Classical conditioning is a type of learning in which an organism comes to associate a previously neutral signal with a meaningful event, so that the signal alone eventually triggers a response it never triggered before. The concept traces back to Ivan Pavlov’s work with dogs in the late 1800s, but its influence extends far beyond salivating animals. Classical conditioning helps explain everything from phobias and post-traumatic stress to brand loyalty, placebo effects, and even immune system changes, making it one of the most practically consequential ideas in psychology and neuroscience.
How Pavlov’s Experiment Actually Worked
Pavlov was studying digestion, not learning. He noticed that his dogs, which initially salivated only when food was placed in front of them, gradually started salivating in response to sounds they had come to associate with feeding, like the noise of the food cart approaching. To investigate, he introduced a bell sound shortly before delivering food. At first, the bell produced no salivation. After repeated pairings, the dogs began to salivate at the sound of the bell alone.1PubMed. Classical Conditioning
The vocabulary that came out of this experiment is still used today, so it helps to know the four core terms in plain language:
- Unconditioned stimulus: something that naturally triggers a response without any learning required. In Pavlov’s case, this was the food.
- Unconditioned response: the automatic reaction to that natural trigger. The dogs salivated when they tasted food, no training needed.
- Conditioned stimulus: a formerly neutral signal (the bell) that, after being paired with the unconditioned stimulus, starts to trigger a response on its own.
- Conditioned response: the learned reaction to the conditioned stimulus. The dogs salivated at the bell, even with no food present.
The key insight is that the animal did not choose to respond to the bell. The association formed automatically through repeated pairing. This involuntary quality is what separates classical conditioning from other forms of learning, like deliberately practicing a skill or learning that pressing a lever delivers a treat.
Why Timing Between Signals Changes Everything
Not all classical conditioning works the same way. One of the biggest variables is when the conditioned stimulus appears relative to the unconditioned stimulus, and even small changes in timing engage different brain regions.
In delay conditioning, the conditioned stimulus stays on until the unconditioned stimulus arrives, so there is overlap. A tone plays, and while the tone is still sounding, a puff of air hits your eye. This kind of conditioning is relatively simple, and in eyeblink studies it depends heavily on the cerebellum, a brain structure at the back of the head typically associated with motor coordination.2PubMed Central. Neural circuitry and plasticity mechanisms underlying delay eyeblink conditioning You do not need to be consciously aware of the association for delay conditioning to work.
In trace conditioning, there is a gap between the conditioned stimulus ending and the unconditioned stimulus arriving. The tone plays, goes silent, and then a moment later the air puff comes. Bridging that silent gap requires the hippocampus, a brain region central to forming conscious memories.3PubMed Central. Trace conditioning and the hippocampus: the importance of contiguity Brain imaging in humans confirms this pattern: the cerebellum is active during both types of eyeblink conditioning, but the hippocampus shows significantly more activity during trace conditioning than during delay conditioning.4PubMed Central. Neural substrates underlying human delay and trace eyeblink conditioning
The same delay-versus-trace distinction appears in fear conditioning. Hippocampal damage interferes with trace fear conditioning but leaves delay fear conditioning largely intact, and different subregions of the hippocampus show distinct patterns of activation depending on which type of conditioning was used.5PubMed. Subregion-specific differences in hippocampal activity between Delay and Trace fear conditioning: an immunohistochemical analysis The practical takeaway is that the brain does not have a single “conditioning center.” What looks like one unified learning process from the outside actually recruits different neural hardware depending on the exact circumstances.
Where Fear Memories Are Built
Fear conditioning is the most intensely studied form of classical conditioning in neuroscience, and its circuitry is increasingly well mapped. The amygdala, a small almond-shaped structure deep in the brain, plays a central role. Converging evidence shows that the amygdala is involved in acquiring, storing, and expressing conditioned fear memories.6PubMed Central. Neural circuits and mechanisms involved in Pavlovian fear conditioning: a critical review
Within the amygdala, a subregion called the basolateral complex has received the most attention. The evidence linking changes at its synapses to fear learning is strong, though researchers acknowledge that no single mechanism fully accounts for all the changes that underpin conditioning.7PubMed Central. Fear conditioning and the basolateral amygdala More recent work has expanded the picture: plasticity in the central nucleus of the amygdala also appears to be critical for acquiring conditioned fear, and various types of inhibitory neurons within the amygdala contribute to distinct stages of how stimuli are processed and memories are formed.8PubMed Central. Synaptic encoding of fear memories in the amygdala
Motor conditioning, like the eyeblink reflex, runs on different hardware. That kind of learning depends on the cerebellum, specifically through changes at two sites: one in the cerebellar cortex and another in a deep cerebellar nucleus. Sensory signals about the conditioned stimulus and the unconditioned stimulus converge in these cerebellar structures, and plasticity at both sites appears necessary for the conditioned blink to emerge.9PubMed Central. Neural circuitry and plasticity mechanisms underlying delay eyeblink conditioning
Extinction Is Not Forgetting
If you stop pairing the bell with food, the dog eventually stops salivating to the bell. Pavlov called this extinction, and for decades many researchers assumed it meant the original association had been erased. It has not. There is now extensive evidence that extinction does not destroy the original learning. Instead, it creates a second, competing memory that suppresses the first one.10PubMed. Context, ambiguity, and unlearning: sources of relapse after behavioral extinction
One telling sign is spontaneous recovery: wait long enough after extinction, and the original conditioned response often comes back on its own. Another is renewal: if you extinguish the response in one room and then test the animal in the original training room, the response returns. These phenomena make sense if the conditioned stimulus now has two available meanings, one from the original learning and one from extinction, and the current context determines which one wins. The extinction memory turns out to be particularly context-dependent, more so than the original fear memory, which is why conditioned fears can reappear in new environments even after seemingly successful extinction training.
This has serious implications for therapy. Exposure-based treatments for phobias and anxiety disorders essentially try to produce extinction of conditioned fear. Understanding that the original fear memory is still there, just inhibited, helps explain why relapse happens and why therapy needs to include strategies for making extinction memories more robust across different settings.
Learning Runs on Surprise
One of the most influential ideas in conditioning theory is that learning is driven by prediction error: the difference between what an organism expected and what actually happened. The Rescorla-Wagner model, proposed in the 1970s, formalized this insight. According to the model, both acquiring fear and reducing fear are governed by a common signal reflecting how surprising the outcome was. When the unconditioned stimulus is more intense or more surprising than predicted, learning increases. When it is fully predicted, learning stops, even if the stimulus keeps showing up.11PubMed. The Rescorla-Wagner model, prediction error, and fear learning
This idea turned out to have a direct biological counterpart. Dopamine neurons in the midbrain fire in a pattern that closely mirrors the mathematical prediction error described by learning models. They fire more than their baseline rate when a reward is better than expected, stay quiet when a reward is fully predicted, and drop below baseline when a reward fails to appear.12PubMed Central. Dopamine reward prediction error coding This finding, observed in humans, monkeys, and rodents, was a landmark for understanding how the brain implements learning at the cellular level. It means classical conditioning is not just a psychological observation; it maps onto a concrete neural mechanism for updating predictions about the world.
From Short-Term to Long-Term Memory
A single pairing of a conditioned and unconditioned stimulus can produce a brief change in behavior, but that change fades quickly. Converting a short-lived memory into a lasting one requires fundamentally different cellular machinery. Much of what we know about this comes from research on the sea slug Aplysia, whose large, accessible neurons made it a model organism for studying memory at the molecular level.
A single pulse of the chemical messenger serotonin, mimicking one learning event, produces a temporary strengthening of synaptic connections lasting minutes. This short-term change involves modifications to existing proteins and does not require new gene activity. By contrast, repeated pulses of serotonin trigger new gene expression and the production of new proteins, resulting in strengthened connections that last more than 24 hours.13PubMed Central. Transcriptional regulation of long-term memory in the marine snail Aplysia The distinction between short-term and long-term memory is not just a matter of degree; it reflects a qualitative shift in what the cell is doing. This is one reason why spaced repetition, spreading learning events over time, tends to produce stronger long-term conditioning than cramming them together.
Conditioning the Immune System
Perhaps the most surprising extension of classical conditioning is that immune responses can be conditioned. This was first demonstrated in animals: a taste paired with an immunosuppressive drug produced, upon later exposure to the taste alone, measurable reductions in immune function. Animals given the conditioned taste stimulus showed reduced thymus and spleen weights and a significant drop in the ability of their immune cells to proliferate.14PubMed. Conditioned taste aversion produced by cyclosporine A: concomitant reduction in lymphoid organ weight and splenocyte proliferation
This is not just a laboratory curiosity. Both cellular and antibody-based immune functions can be shifted through associative learning, and these learned immune responses are clinically relevant because they can affect the development and progression of immune-related diseases. The effect also works in humans.15PubMed. Pavlovian Conditioning of Immunological and Neuroendocrine Functions
A proof-of-concept study with 30 kidney transplant patients demonstrated this in a clinical setting. Patients went through a taste-conditioning paradigm in which their immunosuppressive drugs were paired with a distinctive flavor. When the flavor was later presented alone, their T cell activity dropped compared to their baseline immune function under routine drug intake.16PubMed Central. Learned immunosuppressive placebo responses in renal transplant patients The implication is striking: conditioning could one day serve as a dose-reduction strategy, allowing patients to maintain immune suppression while taking less medication and experiencing fewer side effects. The research is still early, but it suggests that the brain-immune connection is far more trainable than most people assume.
Classical Conditioning and PTSD
The connection between classical conditioning and post-traumatic stress disorder is direct. In PTSD, a traumatic event acts as a powerful unconditioned stimulus, and cues present during the trauma, such as sounds, smells, or visual scenes, become conditioned stimuli that can trigger intense fear, flashbacks, and physiological arousal for years or decades after the danger has passed.17PubMed Central. From Pavlov to PTSD: the extinction of conditioned fear in rodents, humans, and anxiety disorders
What makes PTSD distinct from ordinary fear conditioning is the failure of extinction. Most people exposed to trauma gradually stop reacting to reminders as they encounter those reminders without further harm. In PTSD, this process stalls. The fear-conditioning framework has generated a set of experimental paradigms for studying why extinction fails, including tests of fear generalization (whether the fear response spreads to stimuli that merely resemble the original conditioned stimulus), fear inhibition (the ability to suppress fear when safety signals are present), and conditioned avoidance (the tendency to avoid situations associated with the feared stimulus).18PubMed. The clinical applications and practical relevance of human conditioning paradigms for posttraumatic stress disorder Each of these maps onto a recognizable symptom cluster in PTSD, which is why conditioning research has become a backbone of trauma psychology.
Higher-Order Conditioning and Chains of Association
Classical conditioning does not stop at one link in the chain. In second-order conditioning, a first stimulus is paired with a meaningful event (say, a tone paired with food), and then a second stimulus is paired with the first one. The second stimulus can come to trigger a response even though it was never directly paired with food. This shows that conditioned stimuli themselves can serve as reinforcers, allowing associative chains to extend outward from the original experience.19PubMed. The neural substrates of higher-order conditioning: A review
A related phenomenon called sensory preconditioning works in the opposite order: two neutral stimuli are paired first, and then one of them is paired with a meaningful event. Remarkably, the other neutral stimulus, which was never directly paired with anything motivationally significant, also comes to produce a conditioned response. These higher-order effects help explain how conditioning can create complex webs of association in daily life. A song that was playing in a restaurant where you had a great meal might make you think of that restaurant, which makes you feel happy, even though the song itself has no direct connection to anything rewarding.
How Brands Borrow From Pavlov
Evaluative conditioning is a term used in social and consumer psychology for a phenomenon closely related to classical conditioning: your attitude toward something changes simply because it repeatedly appears alongside something you already like or dislike. A brand logo shown next to attractive images, upbeat music, or celebrities can shift how positively you evaluate that brand, sometimes without you being aware of the pairing.20PubMed Central. Evaluative Conditioning: The “How” Question
Whether evaluative conditioning works through exactly the same mechanism as Pavlovian conditioning is actually debated. Some experiments show patterns consistent with signal learning, the classical Pavlovian process, while others show features that do not quite fit, such as effects that persist despite awareness that the pairing was arbitrary, or effects that emerge from a single pairing rather than from repeated exposure.21PubMed. Evaluative Conditioning: Past, Present, and Future The current consensus is that evaluative conditioning probably involves multiple processes, not a single mechanism. Still, the Pavlovian framework gave researchers the vocabulary and experimental logic to study it. Every time an advertisement pairs a product with a pleasant image, the underlying bet is that your brain will form an association, voluntary or not.
Avoidance and the Limits of a Pure Conditioning Account
Classical conditioning elegantly explains how fears are acquired and how involuntary responses develop. Where it runs into trouble is explaining avoidance. Mowrer’s two-factor theory, proposed in the mid-twentieth century, attempted to combine classical and operant conditioning: first, a stimulus becomes feared through classical conditioning; then, the organism learns to avoid that stimulus through operant conditioning because avoidance reduces the fear. The theory remains influential, but it struggles with a well-known empirical finding: animals continue to avoid long after the fear response itself has apparently extinguished.22PubMed. Two-factor theory, the actor-critic model, and conditioned avoidance If the avoidance is supposed to be maintained by fear reduction, it should weaken as the fear weakens. The fact that it does not suggests additional processes are at work, perhaps habit-based mechanisms that operate independently of the conditioned emotional response.
Classical Conditioning in Robotics
The logic of classical conditioning has also been implemented in artificial systems. Researchers have built brain-inspired models that allow robots to acquire, extinguish, and reacquire conditioned associations, mirroring the phases of biological conditioning. One such model was validated on a humanoid robot, demonstrating that the robot could learn to anticipate events based on conditioned signals and could generalize its learned responses to new stimulus speeds.23PubMed Central. Brain-inspired classical conditioning model The interest here is not in making robots salivate. It is in giving machines a biologically plausible way to learn simple predictive associations from their environments without being explicitly programmed for each scenario. Classical conditioning, stripped down to its computational essence, turns out to be a useful blueprint for adaptive behavior in agents that need to operate in changing, unpredictable settings.