Operant Conditioning

What is operant conditioning?

By Ryan Martinson·Updated September 12, 2026·26 min read

The machine you’re already in

Sometime in the last hour you picked up your phone without deciding to. You weren’t looking for anything in particular. You just checked. Then, probably, you kept scrolling.

Those are two different mechanisms, and behavior analysts have a name for each. Checking pays off on a time basis — either something arrived since you last looked or it didn’t, and pressing the button twice as often will not make messages arrive twice as fast. That is a variable-interval schedule, and it produces a moderate, stubbornly steady rate of responding. It is why you check again four minutes later.

Scrolling is the other kind. Each swipe is a response, and whether it pays depends on how many swipes you make rather than on how long you wait. That is a variable-ratio schedule, and of the five basic schedules — catalogued, with dozens of combinations, in a 700-page book Ferster and Skinner published in 1957 — it is the one that produces the highest and steadiest rate of responding, and the one that keeps behavior going longest after the payoffs stop. It is what a slot machine is. It is what a loot box is. It is what an infinite feed is.

None of which is a character flaw, and in most cases it is not addiction in any clinical sense. It is a schedule doing what seventy years of data say a schedule will do.

Which is the genuinely useful part. A mechanism strong enough to keep you swiping against your own stated wishes is strong enough to aim somewhere else, and aiming it is a skill rather than a personality trait. The rest of this page is that mechanism from the ground up: what a consequence has to do to count as one, the four things it can be, the five ways to time it, and what happens when it stops.

Watch: operant vs. classical conditioning in four minutes

Start here if you are new to the topic. This TED-Ed lesson by Peggy Andover walks through Pavlov's dogs, then shows how Skinner's operant conditioning differs — behavior first, consequence second — and how reinforcement and punishment change what an animal (or a person) does next.

Play video: TED-Ed · Peggy Andover · about 4 minutes The difference between classical and operant conditioning Press play — nothing loads from YouTube until you do

Operant conditioning in one paragraph

Every organism that can learn is constantly running the same experiment: do something, notice what happens next, adjust. A rat presses a lever and a food pellet drops, so it presses again. A toddler says "please" and gets the cookie, so "please" becomes a habit. You check your phone, a notification rewards you, and the checking becomes automatic. In each case the behavior operates on the environment (hence "operant") and the environment answers back with a consequence. Operant conditioning is the study of how those consequences select which behaviors survive and which fade away.^[3]

Three ideas do most of the work:

• Consequences are defined by their effect, not their appearance. A "reward" that doesn't increase behavior is not a reinforcer. A "punishment" that doesn't decrease behavior is not a punisher. This is a functional definition, and it is the single most important thing to understand about the whole field.
• Behavior is selected over time, the way evolution selects traits. Skinner called this "selection by consequences." Reinforcement doesn't teach a rule; it shifts probabilities.^[4]
• The unit of analysis is the three-term contingency: an antecedent sets the occasion, a behavior occurs, a consequence follows. Learn to see the A-B-C pattern and you can read almost any behavior.

How operant conditioning works

Skinner's central insight was that behavior is not just triggered by what comes before it (as in Pavlov's reflexes), it is shaped by what comes after it. He formalized this as the three-term contingency, often written A → B → C:^[3]

Two more variables determine how much a consequence matters:

• Contiguity (timing). Consequences that follow within seconds are far more effective than delayed ones. In laboratory studies the strength of learning drops off steeply as the delay grows to even a few seconds — one reason a paycheck at the end of the month is a poor reinforcer for any specific behavior on a Tuesday morning.^[5]
• Contingency (dependability). The consequence must actually depend on the behavior. If food arrives whether or not the rat presses the lever, lever-pressing doesn't get learned — and, as Skinner showed in his famous "superstition" experiment, pigeons fed on a timer developed odd ritual behaviors that happened to precede the food by chance.^[6]

Inside the Skinner box

The apparatus that made all of this measurable was Skinner's operant conditioning chamber — the "Skinner box." Thorndike had timed cats escaping from puzzle boxes one trial at a time; Skinner's innovation was a box the animal never had to leave, so it could respond whenever it liked and the rate of responding could be recorded continuously.^[2] The Skinner box in full: every part, and why it was built that way ›

A typical experiment runs in four steps. The rat is kept mildly hungry so food works as a reinforcer. It first learns that the click of the food dispenser means a pellet has arrived (a conditioned reinforcer). Then the experimenter shapes lever-pressing, reinforcing closer and closer approximations until the rat presses on its own. Finally the schedule is thinned — every press, then every fifth, then an unpredictable number — and the recorder shows how the pattern of responding changes. Pigeons peck a lit key instead of pressing a lever; the logic is identical. More on Skinner and the box ›

Run your own Skinner box

At the top of this page you were the rat. Here you hold the pellet button. Reading about shaping is one thing; doing it is another. This lab puts a hungry, untrained rat in a chamber and hands you the pellet button. Magazine-train it, shape a lever press one approximation at a time, put the press on a schedule, then take the food away and watch extinction — all in about three minutes. The counters under the chamber track every pellet you deliver and every press the rat makes, from the first step on.

The rat is a stylized model — its tendencies shift with what you reinforce, drift back when you don't, and follow the schedule patterns Ferster and Skinner documented — not a replay of real data. Real shaping takes longer and real rats are more surprising.

Want the lab on its own page to link, share, or assign? Open the virtual Skinner box ›

The four quadrants: reinforcement and punishment

Every consequence can be sorted along two questions. Did the behavior increase or decrease? (That tells you whether it was reinforcement or punishment.) Was a stimulus added or removed? (That tells you whether it was "positive" or "negative.") Crucially, in this vocabulary positive and negative mean plus and minus — added and removed — not good and bad.

The four types of operant conditioning at a glance

A fifth process, extinction, is not a quadrant: the reinforcer that used to follow the behavior simply stops arriving, and the behavior fades.

Which quadrant is it? An interactive check

Almost everyone gets one pair backwards the first time, and it is nearly always negative reinforcement mistaken for punishment. Answer the two questions about any scenario and the tool will classify it.

Five mistakes almost everyone makes

1. Treating "negative reinforcement" as a polite word for punishment. It is the opposite: negative reinforcement makes a behavior more likely by taking something unpleasant away. Taking an aspirin to end a headache is negative reinforcement of aspirin-taking.
2. Reading "positive" as good and "negative" as bad. They mean added and removed. A spray of water in a dog's face is positive punishment.
3. Calling something a reinforcer because it seems nice. A reinforcer is anything that increases the behavior it follows — it is defined by its effect. Scolding that a bored child finds attention-worthy is a reinforcer; a sticker a teenager finds embarrassing is not.
4. Mixing up operant and classical conditioning. If the key event comes before the response and the response is a reflex (salivating, flinching), it is classical. If the key event comes after a voluntary behavior, it is operant.
5. Confusing extinction with punishment. Extinction means the reinforcer simply stops arriving; nothing is added or taken away as a consequence. The behavior fades — usually after a brief burst — rather than being suppressed.

Try three

Reinforcement versus punishment: what the evidence says

Skinner believed punishment was a poor way to change behavior, in part because an early experiment by his student W. K. Estes suggested that punishment only temporarily suppressed responding.^[7] Later research complicated that picture: punishment can produce lasting decreases when it is immediate, consistent, and sufficiently intense from the outset.^[8] But those same studies documented why practitioners still prefer reinforcement:

• Punishment teaches what not to do without teaching what to do instead. Reinforcement builds a replacement.
• Punishment tends to produce escape and avoidance — of the punisher as much as the behavior. (The child learns not to get caught.)
• It can elicit aggression and emotional side effects, and it models the use of aversive control.
• It works best at intensities and consistencies that are ethically or practically unavailable in most human settings.

In parenting specifically, a large body of research links corporal punishment to worse, not better, long-term behavioral outcomes.^[9] Modern applied behavior analysis therefore treats reinforcement-based procedures as the default and reserves punishment for narrow, supervised cases where reinforcement alone has failed and the behavior is dangerous.^[10]

Reinforcement: the two types, kinds of reinforcers, and what makes it work ›
Punishment: the two types, the side effects, and the alternatives ›

Schedules of reinforcement

Once a behavior is learned, how often it gets reinforced changes both how fast the organism responds and how long the behavior persists when reinforcement stops. Ferster and Skinner catalogued these patterns in a 700-page 1957 volume, and the basic findings have held up for seventy years.^[11]

The practical rule: use continuous reinforcement to build a behavior, then thin to an intermittent schedule to make it durable. The variable-ratio schedule is why gambling and infinite-scroll apps are so hard to quit — and why a behavior you reinforce only sometimes can end up stronger than one you reinforce every time. Run the interactive schedule simulator › · How organisms choose between schedules: the matching law ›

Extinction, shaping, and stimulus control

Extinction

When a previously reinforced behavior stops producing reinforcement, it gradually declines. But not immediately: there is usually an extinction burst — a temporary spike in the frequency, intensity, and variability of the behavior — before it fades. (Push the elevator button; nothing happens; you push it harder and faster before giving up.) Behavior that has been extinguished can also show spontaneous recovery after a rest period. More on extinction ›

Shaping

Complex behavior is rarely emitted fully formed, so it can't simply be reinforced. Shaping solves this by reinforcing successive approximations — first any movement toward the lever, then touching it, then pressing it. Skinner used shaping to teach pigeons to play ping-pong; trainers use it to teach dolphins to jump through hoops; speech therapists use it to build words from sounds. How shaping works ›

Stimulus control and the antecedent

A behavior reinforced in one context and not in another comes under stimulus control: it appears when the "discriminative stimulus" (S^D) is present and not otherwise. The rat presses only when the light is on. You swear with friends and not with your grandmother. This is the "A" in A-B-C, and it is the most under-used lever in self-improvement — changing the cue is often easier than willing a new response. Stimulus control in depth › · The ABC model ›

Operant vs. classical conditioning

The two great forms of associative learning are often confused. The clean distinction is what gets associated with what:

In real life the two run together. The sound of the treat bag classically conditions excitement in a dog and operantly reinforces running to the kitchen. Full comparison with examples ›

Can you spot it?

Eight scenarios, instant explanations, no score kept. When you can call all eight without hesitating, you have it — and there is a longer set of twenty.

What happens in the brain

Reinforcement has a physical address. In 1953 James Olds and Peter Milner found that a rat would press a lever thousands of times an hour for a pulse of electricity to its own brain, and in 1997 Wolfram Schultz and colleagues showed what the relevant neurons are doing: midbrain dopamine cells fire when a reinforcer is better than expected, fall silent when it is exactly as expected, and dip when an expected one fails to arrive.^[21][16] That reward prediction error is the brain's teaching signal, and it explains why unpredictable reinforcers hold behavior so well and why a fully predictable one stops teaching. It is not a pleasure signal: Kent Berridge and Terry Robinson showed that animals without dopamine still like sugar but no longer want it.^[22] The neuroscience of operant conditioning, in depth ›

A brief history

• 1898

  Thorndike's puzzle boxes

  Edward Thorndike times cats escaping from latched boxes. Escapes get faster, and he proposes the law of effect: responses followed by satisfaction are "stamped in"; those followed by discomfort are "stamped out."^[1]

• 1930–1938

  Skinner's operant chamber

  At Harvard, as a graduate student and then a junior fellow, Skinner builds the apparatus later nicknamed the "Skinner box" and invents the cumulative recorder; at Minnesota he coins "operant" (1937) and lays out the science of operant behavior in The Behavior of Organisms (1938).^[12][2]

• 1948–1957

  Superstition, schedules, and Verbal Behavior

  The pigeon "superstition" study (1948), Science and Human Behavior (1953), Schedules of Reinforcement with Ferster (1957), and Verbal Behavior (1957) extend the analysis to society and language.

• 1968

  Applied behavior analysis is born

  Baer, Wolf, and Risley publish the founding paper of ABA in the first issue of the Journal of Applied Behavior Analysis.^[15]

• 1997

  The dopamine connection

  Schultz, Dayan, and Montague show that midbrain dopamine neurons encode a reward prediction error — a biological implementation of the learning signal operant theory had assumed.^[16]

The full history of operant conditioning › · B. F. Skinner: life, work, and the Skinner box › · The original books, full text, in the library ›

Examples of operant conditioning in everyday life

Once you know the pattern you see it everywhere:

• Positive reinforcement: A barista is thanked warmly for a latte-art heart and starts making them on every cup. A student's essay earns praise and she writes more. Your phone lights up with a like.
• Negative reinforcement: You clean the kitchen to end your partner's nagging. A student finishes homework early to escape the anxiety of a deadline. A driver takes the side street to avoid the traffic jam.
• Positive punishment: A dog gets sprayed with water for jumping on the couch. A late invoice draws a fee. You bite into a moldy strawberry.
• Negative punishment: A child loses screen time for hitting a sibling. A driver loses points from their license. A team member loses a project after missing deadlines.
• Schedules in the wild: Slot machines (variable ratio). A bakery loyalty card (fixed ratio). Watching the clock in the last minutes of a shift (fixed interval). Fishing (variable interval).

50+ examples, sorted by quadrant and setting ›

Operant conditioning in pop culture

Two scenes worth watching a second time. Both are embedded from YouTube and load only when you press play.

Clips are uploaded by third parties and may disappear; NBC hosts the official Office clip.

Key terms to know

The twelve terms that do most of the work, defined the way behavior analysts define them. Hover or tap any underlined term anywhere on this page for its definition; the full glossary has 128 entries.

Operant

A class of behavior defined by its effect on the environment (what it accomplishes), not by its exact form. Lever-pressing with the left paw or the right paw is the same operant.

Reinforcer

Any consequence that increases the future frequency of the behavior it follows. Positive reinforcers are added; negative reinforcers are removed.

Punisher

Any consequence that decreases the future frequency of the behavior it follows.

Primary vs. secondary reinforcer

Primary reinforcers work without learning (food, water, warmth). Secondary (conditioned) reinforcers acquire their power by being paired with primary ones — money, praise, grades, a clicker.

Three-term contingency

Antecedent → Behavior → Consequence: the basic unit of analysis. The ABC model ›

Discriminative stimulus (S^D)

A cue that signals a behavior will be reinforced. When a behavior reliably occurs in its presence and not otherwise, the behavior is under stimulus control.

Schedule of reinforcement

The rule for which responses get reinforced: continuous, fixed ratio, variable ratio, fixed interval, or variable interval. Schedules ›

Extinction

Withholding the reinforcer that maintained a behavior, so the behavior declines — often after an extinction burst, a temporary increase. Extinction ›

Shaping

Building a new behavior by reinforcing successive approximations of it. Shaping ›

Motivating operation

A condition such as deprivation or satiation that changes how effective a reinforcer is. Food reinforces a hungry rat, not a full one.

Premack principle

A more probable behavior can reinforce a less probable one — "finish your homework, then you can play." Premack ›

Law of effect

Thorndike's 1898 principle that responses followed by satisfying consequences are strengthened and those followed by discomfort are weakened — the ancestor of operant conditioning.

Criticisms and limitations

An honest account includes what operant conditioning does not explain well.

• Biological constraints. Organisms are not blank slates. The Brelands found that raccoons trained to deposit coins would instead "wash" them — an instinctive food-handling behavior that intruded even though it delayed reinforcement, a drift toward species-typical behavior they called instinctive drift.^[14] Reinforcement works with an animal's evolved tendencies, not against them.
• Cognition and language. Chomsky argued that reinforcement cannot explain how children acquire grammar from limited input;^[13] Tolman's rats learned mazes without obvious reinforcement, and Bandura's children learned by watching.^[17] Operant learning is one powerful process among several, not a complete theory of mind.
• Rewards and intrinsic motivation. Expected, tangible rewards for an activity someone already enjoys can reduce their interest once the rewards stop — the overjustification effect. A large meta-analysis found it; a rival meta-analysis found it small and narrow.^[18][19] The fair reading: praise and feedback rarely undermine motivation; paying people for things they already love sometimes does.
• Ethics of control. Skinner's Beyond Freedom and Dignity (1971) argued that since behavior is always controlled by its environment, we should design that environment deliberately. Critics saw a road to manipulation; the ethics codes of behavior analysis answer with consent, least-restrictive procedures, and the client's own goals.^[10]

What survived every critique is the core: the law of effect, schedule effects, extinction bursts, stimulus control, and shaping replicate across species and remain the working toolkit of clinicians, teachers, and trainers. Reinforcement learning — the branch of AI behind game-playing systems and the tuning of language models — is a mathematical descendant of the same ideas.^[20]

How to use operant conditioning on yourself

The same contingency that trains a pigeon can be turned inward, and most self-improvement advice ignores two-thirds of it: it obsesses over motivation, which is not a term in the equation, and neglects antecedents and consequences, which are. The protocol is short. Attach the new behavior to a cue that already happens every day; shrink the behavior until it is almost embarrassing; deliver a small reinforcer within seconds, every time at first; then thin the schedule so the habit survives missed days. The full seven-step protocol, with the evidence on how long habits take ›

Design your own contingency

Fill in the three terms for a habit you actually want. The tool checks each against the science — is the cue stable, is the behavior really a behavior, is the consequence immediate — and gives you a card to print. Nothing you type leaves your browser. Full guide to building habits ›