Do the pocket test tonight. Ask your dog to sit twice: once with a treat visibly in your hand, once with both hands empty and the treats in a cupboard. If you get a fast sit the first time and a long stare at your hands the second, you do not have a stubborn dog. You have a dog who has learned exactly what you taught him, which is that the sight of food is part of the instruction.
This is probably the most common complaint about reward-based training, and it gets answered badly in two opposite directions. One camp says ditch the treats, he should work to please you. The other says just keep feeding forever. The research contradicts the first, and the second is unnecessary. The actual fix is about timing, and it takes about two weeks.
The short version
- A treat shown before the behaviour is a lure. A treat that appears after is a wage. Dogs that “only work for treats” have usually been lured for months, so the food became part of the cue.
- “He should do it to please you” fails on the evidence. In choice experiments with 114 dogs, vocal praise alone held dogs’ interest no better than being ignored.
- The fix is not removing food. It is hiding it until after the behaviour happens, then paying unpredictably.
- Rewards do not have to be food forever: door opening, lead clipping on, a thrown ball and sniffing time all work as pay once the behaviour is solid.
- If he ignores you even with food in your hand, the problem is not motivation and this article is not the fix.
On this page
What actually went wrong
Think about how the sit was taught. Treat in hand, hand above the nose, bottom goes down, treat goes in. That is luring, and it is a perfectly good way to get a behaviour started. The problem is what happens when you keep doing it for months: the dog is not learning “sit means put my bottom down”. He is learning “the full picture of a person holding food, saying a word, standing in the kitchen” means pay is available.
Remove the food from the picture and, from his side, the instruction is incomplete. He is not refusing. He is waiting for the rest of the sentence.
There is a second habit that feeds this: reaching for the treat pouch before you give the cue. Dogs read hands far better than they parse words. The reach itself becomes the real signal, and a cue given without the reach is a different, weaker signal.
The praise myth, measured
The usual advice at this point is that the dog should work for your approval instead. That has been tested, and it is worth knowing what happened.
Researchers Erica Feuerbacher and Clive Wynne ran a series of choice experiments in which dogs could spend time with people offering different things. In the 2015 study, run across two experiments with 114 dogs of mixed breeds, dogs strongly preferred a person petting them to a person praising them. The sharper finding sat underneath: dogs spent no more time with the person delivering vocal praise than with a person offering no interaction at all.
Their earlier work pointed the same direction: most dogs, given the choice, preferred food to petting, and praise only acquires value when it is repeatedly paired with something the dog already wants. The American Veterinary Society of Animal Behavior’s review of the work draws the practical conclusion: praise is a marker you can build value into, not a currency that starts with any.
So “he should just want to please me” is not a training plan. It is a wish. Your voice can become valuable to your dog, but only the way money becomes valuable: by being reliably exchangeable for something real.
Bribe or wage: the difference in practice
| Bribe (lure) | Wage (reinforcement) | |
|---|---|---|
| When the dog sees food | Before he does anything | Only after the behaviour happens |
| What the dog learns | “Food visible = worth working” | “This behaviour makes food appear” |
| Empty hands | Response collapses | Response holds, because empty hands were always part of the picture |
| Long term | You are stuck carrying food forever | Food can become occasional |
Every professional trainer uses food. The ones whose dogs work with empty hands moved from the left column to the right one early, usually within days of teaching a new behaviour. Most owners never make the switch because nobody tells them it exists.
The two-week fix
- Get the food off your body and out of sight. A pot on a shelf in each room you train in. The picture the dog sees at cue time must contain no food and no pouch.
- Cue once, with still hands. Say “sit” and do nothing. Give him a few seconds to think. The first empty-handed responses are slow. That is the dog working it out, not ignoring you.
- Mark, then move to the pay. The instant he sits, say “yes”, then walk to the pot and pay. The gap between behaviour and food is bridged by the word, and the word is what starts acquiring value.
- If he fails twice, make it easier, not louder. Repeating a cue five times teaches him the real cue is the fifth one. Go to a quieter room, ask for something simpler, rebuild.
- Keep sessions short. Three minutes, several times a day, beats one long session where you both get frustrated.
- Do not test him in the park in week one. Empty-hand responses need to be boringly reliable at home before they meet distractions. Outdoors is a different problem, which we covered in why your dog listens at home and ignores you outside.
Fading food without losing the dog
Once the behaviour is reliable with empty hands, you can stop paying every time. The way you do it matters.
Pay unpredictably rather than on a schedule the dog can learn. Sometimes the third sit earns food, sometimes the first, sometimes a jackpot of several pieces for a particularly fast response. Unpredictable pay is what keeps people pulling slot machine levers, and it keeps behaviour steady in dogs for the same reason. What kills behaviour is a predictable slow drain, where the dog can feel the pay getting rarer week by week until working stops being worth it.
And widen what counts as pay. The lead going on is pay for sitting still. The door opening is pay for not barging it. A released “go sniff” is pay for walking nicely past the lamppost. These are rewards you are already carrying, and using them means food becomes one currency among several rather than the only one.
When it is not about treats at all
Two situations look like a treat problem and are not.
If your dog ignores you even when you are holding food, motivation is not the issue. Either the environment is too hard, the dog is anxious or in pain, or, if he is between about nine and fifteen months, you are in the developmental window where responsiveness to the owner dips on its own. Fear or reactivity that appears in that window needs help rather than patience.
And if he takes the treat and then immediately blows you off, the pay is fine but the rate is wrong for the setting. A piece of kibble is fair pay in the kitchen and an insult in a field full of rabbit smells. Match the wage to the difficulty of the job, not to what is convenient in your pocket.
Common questions
Is using treats just bribing my dog?
Showing food first and hoping is a bribe. Food appearing after the behaviour is a wage. The distinction is the entire subject of this article, and it is the difference between a dog who checks your hands and one who does not.
Will my dog get fat from training treats?
Not if you count them. Use part of his daily food ration for training, cut pieces to fingernail size, and save the richer stuff for hard environments. The quantity per response can be tiny; dogs respond to the event more than the volume.
How long before I can train without food entirely?
Reframe the goal. You will always pay your dog with something: food sometimes, play, access, sniffing, petting. What disappears is food being visible before the behaviour, and that can be gone within a couple of weeks.
My trainer says a dog that needs treats does not respect you. True?
Respect is not a mechanism dogs run on, and the praise research shows approval alone carries almost no value that has not been deliberately built. A dog working for reliable pay is not disrespecting you. He is behaving like every organism that has ever been studied.
He does it for my partner without treats but not for me. Why?
Look at the history each of you has with the dog. Whoever paid unpredictably, or whose cues have been backed up more reliably, gets the better response. It is bank balance, not favouritism.
The bottom line
A dog who only works when he can see food learned that from how the food was used, and he can unlearn it in about two weeks: food out of sight, cue once, mark the behaviour, then pay. Keep the pay unpredictable, let doors and walks and sniffing count as wages, and stop expecting praise to fund the operation. The evidence says it never could.
Sources
- Feuerbacher EN, Wynne CDL. Behavioural Processes, 2015. Dogs preferred petting to vocal praise across two experiments with 114 dogs; praise alone matched no interaction.
- Feuerbacher EN, Wynne CDL. Journal of the Experimental Analysis of Behavior, 2014. Most domestic dogs preferred food to petting in concurrent choice.
- American Veterinary Society of Animal Behavior. Research review of the petting and praise studies.
