Psychological Conditioning

Abstract

Psychological conditioning is a form of learning: the process by which experience of the relations between events changes behaviour. This article treats conditioning as an umbrella process with two experimental branches — classical (Pavlovian) conditioning, in which a neutral stimulus paired with a biologically significant one acquires a conditioned response, and operant (instrumental) conditioning, in which a response is strengthened or weakened by its outcome. It sets out the shared phenomena that any account must explain — acquisition, extinction, spontaneous recovery, generalization, and discrimination — and the two findings that overturned the simple contiguity view: that conditioning tracks predictive contingency rather than mere pairing, and that it is constrained by biological preparedness rather than being equipotential across stimuli. A worked example derives the acquisition and extinction of a conditioned response trial by trial.

Keywords: conditioning, classical conditioning, operant conditioning

Conditioning is the oldest experimental subject in the science of learning, and for a century it has been the laboratory model of how an organism adapts to the regularities of its world. Pavlov's salivating dogs and Thorndike's cats in puzzle boxes established its two forms; Watson carried it to human emotion, and Skinner built an entire experimental analysis of behaviour on the second. This article follows conditioning from those origins through the mid-century discovery that it obeys contingency and preparedness rather than raw contiguity, to its modern life as a translational model of fear, extinction, and the neural code for prediction.

What Psychological Conditioning Is

Psychological conditioning is the process by which an organism comes to change its behaviour in response to the arranged relations between events — between one stimulus and another, or between a response and its outcome. Like all learning, it is inferred rather than observed: the conditioning is a theoretical change of state read off a measured difference in behaviour when a contingency between events is imposed and then encountered again. What distinguishes conditioning from other forms of learning is the tightness of the arrangement it studies — a defined signal, a defined outcome, and a controlled relation between them — which is precisely what made it the first part of learning to become a quantitative laboratory science.

The conditioned response is not a mechanical copy of the reflex the outcome evokes. On a functional-behavioural analysis, it is an adaptation that prepares the organism for the biologically important event the signal predicts, so its form reflects what the animal must do about the outcome rather than simply reproducing the reaction the outcome itself provokes (Domjan, 2005). That functional reading anticipates the central modern claim, developed below, that conditioning is the learning of predictive relations among events — the acquisition of information about what signals what — and not the stamping-in of a bond by mere temporal pairing (Rescorla, 1988).

Types of Psychological Conditioning

MeSH classifies psychological conditioning beneath learning and files several subtypes under it. These categories are an indexing classification for the biomedical literature, not a theory of distinct mechanisms: the branches overlap, share their core phenomena, and are better read as different experimental arrangements of the same underlying process than as separate faculties. The two major branches — classical and operant conditioning — are distinguished by what the contingency relates (a stimulus to an outcome, or a response to a consequence), and they routinely operate together in any real episode of learning.

• Classical conditioning — Pavlovian conditioning, in which a neutral stimulus repeatedly paired with a biologically significant one comes to evoke a conditioned response. The contingency is between two stimuli.
• Operant conditioning — instrumental conditioning, in which a response is strengthened or weakened by its consequences, following the law of effect. The contingency is between a response and its outcome.
• Eyelid conditioning — a well-studied classical paradigm in which a tone or light is paired with an air puff to the eye until the tone alone elicits a protective blink; its precise timing makes it a standard model of the cerebellar circuitry of associative timing.
• Automatism — behaviour executed without conscious control or awareness; the term also names the highly automatized, involuntary end state that heavily overtrained conditioned responses can approach.

Classical Conditioning

Pavlov's programme established the first branch. A neutral conditioned stimulus (CS) — a tone, a light — is presented just before an unconditioned stimulus (US) such as food, which already evokes an unconditioned response (UR) of salivation. After repeated pairings the CS alone evokes a conditioned response (CR): the association between CS and US is inferred from the response the CS acquires (Pavlov, 1927). Watson and Rayner then showed that the same procedure reaches human emotion, conditioning a fear response in an infant by pairing a white rat with a startling noise — the demonstration that emotional reactions, not just reflexes, can be conditioned (Watson & Rayner, 1920).

The strength of the CR grows across trials in a negatively accelerated acquisition curve and, when the CS is later presented without the US, declines again in extinction. Neither the rise nor the fall is a simple record of pairing frequency, as the sections below show: what is learned is the predictive relation between the CS and the US. The demonstration below traces both phases, letting the reader run acquisition and then extinction and watch the conditioned response build and subside.

Set how many acquisition trials the conditioned stimulus is paired with the unconditioned stimulus; the CS is then presented alone for 10 extinction trials. Associative strength rises toward its asymptote as the prediction error shrinks, then decays once the outcome is withheld. Extinction drives the response back toward zero — but as new inhibitory learning layered over the original association, not as erasure of it.

Because each trial closes a fixed fraction of the remaining gap, the first few pairings produce the largest gains and the curve bends over as it nears the asymptote — the negatively accelerated learning curve Pavlov first charted. Withholding the US reverses the sign of the error and the response fades, but the residual strength never quite reaches zero, and rest alone can restore part of it.

Operant Conditioning

The second branch begins with Thorndike, who found that cats escaping a puzzle box came to make the successful response more quickly over trials, and named the principle the law of effect: responses followed by a satisfying consequence become more strongly connected to the situation in which they occur (Thorndike, 1911). Skinner turned this into a full experimental analysis of behaviour, in which a freely emitted operant is shaped and maintained by its consequences — reinforcement strengthens it, punishment weakens it — and the schedule on which reinforcement is delivered controls the characteristic pattern and rate of responding (Skinner, 1963).

The power of the response-outcome contingency is shown most starkly when it is illusory. When Skinner delivered food to pigeons on a fixed schedule regardless of what they were doing, the birds nonetheless acquired stereotyped, apparently superstitious rituals — whatever they happened to be doing when food arrived was strengthened, as if the accidental temporal pairing signalled a real contingency (Skinner, 1948). The demonstration below shows how the four basic reinforcement schedules — fixed and variable, ratio and interval — generate their distinctive cumulative response records.

On a cumulative recorder the pen only ever climbs; its slope is the response rate and each tick marks a delivered reinforcer. Choose a schedule and read its signature: ratio schedules (reinforce after a count of responses) drive high rates, interval schedules (reinforce the first response after a delay) drive lower ones, and the fixed versions add the pause or the scallop their variable counterparts smooth away.

The steepest, straightest records belong to the variable-ratio schedule — reinforcement tied to a count of responses, delivered unpredictably, which is exactly why it is the most resistant to extinction and the engine behind gambling. The fixed-interval scallop, by contrast, shows behaviour tracking when reinforcement becomes available rather than how much is worked.

Contingency, Not Contiguity

The classical view held that conditioning is built by contiguity — the sheer co-occurrence of CS and US in time. Rescorla's contingency experiment dismantled it. Holding the number of CS-US pairings constant, he varied the probability of the US in the absence of the CS; when the US was as likely without the CS as with it, no conditioning occurred despite the pairings, and when the CS lowered the probability of the US it became a conditioned inhibitor (Rescorla, 1968). What is learned is the CS's predictive relation to the US, not the tally of times they coincided.

The same lesson emerges from blocking: a cue already predicting an outcome prevents a redundant added cue from being conditioned, because the outcome is no longer surprising (Kamin, 1969). The Rescorla-Wagner model made this quantitative, holding that associative strength changes in proportion to the prediction error — the discrepancy between the outcome and the outcome the present cues already predict: ΔV = αβ(λ − ΣV) (Rescorla & Wagner, 1972). Learning is fast when the outcome is surprising and slows to nothing as it comes to be predicted, and Rescorla later drew the general moral that conditioning is the learning of relations among events rather than the transfer of a reflex by contiguity (Rescorla, 1988).

Biological Preparedness

Conditioning is also not equipotential: not every stimulus is equally associable with every outcome. Garcia and Koelling showed that rats readily learn to associate a novel taste with later illness but not with an immediate shock, and readily associate an audiovisual cue with shock but not with illness — the pairings that matter biologically are learned in a single trial across long delays, while the biologically arbitrary ones are barely learned at all (Garcia & Koelling, 1966). Seligman generalised the finding as a preparedness continuum: evolution has tuned each species to acquire some CS-US relations far more readily than others, so the laws of learning are not fully general but shaped by the organism's biology (Seligman, 1970).

Preparedness explains why the equipotentiality assumption behind early behaviourism was wrong, and why conditioned responses have the adaptive form a functional analysis predicts (Domjan, 2005). The core phenomena that any account of conditioning must reproduce — whichever branch and whatever the prepared constraints — are summarised in Table 1.

The gradient in the fourth row is itself directly measurable, and the demonstration below plots it: a conditioned response spreads to stimuli near the trained value and narrows when a nearby value is explicitly not reinforced.

After a response is conditioned to one value on a stimulus dimension (the CS+, here at 5), it spreads to neighbouring values in a smooth gradient. Turn on discrimination training — in which a nearby value (the CS−, at 3.5) is explicitly non-reinforced — and the gradient sharpens and its peak shifts away from the CS−, the classic peak-shift effect. Move the test stimulus to read the generalized response.

A broad gradient is adaptive when the world is noisy: a stimulus close to the trained one is likely to share its consequences. Discrimination training tells the animal the two values differ, and the gradient narrows around what now matters — often overshooting to a value more extreme than the CS+ itself, evidence that organisms learn the relation between stimuli, not just the absolute trained value.

Worked Example

The rise and fall of a conditioned response can be worked out by hand from the single-cue form of the Rescorla-Wagner rule, ΔV = αβ(λ − V), with only one CS present. Take a combined learning rate αβ = 0.30 and, during acquisition, an outcome of asymptotic strength λ = 100. Because only one cue is present, strength follows V_n = 100(1 − 0.7ⁿ): V₁ = 30, V₂ = 51, V₄ = 76.0, and after eight acquisition trials V = 100(1 − 0.7⁸) = 94.2. The gain shrinks each trial because the term (λ − V) — the prediction error — is consumed as the CS comes to predict the US, which is why the acquisition curve bends over toward its ceiling.

Now switch to extinction: present the CS alone, so the outcome the CS predicts is λ = 0. The rule becomes ΔV = 0.30 × (0 − V) = −0.30V, so strength decays geometrically from its acquired value, V_n = 94.2 × 0.7ⁿ. The first extinction trial drops the CR from 94.2 to 66.0, the third to 32.3, and the fifth to 15.8. Extinction is fast, but note what the arithmetic does not say: the residual strength approaches zero only asymptotically, and the model's excitatory link is never negative — real extinction leaves an inhibitory memory that competes with the original, which is why the response recovers with time and context (Bouton, 2004). Figure 1 plots the eight acquisition trials and the five extinction trials as a single curve.

Discussion

Conditioning has proved one of the most durable units of explanation in psychology, but its modern form is far from the reflex-substitution it began as. Pavlov and Thorndike established that stimulus-stimulus and response-outcome relations can be measured (Pavlov, 1927); (Thorndike, 1911); Watson showed the procedure reaches human emotion (Watson & Rayner, 1920); and Skinner built the experimental analysis of behaviour on the response-outcome contingency and its schedules (Skinner, 1963). Two mid-century findings then dismantled the simple contiguity account on which all of this had rested: conditioning depends on predictive contingency, not the count of pairings (Rescorla, 1968), and it is constrained by biological preparedness rather than being equipotential (Garcia & Koelling, 1966); (Seligman, 1970).

Together these reframed conditioning as the learning of relations among events — the acquisition of information about what predicts what — and the Rescorla-Wagner model gave that idea a precise, error-driven form that explains blocking and the shape of the acquisition curve (Rescorla & Wagner, 1972); (Rescorla, 1988). What conditioning is not, on the modern view, is the automatic imprinting of a habit by repetition; it is a sensitive, adaptive registration of the causal texture of the environment, and its two branches are windows onto the same process rather than separate faculties.

Current Directions

The most active work treats conditioning as a translational model of clinical fear and its treatment. Understanding extinction as new, context-dependent inhibitory learning rather than erasure explains why treated fears relapse, and reframes exposure therapy as the task of making the new, safe learning win the retrieval competition across contexts (Craske, Hermans, & Vervliet, 2018). The behavioural and neurobiological mechanisms of extinction — the recovery phenomena and the amygdala-prefrontal-hippocampal circuits that suppress a learned response without deleting it — have been synthesised across Pavlovian and instrumental conditioning (Bouton, Maren, & McNally, 2021), and the way stress modulates those same circuits is now a central concern for the disorders in which extinction fails (Maren & Holmes, 2016).

A second front is computational and neural. The prediction-error idea at the heart of the contingency view has been tied to the brain's construction of predictive models in the orbitofrontal-amygdala circuit, linking conditioning to model-based reinforcement learning rather than to the accretion of simple bonds (Sharpe & Schoenbaum, 2016). Running through both fronts is the older, unresolved question of how much of conditioning is a cached associative strength and how much is an inference about causal structure — a question the functional analysis of conditioned responding keeps in view (Domjan, 2005).

Common Misconceptions

Conditioning is just automatic pairing: put two things together often enough and a link forms.

Mere co-occurrence is neither necessary nor sufficient. Rescorla's contingency experiment showed that pairings produce no learning when the outcome is just as likely without the cue, and blocking shows a redundant cue is barely learned however often it is paired — conditioning tracks predictive contingency, not the tally of pairings (Rescorla, 1968); (Kamin, 1969).

Extinction erases what was learned.

Presenting the signal without its outcome suppresses the response but leaves the original learning intact: it returns with a change of context, the passage of time (spontaneous recovery), or a reminder. Extinction is new inhibitory learning layered over the old link, not its deletion (Bouton, 2004); (Bouton et al., 2021).

Any stimulus can be conditioned to any response.

Conditioning is not equipotential. Rats associate taste with illness and audiovisual cues with shock, but not the reverse, and some prepared relations are learned in a single trial across long delays — biology constrains which associations form readily (Garcia & Koelling, 1966); (Seligman, 1970).

Glossary

Acquisition.

The phase in which a conditioned response develops and strengthens as the signal is paired with, or the response is followed by, its outcome.

Classical (Pavlovian) conditioning.

Conditioning in which a neutral stimulus paired with a biologically significant one comes to evoke a conditioned response; the contingency is between two stimuli.

Conditioned response (CR).

The learned response evoked by a conditioned stimulus after it has been paired with an unconditioned stimulus.

Conditioned stimulus (CS).

An initially neutral stimulus that, through pairing with a biologically significant one, comes to evoke a learned response.

Contingency.

The predictive relation between two events — how much one changes the probability of the other; the true content of what conditioning learns, as distinct from mere contiguity.

Discrimination.

The narrowing of a conditioned response to a specific signal when a similar signal is explicitly not reinforced, sharpening the generalization gradient.

Extinction.

The reduction of a conditioned response when the signal is presented without its outcome; new inhibitory learning that suppresses the original association rather than erasing it.

Generalization.

The spread of a conditioned response to stimuli resembling the trained one, in proportion to their similarity along a stimulus dimension.

Law of effect.

Thorndike's principle that responses followed by a satisfying consequence become more strongly connected to the situation in which they occur; the basis of operant conditioning.

Operant (instrumental) conditioning.

Conditioning in which a response is strengthened or weakened by its consequences; the contingency is between a response and its outcome.

Prediction error.

The discrepancy between the outcome that occurs and the outcome the present cues already predict; the driving term of the Rescorla-Wagner model, so learning is proportional to surprise.

Preparedness.

The evolved bias that makes an organism acquire some CS-US or response-outcome relations far more readily than others, so conditioning is not equipotential.

Reinforcement.

A consequence that strengthens the response it follows; punishment is the complementary consequence that weakens it.

Schedule of reinforcement.

The rule relating responses to reinforcers — fixed or variable, ratio or interval — which controls the characteristic pattern and rate of operant responding.

Spontaneous recovery.

The return of an extinguished conditioned response after a rest interval, showing that extinction did not erase the original learning.

Unconditioned stimulus (US).

A stimulus that evokes a response without prior learning; the biologically significant event whose association a conditioned stimulus comes to signal.

Key Researchers

Mark E. Bouton. Established extinction as new, context-dependent inhibitory learning and characterized renewal, spontaneous recovery, and reinstatement across Pavlovian and instrumental conditioning. Google Scholar - Faculty page

Michelle G. Craske. Translates extinction learning into exposure-based treatment for fear and anxiety, reframing therapy as inhibitory learning that must win a retrieval competition. Wikipedia - Wikidata - Google Scholar - Faculty page

John Garcia (1917-2012). Discovered conditioned taste aversion — the Garcia effect — showing that conditioning is constrained by biological preparedness rather than being equipotential. Wikipedia

Stephen Maren. Studies the neurobiology of Pavlovian fear conditioning and extinction, and how stress modulates the circuits that suppress a learned association without deleting it. ORCID - Google Scholar - Faculty page

Ivan Pavlov (1849-1936). Founded the experimental study of classical conditioning, demonstrating that a neutral stimulus paired with a biologically significant one comes to evoke a learned, conditioned response. Wikipedia - Wikidata

Robert A. Rescorla (1940-2020). Showed that conditioning tracks predictive contingency rather than contiguity and co-authored the Rescorla-Wagner prediction-error model of associative learning. Wikipedia - Wikidata

Martin E. P. Seligman. Introduced the preparedness continuum, generalising Garcia's finding into the claim that the laws of learning are shaped by each species' biology. Wikipedia - Google Scholar - Faculty page

B. F. Skinner (1904-1990). Founded operant conditioning and the experimental analysis of behaviour, showing that responses are shaped and maintained by their consequences under schedules of reinforcement. Wikipedia - Wikidata

Edward L. Thorndike (1874-1949). Formulated the law of effect from the puzzle-box experiments, establishing instrumental conditioning as the complement to Pavlovian conditioning. Wikipedia - Wikidata

John B. Watson (1878-1958). Founded behaviourism and, with Rosalie Rayner, demonstrated conditioned emotional reactions in the Little Albert study, extending conditioning to human emotion. Wikipedia

Frequently Asked Questions

What is psychological conditioning? Psychological conditioning is a form of learning in which experience of the relation between events changes behaviour, so that a signal comes to evoke a response or an action comes to be governed by its consequences. It is inferred from a measured change in responding when a contingency between events is arranged (Domjan, 2005).

What is the difference between classical and operant conditioning? Classical (Pavlovian) conditioning links two stimuli: a neutral cue paired with a biologically significant one comes to evoke a conditioned response. Operant (instrumental) conditioning links a response to its outcome: an action followed by a satisfying consequence is strengthened, following the law of effect (Pavlov, 1927); (Thorndike, 1911).

Does conditioning happen just because two things occur together? No. Rescorla's contingency experiment held the number of pairings constant and varied how often the outcome occurred without the cue; when the outcome was as likely without the cue as with it, no conditioning occurred. Conditioning tracks the predictive contingency between events, not the count of pairings (Rescorla, 1968).

What is the Rescorla-Wagner model? It holds that associative strength changes in proportion to the prediction error, the gap between the outcome that occurs and the outcome the cues present already predict. Learning is fast when the outcome is surprising and slows as it becomes predicted, which explains the acquisition curve and blocking (Rescorla & Wagner, 1972); (Kamin, 1969).

Can any stimulus be conditioned to any response? No. Garcia and Koelling showed that rats associate taste with illness and audiovisual cues with shock, but not the reverse. Organisms are biologically prepared to acquire some relations far more readily than others, so conditioning is not equipotential (Garcia & Koelling, 1966); (Seligman, 1970).

Does extinction erase a conditioned response? No. Presenting the signal without its outcome reduces the response, but it returns with a change of context, the passage of time, or a reminder of the outcome. Extinction is new, context-dependent inhibitory learning laid over the original link rather than its deletion (Bouton, 2004); (Bouton et al., 2021).

How does conditioning explain superstitious behaviour? When Skinner delivered food to pigeons on a fixed schedule regardless of their behaviour, the birds acquired stereotyped rituals: whatever they happened to be doing when food arrived was accidentally strengthened, as if the temporal coincidence signalled a real response-outcome contingency (Skinner, 1948).

Why is conditioning important for treating fear and anxiety? Exposure therapy is extinction learning, and understanding extinction as inhibitory rather than erasive explains why treated fears relapse and how to make the new, safe learning generalise across contexts. This makes conditioning a working translational model for anxiety-related disorders (Craske et al., 2018); (Maren & Holmes, 2016).

References