The Skinner Box

What is a Skinner box?

Most students meet the Skinner box as a picture: a rat in a metal cage pressing a bar. The picture is accurate but misses the point. The chamber was never the discovery; it was the instrument that made a kind of discovery possible. It let the animal respond whenever and as often as it liked and recorded every response automatically. That made how often an animal did something a measurable quantity, and almost everything in operant conditioning — reinforcement, extinction, schedules, shaping, stimulus control — was first seen as a change in that quantity.

By Skinner's account the box evolved by simplification: a runway along which a rat ran for food, whose timing proved surprisingly orderly, was shortened step by step until the rat needed only to press a horizontal bar to work a food dispenser.^[2] The Behavior of Organisms (1938), which laid out the resulting science, is built almost entirely on records from it.^[1]

Why Skinner never called it a Skinner box

The name was not his. It came out of Clark Hull's laboratory at Yale, which adopted a version of the apparatus in the 1930s; Skinner never used it, and the field settled on operant conditioning chamber.^[3] The eponym misleads in two ways. It makes a method sound like a gadget, when the box was only a way of getting at the method — measuring the rate of a free operant. And within a few years the name attached itself to something else entirely: the enclosed crib Skinner built for his daughter, the subject of the most durable myth about him (below).

The parts of a Skinner box

How a Skinner box experiment runs

A chamber with an untrained rat in it produces nothing: the rat sniffs the corners, grooms, and occasionally leans on the lever by accident. A working session is built in stages, each depending on the one before.

1. Deprivation. The animal is kept mildly hungry — typically around 80 to 85 percent of its free-feeding weight — so that food functions as a reinforcer. Without this motivating operation, nothing else works.
2. Magazine training. Pellets arrive on a timer, no response required. Within a few dozen deliveries the rat goes to the tray the instant it hears the click, now a conditioned reinforcer that can be delivered from across the chamber.
3. Shaping. The experimenter delivers a pellet by hand switch whenever the rat does something closer to a lever press: facing it, approaching, rearing, touching. Each approximation is reinforced until frequent, then dropped for the next. How shaping works ›
4. Continuous reinforcement, then schedules. Every press pays at first. Then the requirement is thinned — every fifth press, an unpredictable number, the first press after a fixed time — and the recorder shows each schedule's signature.
5. Discrimination training. Presses pay only while a light is on. Pressing comes under stimulus control — high in the light, near zero in the dark — the laboratory version of the antecedent in the A-B-C model. Stimulus control in depth ›
6. Extinction. Food stops. The record shows a burst of rapid, variable pressing, then a decline to almost nothing, with some recovery at the next session. What extinction looks like ›

Later experiments are built on this base: punishment (a brief shock for a press that is still reinforced), choice (two keys on different schedules, the procedure behind the matching law), drugs (a dose before the session, with the change in rate as the measure). You can run the magazine-training, shaping, schedule, and extinction stages in the virtual chamber below.

Run a Skinner box yourself

This lab puts an untrained, mildly hungry 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 stop the food and watch extinction. The cumulative record under the chamber draws every press.

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 ›

What a Skinner box measures, and why that mattered

Edward Thorndike's cats, in 1898, were placed one at a time in a latched "puzzle box" with food outside; Thorndike timed each escape and plotted the times across trials.^[4] The mazes and runways that dominated animal psychology for the next forty years worked the same way. These are discrete-trial methods: the experimenter decides when the animal may behave, each trial yields one number, and learning appears as a curve across trials.

In the operant chamber the lever is always there. The animal decides when to respond and how often — a free operant — and the natural measure becomes the rate of response, presses per minute, moment by moment. Skinner argued that rate was the datum psychology had been missing: it changes continuously rather than in trial-sized lumps, it can be read at a glance from a cumulative record, and it proved sensitive to almost every variable an experimenter could manipulate.^[1][5]

Some of the field's landmarks were accidents of the apparatus, and Skinner said so. The cumulative record began as a lucky by-product of the way his early food magazine was built. The first extinction curve he ever saw appeared when the pellet dispenser jammed and the rat went on pressing. And the first intermittent schedule was born when, running short of the pellets he made by hand, he reinforced only one press a minute to save them — and found the record perfectly orderly.^[2] In the same kind of chamber, feeding pigeons every fifteen seconds regardless of what they did produced "superstitious" rituals in six of eight birds, the classic demonstration that accidental contingencies shape behavior.^[6] The papers describing the method and these discoveries are collected in Cumulative Record.^[7]

Pigeons vs. rats

The Behavior of Organisms is a book about rats. Skinner switched to pigeons during the Second World War, when Project Pigeon trained them to guide a missile by pecking at a target image, and largely stayed with them.^[8] The bird pecks quickly and can sustain thousands of responses an hour, has sharp color vision (ideal for work on stimulus control), is cheap and hardy, and lives for a decade or more, so one subject can be studied for years. Schedules of Reinforcement, the 1957 catalog of schedule effects, is built mostly on pigeon records.^[9]

Rats never went away. They are nocturnal and nearly color-blind by comparison, but their physiology, genetics, and brain are far better mapped, which is why the rat (and now the mouse) chamber is standard in pharmacology and neuroscience. The logic is the same for any species — a defined response, an automated consequence, a continuous record of rate — and chambers have been built for monkeys, fish, and humans.

How to read a cumulative record

Skinner's second invention made the data visible. In the cumulative recorder, a strip of paper moves under a pen at constant speed; each response steps the pen a small fixed distance upward, and each reinforcer is marked with a short diagonal tick, or "pip." When the pen reaches the top of the paper it drops back to the bottom and continues. The cumulative record is the field's native graph, and every schedule leaves a signature on it.^[9]

• Slope is rate. Steep means fast responding; flat means the animal has stopped. A change in slope is a change in behavior, visible the moment it happens.
• Pauses. A flat stretch right after a tick is the post-reinforcement pause, which lengthens as ratios and intervals grow. The fixed-ratio record is "break and run": pause, then a burst at full speed.
• Scallops. On a fixed-interval schedule the line curves — nothing just after a reinforcer, then accelerating responding as the interval runs out. Variable schedules erase both pause and curve and draw a steady, nearly straight line.
• Extinction. When reinforcement stops, the record rises steeply for a while (the extinction burst) and then bends toward horizontal.

Ferster and Skinner's book contains hundreds of such records; the reader is expected to see the effects, not compute them. Draw your own records in the schedule simulator ›

The "baby in a box": the Skinner box myth

In 1944, before his second daughter Deborah was born, Skinner built what he called a baby tender and later an air crib: an enclosed crib with a large safety-glass front and a stretched canvas floor, supplied with filtered, warmed air, so that a baby could sleep and play in a diaper alone — no blankets, layers of clothing, or bars — at a constant temperature. It was a bed with climate control. Deborah was taken out to be fed, changed, held, and played with like any other infant. Ladies' Home Journal published Skinner's article about it in October 1945 under a headline he did not write, "Baby in a Box," and a number of families later used commercial versions.^[10]

The myth survives because it fits a picture of Skinner as a cold manipulator. The facts of the box cut against it: it was a measuring instrument, its reinforcers were overwhelmingly food rather than shock, and Skinner spent fifty years arguing against punishment. B. F. Skinner: life, work, critics, and myths ›

Modern descendants of the Skinner box

The chamber remains standard laboratory equipment, though it rarely looks like Skinner's.

• Touchscreen and home-cage chambers. The lever becomes a screen a rat or mouse touches with its nose, so tasks used to test memory and attention in human patients can be run in animals. In home-cage systems the operandum lives inside the animals' enclosure; they identify themselves electronically and work when they choose, around the clock, without handling stress.
• Drug self-administration. Since James Weeks fitted rats with an intravenous line in 1962 so that a lever press delivered a dose of morphine, the operant chamber has been the principal animal model of addiction: the drugs people abuse are the ones animals will work for.^[12]
• Brain stimulation and optogenetics. In 1954 James Olds and Peter Milner showed that a rat would press a lever for a brief pulse of current to a region of its own brain — the first direct evidence that reinforcement has a physical address.^[13] In the current version, the lever fires a laser that switches on a genetically targeted set of neurons, so the reinforcing effect of specific dopamine cells can be tested press by press. The neuroscience of reinforcement ›

The human "Skinner box"

The phrase now describes any product engineered to keep people responding: slot machines, infinite feeds, loot boxes. The analogy has substance. Skinner himself pointed to the gambling device as a variable-ratio schedule, the arrangement that produces the highest and most persistent rates in the laboratory, and the same schedule runs the pull-to-refresh feed.^[14]

Criticisms and limitations of the Skinner box

Artificiality. The chamber studies a hungry animal, a bare environment, and a single arbitrary response, and critics have argued from the start that this tells us about rats in boxes rather than organisms in the world. Two findings from inside the operant tradition gave the objection teeth. Keller and Marian Breland, who left Skinner's laboratory to train animals commercially, reported that trained behavior drifted toward species-typical patterns — raccoons "washing" the coins they were supposed to deposit — whatever the contingencies said.^[15] And the pigeon's key peck, the chamber's signature response, turned out to be partly Pavlovian: a pigeon will begin pecking a lit key that merely precedes food, even though pecking is not required and changes nothing.^[16] The box does not create behavior from nothing; it selects from what the species brings.

Generalization to humans. Noam Chomsky's charge was that terms precise in the pigeon laboratory become loose metaphors when stretched over human language and thought.^[17] Skinner's later work conceded part of the point: humans follow rules and instructions, and rule-governed behavior can look quite different from behavior shaped directly by contingencies.^[18] Human volunteers on laboratory schedules, who arrive with instructions and hypotheses of their own, often fail to show the animal patterns for that reason.

The reply. The chamber is a method for isolating variables, as a physicist's frictionless plane is, and its findings have to be tested outside it. They were: reinforcement, extinction, schedule effects, shaping, and stimulus control replicated across species and then across classrooms, clinics, zoos, and workplaces, where applied behavior analysis measures them against real behavior. Whether that vindicates the philosophy Skinner built on the box is a different, and still open, question. The history of operant conditioning, including the cognitive critique ›

Key takeaways

• The chamber was never the discovery; it was the instrument. It isolates one response and one consequence, with nothing else changing, so their relation can be measured for hours without an experimenter in the room.
• Its measure is the rate of a free operant. Unlike Thorndike's puzzle box or a maze, where the experimenter starts each trial and gets one number, the animal decides when and how often to respond, and every response is recorded.
• On a cumulative record, slope is rate, ticks are reinforcers, and flat stretches are pauses. Each schedule leaves a signature: the fixed-ratio break and run, the fixed-interval scallop, the steady line of variable schedules, and the burst-then-flatten of extinction.
• A session is built in stages that each depend on the one before: deprivation, magazine training, shaping, continuous reinforcement then schedules, discrimination training, and extinction. Punishment, choice, and drug experiments are built on that base.
• The "baby in a box" was an air crib, a climate-controlled bed with no lever, dispenser, or contingencies; no experiment was ever run in it. The chamber's real limitation is artificiality: it selects from what the species brings rather than creating behavior from nothing.

Check yourself

On a cumulative record, the line goes flat for a stretch right after every reinforcer tick, then shoots up steeply. A student concludes the pellet must be punishing the rat, since pressing stops each time one arrives. What is wrong with that reading?

The flat stretch is the post-reinforcement pause, the first half of the fixed-ratio "break and run" pattern, and the steep run that follows shows pressing is as strong as ever. A behavior that has actually stopped draws a line that stays horizontal; a pause followed by a burst at full speed is the schedule's signature, not evidence of punishment.

Thorndike's cat and Skinner's rat both learn to work a mechanism to get food. Why did Skinner treat his chamber as a different kind of measurement?

The puzzle box is a discrete-trial method: the experimenter starts each trial and gets one number, the time to escape. In the chamber the lever is always there, so the animal decides when and how often to respond, and the measure is the rate of a free operant, recorded continuously and read from the slope of the record.

A pigeon begins pecking a lit key that comes on just before free food, even though pecking is not required and changes nothing. Does this show operant conditioning at work?

Not on its own. Because pecking has no consequence, the light merely preceding food is enough to produce it, which is why the key peck is described as partly Pavlovian. It is one of the findings behind the criticism that the box does not create behavior from nothing but selects from what the species brings.

An untrained rat is placed in a chamber with the lever already wired to the pellet dispenser. Why does nothing happen, and which two stages have to come first?

The rat has no reason to press and presses only by accident. It must first be mildly hungry, so that food functions as a reinforcer, and then magazine-trained, so that the dispenser's click becomes a conditioned reinforcer that can be delivered from across the chamber. Only then can shaping reinforce closer and closer approximations to a press.

Explain it to a friend. Explain what a Skinner box measures that Thorndike's puzzle box could not, without using the words "rate" or "trial."