The Four Quadrants of Dog Training, Explained Simply

The confusing bit first: positive and negative are arithmetic

The four quadrants come from operant conditioning, and the two words that trip everyone up were borrowed from arithmetic, not from ethics. Positive means something was added. Negative means something was taken away. Neither word says anything about whether the method is kind or cruel.

The second word tells you what happened to the behavior. Reinforcement makes a behavior more likely to happen again. Punishment makes it less likely. That is the whole definition, and it is measured by what the dog does next, not by what you meant. If you shout at a dog for barking and the barking gets worse, you have not punished anything, whatever it felt like at the time. You may well have reinforced it with attention.

Put the two pairs together and you get four combinations: add something good, remove something bad, add something bad, remove something good. Every training method you have ever seen sits in one of them, and most sit in two or three at once.

Practical take: Judge a method by what the dog is doing a week later, not by the label on the tin. Plenty of things sold as corrections are quietly reinforcing the behavior they are aimed at.

Positive reinforcement: you add something the dog wants

The dog does something, you add something they want, and the behavior becomes more likely. Your dog sits, a piece of chicken arrives, sitting gets more common. That is the entire mechanism.

The reinforcer does not have to be food. A thrown ball, the back door opening, a scratch at the base of the tail, permission to go and sniff a lamp post, being let off the lead: any of these work if the dog wants them at that moment. Only the dog decides. Praise is a powerful reinforcer for some dogs and background noise for others, and if the behavior is not increasing, the thing you are handing over is not doing the job.

This is the quadrant nearly all skill-building happens in, and it is worth learning properly rather than approximately. Our guide to positive reinforcement dog training covers choosing reinforcers, timing them and fading them out.

Negative reinforcement: you take away something the dog dislikes

Something unpleasant is already happening, the dog does something, it stops, and that behavior becomes more likely. The everyday example is steady pressure: a dog feels a constant pull on the lead, steps back toward you, the pull stops, and stepping back gets more common because it switches the pressure off. So does a head collar that tightens while the dog forges ahead and slackens when they drop back.

The quiet problem is structural. Negative reinforcement needs something aversive running in the background so that there is something to remove, which means the dog has to be uncomfortable before the learning can start. That is why it sits much closer to punishment than the tidy arithmetic suggests, and why it rarely appears on purpose in a modern plan. If lead pressure is your main teaching tool, our guide on loose-leash walking works through the reward-based version of the same problem.

Positive punishment: you add something the dog dislikes

The dog does something, you add something they dislike, and the behavior becomes less likely. A shout, a leash jerk, a squirt bottle, a shock from an electronic collar, a knee in the chest for jumping. This is the quadrant most people picture when they hear the word training, and it is the one carrying the clearest evidence against routine use.

It also has awkward technical requirements. To suppress a behavior cleanly, the unpleasant thing has to arrive immediately, arrive every single time, and be strong enough to work on the first try. Miss any of those and you get a dog who is anxious and still jumping. Meet all three and you have, by definition, frightened or hurt the animal you live with. Why punishment-based training backfires goes through the fallout in detail.

Negative punishment: you take away something the dog wants

The dog does something, you remove something they want, and the behavior becomes less likely. The dog jumps up for a greeting, you turn your back and the attention vanishes, jumping stops paying. The puppy's teeth close on your hand, the game ends for ten seconds. The lead tightens on the way to the park, you stand still, and the park stops getting closer.

This is the honest home of most corrections a reward-based trainer uses, and it deserves naming out loud, because plenty of people claim to work in one quadrant while using this one all day. It costs the dog nothing except a missed opportunity, which is what makes it defensible.

It also has a real limit. Withdrawing a reward only works when the thing you withdraw is genuinely what the dog is working for. That is why it handles jumping and attention-seeking so neatly, and does almost nothing for a dog barking at a noise in the street, where the payoff is the noise going away and you never had control of it. The alternatives for those cases are in eight ways to stop unwanted behavior without punishment.

Which quadrant owners actually reach for

Hiby, Rooney and Bradshaw surveyed 364 dog owners for Animal Welfare in 2004 and asked what they really did at home. Sixty-six percent used vocal punishment. Sixty percent used praise. Fifty-one percent used food rewards. The most common training tool in the average house is a raised voice, and almost everyone is mixing quadrants without a plan.

The same survey found that the more tasks an owner trained using punishment, the more problem behaviors they reported. Be careful with that one. It is a correlation in a questionnaire, not a controlled trial, and it cannot tell you whether the punishment produced the problems or the problems provoked the punishment. It is still worth knowing, because the direction people usually assume, that punishing more reduces problems, is not the direction that showed up.

There is a related wrinkle from Cooper and colleagues in 2014: owners who used aversive methods reported higher confidence in them. Feeling certain that a method is working is not evidence that it is.

What the controlled evidence says about outcomes

Two trials are worth knowing by name, because naming them is how you tell a claim from an opinion.

China, Mills and Cooper published a controlled trial in Frontiers in Veterinary Science (7:508, 2020) comparing 63 dogs across three matched groups: reward-based trainers against trainers using electronic collars. Dogs trained with rewards responded to the first recall cue 82.5% of the time, against 71.0% for the e-collar group (p = 0.015), and sat faster after the recall, 1.36 seconds against 1.67. The aversive method was not winning even on plain obedience, which is the ground it is normally defended on.

Vieira de Castro and colleagues, in PLOS ONE 15(12) (2020), followed 92 companion dogs attending aversive, mixed and reward-based training schools. The stress measures separated the groups sharply: mean lip-licking counts of 55.9 in the aversive group against 4.1 in the reward-based group, an effect size of d = 3.9, alongside elevated cortisol. Weeks later, in a cognitive-bias test, the aversive-group dogs read an ambiguous situation more pessimistically.

Now the honest caveat, because the internet usually skips it. Those trials covered obedience tasks, recall and sit, in ordinary companion dogs. They do not show that reward-based training resolves severe aggression, and anyone who tells you the literature has settled that question is stretching it past what was measured. If you are dealing with growling, snapping or a dog who frightens you, read up on what drives aggression in dogs and get a qualified behaviorist involved.

Why most modern trainers live in one quadrant

Put the survey and the trials together and the working position most current trainers hold is not really a moral one. It is what falls out of the measurements. Build behavior with positive reinforcement. Manage the environment so the dog rarely gets to rehearse what you do not want. Keep negative punishment, withdrawing the reward, for the small set of behaviors the dog is doing to get something you control. Leave the other two quadrants alone, because one requires the dog to be uncomfortable first and the other has to hurt to work.

Karen Pryor made this argument decades before the trials existed. In Don't Shoot the Dog! (Bantam, 1985) she sets out eight methods of ending an unwanted behavior and rates only four of them kind or reliable. Punishment is not among the four. Her fifth method, training an incompatible behavior, is the one that does most of the work in real life: teach the dog to do something that physically cannot happen at the same time as the problem. A dog holding a sit by the door is not jumping on the visitor.

The quadrants are a map, not a method. They tell you what category a technique falls into; they do not tell you what to teach first, or how long a session should run, or what to do when your dog can sit in the kitchen and not in the park. If you want that translated into an actual order of skills, Tavi's training module builds the sequence around your dog rather than around the theory.

Frequently asked questions

What are the four quadrants of dog training?

They are the four combinations of adding or removing something with increasing or decreasing a behavior. Positive reinforcement adds something the dog wants, such as food for a sit. Negative reinforcement removes something the dog dislikes, such as lead pressure stopping when the dog moves back. Positive punishment adds something the dog dislikes, such as a leash jerk. Negative punishment removes something the dog wants, such as attention disappearing when the dog jumps. Positive means added and negative means removed, nothing more.

Is negative reinforcement the same as punishment?

No, and this is the most common mix-up. Negative reinforcement increases a behavior by removing something unpleasant, so the dog repeats whatever switched the discomfort off. Punishment reduces a behavior. They are opposites in effect. What confuses people is that negative reinforcement still requires something aversive to be present in the first place, so in practice it feels close to punishment even though the technical label is reinforcement. That is a fair reason to be cautious about it, but it is not the same quadrant.

Which quadrant should I use to train my dog?

Positive reinforcement for almost everything, with negative punishment for the few behaviors your dog is doing to get something you control, such as jumping for attention. The controlled evidence supports this. In China, Mills and Cooper's 2020 trial in Frontiers in Veterinary Science, reward-trained dogs recalled on the first cue 82.5% of the time against 71.0% for dogs trained with electronic collars. Reward-based training was not a compromise on results; it simply came without the stress findings the aversive groups showed.

Is positive punishment ever justified?

It is hard to justify for ordinary training problems. To suppress a behavior cleanly, positive punishment has to be immediate, consistent and strong enough to work first time, which means it has to genuinely frighten or hurt the dog. DK's Complete Dog Breed Book describes harsh handling as simply not effective and likely to cause more problem behaviors, including aggression and anxiety. In a genuine safety emergency people will physically intervene, but that is interrupting a moment, not a training method you build on.

What is an example of negative punishment with a dog?

Your dog jumps up to greet you, and instead of pushing them down you turn away and the attention stops. Nothing was added; something the dog wanted was removed, so jumping pays less. The same shape applies when a play session stops the moment puppy teeth land on skin, or when you stand still because the lead went tight. It only works when you actually control the thing the dog is after, which is why it fixes attention-seeking and not barking at street noise.

Tavi's guides are informational and not a substitute for professional veterinary advice, diagnosis or treatment. When in doubt, call your veterinarian.

Sources

• Hiby, Rooney & Bradshaw, Animal Welfare 13 (2004)
• China, Mills & Cooper, Frontiers in Veterinary Science 7:508 (2020)
• Vieira de Castro et al., PLOS ONE 15(12) (2020)
• Cooper et al. (2014)
• Karen Pryor, Don't Shoot the Dog! (Bantam, 1985)
• DK, The Complete Dog Breed Book, 2nd edn (2020)