Ivan Pavlov’s dog experiments, conducted in the 1890s and early 1900s at his laboratory in St. Petersburg, demonstrated that animals could learn to associate a neutral signal with a meaningful event and respond to the signal alone. Pavlov rang a bell (or, more accurately, used a metronome or buzzer) before feeding his dogs, and after repeated pairings the dogs began salivating at the sound itself, before any food appeared. This process, which Pavlov called “conditional reflexes,” became known as classical conditioning and turned into one of the most consequential discoveries in the history of psychology and neuroscience. Its reach extends well beyond drooling dogs, touching everything from anxiety therapy to drug overdose to how your brain responds to a brand logo.
What Pavlov Actually Did
Pavlov was not a psychologist. He was a physiologist who had already won the 1904 Nobel Prize in Physiology or Medicine for his work on digestion. His laboratory studied how the digestive system regulated itself, and dogs were his primary subjects because he could surgically redirect their salivary ducts to the outside of the cheek, allowing precise measurement of saliva output. He noticed that the dogs’ salivary glands were finely tuned to the food they received: dry, hard food triggered large amounts of saliva, while watery food produced much less, and plain water or saline triggered none at all, because saliva served no purpose for liquids.
The real breakthrough came from what Pavlov initially considered a nuisance. His dogs began salivating before the food arrived, in response to the footsteps of the lab assistant or the sight of the feeding dish. Pavlov called these “psychic secretions” and, rather than dismissing them, spent the next several decades systematically studying them. He would present a neutral stimulus, such as the ticking of a metronome, immediately before delivering food. After enough pairings, the metronome alone produced salivation. The neutral stimulus had become a “conditional stimulus” (often translated as “conditioned stimulus”), the food was the “unconditional stimulus,” the natural drooling at food was the “unconditional response,” and the new drooling at the metronome was the “conditional response.”
How the Learning Takes Shape
Classical conditioning is not simply about two things happening at the same time. Timing and predictability are everything. The conditioned stimulus has to reliably predict the unconditioned stimulus, and the gap between them matters. Research on how quickly animals acquire conditioned responses has shown that learning depends on the temporal relationship between the signal and the reward or threat, specifically the ratio of time between trials versus the duration of the signal itself. When that ratio is large, meaning the signal is a strong predictor because it stands out against a long background of no-signal time, learning happens faster.
This insight contradicts the simple “two things occur together” version of conditioning that many people learn in introductory courses. A tone and food can occur at roughly the same time, but if the tone is no better at predicting food than the general background, the animal won’t develop a conditioned response. The brain is tracking informativeness, not mere coincidence.
The Rescorla-Wagner model, one of the most influential theoretical frameworks in learning science, formalized this idea. It holds that learning is driven by prediction error: the difference between what you expect and what actually happens. When a stimulus is new and an unexpected reward or shock follows, the prediction error is large and learning proceeds quickly. As the animal comes to expect the outcome, the error shrinks and learning slows. This same principle governs both the acquisition of fear and its reduction during therapy, a point with direct clinical relevance.
What Happens in the Brain
Pavlov worked before neuroscience had the tools to peer inside a living brain, so he could only infer what was going on. Modern research has mapped the circuits with considerable precision, and the picture depends on what kind of conditioning you’re talking about.
For fear conditioning, where an animal learns that a tone predicts a shock, the amygdala is the central hub. Specifically, the basolateral amygdala processes the association between the conditioned stimulus and the threat. Studies in rodents show that fear learning results from a strengthening of the neural connections that carry information about the conditioned stimulus into the amygdala, through a process called long-term potentiation.
The amygdala’s role goes beyond just forming the memory. When a fear-conditioned stimulus is encountered again later, the basolateral amygdala can alter how the hippocampus, the brain’s spatial mapping region, represents the environment. Research has found that presenting a fear-conditioned stimulus inside a place cell’s firing area destabilizes that cell’s representation of where the animal is, and that temporarily shutting down the basolateral amygdala prevents this disruption.
Once a fear memory has been formed, it can be destabilized and then re-stabilized each time it is recalled, a process called reconsolidation. Studies have shown that blocking a specific molecular pathway in the basolateral amygdala right after a fear memory is reactivated can weaken the memory.
For simpler motor conditioning, like the classic eyeblink paradigm where a tone predicts a puff of air to the eye, the cerebellum takes center stage. Purkinje cells in the cerebellar cortex learn to pause their constant inhibitory firing at precisely the right moment, allowing the blink to occur just before the air puff arrives. The timing is remarkably precise. A computational model of this process describes the Purkinje cell as learning to lift its brake on a motor pathway with high temporal accuracy so the blink happens at the right instant.
Blocking activity in the interpositus nucleus, a deep cerebellar structure that receives input from Purkinje cells, prevents the animal from acquiring conditioned eyeblink responses altogether, confirming that the cerebellum is not just involved but required for this type of learning.
Extinction Is Not Forgetting
One of the most practically important discoveries about classical conditioning is what happens when you stop pairing the conditioned stimulus with the unconditioned stimulus. If Pavlov rang the metronome over and over without delivering food, the dogs eventually stopped salivating. This is called extinction, and it looks like the learning has been erased. But it hasn’t.
Pavlov himself noticed that extinguished responses could reappear after a rest period, a phenomenon called spontaneous recovery. Modern research has confirmed that extinction does not delete the original association. Instead, it creates a new, competing memory that suppresses the old one. And that new memory is fragile in a way the original is not: it’s highly dependent on context. If you extinguish a conditioned response in one environment and then test the animal in the original training environment, the conditioned response comes roaring back. This phenomenon, called renewal, has been demonstrated across species from rodents to insects. A recent study provided the first evidence of renewal in crickets, showing that extinction in those animals is context-specific, consistent with the theory that extinction is new inhibitory learning layered on top of the original excitatory association.
This matters enormously for therapy. Exposure-based treatments for anxiety and phobias are, at their core, extinction procedures. A person with a spider phobia might be gradually exposed to spiders in a therapist’s office until the fear subsides. But the fear was learned in the real world, and the extinction happened in the clinic. Return to the original context, or encounter the feared object under stress, and the old fear can return. One analysis noted that roughly one in five patients with social anxiety disorder relapse after initially successful extinction-based exposure therapy.
Higher-Order Conditioning
Pavlov also discovered that conditioning could build on itself in layers. Once a dog had learned to salivate at a bell (because the bell predicted food), a new neutral stimulus, say a light, could be paired with the bell alone, without any food. After enough pairings, the light itself produced salivation. The light had never been paired with food directly; its power came entirely from its association with the bell, which had been associated with food. This is called second-order conditioning.
A related phenomenon, sensory preconditioning, works in the opposite order. Two neutral stimuli are paired together first, and then one of them is paired with a meaningful outcome. When the other stimulus is presented later, it also produces a conditioned response, even though it was never directly associated with anything motivationally significant.
These phenomena dramatically extend the reach of classical conditioning, because they mean that new learning can chain off of earlier learning without the original reward or threat ever being present again. In the real world, this helps explain how complex emotional reactions can form to things that were never directly associated with anything harmful or rewarding.
Conditioned Taste Aversion, the Exception That Proves the Rule
Most classical conditioning requires many pairings and works best when the conditioned and unconditioned stimuli are separated by only a few seconds. Conditioned taste aversion breaks both rules. If you eat something unfamiliar and get violently ill hours later, you’ll likely develop an intense aversion to that food after just a single experience. Researchers have shown that this involves a learned reduction in the perceived pleasantness of the taste, not just avoidance of it. The brain actually revalues how the food tastes.
This “one-trial learning” with a long delay was initially controversial because it didn’t fit the standard model. But it makes sense from an evolutionary perspective: an animal that needs twenty exposures to learn that a berry makes it sick won’t survive long. The discovery of taste aversion helped push researchers to recognize that the rules of conditioning are not entirely universal. Different types of associations have different constraints, shaped by the biological relevance of what is being learned.
Drug Tolerance, Overdose, and the Environment
One of the most striking and consequential real-world applications of classical conditioning involves drug tolerance and overdose. When a person regularly uses a drug like heroin in a specific environment, the body learns to associate the sights, sounds, and routines of that environment with the drug. In response, it begins mounting compensatory physiological responses, essentially pre-adjusting in the opposite direction, before the drug even enters the bloodstream. These conditioned compensatory responses help the body counteract the drug’s effects and are a major contributor to tolerance, the need for increasing doses to achieve the same effect.
Research in rats has confirmed this link directly. Drug-associated environmental cues contribute significantly to tolerance. Animal experiments show that conditioned drug-anticipatory responses, alongside the drug’s pharmacological properties, affect mortality from heroin. Further work has demonstrated that blocking specific receptors involved in these compensatory responses can reduce both the expression of opiate tolerance and the conditioned compensatory response itself.
The life-or-death implication is this: if someone who has built up tolerance in a familiar setting uses the same dose in a new environment, the conditioned compensatory responses don’t fire because the environmental cues are absent. Without that physiological pre-adjustment, a dose that was manageable yesterday becomes lethal today. A case report documented exactly this scenario, describing a young man who died from a heroin dose identical to what he had taken the day before in his usual surroundings, simply because the new environment failed to trigger his conditioned tolerance.
This finding has reshaped how harm-reduction programs think about overdose risk. A person returning to drug use after a period of abstinence, or using in a new location, or using under unusual circumstances, faces elevated danger not just because of reduced pharmacological tolerance but because conditioned environmental tolerance has degraded or cannot activate.
Training the Immune System
Perhaps the most surprising frontier of classical conditioning research is its application to the immune system. If you repeatedly pair a distinctive-tasting drink with an immunosuppressive drug, and then later give the drink alone, the immune system shows a measurable suppressive response, as if the drug were present. This has been demonstrated in animals, healthy human volunteers, and patients with immune-related diseases.
The clinical potential is significant. In animal models, behaviorally conditioned immunosuppression has been shown to slow the progression of autoimmune diseases, prolong the survival of transplanted organs, and modify allergic responses. The underlying principle is that associative learning protocols can modulate both cellular and humoral immune functions, and these learned responses are clinically relevant because they affect how immune-related diseases develop and progress.
Human studies have drilled into the mechanism further. In one experiment, researchers found that repeated re-exposures to a conditioned stimulus were necessary to suppress immune cell activity; a single re-exposure wasn’t enough. Equally telling, merely expecting to receive an immunosuppressive drug, without any conditioning history, did not affect immune function. The immune system seems to respond to learned associations rather than conscious expectations, which distinguishes this phenomenon from the purely cognitive side of placebo effects.
The hope driving this research is that conditioning protocols could eventually be used alongside standard medications to reduce drug doses while maintaining therapeutic benefit. A patient with an autoimmune disease or organ transplant might take full-dose immunosuppressive medication during a conditioning phase and then switch to a partial-dose regimen supplemented by conditioned stimuli, reducing side effects while preserving the drug’s effectiveness.
Classical Conditioning in Marketing
Advertisers have been exploiting classical conditioning for decades, though they usually call it “evaluative conditioning.” The logic is straightforward: pair a brand with something pleasant, be it a celebrity, an upbeat song, or a beautiful landscape, and the positive feeling transfers to the brand. Research has identified two pathways through which this can work. Attitude change can result from establishing a memory link between the brand and the pleasant stimulus, or from a more direct transfer of emotion from the stimulus to the brand.
What makes this research interesting is the gap between what people say and what their brains show. Studies using EEG have found that conditioning can alter cortical activity toward both liked and disliked brands, even when participants’ self-reported attitudes don’t budge. In one series of experiments testing audio conditioning with brand names, explicit self-reports and implicit reaction-time tests showed no change, but EEG recordings revealed clear shifts in brain activity at frontal electrode sites after conditioning. The brain was picking up on the associations even when the person wasn’t aware of it, and the neural effects were stronger for negative conditioning than for positive.
This hints at why negative publicity can be so damaging to a brand. A single vivid association between a company and something unpleasant may leave a deeper neural trace than many pleasant associations, even if consumers insist their opinion hasn’t changed.
The Little Albert Controversy
The most famous human demonstration of classical conditioning is the Little Albert experiment, conducted in 1920 by John B. Watson and Rosalie Rayner. They conditioned a young infant to fear a white rat by pairing the rat with a loud, startling noise. The experiment has been a textbook staple for over a century, but who “Albert B.” actually was has remained a mystery, and the answer has turned into one of psychology’s more contentious detective stories.
For years, researchers identified Albert as Douglas Merritte, the son of a wet nurse at the hospital where Watson worked. But a competing investigation presented evidence for a different candidate, Albert Barger, arguing that he matched the described characteristics of Little Albert more closely. The question remains unresolved, though it speaks to the often-sloppy record-keeping of early experimental psychology. The experiment itself is ethically indefensible by modern standards: Watson never deconditioned the child, and the long-term effects of the procedure were never followed up.
Conditioning Across the Animal Kingdom
Classical conditioning is not a quirk of mammalian brains. It appears across an enormous range of species, from sea slugs to humans, suggesting it is one of the most fundamental forms of learning that evolution has produced. In the sea slug Aplysia, researchers have traced the cellular basis of classical conditioning down to individual synapses, finding that a form of synaptic strengthening that depends on specific receptor types is necessary for conditioning to occur.
At the other end of the complexity spectrum, research on fruit flies has shown that the capacity for associative learning itself responds to natural selection. When researchers selectively bred fruit flies for better associative learning of color stimuli over multiple generations, they found that the selected lines not only improved at the trained task but also showed better associative learning for stimuli and rewards they had never been selected on. By the third generation, the selected lines consistently outperformed the control lines. This correlated response suggests that associative learning is underpinned by a general biological mechanism rather than separate, stimulus-specific systems.
The Soviet Politicization of Pavlov
Pavlov’s legacy carries an unusual political dimension. After his death in 1936, the Soviet government elevated his work to something approaching state doctrine. In 1950, Stalin ordered a joint session of the Soviet Academy of Sciences and the Academy of Medical Sciences, known as the “Pavlovian session,” which declared that all of physiology and psychology should be divided into “pre-Pavlovian” and “Pavlovian” stages. The pre-Pavlovian stage was dismissed as idealistic philosophy, while the Pavlovian stage was declared materialistic at its core.
This had real consequences for Soviet science. Researchers who didn’t frame their work in Pavlovian terms faced professional ruin. The irony, as historians have pointed out, is that this kind of intellectual absolutism was in direct contradiction to Pavlov’s own approach. He was a relentless empiricist who revised his ideas when the data demanded it. The man who built his career on careful observation of what organisms actually do would likely have been horrified to see his name turned into an ideological litmus test that suppressed the very scientific freedom he valued.