Positive Punishment

How positive punishment works

Positive punishment is the mirror image of positive reinforcement: the behavior occurs, a stimulus appears that was not there before, and the future rate of the behavior drops.

• A rat presses a lever → a brief shock follows → pressing decreases.
• A toddler touches the stove → it burns → stove-touching decreases (a natural punisher).
• A driver runs a red light → a ticket arrives → red-light running decreases, at least at that intersection.

Like reinforcement, punishment depends on contingency and contiguity. A ticket that arrives three weeks later is a weak punisher for a specific act of speeding, which is why cameras and points systems aim to make the consequence certain rather than severe. The same A-B-C analysis applies: antecedent, behavior, and a consequence that changes the future probability of that behavior in that context.

Check any scenario in two questions

Not sure whether a consequence is this quadrant or a neighbor? Answer the two questions and the checker names it.

Examples of positive punishment

Notice that the natural punishers — burns, nausea — produce fast, durable learning, while many human examples carry a qualifier: "when the owner is present," "at that spot." Contrived punishment tends to suppress behavior in the punisher's presence rather than eliminate it. Examples for every quadrant ›

What the research says about punishment

Thorndike and Skinner: the early doubts

Thorndike's original law of effect (1898) gave punishment equal billing: responses followed by discomfort were "stamped out." By 1932 he had concluded that annoying consequences did not weaken connections nearly as directly as satisfying ones strengthened them.^[2] Skinner reached the same position from the laboratory. His student W. K. Estes trained rats to press a lever for food, then briefly punished pressing with shock during extinction. Pressing dropped sharply while the shock was in effect, then recovered; by the end of extinction the punished rats had made about as many responses as unpunished ones.^[3] Skinner concluded that punishment merely suppresses behavior temporarily, and argued against it for the rest of his career.^[1]

Azrin and Holz: when punishment does work

The picture changed in the 1960s, when Nathan Azrin and colleagues ran punishment through the same parametric analysis Skinner had applied to reinforcement. Their landmark 1966 chapter concluded that punishment can produce complete and lasting suppression — but only under specific conditions.^[4][5]

The same research documented the costs: escape and avoidance of the punishing agent, aggression elicited by aversive stimulation, and the odd finding that a punisher can become a signal for reinforcement, so that punishment paradoxically increases behavior when it predicts reward.^[4] Russell Church's review reached a similar verdict: punishment is a lawful process, but its effects depend on parameters most users never control.^[6]

The applied literature

In 2002, Dorothea Lerman and Christina Vorndran reviewed what applied behavior analysts actually knew about punishment: laboratory findings on intensity, immediacy, and schedule had rarely been tested with people in clinical settings; punishment sometimes remained necessary when reinforcement-based treatments failed for dangerous behavior; and the field needed better data on using it with the least intensity and fewest side effects.^[7] Punishment is neither the reliable tool folk wisdom assumes nor the useless one Estes's rats suggested.

Side effects of positive punishment

Even when punishment suppresses a target behavior, it produces collateral effects that reinforcement does not.^[4][7]

• Escape and avoidance. The learner avoids the punisher — and the person, place, or task associated with it. The child punished for spilling hides spills; the employee stops reporting mistakes; the dog steals food only when no one is in the room.
• Aggression. Aversive stimulation elicits fighting. Rats shocked together attack each other, and the same reflexive aggression appears across species.^[8] Punished children and animals lash out, often at an unrelated target.
• Emotional responding. Fear, crying, freezing, and disruption of behavior you wanted to keep. A punished student may stop talking out of turn and also stop participating.
• Modeling. Punishment teaches that adding aversive consequences is how you change people. Children who are hit are more likely to hit.
• Suppression, not elimination. The behavior returns when the punisher is absent or the motivation rises. Punishment does not teach what to do instead.
• The punisher becomes a conditioned aversive stimulus. Anything reliably paired with punishment — the parent's raised voice, the classroom, the training collar — acquires aversive properties of its own, spreading escape and avoidance further.

Corporal punishment: what the evidence shows

Spanking is the most studied form of positive punishment in humans. A 2002 meta-analysis of 88 studies found corporal punishment associated with immediate compliance and with worse outcomes on essentially every other measure — aggression, antisocial behavior, mental health, the parent–child relationship.^[9] A 2016 meta-analysis by Elizabeth Gershoff and Andrew Grogan-Kaylor addressed the objection that earlier work had lumped spanking together with abuse. Restricting the analysis to ordinary open-handed spanking across 75 studies and more than 160,000 children, they found spanking significantly associated with 13 of the 17 outcomes examined, all in the harmful direction; none favored spanking.^[10]

The evidence is largely correlational, and it is fair to ask whether difficult children simply get spanked more. But longitudinal studies that control for prior behavior still find spanking predicting later increases in problem behavior, and no comparable evidence shows benefits. In 2018 the American Academy of Pediatrics recommended that parents not spank, hit, or otherwise physically punish children, nor use verbal abuse that shames or humiliates, and instead rely on positive reinforcement, limit-setting, and brief time-out.^[11] The operant analysis predicts exactly this: spanking as actually practiced — inconsistent, delayed, escalating, delivered by a person the child depends on — violates every condition on Azrin and Holz's list.

Aversives in dog training: what the evidence shows

The same question has been studied in companion dogs, with the same answer. An early owner survey found reward-based methods associated with higher obedience and punishment-based methods with more problem behaviors.^[12] A survey at a veterinary behavior clinic found that confrontational techniques — hitting, alpha rolls, leash jerks — frequently provoked an aggressive response from the dog.^[13] A 2017 review concluded that aversive methods carry welfare risks and no advantage in effectiveness.^[14] And a 2020 study of dogs at aversive-based versus reward-based training schools found the aversively trained dogs showed more stress behaviors, higher salivary cortisol, and a more pessimistic bias in a cognitive test outside training.^[15] Positive punishment trades short-term suppression for long-term stress, which is why the reward-based methods on our dog training page are the standard recommendation of veterinary behavior organizations.

When is punishment used in applied behavior analysis?

Rarely, and only under conditions. In 1988 a task force of the Association for Behavior Analysis affirmed that clients have a right to effective treatment, which may in rare cases include restrictive procedures when less intrusive ones have failed and the behavior is dangerous.^[16] The Behavior Analyst Certification Board's ethics code operationalizes that position: prioritize reinforcement-based procedures; recommend punishment only when the severity of the behavior or the failure of less intrusive procedures warrants it; always include reinforcement of alternative behavior; and monitor data and side effects continuously.^[17] In practice a punishment procedure in ABA follows a functional assessment, sits inside a plan that teaches a replacement, uses the least intrusive procedure that works, requires informed consent and oversight, and is faded as soon as possible.^[18] The contrast with everyday punishment — improvised, delayed, escalating, standalone — could hardly be sharper.

Natural vs. contrived punishers

A natural punisher is produced by the behavior itself: the burn from the stove, the fall from tipping a chair. A contrived punisher is delivered by another person: the reprimand, the fine, the leash jerk. Natural punishers teach efficiently because they are immediate, perfectly consistent, impossible to escape, and there is no punishing agent to become a conditioned aversive stimulus. "Let natural consequences happen" is sound advice whenever the consequence is safe: a child who forgets a coat gets cold, and cold is a better teacher than a lecture.

Reprimands and overcorrection

Reprimands

A reprimand is a brief, firm verbal statement delivered immediately after the behavior. Classroom research found that soft reprimands, audible only to the target student, reduced disruptive behavior more than loud public ones, which can function as attention and entertainment for the class.^[19] Reprimands work better delivered up close, with eye contact, than shouted across the room.^[20] They lose their effect when used constantly or when the student is reinforced by the reaction.

Overcorrection

Overcorrection requires the person to repair the effects of the behavior beyond the original state (restitutional — clean the whole table, not just your spill) or to practice a correct alternative repeatedly (positive practice — walk calmly to the door five times after running). Foxx and Azrin developed it in the early 1970s for severe behavior in institutional settings.^[21] It is effortful, can provoke aggression when the person is physically guided, and is used far less today than reinforcement-based procedures.

Alternatives to positive punishment

Because punishment does not teach a replacement, the most effective way to reduce a behavior is usually to build something else in its place.

Differential reinforcement paired with extinction is the default treatment package in ABA and the core of evidence-based parent training. Negative punishment — removing a reinforcer rather than adding an aversive — carries fewer side effects and is what pediatricians mean by time-out. Antecedent strategies are the most under-used option: the easiest way to reduce a behavior is often to change the situation that occasions it, as our guide to the ABC model explains.

Positive punishment vs. negative punishment vs. negative reinforcement

Positive punishment adds an aversive to decrease behavior (a fine for speeding). Negative punishment removes a reinforcer to decrease behavior (losing the car keys for speeding). Negative reinforcement removes an aversive to increase behavior (slowing down ends the passenger's complaints). "Positive" and "negative" only ever mean added and removed; whether it is punishment or reinforcement depends on which way the behavior moves.

Key takeaways

• "Positive" means added, not good. A stimulus is a punisher only if it reduces the behavior it follows; "I punished him and it didn't work" is a contradiction in terms.
• Punishment can produce complete suppression, but only when it is immediate, at full intensity from the start, consistent, impossible to escape, and paired with a reinforced alternative. Everyday punishment is usually delayed, escalating, and inconsistent, which is why it mostly suppresses behavior in the punisher's presence.
• Its side effects are escape and avoidance of the punisher, aggression, fear, modeling of aversive control, and a punisher that becomes a conditioned aversive stimulus. Punishment does not teach what to do instead.
• Spanking is associated with worse outcomes on nearly every measure studied and with no long-term benefits, and pediatricians recommend positive reinforcement, limit-setting, and brief time-out instead. Aversive dog-training methods show the same pattern: more stress and no advantage in effectiveness.
• The most effective way to reduce a behavior is usually to build something else in its place: differential reinforcement, extinction, antecedent changes, or negative punishment, which carries fewer side effects.

Check yourself

A teacher yells at a class clown after every joke. Over the month, the joking increases. What was the yelling?

A positive reinforcer. Consequences are classified by their effect, not their intent: the behavior went up and something was added. If the yelling is the most attention the student gets all day, it reinforces the clowning it was meant to stop.

A dog is sprayed with water whenever it jumps on the couch, and it stops jumping — while the owner is home. What has the dog learned?

The behavior has been suppressed in the punisher's presence, not eliminated. Contrived punishment tends to work only where the punisher is, and the dog has found an escape route: jumping when no one is in the room. Nothing has taught it what to do instead.

A parent scolds mildly, then louder, then more firmly as a behavior continues. Why does this escalation tend to fail?

A punisher that starts mild and is gradually increased produces adaptation; each step trains tolerance of the next. Punishment suppresses behavior when it is introduced at full intensity from the start, which is one reason it is so hard to use well outside a laboratory.

One child forgets a coat and gets cold. Another is lectured about forgetting coats. Which consequence teaches better, and why?

The cold. It is a natural punisher: immediate, perfectly consistent, impossible to escape, and with no punishing agent to become a conditioned aversive stimulus. The lecture is a contrived punisher delivered by a person, and that person can become something the child learns to avoid.

Explain it to a friend. Explain why punishment that seems to work in the moment usually fails in the long run, in two sentences a twelve-year-old would follow.