Would you raise a dog, or a child, and never teach them "no"?
Asked plainly, almost no one says yes. A boundary isn't cruelty — it's information. And yet ruling out the word "no" is the single commitment at the centre of force-free dog training. This page takes that idea seriously. It starts with everything the force-free side gets right, and follows the logic to where it runs out.
Every number on this page has a source you can check. Where the research is thin, the page says so.
Let's start with what everyone agrees on
Most of the argument about dog training is fought over a small corner of it. The centre is settled, and the force-free side is right about most of it.
For most pet-dog problems, reward-based training works.
When reward-based and mixed methods were compared on teaching a new task, the reward-based group did at least as well.
Source: Vieira de Castro et al. 2021, PLOS ONE, efficacy study
Training built on aversives measurably stresses dogs.
Dogs from aversive-based schools showed more stress behaviours in class, higher cortisol after sessions and a more pessimistic bias on a judgement task. Dogs from mixed-method schools showed more stress behaviours than reward-only dogs too.
Source: Vieira de Castro et al. 2020, PLOS ONE, 92 companion dogs across reward, mixed and aversive schools
A well-run e-collar programme is not more effective than a well-run reward programme, even for recall and chasing.
Owners in all three groups reported improvement at about the same rate. The e-collar dogs spent more time tense, yawned more and explored their surroundings less.
Source: Cooper et al. 2014, Defra-funded UK study, 63 dogs in three groups
In ordinary hands, tools go wrong.
In a French owner survey, e-collar results fell well short of what trainers achieve in studies, and 7% of the dogs had burns, most from automatic bark collars.
Source: Masson et al. 2018, owner survey, France
The veterinary-behaviour bodies recommend reward-based methods only.
The AVSAB position statement recommends reward-based training and says it has seen no evidence that aversives are needed, even for severe cases. Germany, Denmark and Wales ban e-collars. These are policy positions rather than experiments, and this page treats them that way.
Source: AVSAB Position Statement on Humane Dog Training, 2021
None of this is in dispute. The question is what happens at the edges.
A reward can say yes. It can't say no.
Force-free isn't a style that happens to be gentle. Its founding rule is specific: the word "no" is off the table. In its own framework, telling a dog "no" is filed under punishment — and punishment is the thing the whole approach is built to avoid. So a force-free trainer isn't choosing not to say no. They've ruled it out.
Force-free keeps one kind of "no": the treat goes away, the game ends, the door stays shut. On paper that's a consequence. In practice it's the weakest one there is. A child connects it across time — no homework, no PlayStation tonight — because a child can hold "because of that, this." A dog mostly can't. For the withdrawal to teach anything, it has to land in the same second as the choice, every time, and read as tied to the act — one of the hardest things in all of training to execute, and, not by accident, the tool working trainers reach for least. So the honest picture isn't that force-free has a quiet, gentle "no." Its only "no" is one that rarely arrives in a form the dog can use.
An everyday scene
The dog jumps on guests at the front door.
With rewards only, the trainer can
- Reward four paws on the floor
- Teach a sit at the door and pay it well
- Put the dog on a leash before guests arrive
- Ask guests to turn away until the dog settles
- Repeat until the jump fades
What the trainer cannot do
- Tell the dog, at the moment of the jump, that the jump itself was the mistake
For a lot of dogs this works, and that is fine. Turning away is a consequence, and jumping stops paying. The problem starts when the behaviour pays the dog better than the trainer does.
When the world pays the dog
Some behaviours carry their own reward. The chase is the fun. The bin is the meal. The other dog is the whole point of the pull. Nothing the trainer withholds changes that, because the trainer was never the one paying. In the studies we have, skilled reward-based trainers did fix recall and chasing in typical pet dogs. The hard cases are dogs with a strong predatory drive and live prey in front of them. That is where the evidence is thinnest and where a reward-only toolbox has the least to offer.
- Chasing wildlife or cyclists
- Scavenging on walks
- Pulling towards another dog
- Counter-surfing
You can't out-pay a squirrel.
Source: Cooper et al. 2014, 63 dogs: 92% of owners in all three groups reported improvement in recall and chasing, with no difference in efficacy; the e-collar group showed more tension. Dale et al. 2013, 120 dogs: one session of high-level aversion training kept 87% of dogs away from kiwi a year later.
Three words a balanced trainer teaches
- Yes
A marker for the right choice. The reward follows.
- Free
A release. The exercise is over and the dog may move.
- No
A marker for the wrong choice. A consequence the dog already understands follows.
A trainer who can only reward is working with half a vocabulary.
What happens after "we'll manage it"?
When prevention becomes the plan
Management means arranging the dog's world so a behaviour cannot be practised: a leash, a gate, a crate, a muzzle, distance from triggers. Both sides use it, and for a difficult dog both sides start with it. The question is whether it is a bridge or a destination.
- First step
Leash and distance
The dog is kept below threshold so training can begin. Sensible, and both sides do it.
- If the behaviour holds
Long line and a secure fence
Off-leash time is replaced by controlled space. The walk gets shorter and more planned.
- If it still holds
Muzzle on every walk
The muzzle becomes part of the walking kit. Guests are met behind a gate, or not at all.
- If nothing changes
A smaller life
The dog is safe. It is also crated, muzzled and walked at quiet hours, and the owner is managing it for the rest of its life.
Management that works is not failure. Management that never ends is a plan that stopped, and the equipment stays on the dog.
No study tracks how much equipment a dog wears over its life by training method. This ladder describes what careful reward-only plans for hard cases tend to contain: a long line, secure fencing, a muzzle that may stay on for good, and no guest greetings for months. The order and the ending are what we see in practice.
We have asked this question before, about children.
We already ran this experiment on children
This page opened with a question: raise anything — a child, a dog — and never teach it what "no" means, and see what you get. Force-free has an honest answer. Don't say no; redirect — reward the behaviour you'd rather see instead. Often that works; you can build a better habit in place of the old one. But redirection is a detour, not a boundary. It says "do this instead." It never says "that — don't." And some moments don't have a detour. The dog is already through the door.
Dogs of authoritative owners, high expectations and high responsiveness, did best.
Across 48 owner and dog pairs, their dogs showed the highest rate of secure attachment, were the most socially responsive, and were the most persistent and successful on a solvable puzzle task.
Source: Brubaker & Udell 2022, Animal Cognition, Oregon State University, 48 owner–dog pairs
Permissive
Low expectations, high warmth
The parent responds to everything and asks for little. Rules exist but rarely hold.
In the human literature: poorer self-control and more difficulty with limits. The dog study did not test this.
Trainers compare this to reward-only training that drifts into never saying no. That is an analogy, not a finding of the study.
Authoritarian
High expectations, low warmth
Rules are enforced. Little is explained and little is rewarded.
In the human literature: compliance with more anxiety and less initiative. The dog study did not test this.
Trainers compare this to correction-heavy, old-style compulsion training. The study measured expectations and responsiveness, not tools.
Authoritative
High expectations, high warmth
Clear rules, explained and held, by a parent who pays attention and responds.
In the dog study: the most secure attachment and the best puzzle persistence.
Trainers compare this to balanced training: mostly rewards, with limits that hold. The study supports the description of the owner, not any particular tool.
The study says nothing about tools. It says dogs do best with owners who expect a lot and respond a lot. That is a fair description of what the force-free side gets right, and of what the permissive drift gets wrong.
Most dogs are fine either way. Some are not.
The dogs where it matters most
Most of the argument is about ordinary pet dogs, and for them it barely matters who is right. The dogs that decide it are the ones that bite, chase, or work for a living. Here the choice of method stops being a matter of taste and turns into something with life-or-death consequences for the animal: rule a tool out on principle, and it is not there on the day management fails and a dog puts a bite through skin. No trial proves that one method sends more dogs to that day than another — the honest core of this beat is that the study has not been done. But when a rule about tools meets a genuinely dangerous dog, it is the dog, not the rule, that pays for the mismatch. Read the source under each claim; this is where the evidence is thinnest.
No controlled study compares outcomes for severe aggression by training method.
Both camps fill the gap with case stories, and neither can point to a trial. A balanced trainer worth hiring does not correct a dog that is already over threshold. The work happens below it.
Source: Gap confirmed across every source reviewed for this page; see the research paper
A well-timed aversive can stop predatory behaviour, and the effect can last.
In a New Zealand conservation programme, 120 hunting dogs had a single session of aversion training around kiwi, using a brief, high-level stimulus. All of them avoided kiwi at once. A year later, 87% still did, in new places, with no collar on. This is conservation aversion with a strong stimulus, not everyday pet training.
Source: Dale et al. 2013, Applied Animal Behaviour Science, 120 dogs, one-year follow-up
Unskilled or unpredictable corrections make dogs worse.
Belgian military handlers who used frequent, unpredictable aversives had lower-performing dogs, with a spiral of more punishment and more suppressed posture. In a separate lab study, dogs that could predict and control an aversive showed no lasting stress, while dogs that could not showed large cortisol spikes.
Source: Haverbeke et al. 2008, Applied Animal Behaviour Science; Schalke et al. 2007, Applied Animal Behaviour Science
Recall and chasing were fixed with and without the collar.
Experienced trainers who normally use e-collars got the same improvement when they worked without one. Their dogs also showed lower cortisol than the reward-only group. The honest reading is that trainer skill did the work.
Source: Cooper et al. 2014, 63 dogs; group B, the same trainers without a collar
The survey did not record training method and does not link euthanasia to force-free or balanced training. Most of these dogs were being managed by someone. Management that holds for years and fails once is the pattern the numbers describe, whichever method was in use.
Where the dog cannot fail
Police, military and detection dogs are trained under written standards, and those standards tell us what the people responsible for them permit. They do not tell us that every elite programme uses corrections, and no census of programmes exists. What they show is that the bodies certifying working dogs have not ruled corrective tools out, while the same field's own research shows that badly timed corrections cost performance.
USPCA and NAPWDA
Certification rules allow prong and remote collars to be worn during testing, though not activated.
Lackland Air Force Base
Supplies roughly 900 military working dogs a year to the US armed forces.
Source: USPCA and NAPWDA certification rules; Whole Dog Journal on Lackland AFB; Haverbeke et al. 2008
This is the section where the research is thinnest. No trial compares aggression outcomes by method, and the working-dog evidence is about standards, not results.
So what does the other side actually do?
What balanced training actually looks like
Balanced trainers use rewards to teach and a consequence to hold a behaviour the dog already knows. Most of the work is rewards. Tools, where used at all, come after the behaviour is taught. That is the position. Nobody has measured the proportion.
The ladder: least intrusive first
The LIMA principle, Least Intrusive, Minimally Aversive, asks a trainer to work down a list and stop at the first step that works. Balanced and force-free trainers both cite it. The IAABC endorsed it for years and has since moved to a stricter standard. The two sides disagree about whether the bottom rung should exist at all.
- Health and wellbeingRule out pain, illness and unmet needs before training anything.
- Arrange the environmentChange what the dog meets so the behaviour is not triggered.
- Reward the right behaviourTeach and pay the behaviour you want until it is fluent.
- Remove a rewardWithdraw attention, a game or access when the wrong behaviour appears.
- A correction the dog understandsOnly for a behaviour already taught, applied once, clearly, at the moment it happens.
Source: IAABC, Why animals need trainers who adhere to the least intrusive principle; IAABC Standards of Practice
A correction the dog can predict does not produce lasting stress. A random one does.
Dogs that could see the correction coming and avoid it through their own behaviour showed no notable or persistent stress. Dogs corrected unpredictably showed large cortisol rises. Small laboratory sample.
Source: Schalke et al. 2007, Applied Animal Behaviour Science
Frequent, unpredictable aversives made working dogs perform worse.
The more a handler punished without a pattern, the lower the dog's performance and the more suppressed its posture.
Source: Haverbeke et al. 2008, Applied Animal Behaviour Science, Belgian military handlers
In owners' hands, the tool causes harm.
Owner-reported e-collar success in France was far below what trainers achieve in studies, and 7% of dogs had burns. Timing and skill make the difference, and most owners have neither.
Source: Masson et al. 2018, Journal of Veterinary Behavior; Vieira de Castro et al. 2020, PLOS ONE
On the numbers
You may have read that balanced training is 95% reward, or that reward-only training tops out at 80% reliability. Neither figure has a source. The 95% is how balanced trainers describe their own work. The 80% is a slogan. This page does not use either as a statistic.
Where balanced training fails
The best-evidenced finding on this page is against unskilled tool use. Badly timed corrections stress dogs, suppress them and lower their performance. The balanced community sells tools more readily than it teaches timing, and most pet dogs never need a prong or an e-collar at all.
The evidence is not for the tool. It is for the timing, and against anyone who has the tool without it.
What a reliable dog gets to do
Training is not for obedience. It is for the size of the dog's life. Here is what a managed dog and a trained dog each get to do, as we see it in practice.
| Activity | Managed dog | Trained dog |
|---|---|---|
| Off-leash walk in the woods | Managed dogLong line only | Trained dogAvailable |
| A cafe with friends | Managed dogLeft at home | Trained dogAvailable |
| Family gatherings | Managed dogMuzzled, or behind a gate | Trained dogAvailable |
| Beach and open fields | Managed dogLong line, quiet hours | Trained dogAvailable |
| Meeting a new dog | Managed dogAvoided; cross the street | Trained dogAvailable |
| Guests at the door | Managed dogCrated until they leave | Trained dogAvailable |
| Holiday with the family | Managed dogKennel or a pet-sitter | Trained dogAvailable |
| Children in the garden | Managed dogSupervised and muzzled | Trained dogAvailable |
No study measures a dog's freedom over a lifetime by training method. This table is what we see in practice. The one measured signal in the sources cuts the other way too: in Cooper et al. 2014, the e-collar dogs explored their surroundings less. A tool, badly timed, shrinks the dog's world as well.
How to tell a good trainer from a bad one
The studies and the debates land in the same place. The divide that predicts outcomes is skilled versus unskilled, not positive versus balanced. Predictable, well-timed consequences did not stress dogs, and random ones did. Experienced trainers fixed chasing with and without the collar. Owners with a tool and no timing hurt their dogs.
Whatever a trainer calls their method, these are the things to check.
Starts with rewards, and can show it
Ask to watch a session. Most of what you see should be the dog being paid for getting it right.
Can say in advance when they would use a consequence, and what it would be
A good answer names the behaviour, the moment and the exact consequence. A vague answer means it will be improvised on your dog.
Measures results outside the training room
The test is your kitchen, your street and your in-laws' garden, not the trainer's field.
Refuses cases beyond their skill, and says so
A trainer who takes every aggression case is a warning sign. Severe cases start with a veterinary behaviourist.
Does not reach for a tool first
A prong or an e-collar in the first session, before the behaviour has been taught, is the wrong order whatever the method is called.
Keeps you involved
The dog goes home with you. A method you cannot carry out yourself is not a method.
Judge a trainer by what their dogs can do a year later, and by how they answer when they do not know.