You ask for a sit with a treat in your fingers and get it instantly. You ask with an empty hand and your dog looks past you, at the cupboard.
Everyone will tell you your dog is holding out for a bribe. What has actually happened is smaller and more fixable than that.
Your dog is reading your hand, not your word
If the food is visible before the behaviour, it stops being a reward and becomes part of the cue. The whole picture your dog learned is hand up, food showing, then sit.
Take the hand away and most of the cue has gone with it. Nothing was lost here. The plain version was never learned in the first place.
Why it falls apart now
Around the onset of adolescence, dogs respond less to the cues of the person who raised them, and that responsiveness returns later in the stage.
So a thin cue that held at 5 months old stops holding at 10 months old. The food was carrying more of the weight than you could see, and now the margin is gone.
A new place resets how easy a cue is, too. Learning happens below the level at which your dog cannot cope, which is why the kitchen version survives and the park version does not.
Safety note
Food is not a crutch you are meant to outgrow. Reward-based training is associated with higher reported responsiveness and fewer reported problem behaviours, and it is what the veterinary behaviour guidance recommends. The thing to change is the order, not the reward. Taking the food away and reaching for a correction instead tends to produce a quieter dog who offers less. Punishment-based approaches are associated with fear and avoidance.
Change the order, this week
Keep your hands empty and the food in a pocket or a pouch behind your hip. Your dog should not be able to see it when you ask.
Say the word once. Then wait 3 seconds, saying nothing and moving nothing. The waiting is the exercise.
The moment your dog does it, reach for the food and pay. The reward arrives after the behaviour, which is where it has always belonged.
Do 8 reps twice a day in one quiet room. Then take the same 8 reps to the hallway, the garden, and the pavement outside.
If your dog cannot do it with an empty hand, make the behaviour easier rather than showing the food again. Ask for a sit rather than a down.
Then make the pay less predictable
A reward that arrives every single time is easy to read. One that arrives often, sometimes twice, and sometimes a beat late holds up better in hard places.
Pay every rep for the first week. Once 8 out of 8 are clean, pay roughly 3 in every 4, picked at random rather than in a pattern.
Vary what the pay is. Food, a thrown ball, a chase game, a sniff at the lamppost, or being let off the long line.
Keep the hardest things on full pay, for years. Recall away from another dog gets food every time, for as long as you own your dog.
Measure the training food out of your dog's daily ration, so 16 reps a day does not become a weight problem.
What not to do
Do not hold food at your dog's nose to start the behaviour off. That is a lure, and a lure that never fades becomes the cue.
Do not say the word 4 times while your dog thinks about it. You are teaching the cue to be the fourth one.
Do not test the empty hand at the park first. Test it in the kitchen, where the plain version is easy, then move outward.
Do not go cold turkey on food to prove your dog can work without it. Reduce how predictable it is, not whether it exists.
Behaviour change is built from management plus reinforcement together. Here the management is where you keep the food, and the reinforcement is what happens after the behaviour.
What to expect
Expect a week or two of slightly worse-looking responses while the lure comes out. That dip is the plain cue being built, rather than the training failing.
If this is true of every cue rather than one, the Assessment covers all of them. It is free, takes about 3 minutes, and asks for no email address.
Today's version, in one line: food in your pocket, one word, 3 seconds of silence, then pay.