Positive Reinforcement

Positive reinforcement is an operant conditioning technique in which a desirable stimulus is added after a behavior, making that behavior more likely to happen again. B. F. Skinner developed the concept, showing that behavior is shaped by its consequences rather than by conscious intention.

How Does It Work?

In this term, “positive” describes an operation, not a value judgment.

It means a stimulus is added after a behavior, whether or not anyone thinks the stimulus is good. The contrast is with negative reinforcement, where something unpleasant is removed instead.

“Reinforcement” names an effect on behavior. A procedure counts only if the behavior becomes more likely.

It works for new behaviors too. It can also strengthen ones that already exist.

Operant conditioning is the theory that underlies Skinner’s technique. It grew from Thorndike’s law of effect, the idea that behavior followed by a satisfying outcome tends to be repeated (Thorndike, 1898).

In essence, one can modify behavior by controlling the consequences that follow it (Skinner, 1938).

Skinner argued that learning is an active process. When humans and animals act on their environment, consequences follow. He believed these external, observable consequences explained behavior, not internal thoughts or motivations (Staddon & Cerutti, 2003).

A consequence that makes a behavior more likely to happen again is called reinforcement. One that makes a behavior less likely is called punishment. Both can work by adding something or by taking something away, a distinction covered later in this article.

Premack Principle

The Premack principle says that, of any two activities, the one a person does more often will reinforce the one they do less often. It solves a practical problem.

How can you know, in advance, what will reinforce a particular person?

Psychologist David Premack found that this works in either direction, depending on which activity is momentarily more probable. It remains one of the most useful ideas in applied behavior analysis.

Premack’s Original Study

Premack (1962) later showed the relationship reverses. Depriving rats of a chance to run made running reinforce drinking, even though drinking is normally the more probable activity.

Grandma’s Rule in Everyday Life

This principle is the empirical basis of “Grandma’s rule”: first eat your vegetables, then you may have dessert. It explains why a preferred activity, such as screen time, can reinforce a non-preferred one, such as homework or tidying a bedroom.

Timing matters here. A common parenting mistake reverses the logic: offering the reward before the behavior happens turns reinforcement into a bribe.

Evaluation: Useful but Incomplete

The Premack principle is genuinely predictive and widely used. Later work refined it. Timberlake and Allison (1974) proposed the response-deprivation hypothesis: what matters is not relative probability, but whether a schedule restricts a response below its normal baseline.

Under this account, even a lower-probability activity can reinforce a higher-probability one, provided access to it has been restricted.

The principle has practical limits, too. Measuring baseline probabilities is not always possible outside the laboratory. And like Skinner’s functional definition, it describes what will reinforce without explaining why.

Examples

Animal training

A classic use of positive reinforcement is in animal training and behavior. The general adage of animal training is to reward positive behaviors and ignore undesirable ones.

This approach dates to B. F. Skinner’s early psychological research.

Skinner devised a method of rewarding positive behavior called the Skinner box (Dezfouli & Balleine, 2012).

Essentially, this skinner box contains a lever or button that delivers a reward, such as food or water, when pressed correctly. A tracker then records each response.

Skinner did not wait for chance. He used shaping: reinforcing small steps that increasingly resembled the goal behavior, and raising the bar as the animal improved.

A pigeon, for example, might be reinforced in stages:

1. Turning toward the lever
2. Approaching it
3. Pressing it, the only response still reinforced

Small steps add up. Trainers add a marker, such as a click that signals the exact instant of the correct response, then follow it with food. This marker or clicker training now underlies much of modern guide-dog and police-dog training (Skinner, 1951; Skinner, 1953).

Shaping works for people too. A pupil might first be reinforced for ten minutes on task, then twenty, then a whole lesson.

Chaining extends the same logic to sequences, such as getting dressed. Each link is reinforced in order, and each completed link often serves as the conditioned reinforcer for the one before it.

Both are used to teach self-care skills.

Animals learn this way too. They perform a task in a particular way to receive a reward, an example of tangible reinforcement (Dezfouli & Balleine, 2012).

In the workplace

Positive reinforcement can also be used in the workplace to encourage desirable behaviors. For example, an employer may give employees a bonus for meeting or exceeding sales targets.

This is often called “performance-related pay” or “pay for performance.”

In another example, an employee who always arrives at work on time may receive a monthly gift card that reinforces timeliness.

Alternatively, an employer might give employees extra paid vacation days as a way of reinforcing good attendance (Ackerman, 2022).

Timing matters at work too. Organizational behavior management applies operant principles to safety, productivity, and service quality, and stresses immediate, specific feedback over distant annual rewards.

A bonus paid months after the performance it targets is a weak reinforcer, because the link between behavior and reward is lost.

Positive reinforcement is not the same as bribery. A bribe is offered before the behavior, to induce it; reinforcement is delivered after the behavior has already happened.

For example, an employee who already arrives at work on time every day cannot be bribed to keep doing so. That employee is only eligible for positive reinforcement, delivered as a reward after the fact (Ackerman, 2022).

Social Media

Social media use has skyrocketed in recent years, and positive reinforcement may be one reason why. Platforms such as Facebook, Twitter, and Instagram use likes, followers, and other rewards to keep users engaged.

The pattern behind this is no accident.

Likes and replies arrive on an unpredictable schedule. Sometimes a post gets many; sometimes it gets almost none. This unpredictability is itself the hook. Psychologists call it a variable-ratio schedule, the same pattern that makes slot machines hard to resist, and it produces the steadiest responding of any pattern.

For example, a person who posts a photo on Instagram is likely to receive likes from friends and followers, which encourages them to keep posting. Frequent tweeters who get replies and retweets keep tweeting for the same reason.

Survey data agree. In a survey of 379 smartphone users, reinforcement rewards (perceived enjoyment and mood regulation) drove addiction, and automatic habit strengthened that effect (Chen et al., 2019).

Not every platform uses the same reward.

Instagram and Facebook use a likes system. Snapchat uses streaks instead: the number of consecutive days that two people have Snapchatted each other, which encourages users to keep messaging to protect the streak (Ackerman, 2022).

Token Economies: The Ayllon and Azrin Study

A token economy rewards behavior with tokens, such as plastic chips or points, that can later be exchanged for real rewards. Tokens work instantly. They solve the timing problem that a slower, real reward cannot.

Tokens also resist satiation, when a learner has had enough. Because they can buy many different backup rewards, a person rarely tires of them the way they tire of one food or treat.

Everything turns on contingency: the token must depend on, and follow, the specific behavior it is meant to strengthen.

Evaluation: The reversal design gives the study strong internal validity, and the model has since been used in classrooms, prisons, and developmental services. Two problems remain: gains can fade once the program ends, and controlling access to ordinary comforts raises ethical questions.

Positive reinforcement is not the opposite of negative reinforcement

Both positive and negative reinforcement make a behavior more likely to happen again. They differ only in the type of consequence involved.

• Positive reinforcement adds a desirable consequence after a behavior, increasing the chance it repeats.
• Negative reinforcement removes an unpleasant condition after a behavior, which also increases its future occurrence (Dozier, Foley, Goddard, & Jess, 2019).

A smoker illustrates both. Relief after a cigarette makes smoking more likely to happen again. If the same smoker instead got a candy bar for smoking, that would be positive reinforcement doing the same job through an added reward.

A popular claim holds that positive reinforcement works faster, while negative reinforcement produces longer-lasting change.

This is not well supported.

What actually governs how long a behavior lasts is the schedule of reinforcement, not whether it is positive or negative (Ferster & Skinner, 1957).

Not everyone responds the same way. Some people respond strongly to one type of reinforcement and barely at all to the other. Even a single behavior may respond to both, depending on the person. The right choice depends on the behavior, the individual, and the desired outcome.

Effectiveness

Numerous studies confirm that positive reinforcement works (Ackerman, 2022). It is widely used in research and everyday life, from training dogs to raising children and running a classroom.

Clinics use it too. Contingency-management programs reinforce verified abstinence in addiction treatment, as well as medication adherence.

Several factors determine how well it works, and three matter most: contingency, immediacy, and a learner who currently wants the reinforcer. Most failures in practice trace back to one of them.

• Contingency. The reinforcer must depend on the behavior and follow it. One offered beforehand to induce the behavior is a bribe, not reinforcement.
• Timing. Deliver the reinforcer as soon as possible after the behavior. A longer gap weakens the connection and lets an unrelated behavior get reinforced instead.
• Motivation. A reinforcer only works when the learner currently wants it.
• Magnitude and quality. Larger, better reinforcers sustain more responding, with diminishing returns and more satiation (the learner having had enough). Small, frequent ones often beat one large one.
• Simplicity. Positive reinforcement never requires removing anything or introducing an unpleasant consequence.
• Positive associations. Learning linked to good feelings tends to be remembered well beyond the point where reinforcement stops (Ackerman, 2022).

Motivation deserves a closer look. Michael (1982) called this a motivating operation: deprivation increases how much a reinforcer is worth, and satiation reduces it to nothing.

Food reinforces a hungry animal, not a full one. This is why the same reward can work powerfully one day and fail completely the next.

This simplicity tends to improve morale and motivation, since it is generally easier to encourage a behavior than to discourage one (Kamery, 2004).

Common Mistakes When Using Positive Reinforcement

Because positive reinforcement feels intuitive, it is easy to apply badly. Five mistakes recur.

• Non-contingent delivery. Rewards given regardless of behavior, or beforehand as a bribe, do not strengthen the target response. A parent who gives in to end a tantrum reinforces it.
• Reinforcing the wrong behavior. Attention is a powerful reinforcer, so attending to misbehavior can strengthen it. Ayllon and Azrin (1968) documented staff inadvertently reinforcing withdrawn behavior this way.
• Adventitious (accidental) reinforcement. A chance pairing of a response and a reward can strengthen an irrelevant behavior. This is the mechanism behind Skinner’s superstitious pigeons (Skinner, 1948) and human lucky rituals.
• Ratio strain. Thinning a schedule too abruptly, so that far more responses are needed per reinforcer, makes responding break down.
• Over-reliance on tangibles. Sweets or money used where praise or natural reinforcers would serve risk satiation and can undermine interest in the task itself.

Good programs avoid these traps. They reinforce the desired alternative behavior, withhold attention from the unwanted one, fade schedules gradually, and rotate reinforcers.

Evidence Check: Do Token Economies Hold Up?

Verdict: Supported with qualifications.

Meta-analyses of classroom token economies report large behavioral gains. But they rest on single-case and small-study designs, and they test the whole token package rather than the generalized-reinforcer mechanism credited earlier in this article.

Classroom Token Economies: Large Effects, Single-Case Evidence

Level of evidence: Meta-analysis or systematic review; partly replicated.

The claim is that giving tokens for target behaviors, to be exchanged later for backup rewards, increases those behaviors in classrooms.

Soares, Harrison, Vannest, and McClelland (2016) pooled peer-reviewed single-case research published between 1980 and 2014. Single-case designs follow a few individuals through an intervention. The review covered 88 baseline-to-intervention (AB) contrasts from 28 studies and 90 participants.

Using an effect size that controls for baseline trend, they obtained a weighted mean of 0.82 (SE = 0.03, 95% CI 0.77 to 0.88). The CI, or confidence interval, is the likely range.

Keeping only studies with strong demonstrations of a functional relationship gave 0.85 (95% CI 0.74 to 0.97). The result barely moved.

Token economies were slightly more effective for youth aged 6 to 15 than for children aged 3 to 5, and for behavioral than for academic goals. No difference appeared between general and special education settings. Adding response cost (taking tokens away) or verbal cueing made no difference either.

Kim et al. (2022), described above, reviewed 24 studies from kindergarten to fifth grade and also reported large effects. Ivy, Meindl, Overley, and Robson (2017) reviewed 96 token-economy articles published between 2000 and 2015. Only 18 (19%) described every primary component in replicable detail.

So the evidence supports the practical claim. A token system delivered contingently raises target behavior in classrooms, and two reviews by different author teams reach the same large-effect conclusion.

The weight rests on Soares et al. (2016). It tested the claim directly, separated strong from weaker demonstrations, and tested moderators.

The claim still has limits. Single-case contrasts show change within participants. They cannot rule out that attention, novelty, or classroom routine contributed.

The reviews also test the token package, not the mechanism. They do not show that tokens work specifically as generalized conditioned reinforcers, or that gains last once the system is withdrawn.

How far can the evidence be trusted? Soares et al. (2016) pooled contrasts with no randomised comparison group, from a small sample of published classroom studies. It cannot show how large the effect would be across schools or for unpublished null results.

Its date range overlaps Kim et al. (2022) for 2000 to 2014, so some studies may be shared. The reviews are not fully independent.

Ivy et al. (2017) adds that incompletely reported procedures limit how far any pooled estimate can guide a teacher.

The effect is therefore real and large as pooled. How much of it comes from token reinforcement itself, rather than the surrounding program, and for how long, is not established by this evidence.

Critical Evaluation of Positive Reinforcement

Positive reinforcement is powerful and well replicated, but it has real limits. The strengths below explain why it is so widely used, and the limitations explain when it can fail or even backfire.

Strengths

• Objective and testable. A reinforcer is defined by its effect on behavior, not by whether it seems pleasant, so positive reinforcement can be tested, not assumed (Skinner, 1938).
• Predictive and practical. The Premack principle lets practitioners predict what will reinforce someone in advance. The underlying technology, including shaping and token economies, is backed by meta-analytic evidence (Premack, 1959; Kim et al., 2022).
• Humane relative to aversive control. It strengthens behavior without the fear or aggression that punishment can produce, which is why it is the first-line strategy in education and animal training.

Because a reinforcer is defined by its measured effect, practitioners can test a candidate reward directly instead of guessing. That limits wasted effort in classrooms and clinics.

The toolkit built on this logic, including shaping, token economies, and contingency management, now rests on meta-analytic evidence rather than isolated demonstrations (Kim et al., 2022). It transfers across ages and settings.

Limitations

• Reward can backfire. Rewarding an already-enjoyable activity can undermine later interest in it, though the effect is bounded, not universal (Lepper et al., 1973; Deci et al., 1999).
• Fragile transfer. Gains often fail to generalize beyond the program that produced them and can fade once reinforcement stops (Ayllon & Azrin, 1968).
• Reinforcers are personal and short-lived. What reinforces someone depends on their individual history and current state, so no single reward works for everyone or forever.
• Under-explains human cognition. In humans, psychologist Albert Bandura argued that reinforcement works partly through conscious expectation and observing others rewarded, not just automatic strengthening (Bandura, 1977).
• Ethical limits. Used covertly or for an institution’s benefit, reinforcement can override autonomy, as with engagement-maximizing “like” schedules.

None of these limitations is fatal, but each raises the cost of doing positive reinforcement well. The backfire risk is largely confined to salient, expected rewards for tasks people already enjoy.

It is mostly spared, or even reversed, by praise and by rewards tied to meeting a standard rather than to mere participation (Cameron et al., 2001).

Fragile transfer can be planned for. Practitioners program generalization deliberately by thinning the schedule, recruiting reinforcers already present in daily life, and practicing across varied settings, rather than depending on one artificial contingency.

Because reinforcers are personal and can stop working once someone is satiated on them, well-run programs also need ongoing reassessment and rotation of what is actually being offered.

Ethics sets a further limit. Reinforcement used covertly, or to serve an institution rather than the individual, raises questions of autonomy. Examples include “like” schedules engineered to maximize engagement, loyalty programs that encourage unnecessary spending, and ward regimes that reward compliance for staff convenience.

Best practice stresses transparency, informed consent, and minimal control over ordinary entitlements. It also balances extrinsic reinforcement with intrinsic motivation.

The Overjustification Effect

The most serious challenge to positive reinforcement is that rewarding an activity someone already enjoys can reduce later interest in it. Psychologists call this overjustification.

Evaluation: This is a robust, much-replicated finding, and it feeds directly into self-determination theory’s account of how controlling rewards can undermine autonomy. Its scope, however, is genuinely contested.

A meta-analysis of 128 experiments found tangible, expected rewards reduce free-choice intrinsic motivation, while verbal praise enhances it (Deci et al., 1999). A later re-analysis argued the effect is limited, not pervasive (Cameron et al., 2001). Rewards for meeting a high standard can raise interest rather than harm it.

The effect is real but bounded. Read together, the safest approach targets specific behavior or quality, such as behavior-specific praise, and is used carefully on tasks people already enjoy.

Rival Accounts of Why Reinforcement Works

Skinner defined reinforcement by its measured effect and stayed silent on mechanism. Two long-standing rivals try to say why consequences change behavior.

Cognitive Expectancy and Incentive Motivation

Bolles (1972) reviewed the difficulties facing reinforcement as the central idea in learning. He proposed two alternatives.

In the first, an organism learns that a response is followed by an outcome and behaves accordingly. In the second, an anticipated outcome creates a motivational state that pulls behavior toward it.

Neither treats reinforcement as an automatic strengthener of a stimulus-response link. Expectation does the causal work instead.

Brain research points the same way. Midbrain dopamine neurons signal a reward-prediction error, the gap between the reward received and the reward expected (Schultz, Dayan, & Montague, 1997).

They fire to an unexpected reward, shift to a cue that reliably predicts reward, and dip below baseline when an expected reward is omitted. That presupposes an expectation.

The two accounts make the same behavioral predictions: contingency matters, and unpredictable outcomes are especially potent. Behavioral data alone have never settled the debate.

Cognitive-Social Mediation

Bandura (1977) pushed the same challenge into human social behavior. He argued that reinforcement works largely by conveying information about which behavior pays off and what the incentive is worth, mediated by expectation and self-evaluation.

It is not an automatic, mechanical strengthener.

Humans also learn from vicarious reinforcement, by watching someone else rewarded, with no consequence to themselves. This account keeps the central finding that consequences change future behavior. The mechanism shifts from automatic strengthening to cognitive mediation.

None of these rivals overturns the testable core of the operant account. Each still predicts that consequences change future behavior, and interventions built on positive reinforcement work whichever mechanism is correct.

What is disputed is Skinner’s silence on mechanism. Where the operant analysis stops at the observed contingency, the rivals try to say why the contingency works.

Contemporary Research

Research on positive reinforcement has moved from single demonstrations toward large syntheses. Two recent meta-analyses stand out.

Token Economies Still Work, and the Details Matter

Teacher Praise Is Cheap, Evidence-Based, and Under-Used

Teacher praise is the cheapest form of positive reinforcement. The evidence for it keeps accumulating.

A meta-analysis of school-based coaching found that behavior-specific praise was among the practices teachers adopted most readily. Coaching reliably increased how often teachers used it and improved student outcomes (LaBrot et al., 2024).

Whether the reinforcer is a token or a specific word of praise, well-designed positive reinforcement produces reliable gains, and long-term maintenance remains the open frontier.

Further Information

• Sprouls, K., Mathur, S. R., & Upreti, G. (2015). Is positive feedback a forgotten classroom practice? Findings and implications for at-risk students. Preventing School Failure: Alternative Education for Children and Youth, 59(3), 153-160.