
Definition
Classical conditioning is learning that one stimulus predicts another, so that a reflex-like response comes to be triggered by a previously neutral cue (Pavlov's dogs salivating to a signal). Operant conditioning is learning that a behavior has consequences, so that behavior becomes more or less frequent depending on what follows it (a rat pressing a lever for food).
A student is studying for a psychology exam. She reviews Pavlov and the dogs that salivated at a signal, then Skinner and the rats pressing levers for food. Each example makes sense on its own. Then the exam asks, "How does classical conditioning differ from operant conditioning?" and she freezes. She knows both involve learning, that one is about stimuli and the other about behavior, but the exact difference slips away. If that sounds familiar, this guide is for you.
The short version
- Classical (Pavlovian) conditioning pairs two stimuli. The learner acquires a response to a cue that predicts something important. The response is elicited (it happens to the learner).
- Operant (instrumental) conditioning links a behavior to its consequence. The learner acts and the environment responds. The behavior is emitted (the learner does it).
- Both show acquisition, extinction, generalization and discrimination, but the procedures differ.
- Real behavior usually involves both at once. Phobias and avoidance are the standard example.
Classical conditioning
In classical conditioning an organism learns that one stimulus signals another. Ivan Pavlov, who won the 1904 Nobel Prize in Physiology or Medicine for his work on digestion, noticed that dogs in his laboratory began to salivate before food arrived. He turned that observation into a research program, described in Conditioned Reflexes (Pavlov, 1927).
A note on the textbook picture: the dog-and-bell image is a simplification. According to Pavlov's biographer Daniel Todes (2014), the laboratory relied mostly on stimuli whose timing could be tightly controlled, such as metronomes and buzzers, and bells were rare. The principle is the same whatever the signal.
The vocabulary you need:
- Unconditioned stimulus (US): naturally produces a response (food).
- Unconditioned response (UR): the natural reaction (salivation).
- Neutral stimulus (NS): initially produces no relevant response (a tone).
- Conditioned stimulus (CS): the neutral stimulus after it has been paired with the US.
- Conditioned response (CR): the learned response to the CS (salivation to the tone).
Timing matters: the CS generally has to come before, or overlap with, the US. A tone that sounds only after the food is typically a poor predictor and produces weak conditioning. More important than sheer closeness in time is information: Rescorla (1967, 1988) showed that what the organism learns is whether the CS predicts the US, a point that reshaped the field (more on this in the section on how the model evolved, below).
For a deeper treatment, see our guide to classical conditioning and Pavlov's experiments.
Operant conditioning
In operant conditioning, behavior is shaped by its consequences. Edward Thorndike's puzzle-box experiments with cats (Thorndike, 1898) led to his law of effect: behaviors followed by satisfying consequences become more likely, and those followed by discomfort become less likely. B. F. Skinner developed the approach systematically, using what became known as the operant chamber, or "Skinner box," and set out the research in The Behavior of Organisms (Skinner, 1938) and Science and Human Behavior (Skinner, 1953).
There are four basic consequences. The words positive and negative mean "add" and "remove," not "good" and "bad."
| Type | What happens | Effect on behavior | Example |
|---|---|---|---|
| Positive reinforcement | Something pleasant is added | Increases | Praise after finishing homework |
| Negative reinforcement | Something unpleasant is removed | Increases | Taking an aspirin removes a headache, so you take it again next time |
| Positive punishment | Something unpleasant is added | Decreases | A speeding ticket |
| Negative punishment | Something pleasant is removed | Decreases | Losing phone privileges for skipping study time |
The core difference
The two models describe different kinds of learning, not two versions of one process.
- In classical conditioning the learner links stimulus with stimulus. The response is largely involuntary (salivation, a racing pulse, fear). The learner does nothing to produce the outcome.
- In operant conditioning the learner links behavior with consequence. The behavior is "voluntary" in the sense that the organism acts on the environment and the environment responds.
A quick test: does the response happen to the learner, or does the learner do something to get an outcome? The first is classical; the second is operant.
A matching test on the stimuli: in classical conditioning the US comes whether or not the organism responds; in operant conditioning the consequence depends on what the organism does.
The methods differ in practice too. Classical conditioning is typically studied by measuring reflex-like responses, such as salivation or skin conductance, while operant conditioning is studied by measuring rates of behavior, such as lever presses per minute. A way to remember the contrast: in classical conditioning the world acts on the organism; in operant conditioning the organism acts on the world.
Side-by-side comparison
| Dimension | Classical (Pavlov) | Operant (Skinner, Thorndike) |
|---|---|---|
| What is associated | Stimulus with stimulus | Behavior with consequence |
| Type of response | Elicited, reflex-like (involuntary) | Emitted, goal-directed (voluntary) |
| Learner's role | Receives stimuli | Acts on the environment |
| What drives learning | Whether the CS predicts the US | The consequence that follows the behavior |
| Classic example | Dog salivates to a signal that predicts food | Rat presses a lever to get food |
| Everyday example | Feeling anxious at the sound of a dentist's drill | Studying because good grades have been rewarded |
| Typical clinical uses | Exposure therapy for fears and phobias | Behavior modification, token economies, parent training |
The key question for telling them apart: is the behavior something that happens to the subject (classical) or something the subject does (operant)?
Shared processes
Both models include processes that exams love to ask about, and the vocabulary is the same even though the procedures differ.
Acquisition is the stage in which the response is learned. In classical conditioning it grows with CS–US pairings; in operant conditioning it grows as the behavior is reinforced.
Generalization
In classical conditioning, generalization occurs when a stimulus similar to the CS also produces the CR: a dog conditioned to one tone may also respond to a similar tone. In Watson and Rayner's (1920) study of Little Albert, the fear he learned toward a white rat was reported to spread to other white, furry objects. In operant conditioning, generalization occurs when a behavior reinforced in one context shows up in similar contexts. A child praised for saying "please" at home may start saying it at school.
Discrimination
Discrimination is the complementary process: the learner comes to respond only to the specific stimulus and not to similar ones. In classical conditioning this is achieved by pairing the CS with the US while similar stimuli appear without it. In operant conditioning it is achieved by reinforcing the behavior in the presence of a discriminative stimulus, a signal that a behavior will pay off, and not in its absence.
Extinction
Extinction is the weakening of a learned response when the learned relation stops. Classically, the CS is presented repeatedly without the US and the CR fades. Operantly, the behavior is emitted without reinforcement and becomes less frequent. Extinction is not forgetting: the original learning remains and the response can return, for example through spontaneous recovery or when the context changes (Bouton, 2004). That is one reason a fear treated with exposure can come back when the context changes.
Clinical applications of classical conditioning
Classical conditioning helps explain how many fear and anxiety responses are acquired. A person who survives a car accident may feel anxious when driving or near vehicles. The accident (US) produced fear (UR), and stimuli present at the time, such as the car and the road, became conditioned stimuli that trigger fear (CR).
Acquisition, generalization and extinction give researchers and clinicians a rigorous framework for how fears develop and persist. Exposure therapy, one of the best-studied treatments for anxiety disorders, can be understood as extinction of a conditioned fear response. Fear learning also follows the prediction-error logic described below: reviewing the Rescorla–Wagner model, Yau and McNally (2023) conclude that learning to fear and learning to reduce fear are governed by a common, signed prediction error.
Textbook applications include:
- Exposure therapy for phobias: presenting the feared stimulus without the harmful outcome until the fear response extinguishes.
- Systematic desensitization: pairing the feared stimulus with a state of relaxation that is incompatible with anxiety, an approach Wolpe (1958) described as reciprocal inhibition.
- Aversion therapy: pairing an attractive but harmful stimulus, such as alcohol, with an unpleasant consequence to reduce its appeal. Textbooks present it mainly as an illustration of counterconditioning.
- The enuresis alarm: described in the note that follows.
A note on classical conditioning in medicine
The enuresis (bedwetting) alarm is a classic application. Mowrer and Mowrer (1938) described a pad-and-bell device that sounds when moisture is detected, and it has long been explained in classical terms, although operant accounts are also discussed. The example shows how a laboratory principle can become a practical tool. Textbooks present it as a historical illustration, not as a recommendation for any individual child.
Clinical applications of operant conditioning
Operant conditioning is the root of behavior modification and a core ingredient of cognitive-behavioral therapy. The central idea is that problem behaviors persist because their consequences reinforce them, and that changing the consequences can change the behavior.
Techniques derived from operant conditioning include:
- Differential reinforcement: reinforcing an alternative behavior, ideally one incompatible with the problem behavior, in its place.
- Token economy: a system in which desirable behaviors earn tokens that can be exchanged for rewards. Ayllon and Azrin (1965) used tokens as conditioned reinforcers on a psychiatric hospital ward; the reinforced behaviors dropped when reinforcement stopped and returned when it resumed. Token systems are also used in classrooms.
- Extinction: withdrawing the reinforcement that maintains a problem behavior. A child who has tantrums to get attention stops getting it when the tantrums occur.
- Shaping: reinforcing successive approximations of a target behavior, which lets complex skills be taught step by step (Skinner, 1953).
Behavior that looks senseless often turns out to be maintained by its consequences. In a classic study of self-injury, higher levels of the behavior were consistently tied to a specific environmental condition, such as demands or social attention, in six of nine participants (Iwata et al., 1994). That logic, now called functional analysis, asks what a behavior gets or avoids. Reviews of rodent research also report sex differences in acquisition and extinction across several classical and operant tasks (Dalla & Shors, 2009), one reason lab findings should be generalized with care.
An everyday example that ties it together
Take someone who cannot stop snacking in the evening even though they know it hurts. Classically, cues such as the kitchen light, the time of day and the sight of the package can trigger craving before a bite is taken. Operantly, eating produces an immediate pleasant consequence, which reinforces the habit, and giving in also ends the discomfort of the craving, which is negative reinforcement. Both processes run at the same time. Removing snacks from the house (stimulus control) works on the cues but leaves the reinforcement loop untouched; changing what happens after each urge works on the consequences. A comprehensive approach identifies both the conditioned stimuli and the consequences that maintain the behavior.
Parenting and classrooms
Operant principles operate whether or not anyone names them. If a tantrum in the grocery store gets a toy, the tantrum is positively reinforced, and a parent who gives in to end the scene is negatively reinforced by the relief. Because intermittent reinforcement resists extinction, occasionally giving in can make the behavior harder to stop than if it had always been rewarded.
Parents use these techniques daily without the technical vocabulary: praising a child for tidying a room is positive reinforcement, and withdrawing screen time for skipped homework is negative punishment.
The catch is that operant conditioning does not distinguish between desirable and undesirable behavior: it strengthens what works. A parent does not have to intend to reinforce a tantrum; if the consequence favors the child, the tantrum is reinforced. Inconsistent consequences amount to an intermittent schedule of reinforcement, the pattern most resistant to extinction, which is one reason consistency matters.
Common mistakes that cost points on exams
- Confusing negative reinforcement with punishment. Negative reinforcement increases a behavior by removing something unpleasant. Punishment decreases a behavior.
- Treating "positive" as "good." Positive means a stimulus is added, negative means it is removed.
- Forgetting the order in classical conditioning. The CS should signal the US, so it normally precedes it. A tone after the food is a weak design.
- Saying extinction erases learning. It produces new, competing learning, which is why responses can return.
- Calling every learned behavior "operant." Emotional reactions to cues (fear at the dentist's chair) are typically classical.
- Forgetting that both can operate together. The avoidance example earns credit. So does a smoker: cigarettes may become linked with certain settings and moods (classical) while the effects of nicotine reinforce the habit (operant).
- Misattributing the examples. Pavlov is classical, Skinner and Thorndike are operant, Watson and Rayner (1920) is classical (Little Albert).
Reinforcement schedules (operant only)
Ferster and Skinner (1957) described how the pattern of reinforcement shapes behavior. Continuous reinforcement means every response is reinforced. Intermittent schedules reinforce only some responses, and behavior learned this way is generally more resistant to extinction.
- Fixed ratio: reinforcement after a set number of responses (pay per ten items assembled). Produces a high rate with short pauses after each reward.
- Variable ratio: reinforcement after an unpredictable number of responses. Produces high, steady responding and strong resistance to extinction. Slot machines are the textbook example: the reward arrives unpredictably, so the behavior keeps going even though most attempts go unrewarded.
- Fixed interval: the first response after a set time is reinforced. Responding drops right after a reward and speeds up as the next one approaches, like studying harder just before a scheduled exam.
- Variable interval: reinforcement after unpredictable time periods. Produces a slow, steady rate, like checking email through the day.
The general rule is that intermittent schedules produce behavior more resistant to extinction than continuous ones. That has direct clinical implications: if a problem behavior is maintained by intermittent reinforcement, such as the occasional attention a child gets for tantrums, it is harder to extinguish than if reinforcement had always been delivered.
Higher-order conditioning
Classical conditioning is not limited to pairing a neutral stimulus with a US. Once a CS is established, it can serve as the basis for new learning. If a dog already salivates to a tone (the CS), a light can be presented just before the tone. After several repetitions the light alone can produce salivation, even though it was never paired directly with food. This is called second-order conditioning, which Pavlov (1927) also described.
Higher-order conditioning helps explain why emotional responses spread to stimuli ever further from the original. Someone who developed a fear of driving after an accident may come to fear cars, highways, bridges and even road maps, each new stimulus linked to the fear through a chain of pairings.
Related work on evaluative conditioning studies how liking for a stimulus changes after it has been paired with positive or negative stimuli (De Houwer et al., 2001). Pairing a product with attractive images or well-liked people is the kind of arrangement studied under that heading.
How the model evolved: from stimulus to prediction error
Classical conditioning theory has changed since Pavlov's original formulation. The Rescorla–Wagner model (Rescorla & Wagner, 1972) proposed that conditioning depends not only on the time relation between CS and US but on prediction error: the difference between what the organism expects and what actually happens.
In the model, learning is proportional to surprise. If the CS fully predicts the US, nothing new is learned; if the US is unexpected, learning is greatest. The principle explains what simple contiguity cannot, such as blocking: if cue A already predicts the US, adding cue B produces little conditioning of B even though B and the US occur together, because the organism is not surprised.
The idea has moved beyond experimental psychology and into the neuroscience of learning. Dopamine neurons recorded in primates respond strongly to an unexpected reward, little to a fully predicted one, and dip when an expected reward fails to arrive (Schultz et al., 1997). Yau and McNally (2023) reviewed how the model applies to fear learning, concluding that learning to fear and learning to reduce fear are governed by a common, signed prediction error, while also noting findings the model cannot explain.
How the two work together
Real-world behavior rarely belongs to just one model. The standard example is avoidance, explained by two-factor theory (Mowrer, 1951). Suppose someone has a panic attack in a crowded elevator. By classical conditioning, elevators become a conditioned stimulus that triggers fear. Then, by operant conditioning, avoiding elevators brings relief, which is negative reinforcement, so avoidance strengthens. That is why exposure therapy often targets both: it presents the feared cue without the feared outcome (extinction of the classically conditioned fear) and prevents the avoidance that has been keeping the fear alive.
The same structure appears in social anxiety treated with cognitive-behavioral therapy. Anxiety can become conditioned to social situations (classical: the situation is the CS that triggers anxiety), while avoidance persists because it lowers anxiety in the short term (operant: avoidance is negatively reinforced). Effective treatment combines exposure, to extinguish the conditioned fear, with prevention of avoidance, to break the cycle of negative reinforcement. Both models are needed to explain why the problem exists and why it persists.
A student's question
"Is the difference just that classical is involuntary and operant is voluntary?" That is a good first pass, but it is incomplete. The more precise difference is what is contingent on what. In classical conditioning the US follows the CS regardless of what the learner does. In operant conditioning the consequence depends on what the learner does. "Voluntary" is shorthand for that dependence on action.
Frequently asked questions
Did Pavlov discover classical conditioning or invent it?
He discovered it. Pavlov was studying digestion, work that earned him the 1904 Nobel Prize in Physiology or Medicine, when he noticed that his dogs salivated before food arrived, and he turned that observation into a research program (Todes, 2014).
What is the difference between reinforcement and punishment?
Reinforcement makes a behavior more likely; punishment makes it less likely. Each can be positive (add something) or negative (remove something).
Does classical conditioning explain phobias?
Partly. It explains how a cue can become associated with fear after an aversive event. But phobias can also be learned by watching others or by being told about danger, and avoidance keeps them going through operant processes.
What is the Rescorla–Wagner model?
A model of classical conditioning in which learning depends on prediction error: you learn when an outcome is more or less than expected, and little when it is fully predicted (Rescorla & Wagner, 1972).
Which is which: Pavlov or Skinner?
Pavlov is classical conditioning. Skinner (building on Thorndike) is operant conditioning.
Does social learning fit either model?
Not entirely. Learning by observing others goes beyond both; see Bandura's social learning theory.
Keep reading
- Classical conditioning: Pavlov's experiments explained — the full model, extinction and exposure therapy.
- Bandura's social learning theory — learning by observation, beyond direct conditioning.
References
- Ayllon, T., & Azrin, N. H. (1965). The measurement and reinforcement of behavior of psychotics. Journal of the Experimental Analysis of Behavior, 8(6), 357–383. https://doi.org/10.1901/jeab.1965.8-357
- Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning & Memory, 11(5), 485–494. https://doi.org/10.1101/lm.78804
- Dalla, C., & Shors, T. J. (2009). Sex differences in learning processes of classical and operant conditioning. Physiology & Behavior, 97(2), 229–238. https://doi.org/10.1016/j.physbeh.2009.02.035
- De Houwer, J., Thomas, S., & Baeyens, F. (2001). Association learning of likes and dislikes: A review of 25 years of research on human evaluative conditioning. Psychological Bulletin, 127(6), 853–869. https://doi.org/10.1037/0033-2909.127.6.853
- Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. Appleton-Century-Crofts.
- Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994). Toward a functional analysis of self-injury. Journal of Applied Behavior Analysis, 27(2), 197–209. https://doi.org/10.1901/jaba.1994.27-197 (Original work published 1982)
- Mowrer, O. H. (1951). Two-factor learning theory: Summary and comment. Psychological Review, 58(5), 350–354. https://doi.org/10.1037/h0058956
- Mowrer, O. H., & Mowrer, W. M. (1938). Enuresis: A method for its study and treatment. American Journal of Orthopsychiatry, 8(3), 436–459. https://doi.org/10.1111/j.1939-0025.1938.tb06395.x
- Pavlov, I. P. (1927). Conditioned reflexes: An investigation of the physiological activity of the cerebral cortex (G. V. Anrep, Trans.). Oxford University Press.
- Rescorla, R. A. (1967). Pavlovian conditioning and its proper control procedures. Psychological Review, 74(1), 71–80. https://doi.org/10.1037/h0024109
- Rescorla, R. A. (1988). Pavlovian conditioning: It's not what you think it is. American Psychologist, 43(3), 151–160. https://doi.org/10.1037/0003-066X.43.3.151
- Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current research and theory (pp. 64–99). Appleton-Century-Crofts.
- Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593–1599. https://doi.org/10.1126/science.275.5306.1593
- Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. Appleton-Century.
- Skinner, B. F. (1953). Science and human behavior. Macmillan.
- Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associative processes in animals. Psychological Review Monograph Supplements, 2(4), i–109. https://doi.org/10.1037/h0092987
- Todes, D. P. (2014). Ivan Pavlov: A Russian life in science. Oxford University Press.
- Watson, J. B., & Rayner, R. (1920). Conditioned emotional reactions. Journal of Experimental Psychology, 3(1), 1–14. https://doi.org/10.1037/h0069608
- Wolpe, J. (1958). Psychotherapy by reciprocal inhibition. Stanford University Press.
- Yau, J. O.-Y., & McNally, G. P. (2023). The Rescorla–Wagner model, prediction error, and fear learning. Neurobiology of Learning and Memory, 203, Article 107799. https://doi.org/10.1016/j.nlm.2023.107799