Classical vs operant conditioning: the difference, and where it blurs

Classical conditioning is learning that one event signals another: a stimulus comes to predict something that matters, and the learner starts responding to the signal. Operant conditioning is learning that a behaviour has a consequence: what follows the behaviour makes it more or less frequent. The quickest test is to ask whether the learner’s behaviour changed what happened. If it did not, the conditioning is classical. If it did, it is operant.

• Classical (Pavlovian, respondent) conditioning links a stimulus to a stimulus. The response is elicited, and nothing the learner does changes what arrives.
• Operant conditioning links a behaviour to its consequence. Reinforcement makes the behaviour more frequent, punishment makes it less frequent.
• B. F. Skinner drew the line himself in 1937: conditioning that depends on a stimulus he called Type S, conditioning that depends on a response Type R.
• Pavlov’s Nobel Prize in 1904 was for his work on digestion, not for conditioning. And the signal in his opening demonstration was a metronome, not a bell.
• The rule “classical is involuntary, operant is voluntary” is a shorthand with known exceptions. Pigeons will learn to peck a key through a purely Pavlovian procedure.

The difference in one table

Most of the confusion comes from the fact that both kinds of conditioning use the words “stimulus”, “response” and even “reinforcement”. The table sets them side by side, with each row answering a question an exam scenario usually hides.

Classical conditioning: a signal that predicts an event

Classical conditioning starts with a reflex the learner already has. Food in a dog’s mouth produces saliva without any training: the food is the unconditioned stimulus (US), the salivation is the unconditioned response (UR). A neutral stimulus that reliably comes before the food becomes a conditioned stimulus (CS), and the salivation it now produces on its own is the conditioned response (CR). The UR and the CR can be the same response. What differs is what triggers it.

The textbook Psychology 2e from OpenStax puts the core of it plainly: in classical conditioning, “something in the environment triggers a reflex automatically”. The learner does not have to do anything for the food to arrive.

What Pavlov actually used

The first demonstration in Pavlov’s second lecture on conditioned reflexes, published in English in 1927, uses a metronome: “when the sounds from a beating metronome are allowed to fall upon the ear, a salivary secretion begins after 9 seconds”. The same lecture describes an electric buzzer and the “loud buzzing of an electric bell”. So a bell was not invented by textbooks, but it was one stimulus among several, and not the one Pavlov chose to open with.

The bell’s appearance is more interesting than the bell itself. In that experiment it was sounded 5 to 10 seconds after the dog was given food, and it failed to become a conditioned stimulus even after 374 combinations. An electric buzzer sounded before the food worked after a single combination. Pavlov’s conclusion was that the conditioned stimulus must begin before the unconditioned one. That order is the whole point: a signal that comes after the event predicts nothing.

One more correction worth carrying into an exam. Pavlov’s 1904 Nobel Prize in Physiology or Medicine was awarded, in the Nobel Foundation’s wording, “in recognition of his work on the physiology of digestion”. The prize was not for conditioning.

Why pairing is not enough

The standard definition says a neutral stimulus is paired with a meaningful one until it produces the same response. In “Pavlovian conditioning: It’s not what you think it is” (1988), the learning researcher Robert Rescorla called that description inadequate, and described conditioning instead as “the learning of relations among events so as to allow the organism to represent its environment”.

His example: a rat hears a tone and receives mild shocks. If the shocks are just as likely when the tone is off as when it is on, the tone and the shock still coincide often, but the tone tells the rat nothing. Rescorla reports that in many conditioning situations the learning follows that information, not the bare coincidence. For a student, the useful version is short: in classical conditioning, the CS works because it predicts the US.

Operant conditioning: a behaviour and its consequence

Operant conditioning starts with something the learner does. What happens next changes how often they do it. The idea is older than the name. Edward Thorndike’s law of effect, set out in Animal Intelligence in 1911, held that responses “accompanied or closely followed by satisfaction” become more firmly connected to the situation, and those followed by “discomfort” have their connections weakened.

The vocabulary came from Skinner. In a 1937 reply to Konorski and Miller he separated two kinds of behaviour. A response made to a specific stimulus, a reflex in the traditional sense, he called “a respondent”. A response that occurs without any stimulus eliciting it he named differently: “I shall call such a unit an operant and the behavior in general, operant behavior.” That is where the name “operant conditioning” comes from, and why classical conditioning is also called respondent conditioning.

The four consequences, in brief

Consequences come in two directions and two operations. Reinforcement makes a behaviour more frequent, punishment makes it less frequent. “Positive” means something is added, “negative” means something is removed. That gives four terms, and the pair students confuse most, negative reinforcement vs punishment, has its own article, including the definition Skinner reversed in 1953.

It is the contingency that is learned

What an operant learner picks up is the relation between their behaviour and what follows it. Seligman and Maier’s 1967 experiment on learned helplessness turned on exactly that. Dogs that could switch a shock off and dogs that received the same shocks with no way to stop them had the same physical experience. Only the second group later failed to escape when escape became possible. What differed between them was whether their behaviour controlled the outcome.

Type S and Type R: a memory aid from Skinner himself

The most reliable way to remember which is which comes from the 1937 paper. Skinner wrote that, “to gain a mnemonic advantage”, he would call conditioning resulting from “the contingency of a reinforcing stimulus upon a stimulus” Type S, and conditioning resulting from “contingency upon a response” Type R.

• Type S (classical): what matters depends on a stimulus. The food follows the metronome, whatever the dog does.
• Type R (operant): what matters depends on a response. In Skinner’s own example, food follows the lever press, and only the lever press.

So ask what the important event depends on. If it depends on a signal, the conditioning is Type S and classical. If it depends on the learner’s own behaviour, it is Type R and operant.

A mnemonic gets the pair into memory, but it does not keep it there. Terms that differ by one word, like CS and US or CR and UR, are exactly the kind that blur when reread and hold when tested. Our guide to memorising psychology terminology covers the methods with the best evidence.

What the two have in common

Both are forms of associative learning, and several processes appear in both.

• Acquisition: the learned response or behaviour builds up over repeated experience of the relation, signal and event in one case, behaviour and consequence in the other.
• Extinction: when the relation stops, the learned response fades. In classical conditioning the CS is presented without the US; in operant conditioning the reinforcement stops. OpenStax notes that how quickly an operant behaviour extinguishes depends on the reinforcement schedule.
• Discrimination: the learner comes to respond to one stimulus and not to a similar one. Skinner’s 1937 paper discusses discrimination in both Type S and Type R conditioning.

The shared vocabulary goes further back than the processes. Jack Michael noted in a 1975 paper in Behaviorism that before Skinner applied the word to consequences, “reinforcement” was “widely used to refer to the unconditioned stimulus of respondent conditioning”, because it strengthened the link between the CS and the response. That is why both chapters of a textbook use the word, and why it is worth asking which sense a question means.

Where the line blurs

The usual shorthand is that classical conditioning governs involuntary responses such as salivation or fear, and operant conditioning governs voluntary behaviour. It works for most exam questions. It is not the definition, and there is a well-known case where it fails.

In 1968, Paul Brown and Herbert Jenkins lit a response key in a pigeon’s chamber for a moment and then delivered food, whether or not the pigeon pecked. The pigeons came to peck the key reliably anyway. According to the paper’s abstract, the pecking depended on the light being followed by food, in that order. The researchers called the effect auto-shaping. Pecking a key looks like exactly the kind of behaviour the voluntary rule would file under operant, yet the procedure was Pavlovian: the food depended on the light, not on the peck.

That is why the Type S / Type R test is more reliable than the voluntary/involuntary one. It asks about the procedure, what the important event depended on, rather than guessing what kind of response it is.

The two processes can also run in the same situation. A phone’s notification sound predicts a message whatever you do, which is a classical relation. Picking up the phone is then followed by whatever the message holds, which is an operant one. Real behaviour rarely comes labelled, and in a scenario like this the useful exam skill is to name which part of the situation each type of conditioning describes.

Classifying exam scenarios

Exam questions describe a situation and ask which kind of conditioning it shows. Run the same two checks on each: did the outcome depend on the learner’s behaviour, and what changed afterwards?

The first row is Little Albert. In Watson and Rayner’s 1920 study, the noise came from a hammer striking a steel bar behind the child’s head just as his hand touched the rat, and the conditioning began when he was 11 months and 3 days old. The rat was the CS, the noise the US, and the fear the CR. Nothing Albert did stopped or started the noise, so it is classical.

The last row is the trap. If a question describes a pigeon pecking and food arriving, check whether the food depended on the peck before answering “operant”.

The behavioural tradition these experiments come from is also the ancestry of the cognitive and behavioural family of psychotherapies, whose names are sorted out in our plain glossary of therapy approaches.

Keeping the terms apart

The vocabulary of this topic is a set of near-synonyms: CS and US, CR and UR, respondent and operant, reinforcement and punishment, positive and negative. In Psychology World you can save them to one list and drill them as flashcards, as multiple-choice questions once the list has four or more words, or by typing the answer out, which is the mode that really tests you. What each study mode does is set out on the home page.