Psychology › Learning & behavior › Operant conditioning
Reinforcement and punishment
· Concept
Well replicatedStrong, repeated evidence.
Consequences that make a behavior more likely (reinforcement) or less likely (punishment) to happen again.
See it on the Psychology map →What it says
Behaviorists sort consequences into four types. Positive reinforcement adds something good (a bonus for closing a lease). Negative reinforcement removes something bad (the car's chime stops when you buckle up); it also strengthens behavior and is not punishment. Positive punishment adds something unpleasant (a fine); negative punishment removes something valued (a lost privilege). Immediate, consistent consequences work best. Rewarding the wanted behavior generally works better than punishing the unwanted one, which tends to stop only while the punisher is watching.
Why it matters
These rules quietly run workplaces, parenting, loyalty programs, apps and incentive pay. Knowing them helps design incentives that reward what you actually want, not what is easy to count.
Think of it like
Lease terms already use it: a free month for signing early reinforces; a late fee punishes paying late.
The evidence
Status: well replicated. Thousands of animal and human experiments since the 1900s; a 2016 meta-analysis linked spanking to worse child outcomes. A limit: rewards can undercut enjoyment of tasks people already like.
Open questions
When rewards backfire: how much paying for an already enjoyable activity lowers inner motivation is still argued between behaviorists and motivation researchers.
Inside this (1)
Connections (5)
▸Dopamine, reward and prediction error signals reward in Reinforcement and punishmentWell replicated
The connection
Dopamine supplies the teaching signal that reinforcement needs. A better-than-expected outcome makes dopamine neurons burst, strengthening whatever behavior came just before; a worse-than-expected one makes them dip. Once a reward is fully predicted, the burst moves to the earliest cue that predicts it. That is the cellular version of Thorndike's and Skinner's rules.
Who showed it and when
James Olds and Peter Milner found rats would press levers to stimulate reward areas at McGill in 1954; Wolfram Schultz recorded monkey dopamine neurons in Fribourg, Switzerland, in the 1980s–1990s.
The evidence
Recordings in monkeys and humans, plus experiments where switching on dopamine cells alone, with light, made mice learn and seek out a location, show the link is causal.
Open questions
Whether dopamine signals reward itself or importance more broadly: some dopamine cells respond to threats or surprises, and 'liking' a reward seems to rely on other chemicals.
▸Thorndike's law of effect stated the principle behind Reinforcement and punishmentWell replicated
The connection
Thorndike's law of effect is the core idea of reinforcement: a satisfying result strengthens the link between a situation and the action that produced it, and an 'annoying' one weakens it, the original version of punishment. Later behaviorists replaced his words 'satisfaction' and 'annoyance' with neutral terms defined only by their effect on behavior.
Who showed it and when
Columbia University, New York: his 1898 dissertation, his 1911 book Animal Intelligence, then experiments with human learners around 1930.
The evidence
Thorndike's published learning curves, and textbooks that cite the law of effect as the origin of reinforcement theory.
Open questions
By the 1930s Thorndike concluded punishment weakens behavior far less than reward strengthens it; how symmetrical the two really are is still argued.
▸Reinforcement and punishment is the engine of Operant conditioningWell replicated
The connection
Operant conditioning runs on consequences. Reinforcement increases how often a behavior occurs; punishment, or simply no reward, decreases it. Skinner defined these terms purely by effect: something counts as a reinforcer only if it actually strengthens behavior, whatever anyone intended. A bonus nobody values reinforces nothing.
Who showed it and when
Formalized in Skinner's labs at Harvard and Minnesota, 1930s–1950s, building on Thorndike's work and borrowing the word 'reinforcement' from translations of Pavlov.
The evidence
Countless experiments that change rewards and measure response rates; the same patterns appear in classrooms, clinics and workplaces.
Open questions
Why some rewards undermine motivation, and what reliably reinforces humans: praise, money, attention and even information can all work, but not predictably.
▸Variable reward schedules shape the power of Reinforcement and punishmentWell replicated
The connection
The same reward can build weak or very stubborn behavior depending on its schedule. Rewarding every response teaches fastest but fades fastest once the reward stops. Intermittent, unpredictable rewards make behavior far more persistent, the 'partial reinforcement extinction effect'. Timing, not just size, sets the strength of reinforcement.
Who showed it and when
Skinner stumbled on intermittent reward at Harvard in the early 1930s, by his account while stretching a short supply of food pellets; he and Charles Ferster catalogued schedules in 1957.
The evidence
Thousands of pigeon and rat experiments, plus human studies; Lloyd Humphreys demonstrated the extinction effect in people in 1939.
Open questions
Why unpredictability makes behavior so persistent: theories cite frustration, difficulty noticing that rewards have stopped, and dopamine responses to uncertainty, without full agreement.
▸Self-determination theory (autonomy, competence, relatedness) limits the reach of Reinforcement and punishmentMixed evidence
The connection
Behaviorism held that rewards strengthen behavior. Deci showed a limit: rewarding people for activities they already enjoy can shift their reasons from interest to payment, so interest drops when rewards stop. Expected, tangible rewards tied to doing a task undermine motivation most; unexpected rewards and sincere praise usually do not. For dull tasks, rewards still raise effort.
Who showed it and when
Edward Deci's 1971 puzzle study; Mark Lepper, David Greene and Richard Nisbett's 1973 study of preschoolers drawing for 'Good Player' awards at Stanford; Deci, Richard Koestner and Ryan's 1999 meta-analysis.
How solid is it?
The 1999 meta-analysis of 128 experiments found the undermining effect reliably; rival reviews by Judy Cameron and David Pierce argued it is narrower than claimed.
Open questions
How much this matters for pay at work, where few tasks are done for fun; incentive pay usually raises output on routine tasks.