Operant Dog Training: The Four Quadrants, Precisely

Operant conditioning is the relationship between a behaviour and its consequence, described in four combinations — something added or removed, making the behaviour more or less likely — and the reason the vocabulary is worth learning is that two of the four are what aversive training actually is, under names that sound neutral.

What operant conditioning is

Operant conditioning is learning from consequences. The MSD Veterinary Manual defines it as "making an association between a behaviour and the consequences of that behaviour", where the result either increases or decreases how likely the behaviour is next time.

It is one of two learning processes running in every training session. The other is classical conditioning, which is about associations between stimuli and creates how the dog feels, and the two are constantly interacting.

The four combinations exist because there are two things a consequence can do — arrive or be taken away — and two directions it can push a behaviour. That is all "the quadrants" means.

The four, defined properly

MSD's convention is worth memorising because it removes the moral colouring the words carry in English. "Positive" means something was applied; "negative" means something was removed. Reinforcement increases a behaviour; punishment decreases it.

Positive reinforcement. Something is added as a consequence, generally something the dog wants, and the behaviour increases. Food after a sit, the game continuing when the toy is dropped.

Negative reinforcement. Something is removed, generally something unpleasant, and the behaviour increases. MSD's own examples are avoidance and escape — and, pointedly, "if the owner puts pressure on a head halter until the desired behaviour is achieved, the release of tension is negative reinforcement."

Positive punishment. Something is applied, generally something unpleasant, and the behaviour decreases. A leash correction, a shout, a shock.

Negative punishment. Something is removed, generally something the dog wants, and the behaviour decreases. Play stopping when teeth touch skin.

There is a fifth thing that is not a quadrant at all. Extinction is what happens when a behaviour that used to be reinforced stops being reinforced, and it behaves differently from all four.

A reward is not automatically reinforcement

This is the sentence that changes how people train, and it is MSD's: "a reward is not synonymous with positive reinforcement. Unless there is a clear relationship (timing, consistency, contiguity) between the behaviour and the reward, the reward does not achieve the goal of positively reinforcing behaviour."

Three requirements, all mechanical. The reward has to arrive close enough in time to the behaviour for the dog to connect them, it has to arrive reliably while the behaviour is being learned, and it has to be contingent on that behaviour rather than on being nearby.

Most "he knows it, he just ignores me when it matters" is a failure in one of those three, plus a cue attached too early. MSD's sequence is deliberate: reward immediately and consistently until the behaviour is reliably repeated, and only then add the word before the behaviour-reward sequence.

Once it is learned, MSD says the behaviour can move to a variable schedule, where the time or number of repetitions before a reward changes. That is what makes a trained behaviour durable — and it is also, uncomfortably, the same schedule that makes counter surfing so hard to shift.

Why positive punishment fails in practice

MSD does not argue that punishment cannot suppress behaviour. It lists what has to be true for it to work: the aversive stimulus must occur "sufficiently close to the onset of the behaviour, with complete consistency and at an appropriate intensity."

Read that as a specification. Every single occurrence, within about a second, at an intensity high enough to stop the behaviour and low enough not to frighten the dog — in a house, with a job, with children, in the rain.

MSD then describes what happens when the specification is not met, and it is not simply "nothing". Punishment paired with exposure to a stimulus "can result in a conditioned fear of the stimulus", so a correction delivered while the dog is looking at another dog teaches the dog something about other dogs. Its list of examples includes pairing another animal with restraint or pain from a choke, prong or shock collar.

Two further limits are worth quoting. "Punishment cannot be used to achieve desirable behaviours, only to stop what is undesirable." And: "if an unpleasant consequence occurs only when the owner is present, the behaviour may continue in the owner's absence."

MSD's summary of the whole question, in its section on canine development, is one line: "positive punishment and negative reinforcement do not help dogs learn alternative, desirable behaviours; instead, these approaches break down trust and increase the risk of fear, anxiety, and aggression."

Extinction, and the burst

Removing reinforcement removes a behaviour, but not smoothly. MSD warns that owners "must be prepared for the intensity of the behaviour to initially increase before it is extinguished", which is called an extinction burst.

Two things make extinction slower: value and intermittency. Valuable rewards, a long history of performance and any intermittent reinforcement all increase resistance, and "even occasional petting of the dog in response to its jumping" counts.

Which gives the practical rule. Giving in halfway through teaches the dog that higher-intensity behaviour achieves the outcome, so a plan you cannot hold for two weeks is worse than no plan.

Negative punishment has its own failure mode along the same lines. If the dog cannot work out which behaviour removed the play or the attention, MSD says the undesirable behaviour "may actually intensify because of frustration" — which is why removing something always has to be paired with paying something else.

What "balanced training" means in this vocabulary

It means using all four quadrants, which in practice means adding positive punishment and negative reinforcement to a reward-based programme. Stated in quadrant language it sounds even-handed; stated in equipment terms it means prong collars, choke chains and electronic collars.

AVSAB's guidance for choosing a trainer lists "balanced training" as a red flag, alongside the tools and the vocabulary of boss, alpha, pack leader, respect, obey and command. Its 2021 position statement recommends reward-based methods for all dog training including the treatment of behaviour problems.

The evidence base behind that is not a single study. Ziv's review of aversive training methods and Vieira de Castro and colleagues' comparison of training schools both report worse welfare indicators in dogs trained with aversive-based methods, and no efficacy advantage that would justify the trade.

Is operant conditioning the same as clicker training?

Clicker training is one way of doing it. The clicker is a marker — a sound classically conditioned to predict food — that solves the timing half of MSD's timing, consistency and contiguity requirement by letting you mark the instant the behaviour happens even when the food arrives two seconds later.

Is negative reinforcement the same as punishment?

No, and MSD flags the confusion directly: punishment decreases behaviour and reinforcement increases it. Negative reinforcement increases a behaviour by removing something unpleasant when the dog performs it, which means something unpleasant had to be present in the first place.

Which quadrant should I use?

Positive reinforcement for what you want, management to prevent what you do not want, and negative punishment sparingly and always paired with a paid alternative. That is AVSAB's recommendation and MSD's practical advice, and it is what every protocol on this site is built from — the applied versions are in puppy training basics and dog trick training.

Does learning theory fix fear or aggression?

Not on its own, and this is where the vocabulary stops being enough. Fear is an emotional state, so the work is classical — changing what a trigger predicts — and MSD's advice for fearful or anxious animals is to prioritise desensitisation and counterconditioning over changing the behavioural response. A dog that has growled, snapped or bitten needs a veterinary assessment first.

Related guides

• heel meaning dog — Heel is a precise position held on cue.
• dog impulse control training — Impulse control is a set of trained behaviours, not a personality trait.
• puppy training basics — The order matters more than the list.
• dog trick training — Tricks are where you build marker timing, rate of reinforcement and shaping.

Every guide we have written is on the dog training guide index.

Sources

1. Treatment of Behavior Problems in Animals — MSD Veterinary Manual (Merck), professional edition, 2025.
2. Glossary of Behavioral Terms for Veterinary Medicine — MSD Veterinary Manual (Merck), professional edition, 2024.
3. Social Behavior of Dogs — MSD Veterinary Manual (Merck), professional edition, 2025.
4. Position Statement on Humane Dog Training — American Veterinary Society of Animal Behavior (AVSAB), 2021.
5. How to Choose a Trainer Position Statement — American Veterinary Society of Animal Behavior (AVSAB), 2026.
6. The effects of using aversive training methods in dogs—A review — Journal of Veterinary Behavior 19:50-60 (Ziv), 2017.
7. Does training method matter? Evidence for the negative impact of aversive-based methods on companion dog welfare — PLOS ONE 15(12):e0225023 (Vieira de Castro, Fuchs, Morello, Pastur, de Sousa & Olsson), 2020.