Classical Conditioning Notes
Introduction
Classical or Pavlovian conditioning is a form of learning in which two or more environmental events become associated, enabling organisms to anticipate and prepare for what will happen next. In Pavlov’s classic setup, a dog salivated to a sound that had been paired with food powder. The food powder is the unconditioned stimulus (US), which evokes an unconditioned response (UR) on its own. The sound began as a neutral stimulus but, after contiguously presenting it with the US, became a conditioned stimulus (CS) that elicited a conditioned response (CR). The term “conditioned” reflects the CS’s acquired association with the US. Over time, researchers recognized that classical conditioning encompasses more than simple reflexes and involves a broader set of responses. A CS is a stimulus that, after training, predicts the US and typically elicits an anticipatory CR. The US is a stimulus that does not require training to evoke a response and is usually relevant to the organism. CSs and USs are also known as cues/predictors and outcomes/consequences, respectively, though these terms are used somewhat interchangeably. Classical conditioning helps animals prepare for environmental events, enhancing their ability to adapt. The following sections review how views of classical conditioning have evolved, the basic preparation and phenomena studied in the lab, influential theoretical developments, and the role of conditioning in adaptation, therapy, and contemporary psychology.
Historical Background
Pavlovian conditioning was discovered independently in Russia and the United States near the end of the 19th century. Edward Twitmyer, at the University of Pennsylvania, studied whether the knee-jerk reflex could be facilitated by a bell signaling an anticipation of knee tapping with a 500 ms delay. He observed that participants began to extend their knees after hearing the bell even without a knee tap, concluding that a stimulus other than the usual reflex elicited the response. Although Twitmyer presented his findings, the impact was limited and he did not pursue conditioning further. Consequently, Ivan Pavlov is credited with the discovery, making conditioning widely known in the early 20th century, especially in the United States, where researchers such as John B. Watson and Karl Lashley conducted conditioning studies and Watson’s Conditioned Reflexes (1927) helped popularize the field. Early 20th-century work treated conditioning as one fundamental learning process alongside operant conditioning, with basic paradigms extended to fear conditioning and various species. Classical conditioning models initially posited a primarily automatic, low-level process driven by temporal contiguity. Throughout the mid-20th century, this view was challenged by findings such as the blocking effect (Kamin, 1969), supporting the idea that conditioning is more complex and involves higher-level processing. Rescorla (1988) described conditioning as learning about environmental relations that enable the organism to represent its world, highlighting that association formation is substantial and informative for predicting events. Clinical applications also emerged, with Watson, Rosalie Rayner, and Mary Cover Jones illustrating how conditioning can account for fear acquisition and its treatment. Joseph Wolpe’s systematic desensitization extended laboratory findings to real-life anxiety treatment. The translation from basic research to clinical application has continued to grow, with contemporary research embracing contingency judgments and causal learning (Dickinson & Shanks, 1985) and the development of mathematical models like the Rescorla–Wagner framework, which formalize learning as prediction errors. The field today remains closely linked with neuroscience and cognitive psychology and is a vibrant area of study.
Major Concepts
Acquisition is the process by which CRs come under the control of the CS, typically after pairing with the US. The CS is initially neutral but comes to evoke a response similar to the US through training. Several factors influence acquisition: more salient or intense stimuli generally produce faster learning; novelty enhances acquisition via latent inhibition, where pre-exposure to a CS without a US slows subsequent conditioning; and the temporal arrangement of CS and US, including inter-stimulus intervals and inter-trial intervals, also modulates learning. Contiguity refers to the spatial and temporal proximity of CS and US, while contingency concerns the probability of the US given the CS relative to its probability when the CS is absent. In delay conditioning, the US starts with the CS and overlaps; in trace conditioning, there is a gap between the CS offset and US onset; simultaneous conditioning presents CS and US together; backward conditioning presents the US before the CS. Contiguity has historically explained conditioning but later views acknowledge contingency and information value as central, especially after Rescorla (1969). Contingency is interpreted as the CS’s information about the US: positive contingency means the US is more likely with the CS; negative contingency means the CS predicts the absence of the US; zero contingency means the CS provides no information about the US. Excitatory conditioning occurs when the CS predicts the US (positive contingency), whereas conditioned inhibition occurs when the CS predicts the absence of the US (negative contingency). Inhibition typically requires an excitatory context and is harder to measure than excitation. Techniques to assess conditioned inhibition include bidirectional responses (where responses can increase or decrease relative to baseline), the summation test (testing an excitatory CS+ in the presence of an inhibitory CS− to see if the response is reduced), and the retardation test (conditioning the CS− with the US and observing slower reacquisition). Extinction—when a CS is no longer paired with the US—reduces CRs, but its effects are often not permanent. Contextual renewal shows that extinction learning is context-specific; renewal types include ABA (acquisition in A, extinction in B, test in A), ABC (A to B to C), and AAB (acquisition and extinction in A, test in B). Spontaneous recovery refers to the return of CR after a time delay; reinstatement occurs when the US is delivered again after extinction; rapid reacquisition occurs when conditioning is re-established faster than initial learning.
Experimental Preparations
Classical conditioning research exploits various preparations to probe learning processes. Fear conditioning pairs a CS with an aversive US (often a mild electric shock) and is measured via physiological and behavioral indices such as skin conductance (galvanic skin response), pupillary dilation, and approach/avoidance in humans, or freezing in rats. The conditioned suppression paradigm uses lever-pressing for food as a baseline and measures suppression when CS+ is paired with shock, with a suppression ratio calculated as ext{Suppression ratio} = rac{R{CS}}{R{CS} + R{preCS}}$, where $R{CS}$ is the response rate during the CS and $R_{preCS}$ the rate during the baseline period. A suppression ratio approaching 0 indicates strong fear; a ratio around 0.5 indicates little or no fear. Conditioned taste aversion demonstrates how Pavlovian conditioning can generate strong aversions to flavors when paired with illness (e.g., lithium chloride); taste aversion can occur after a single CS–US pairing and can be elicited with delays between CS and illness, revealing preparedness for certain associations. Eye-blink conditioning uses a CS paired with an air puff or mild shock to elicit the reflexive eye-blink; it has been extensively studied for insights into learning mechanisms and neural substrates. Sign-tracking and goal-tracking (autoshaping) involve cues that predict rewards: sign-tracking sees animals approach and interact with a CS (e.g., a illuminated keylight) while goal-tracking involves approaching the US delivery location (e.g., food magazine). Contingency and contiguity influence whether sign-tracking or goal-tracking predominates, with temporal/spatial separation favoring goal-tracking and close CS–US proximity favoring sign-tracking. Drug tolerance can reflect associative tolerance, where the context and cues predicting drug effects elicit anticipatory compensatory responses, reducing observed drug effects in familiar settings. Causal learning studies examine how people infer the causal structure among stimuli, such as meals, chemicals, and sickness, and show that acquisition and extinction patterns resemble classical conditioning, including cue competition phenomena. Conditioned analgesia occurs when stress reduces pain and a CS paired with stress subsequently lowers pain responses, illustrating conditioning of pain modulation.
Key Phenomena
Cue competition encompasses several Pavlovian effects where multiple CSs compete for the CR. Overshadowing occurs when two CSs are trained together with a US, and testing shows a stronger CR to the more salient CS than to the less salient one. Potentiation (a common finding in taste aversion) often yields stronger responses to the less salient CS under certain conditions. Blocking—the classic effect identified by Kamin (1969)—happens when a previously trained CS1 blocks learning about a new CS2 when both are paired with the same US; CS2 is less effective at eliciting a CR when tested alone. Interference occurs when two independent trainings share a common element; proactive interference arises when a prior learning affects a later one, while retroactive interference occurs when later learning affects earlier associations. Cue interference involves shared USs with different CSs; outcome interference involves the same CS with different USs or with extinction. Latent inhibition describes slower learning when a CS is pre-exposed without the US.
Theoretical Developments
Early theories, including Pavlov’s, posited that conditioning arose from a CS–US linkage formed during US presentation. Hull and Spence contributed mechanistic views, and Bush and Mostellar emphasized learning as a function of the maximum possible knowledge gained gradually through error correction. Rescorla and Wagner (1972) formalized learning as a function of prediction error, introducing the equation igtriangleup V = \alpha\beta(\lambda - \sumj Vj), where $V_j$ is the associative strength of cue $j$, \lambda\alpha\beta$$ are salience and learning rate parameters. This model can account for learning with multiple CSs and explains blocking and other competition effects. Mackintosh (1975) and Pearce and Hall (1980) proposed attentional models: Mackintosh argued that cues that are good predictors acquire higher associability and attract more attention, accelerating learning for predictive cues; Pearce and Hall proposed the opposite, suggesting associability is higher for cues that are poor predictors and for cues that are inconsistent, so learning to predict a poor predictor happens faster early in training. Hybrid models attempt to reconcile these views by allowing associability to increase in certain contexts. Wagner (1981) offered the Standard Operating Procedures (SOP) model, in which stimuli exist in neural states (inactive, A1, A2) with activation dynamics; excitatory conditioning occurs when CS and US are in A1, and learning can proceed via activation of multiple elements in memory. Other theories emphasize processing of the US as a complex event with many dimensions, suggesting that learning can occur over multiple attributes or as a configurational (or holistic) stimulus rather than as independent cue–US links. Pearce (1987) extended this configurational approach, arguing that organisms learn about the entire configuration of cues present and that generalization decrements depend on the similarity between the training configuration and test configurations. Miller and Matzel (1988) proposed the comparator hypothesis, positing that learning involves comparing the strength of the target cue with other cues present during training, which can lead to changes in response without new acquisition. Gallistel and Gibbon (2000) emphasized temporal intervals and rate of reinforcement, arguing that behavior reflects comparisons of reinforcement rates between the CS and the surrounding context rather than discrete CS–US links. Contemporary work integrates time-based and prediction- error accounts and continues to test these models against a broad array of conditioning phenomena.
Functional Point of View
From a functional perspective, classical conditioning enables organisms to represent their world and adapt to environmental contingencies, improving survival and reproductive success. Conditioning generalizes across species and response systems (fear, pain, taste, sex, etc.), suggesting an evolutionarily conserved process. A notable finding is that naturally related CS–US relationships produce faster and more robust learning than neutral pairings, reflecting an evolutionary predisposition to attend to ecologically meaningful cues. For example, learning that a particular flavor predicts illness is faster than learning that a sound or visual cue predicts illness, a phenomenon linked to preparedness and the ecological validity of stimuli. Conditioning also extends beyond perceptual representation to influence how organisms interact with the US, not merely the CS, illustrating a broader functional role in behavior and survival. A sexual conditioning example shows that pairing a female cue with copulation results in rapid approach and mating responses, and naturalistic CSs often show more robust resistance to extinction and cue competition.
Applications Outside of the Laboratory
Classical conditioning has informed psychopathology and psychotherapy since its early development. Experiments on experimental neuroses in dogs and humans demonstrated how conditioning could yield adaptive and maladaptive emotional responses to novel stimuli. Watson and Rayner’s fear conditioning work with “Little Albert” illustrated how fear could be acquired through CS–US pairings and generalized to related stimuli, while Mary Cover Jones demonstrated counterconditioning to reduce fear, laying groundwork for systematic desensitization. Wolpe systematized systematic desensitization (SD) by combining anxiety reduction through relaxation with gradual exposure to fear-eliciting stimuli, under the principle of reciprocal inhibition. SD involves: (a) establishing an anxiety hierarchy, (b) training an incompatible response such as relaxation, and (c) gradual exposure to stimuli while maintaining relaxation. Exposure therapy, a broader CBT approach, uses graded exposure, flooding, in vivo exposure, imaginal exposure, and increasingly, virtual reality exposure to reduce phobic and anxiety symptoms, PTSD, OCD, and cue reactivity to drugs or foods. Relapse prevention in exposure therapy considers renewal (context changes), spontaneous recovery (time), reinstatement (re-exposure to the US), and rapid reacquisition. Techniques such as massive extinction (extensive exposure), multiple-context exposure, extinction retrieval cues, and deepened extinction (exposure to multiple anxiety-eliciting cues simultaneously) have shown promise in reducing relapse in both animal and human studies.
Current Status of the Study
Neuroscience has linked classical conditioning to neural substrates, including dopaminergic and noradrenergic systems involved in prediction error signaling and learning. Brain structures such as the basal ganglia, cerebellum, and striatum participate in predicting rewards and guiding behavior, with dopamine-based and norepinephrine-based error correction implicated in conditioning processes. The field also intersects with cognitive psychology, where perception, attention, and other cognitive processes influence learning. Connectionist approaches (delta-rule-inspired) model learning as multiple parallel associations among inputs and outputs, echoing classical conditioning principles. Overall, classical conditioning remains a vibrant, interdisciplinary area connecting behavior, cognition, and neuroscience, with ongoing refinement of theories and expanding clinical applications.
References (Selected)
Pioneering and influential works include Pavlov (1927) on conditioned reflexes; Watson & Rayner (1920) on fear acquisition and generalization; Jones (1924) on elimination of fears; Wolpe (1961) on systematic desensitization; Kamin (1969) on predictability and conditioning; Rescorla & Wagner (1972) on a formal learning model; Mackintosh (1975) and Pearce & Hall (1980) on attentional theories; Miller & Matzel (1988) on the comparator hypothesis; Gallistel & Gibbon (2000) on time, rate, and conditioning; Delamater & Oakeshott (2007); Miguez, Laborda, & Miller (2014) on conditioned analgesia; and Domjan (2005) for a functional perspective on Pavlovian conditioning. A broader set of cross-referenced concepts and researchers also appears in the text, including interactions with conditioning theory, ambiguity, and applications across domains.