emitting the response before a deadline cancels an event that would otherwise occur, and over trials the response comes to be made reliably. It is a species of negative reinforcement, the strengthening of a response by the removal or omission of an aversive stimulus, and it sits within operant conditioning because the behavior is controlled by its consequences. The paradigm case is the shuttle box, in which a signal—a light or tone—precedes a shock by a few seconds, and an animal that crosses to the other compartment during the signal both turns the signal off and prevents the shock. Two arrangements are distinguished. In signaled or discriminated avoidance a warning stimulus marks the interval in which the response is effective, and responding rises sharply once the animal begins to respond during the signal. In unsignaled or free-operant avoidance there is no warning stimulus at all; shocks are scheduled to occur at fixed intervals unless a response postpones them, and the animal must space its responses in time (Sidman, 1953). What unites the arrangements, and what makes avoidance a single phenomenon, is the logical oddity already noted: the event that would reinforce learning by its removal is precisely the event that a well-learned response guarantees will not occur. Avoidance is therefore not merely a variety of operant conditioning but the case that most sharply exposes the question of what a reinforcer is, and every major theory of avoidance is at bottom an answer to that question. The behavior is also of unusual reach, because it is the laboratory model of the defensive and anxious behavior that dominates clinical psychology: phobic escape, compulsive checking, and the withdrawal that narrows a life are all, in structure, avoidance responses maintained by the aversive events they successfully prevent (Krypotos et al., 2015). Figure 1 makes the escape-avoidance distinction concrete by placing the two contingencies on a common timeline.

The Two-Factor Theory

Mowrer's two-factor theory is the foundational account of avoidance and the source of nearly every question the field has since pursued (Mowrer & Lamoreaux, 1946). Its problem was the one stated above: a successful avoidance response is followed by no aversive event, so on a simple reinforcement view there is nothing to strengthen it. Mowrer's solution was to hold that avoidance is not one kind of learning but two, working in sequence. First, by classical conditioning, the warning signal that precedes shock becomes a conditioned stimulus for fear; the signal, initially neutral, comes to evoke a conditioned emotional state through its pairing with the aversive unconditioned stimulus. Second, by operant conditioning, the response that terminates the now-fearful signal is reinforced—not by the absence of shock, which is not an event, but by the reduction of fear, which is. The response removes the frightening signal, fear falls, and that fall in an aversive internal state is the reinforcer. Two factors, classical fear conditioning and operant fear reduction, thus combine so that the second learning process is driven by the product of the first. The theory is elegant because it dissolves the puzzle without inventing a new principle: the reinforcer for avoidance is a real, present, removable stimulus—conditioned fear—rather than the metaphysically awkward non-occurrence of shock. It also makes testable predictions. Fear should be acquired before avoidance appears, avoidance should depend on a signal that can carry fear, and manipulations that reduce fear should undermine avoidance. Much of the classic literature is the working-out of these predictions, and the theory's mechanism—an emotional state, conditioned to a cue, motivating and reinforcing instrumental action—became a template for later two-process accounts across learning. The demonstration below builds the two factors in sequence, showing conditioned fear rising to the signal and then the avoidance response strengthening as its termination reduces that fear.

Persistence and Resistance to Extinction

The finding that made avoidance a lasting problem was its extraordinary persistence. Solomon and Wynne trained dogs in a shuttle box with intense shock and found that once the avoidance response was established it continued almost indefinitely, with short and stable latencies, even though the shock was never again delivered because the dogs always responded in time (Solomon & Wynne, 1953). Responses that on the two-factor view should have extinguished—since the fear-conditioned signal was now never paired with shock—instead showed no sign of weakening over hundreds of trials. Solomon and Wynne offered two principles to reconcile this with the theory. The anxiety-conservation principle held that because a fast avoidance response terminates the signal almost immediately, the animal is never exposed to the signal long enough for its fear to extinguish; the very success of avoidance protects the underlying fear from the prolonged, unreinforced exposure that extinction requires. The partial-irreversibility principle held that traumatic conditioning produces fear that is partly permanent. The first principle has proved the more influential, because it identifies a self-perpetuating trap: the response prevents the experience that would teach it to be unnecessary. This has a direct clinical reading. Exposure-based therapy for anxiety works by response prevention—blocking the avoidance response so that the person remains in contact with the feared but safe situation long enough for fear to extinguish, exactly the exposure that successful avoidance otherwise denies (Pittig et al., 2018). Avoidance is thus not merely persistent but self-sealing, and the demonstration below shows why: with the response available, fear is conserved indefinitely, while blocking the response (flooding) allows the signal to lose its fear.

Free-Operant Avoidance

Sidman's procedure removed the element the two-factor theory needed most: the warning signal (Sidman, 1953). In free-operant, or unsignaled, avoidance there is no exteroceptive cue at all. Two intervals define the schedule. The shock-shock interval is the time between successive shocks if the animal does nothing; the response-shock interval is the time by which each response postpones the next shock. An animal that responds steadily, so that its response-shock intervals keep resetting the clock before it expires, can postpone shock indefinitely and receive very few. Sidman showed that rats readily learn this, spacing their responses in time to hold shock frequency low, even though there is no signal to become conditioned to fear and no signal-termination to reinforce the response. This is a serious difficulty for the two-factor theory in its literal form, because the theory's reinforcer—termination of a fear-eliciting signal—does not exist in the procedure. Two lines of repair were offered. One held that time itself, or the animal's own behavior and proprioceptive feedback, serves as the conditioned aversive stimulus, so that passage of time without responding becomes frightening and responding reduces that fear; the two-factor mechanism is preserved by finding an internal signal to replace the missing external one. The other, developed by Herrnstein, abandoned the moment-to-moment fear mechanism and proposed that avoidance is reinforced directly by the reduction in shock frequency over an extended period—a molar reinforcer defined over time rather than a molecular one tied to each response (Herrnstein, 1969). On this view the animal is sensitive to the overall rate of shock and responds because responding lowers it, no mediating fear required. The demonstration below realizes a Sidman schedule, letting the reader set the response-shock interval and a response rate and watch the resulting shock frequency, the quantity Herrnstein's account makes central.

Species-Specific Defense Reactions

Bolles attacked the two-factor theory from a different direction, arguing that it rests on a false picture of how defensive behavior is acquired (Bolles, 1970). The theory assumes that an animal learns an avoidance response by trial and error, gradually strengthening whatever action happens to reduce fear. But Bolles observed that in the wild there is no time for trial-and-error learning about predators: an animal that had to learn to avoid a predator by being caught and reinforced would be eaten before the learning occurred. Natural selection must therefore have equipped animals with innate species-specific defense reactions—flight, freezing, and defensive fighting—that are released immediately by any threatening or novel stimulus, without prior learning. What laboratory avoidance training does, on this account, is not build a new response from scratch but select among these pre-existing defensive reactions: a shuttle-box response is learned quickly when the response the experimenter requires happens to be a species-specific defense reaction, such as fleeing to another location, and slowly or not at all when it does not, such as pressing a lever, which is no part of a rat's defensive repertoire. This explains a robust but awkward fact: rats learn shuttle avoidance and running avoidance in a few trials but learn lever-press avoidance poorly, a difference the two-factor theory, which treats all operant responses as equivalent, cannot easily accommodate. Bolles's account reframed avoidance learning as the tuning of an evolved defensive system rather than the reinforcement of an arbitrary response, and it aligned the study of avoidance with the broader recognition, contemporaneous with Garcia's work on taste aversion, that learning is constrained and shaped by a species' biology. The lesson survives in modern threat-conditioning research, which treats defensive responses as the output of dedicated, evolutionarily ancient circuits rather than as freely conditionable operants (LeDoux, 2014).

Computational and Neural Accounts

The contemporary reformulation of avoidance has two connected strands, one computational and one neural. The computational strand embeds the two-factor idea in reinforcement learning. Maia showed that the actor-critic architecture, a standard reinforcement-learning model, reproduces the two-factor structure with striking fidelity: the critic learns the aversive value of states and signals, computing something like conditioned fear, while the actor learns responses using the critic's value estimates, so that a response that moves the system to a safer, lower-value state is reinforced by the resulting improvement (Maia, 2010). This gives Mowrer's two processes a precise algorithmic form—a critic that does classical fear conditioning and an actor that does operant response selection—and it explains persistence naturally, because once the actor reliably reaches safety the critic's fearful value estimate is never corrected downward. The neural strand locates these computations in identifiable circuits. LeDoux's work established that the amygdala is the hub of a defensive circuit that detects threat and drives both the conditioned fear responses of the critic and the behavioral output, and that the transition from freezing to active avoidance involves a shift in control from central-amygdala reflexive circuits to circuits capable of instrumental action (LeDoux, 2014). The most consequential recent move has been conceptual. LeDoux and colleagues argued that the field had conflated two different things under the word fear: the objectively measurable defensive responses and physiological states that avoidance procedures actually engage, and the conscious feeling of fear, which is a separate, cortically dependent process (LeDoux et al., 2017). On this reconceptualization the reinforcer in avoidance is the change in a nonconscious threat-processing state, not a felt emotion, which both preserves the mechanistic core of the two-factor theory and detaches it from the introspective notion of fear that had made the theory hard to test. Cain has since argued that many long-standing puzzles of avoidance dissolve once the behavior is analyzed in terms of these circuits and the competition between defensive response systems, rather than forced into the single mediating variable of classical fear (Cain, 2019).

Human Avoidance and Anxiety

Avoidance matters clinically because it is the engine of the anxiety disorders, and translating the animal paradigm to humans has become an active research program with its own methods and cautions. Human avoidance is studied with computer tasks in which a signal predicts an aversive outcome—a shock, an unpleasant sound, or a monetary loss—and a response cancels it, allowing the acquisition, generalization, and extinction of avoidance to be measured directly in people (Krypotos et al., 2015). These paradigms show that maladaptive avoidance behaves as the clinic predicts: it generalizes to safe stimuli, it persists after the threat is gone, and it prevents the corrective learning that would otherwise reduce fear, which is why exposure with response prevention is the core of effective treatment (Pittig et al., 2018). But the translation is not automatic, and Krypotos and colleagues have pressed the question of validity: whether laboratory avoidance tasks actually measure the clinically relevant construct or capture something narrower, and whether the costs and contingencies used in the laboratory correspond to those that maintain avoidance in disorder (Krypotos et al., 2018). Their critique is that a task can produce reliable avoidance behavior while still being a poor model of pathological avoidance, because it omits the conflict between avoidance and valued goals that defines the clinical problem—a phobic patient avoids not in a neutral setting but at the cost of things they care about. Recent work therefore builds approach-avoidance conflict into the paradigms, pitting the avoidance response against a reward so that the measure captures the trade-off a patient actually faces (Pittig et al., 2018). The clinical translation thus runs in both directions: the animal work supplies the mechanism—conditioned threat value, negatively reinforced responding, self-sealing persistence—while the human work disciplines it, insisting that a model of avoidance earn its clinical relevance rather than assume it. Table 1 summarizes the major theoretical accounts and the signature finding that motivates each.

Table 1

Theoretical Accounts of Avoidance Learning

Worked Example

The anxiety-conservation principle can be made quantitative, and doing so shows why an avoidance response, once learned, need never extinguish. Represent the fear evoked by the warning signal as a value V that would fall toward zero with continued unreinforced exposure to the signal, but only in proportion to the time the organism actually spends exposed to it. Let extinction reduce fear by a rate of 0.10 per second of exposure, so that after t seconds of continuous signal exposure without shock the fear is V(t) = V₀ · exp(-0.10 · t). Suppose fear begins at V₀ = 1.00. If the animal did not respond and simply endured the signal, then after 20 seconds of exposure per trial its fear would fall to 1.00 · exp(-2.0) = 0.135, and after a few such trials the signal would be nearly neutral—ordinary extinction. Now let the animal make a fast avoidance response that terminates the signal 1.2 seconds after it begins. The exposure per trial is only 1.2 seconds, so fear falls to just 1.00 · exp(-0.10 · 1.2) = 1.00 · exp(-0.12) = 0.887 per trial. Over ten trials the accumulated exposure is 12 seconds, giving 1.00 · exp(-1.2) = 0.301; but the response latency itself shortens as fear stays high and the response is well practiced, so in reality exposure per trial keeps falling and the total accumulated exposure converges to a small finite limit. If latency halves each block so that exposures run 1.2, 0.6, 0.3, 0.15, … seconds, the total exposure summed over all future trials is 1.2 · (1 + ½ + ¼ + …) = 1.2 · 2 = 2.4 seconds, giving an asymptotic fear of 1.00 · exp(-0.24) = 0.787. Fear never reaches zero because the animal never accumulates more than 2.4 seconds of signal exposure across its entire history, no matter how many trials it runs (Solomon & Wynne, 1953). The lesson is that the response caps total exposure at a finite budget, and a finite exposure budget cannot drive an exponential decay to zero—so the fear, and the avoidance it motivates, are conserved indefinitely. Only response prevention, which forces exposure back up toward the 20-second figure, restores the extinction the response had suppressed (Pittig et al., 2018).

Discussion

Avoidance learning has traced the same path as the other core phenomena of conditioning, from a reflexive account toward an inferential and mechanistic one, and its history is unusually legible because a single question—what reinforces a response followed by nothing—runs the length of it. Mowrer's two-factor theory gave the first and most durable answer, dissolving the puzzle by making conditioned fear the present, removable reinforcer and joining classical and operant learning in one behavior (Mowrer & Lamoreaux, 1946). The persistence Solomon and Wynne documented then showed that the theory's own logic implied a trap: successful avoidance protects the fear that drives it from the exposure that would erase it, a self-sealing dynamic that is the behavioral heart of clinical anxiety and the rationale for response prevention (Solomon & Wynne, 1953). Sidman's signal-free avoidance and Herrnstein's molar reinforcer strained the theory's literal mechanism and forced a choice between rescuing it with an internal signal and replacing it with sensitivity to shock frequency over time (Sidman, 1953; Herrnstein, 1969). Bolles cut across the whole debate by denying that avoidance responses are built by reinforcement at all, locating them instead in an evolved repertoire of species-specific defense reactions that training merely selects among (Bolles, 1970). The modern synthesis has not discarded any of this so much as translated it: the actor-critic model gives the two factors an algorithm, the amygdala circuit gives them an anatomy, and the reconceptualization of fear as a nonconscious threat-processing state detaches the mechanism from the introspective emotion that made it hard to test (Maia, 2010; LeDoux et al., 2017; Cain, 2019). Threaded through the modern work is the clinic, which both receives the mechanism and disciplines it, demanding that human avoidance paradigms model the approach-avoidance conflict that defines pathological avoidance rather than merely reproduce avoidance behavior (Krypotos et al., 2018). What began as a paradox about the reinforcement of a non-event is now understood as the negatively reinforced output of an evolved threat system, computed by identifiable circuits and disordered in identifiable ways, and the response that Mowrer struggled to explain remains the through-line connecting the reflex to the disorder.

Current Directions

Contemporary research on avoidance concentrates where its behavioral, computational, and clinical strands meet. The most active clinical program treats maladaptive avoidance as a transdiagnostic mechanism and builds approach-avoidance conflict into its measures, so that the laboratory task captures the trade-off between safety and valued goals that defines pathological avoidance rather than avoidance in a neutral vacuum (Pittig et al., 2018). A companion methodological effort asks whether human avoidance paradigms are valid models of the clinical construct at all, scrutinizing the contingencies, costs, and instructions of the tasks and warning that reliable avoidance behavior in the laboratory does not by itself guarantee clinical relevance (Krypotos et al., 2018). On the computational side, reinforcement-learning accounts have moved beyond the basic actor-critic to ask how model-based and model-free systems interact in avoidance, how the value of avoidance generalizes across stimuli, and how these processes can be estimated from human behavior and neural signals to characterize individual differences in anxiety (Cain, 2019). The neurobiological program pursues the circuit basis of the transition from reactive defensive responses to active, instrumental avoidance, and the conditions under which control passes between them, building on the amygdala-centered account and its extension to cortical and striatal contributions (LeDoux et al., 2017). Threading through all of it is the reconceptualization of fear, whose consequence—that the reinforcer in avoidance is a nonconscious threat-processing state rather than a felt emotion—reframes both the animal experiments and the design of human tasks and therapies, and remains a subject of active debate about how conscious feeling relates to the defensive mechanisms that avoidance procedures engage (LeDoux et al., 2017; Cain, 2019). The enduring aim is a single account in which the shuttle-box response, the actor-critic value estimate, the amygdala circuit, and the self-sealing avoidance of the anxious patient are all expressions of one negatively reinforced process.

Common Misconceptions

Avoidance and escape are the same thing.

They differ in timing and in what is reinforced. In escape the response terminates an aversive stimulus that is already present; in avoidance the response prevents the aversive stimulus from occurring at all. Escape is easy to explain because the aversive event is there to be removed; avoidance is the hard case precisely because a successful response is followed by no aversive event, which is why it required the two-factor theory (Mowrer & Lamoreaux, 1946).

Avoidance responses should extinguish quickly once shock stops.

The opposite is true, and it is the field's most consequential finding. Because a fast avoidance response terminates the warning signal almost immediately, the organism is never exposed to the signal long enough for its conditioned fear to extinguish, so avoidance can persist over hundreds of unreinforced trials. Extinguishing it requires response prevention, blocking the response to force prolonged exposure (Solomon & Wynne, 1953).

Avoidance always requires a warning signal to become conditioned.

Sidman's free-operant procedure showed that animals will avoid with no exteroceptive warning signal at all, spacing their responses in time to postpone shocks scheduled at fixed intervals. This strained the literal two-factor theory and motivated accounts in which time or shock frequency, rather than a signal, controls avoidance (Sidman, 1953; Herrnstein, 1969).

Any response can be learned equally well as an avoidance response.

Bolles showed that responses matching a species' innate defensive reactions—fleeing, freezing—are learned in a few trials, while responses foreign to the defensive repertoire, such as lever pressing in rats, are learned poorly or not at all. Avoidance training selects among evolved defensive reactions rather than building an arbitrary response from scratch (Bolles, 1970).

Glossary

Actor-critic model.

A reinforcement-learning architecture in which a critic learns the value of states and an actor learns responses from that value; it reproduces the two-factor structure of avoidance, the critic computing fear and the actor selecting the response.

Anxiety-conservation principle.

Solomon and Wynne's proposal that a fast avoidance response terminates the warning signal before its conditioned fear can extinguish, so that avoidance protects the very fear that drives it and thereby perpetuates itself.

Approach-avoidance conflict.

A situation in which the same stimulus or action carries both reward and threat, so that avoidance is pitted against a valued goal; increasingly built into human paradigms to model the trade-off that defines pathological avoidance.

Avoidance learning.

The acquisition of a response that prevents an aversive event from occurring, so that the successful response is reinforced by the non-occurrence of harm.

Discriminated avoidance.

Signaled avoidance, in which a warning stimulus marks the interval during which a response will prevent the aversive event; the paradigm case of the two-factor theory.

Escape learning.

Learning a response that terminates an aversive stimulus already present; distinguished from avoidance, in which the response forestalls an aversive stimulus that has not yet arrived.

Free-operant avoidance.

Sidman's unsignaled procedure in which shocks occur at fixed intervals unless a response postpones them; there is no warning signal, and the organism must space its responses in time.

Molar reinforcement.

Reinforcement defined over an extended period, such as the reduction of shock frequency over time; Herrnstein's alternative to the molecular, response-by-response reinforcer of the two-factor theory.

Negative reinforcement.

The strengthening of a response by the removal or omission of an aversive stimulus; the general category to which both escape and avoidance belong.

Response prevention.

Blocking the avoidance response so that the organism remains exposed to the feared but safe stimulus long enough for conditioned fear to extinguish; the mechanism underlying exposure therapy for anxiety.

Response-shock interval.

In a free-operant avoidance schedule, the time by which each response postpones the next shock; steady responding that keeps resetting this interval holds shock frequency low.

Shock-shock interval.

In a free-operant avoidance schedule, the time between successive shocks if the organism makes no response; the baseline rate that responding serves to postpone.

Species-specific defense reactions.

Bolles's term for the innate, unlearned defensive behaviors—flight, freezing, defensive fighting—released by threat; avoidance training selects among these rather than building a new response.

Two-factor theory.

Mowrer's account in which avoidance combines classical conditioning of fear to a warning signal with operant reinforcement of the response by the reduction of that fear when the signal is terminated.

Key Researchers

Tom Beckers. Professor at KU Leuven; his review synthesized the theoretical models of avoidance learning and its recent developments, and his work develops and scrutinizes the human paradigms that connect avoidance to clinical anxiety. ORCID - Google Scholar - Faculty Page

Robert C. Bolles (1928-1994). Professor at the University of Washington; he argued that avoidance training selects among innate species-specific defense reactions rather than building an arbitrary response, reframing avoidance as the tuning of an evolved defensive system. Wikipedia

Michelle G. Craske. Professor at the University of California, Los Angeles; her work on the role of associative fear and avoidance learning in the anxiety disorders links the laboratory paradigm to exposure-based treatment and the approach-avoidance conflict that defines pathological avoidance. Faculty Page - Wikipedia

Richard J. Herrnstein (1930-1994). Professor at Harvard University; his analysis of method and theory in the study of avoidance proposed that responding is reinforced directly by the reduction of shock frequency over time, a molar alternative to the two-factor mechanism. Wikipedia

Joseph E. LeDoux. Professor at New York University; he established the amygdala as the hub of the defensive circuit underlying avoidance and led the reconceptualization of fear as a nonconscious threat-processing state distinct from the conscious feeling. Google Scholar - Faculty Page - Wikipedia

O. Hobart Mowrer (1907-1982). Professor at the University of Illinois; his two-factor theory of avoidance, combining classical fear conditioning with operant fear reduction, is the foundational account of the phenomenon and the source of the questions the field still pursues. Wikipedia

Frequently Asked Questions

What is avoidance learning? Avoidance learning is the process by which an organism acquires a response that prevents an aversive event from occurring, so that the successful response is reinforced by the non-occurrence of harm (Mowrer & Lamoreaux, 1946).

How is avoidance different from escape? They differ in timing: in escape the response terminates an aversive stimulus already present, whereas in avoidance the response prevents the aversive stimulus from arriving at all. Avoidance is the harder case because a successful response is followed by no aversive event to serve as a reinforcer (Mowrer & Lamoreaux, 1946).

What is the two-factor theory of avoidance? Mowrer's two-factor theory holds that avoidance combines two learning processes: a warning signal acquires fear through classical conditioning, and the response that terminates the signal is then reinforced by the resulting reduction of fear (Mowrer & Lamoreaux, 1946).

Why is avoidance so resistant to extinction? Because a fast avoidance response terminates the warning signal almost immediately, the organism is never exposed to the signal long enough for its conditioned fear to extinguish; the success of avoidance protects the fear that drives it, so responding can persist over hundreds of unreinforced trials (Solomon & Wynne, 1953).

What is free-operant avoidance? Free-operant, or Sidman, avoidance is a procedure with no warning signal in which shocks occur at fixed intervals unless a response postpones them; the organism learns to space its responses in time to keep shock frequency low (Sidman, 1953).

What are species-specific defense reactions? They are the innate defensive behaviors (flight, freezing, and defensive fighting) that Bolles argued are released by threat without prior learning; avoidance training selects among these evolved reactions rather than building an arbitrary new response, which is why rats learn fleeing but not lever-pressing as avoidance responses (Bolles, 1970).

How do modern accounts explain avoidance? Computational models cast avoidance as actor-critic reinforcement learning, with a critic that learns aversive value and an actor that selects responses, and neural work locates these computations in amygdala-based threat circuits, reconceiving the reinforcer as a nonconscious threat state rather than a felt emotion (Maia, 2010; LeDoux et al., 2017).

Why does avoidance matter for anxiety disorders? Maladaptive avoidance is the behavior that sustains anxiety, because escaping a feared situation prevents the person from learning it is safe; effective treatment uses exposure with response prevention to block avoidance and allow fear to extinguish (Pittig et al., 2018).

References