Shaping in psychology: Skinner's bowling pigeon and the method of successive approximations

Shaping is an operant conditioning procedure that builds a new behaviour by reinforcing successive approximations to it. Each reinforced response is a little closer to the target than the one before, and the rougher versions stop earning a reward. Shaping is necessary because a complex behaviour may never occur on its own, so there is nothing to reinforce until shaping produces it. The name comes from B. F. Skinner. In 1943, he and two of his students taught a pigeon to knock a ball around a box in a few minutes. It was Skinner’s first shaping by hand.

• The APA Dictionary of Psychology defines shaping as “the production of new forms of operant behavior by reinforcement of successive approximations to the behavior”.
• Every step has two parts: reinforcement of the current approximation, and extinction of the earlier ones. Extinction makes behaviour more variable, and that variation is thought to supply the next approximation.
• Skinner described successive approximation in 1937 and 1938. According to a 2004 history, his first shaping by hand came in 1943. The word “shaping” first appeared in his writing in 1951.
• Shaping changes the form of one response. Chaining joins several responses into a sequence. Autoshaping is not shaping at all: it is a Pavlovian effect.
• Shaping has limits. The animal trainers Keller and Marian Breland saw shaped behaviour drift back toward the instinctive behaviour of the species.

What shaping means

The APA Dictionary of Psychology defines shaping as “the production of new forms of operant behavior by reinforcement of successive approximations to the behavior”. It also lists two other names: approximation conditioning and behavior shaping. The behaviour that the trainer wants at the end is the target behaviour. A successive approximation is a response that is closer to the target than the responses that earned a reward before it.

Shaping is a procedure of operant conditioning, in which the consequences of a behaviour change how often it occurs. Reinforcement is the consequence that makes a behaviour more frequent. But reinforcement can only strengthen a behaviour that already occurs. The open textbook Psychology 2e from OpenStax states the problem directly: shaping “is needed because it is extremely unlikely that an organism will display anything but the simplest of behaviors spontaneously”.

Skinner made the same point in Science and Human Behavior in 1953. The original probability of the finished response, he wrote, “is very low; in some cases it may even be zero”. Waiting for the complete behaviour can take “many hours or days or weeks”, and it works less well: “When we wait for a single complete instance, we reinforce a similar sequence but far less effectively because the earlier steps are not optimally strengthened.”

The technical name for this kind of selection is differential reinforcement: some forms of a response are reinforced and other forms are not. In shaping, the line between the two moves. A response that earned food yesterday can earn nothing today, because the trainer now asks for more.

How shaping works, step by step

OpenStax divides the procedure into five steps. In condensed form:

1. 1.Reinforce any response that resembles the desired behaviour.
2. 2.Reinforce the response that more closely resembles the desired behaviour, and stop the reinforcement of the earlier response.
3. 3.Reinforce a response that resembles the desired behaviour even more closely.
4. 4.Continue to reinforce closer and closer approximations.
5. 5.At the end, reinforce only the desired behaviour.

Skinner’s own example in Science and Human Behavior is a hungry pigeon and a spot on the wall of its box. First, the bird gets food when it turns slightly toward the spot. Then the food is held back “until a slight movement is made toward the spot”. Next, only positions closer to the spot earn food, then only a forward movement of the head. At the end, only contact of the beak with the spot earns food. A bird that is “well adapted to the situation and to the food tray”, Skinner wrote, “can usually be brought to respond in this way in two or three minutes”. He compared the process to sculpture:

The five steps make shaping look mechanical. In practice, the trainer decides when to move the criterion and how large each step is.

When the target behaviour finally occurs, the question changes from how to produce it to how to keep it. OpenStax calls continuous reinforcement, a reward after every response, “the quickest way to teach someone a behavior”. How often to reinforce the behaviour after that is the subject of the schedules of reinforcement.

The day Skinner discovered shaping

In 1943, Skinner was on leave from the University of Minnesota to work on a wartime project for General Mills, Inc. His laboratory was on the top floor of a flour mill in Minneapolis. Two of his students, Keller Breland and Norman Guttman, worked with him. Skinner told the story many times. His 1958 version, in the journal American Psychologist, starts like this:

The pigeon did not swipe. “Though we had all the time in the world, we grew tired of waiting,” Skinner wrote. So the three men reinforced “any response which had the slightest resemblance to a swipe—perhaps, at first, merely the behavior of looking at the ball”. Then they selected responses that were closer to the final form. “The result amazed us. In a few minutes, the ball was caroming off the walls of the box as if the pigeon had been a champion squash player.” In his 1979 autobiography, Skinner added that the food dispenser was “operated by a hand switch”. He wrote: “I remember that day as one of great illumination.”

Why were they amazed? Gail Peterson of the University of Minnesota asked this question in a 2004 paper, A day of great illumination. Skinner had already described successive approximation. A 1937 paper listed steps to teach a rat to press a lever. The Behavior of Organisms (1938) reported experiments that changed the force and duration of lever presses step by step. His rat Pliny had learned a long routine with a marble. But Peterson concluded that none of this was shaping by hand:

• The 1937 lever-press steps were hypothetical. Skinner later admitted: “I do not remember actually shaping lever-pressing in such explicit stages, but I was sure it could be done.”
• Pliny’s routine was built “by making slight changes in the apparatus”, in Skinner’s words, not by a person who watched the rat.

The new element in 1943 was a person who watched the animal and operated the feeder at the right moment. Skinner wrote that they had discovered “how much easier it was to shape behavior by hand than by changing a mechanical device”.

The day also gave the procedure its name. According to Peterson, the word “shaping” does not occur in any paper by Skinner before 1951. The Behavior of Organisms uses it only in a preface written in 1966. It first appears in his 1951 Scientific American article How to Teach Animals, where reinforcement makes it “possible to shape an animal’s behavior almost as a sculptor shapes a lump of clay”. In 1965 a student asked Skinner whether successive approximation was first used in 1943. Skinner replied: “The date 1943 is, I think, correct for the first deliberate use of successive approximation.”

Animal trainers had used the same idea before. Peterson cites the German dog trainer Konrad Most, who described it in a 1910 manual. His argument is that Skinner appears to have been the first to recognise its scientific importance. The day also changed a career. Skinner wrote that the spectacle “so impressed Keller Breland that he gave up a promising career in psychology and went into the commercial production of behavior”.

Why extinction is built into every step

The APA Dictionary and OpenStax both define shaping through reinforcement, but every step also uses extinction. Extinction is what happens when a behaviour that was reinforced stops being reinforced and becomes less frequent. As our article on negative reinforcement and punishment explains, extinction is neither reinforcement nor punishment. In shaping, each time the criterion moves, the old approximation goes on extinction. Skinner’s 1937 description says it directly: “When one step has been conditioned, the reinforcement is withdrawn and made contingent upon the next.” A 1960 paper by Isaacs, Thomas and Goldiamond, described in the next section, defines shaping with the same two parts: the reinforcement of the parts of a response that move in the desired direction, “and the nonreinforcement of those which are not”.

The extinction does more than remove old responses. It also helps to produce the next approximation, because behaviour becomes more variable when reinforcement stops. Neuringer and colleagues, in the 2001 study below, describe an early example. In a 1951 study by John Antonitis, rats were reinforced for a nose poke anywhere along a 50-centimetre opening. During training, the place of their pokes became more and more predictable. When reinforcement stopped, the pokes spread out along the whole opening.

A 2001 study by Allen Neuringer, Nate Kornell and Matt Olufs found the same pattern in sequences of responses by rats. During extinction, rare sequences became more frequent, but the most common sequences stayed the most common. The authors connected this to shaping:

They wrote that “short-term extinction is part of the shaping process”. A pigeon that no longer gets food for a look at the ball does not stop moving. It varies what it does, and the trainer reinforces the new movement that is a step toward the ball. The 2001 study measured extinction, not shaping itself, so this explains why shaping works but does not test it directly.

Shaping in people: the chewing-gum study

Skinner thought that shaping explained much of human development. In Science and Human Behavior he wrote that a child learns to stand, to walk and to grasp objects “through the reinforcement of slightly exceptional instances of his behavior”. Later, he wrote, the child learns to talk “through the same process”. That was a claim, not an experiment. For language, Noam Chomsky rejected the idea that children need careful shaping, in his 1959 review of Skinner’s Verbal Behavior: “It is simply not true that children can learn language only through ‘meticulous care’ on the part of adults who shape their verbal repertoire through careful differential reinforcement.”

Documented cases of shaping in people are narrower. In 1960, Wayne Isaacs, James Thomas and Israel Goldiamond published one in the Journal of Speech and Hearing Disorders. They worked with two hospital patients with a diagnosis of schizophrenia who no longer spoke. Patient A was 40 years old. According to the report, he became completely mute almost immediately after his admission to hospital, 19 years earlier. Patient B had not spoken for 14 years.

During a group session, a packet of chewing gum fell out of the experimenter’s pocket by accident. Patient A’s eyes moved toward it and then back. The experimenter chose this small response as the start. In private sessions three times a week, the procedure continued in stages:

1. 1.Weeks 1 and 2: the experimenter held up a stick of gum and gave it to the patient when his eyes moved toward it.
2. 2.Weeks 3 and 4: the gum came only after a movement of the lips, and then only after a sound. By the end of this stage, the gum produced eye movement, lip movement and “a vocalization resembling a croak”.
3. 3.Weeks 5 and 6: the experimenter said “Say gum, gum” and gave the gum only for sounds that were closer and closer to the word.

At the sixth session, at the end of week 6, the patient said, “Gum, please.” He then answered questions about his name and age. At first he answered only the experimenter. The authors then worked to transfer his speech to other people, and after a month he began to answer the questions of a nurse.

The report has clear limits. The authors called it a “preliminary report”, and it describes two patients with no comparison group. It shows how a procedure can build speech one approximation at a time. It is not evidence that shaping treats schizophrenia. The same principle appears in later rehabilitation research. Edward Taub and colleagues (1994) described a treatment for patients with chronic stroke. For two weeks, the unaffected arm was restrained and the affected arm was trained. Their abstract reports large improvements in motor ability, and “preliminary data suggesting that shaping with verbal feedback further enhances the motor recovery”.

Where shaping fails: instinctive drift

After 1943, Keller Breland and Marian Breland trained animals for displays, fairs and television. In their 1961 paper The misbehavior of organisms, they reported that they had conditioned “thirty-eight species, totaling over 6,000 individual animals”. The paper is about the cases that went wrong.

One case was a pig. It was “conditioned to pick up large wooden coins and deposit them in a large ‘piggy bank’”, usually four or five coins for one reinforcement. The training began with a small step: “Of course, we started out with one coin, near the bank.” The pig learned the task. Then, “in pig after pig, usually after a period of weeks or months”, the behaviour broke down. The pig would “drop it, root it, drop it again, root it along the way”.

A raccoon went through similar steps. It got a reward first for picking up one coin, then for dropping it into a metal box, then for two coins. At the last stage, it spent “seconds, even minutes, rubbing them together (in a most miserly fashion), and dipping them into the container”. The Brelands linked each failure to the food-getting behaviour of the species: rooting in pigs, so-called washing in raccoons. These behaviours delayed the reward, although nothing in the task required them.

The Brelands called this “instinctive drift” and summarised it as “learned behavior drifts toward instinctive behavior”. After 14 years of work, they concluded that the behaviour of any species cannot be understood “without knowledge of its instinctive patterns, evolutionary history, and ecological niche”. The lesson for shaping is that a trainer can move behaviour far, step by step, but the result can drift back toward what the species does naturally.

Shaping vs chaining vs autoshaping

The APA Dictionary defines chaining as “an operant conditioning technique in which a complex behavioral sequence is learned”, with primary reinforcement only after the final response. In backward chaining, the final response is taught first, and its stimulus becomes a conditioned reinforcer for the response before it. In forward chaining, the first step is taught first, then the second, and so on.

The difference between shaping and chaining is the unit. Shaping changes the form of one response: a look at the ball becomes a swipe at it. Chaining joins separate responses into an order. The two procedures often work together. OpenStax notes that shaping “is often used in teaching a complex behavior or chain of behaviors”, and Skinner wrote that every step in Pliny’s chain “had to be worked out through a series of approximations”.

Autoshaping is the term most likely to mislead, because its name contains “shaping”. In a 1968 study by Paul Brown and Herbert Jenkins, pigeons learned to peck a key that lit up before food arrived. The food came whether they pecked or not. No response was reinforced, so no response was shaped. Our comparison of classical and operant conditioning explains why this effect counts as Pavlovian.

Classifying exam scenarios

Exam questions describe a training situation and ask which procedure it shows. Ask two questions. Does the requirement move, so that a rough version that once earned a reward no longer does? Then it is shaping. Are separate actions joined in a fixed order, with the reward at the end? Then it is chaining.

Two traps are common. The first is a requirement that grows in amount rather than form. OpenStax’s example is a child who is rewarded for putting away one toy, then five toys, and at the end the whole room. OpenStax still calls this shaping, because the requirement moves toward the target. The second trap is the word “autoshaping”. If the food arrives whatever the animal does, nothing is shaped.

Keeping the terms apart

Shaping, successive approximation, differential reinforcement, chaining, autoshaping, instinctive drift and extinction are easy to confuse, because each describes how training builds or breaks a behaviour. In Psychology World you can save terms like these to one list. Then you can drill the list as flashcards, as multiple-choice questions once it has four or more terms, or by typing the answer. What each study mode does is set out on the home page. Our guide to memorising psychology terminology explains which study methods have the strongest evidence.