Your children run across the garden. Your 10 month old collie drops into a crouch, sweeps round behind them, and nips at a heel or a trouser leg until they stop.
At 5 months old it looked like a game. It does not look like a game now, and a child who has been nipped is not wrong about that.
What is actually happening
Herding breeds are built to move animals: watch, stalk, circle, and use the mouth at the back end if the animal will not move. Children running and shrieking are the closest thing to livestock your garden contains.
Adolescence sharpens it. The sequence gets longer, faster and more committed. Dogs also respond less to the cues of the person who raised them around the onset of the stage. That responsiveness returns later in adolescence.
It also pays every time. The children stop or scream, which is exactly the result the behaviour is for. That is a rehearsal, and rehearsed behaviour gets easier for your dog to do.
Management today, before any training
Do not let running and children and your dog happen in the same space unsupervised. Behaviour change is built from management plus reinforcement, and this week is mostly management.
Put a house lead on your dog's harness so you can move them without your hands on their body.
Separate before it starts rather than after. When the garden gets fast, your dog goes behind a gate with a chew.
Give the children 1 rule and make it the same every time: if the dog gets busy, stand still and walk indoors. No running away, no shouting at the dog.
Nobody grabs, scruffs, pins or shouts your dog down. Punishment-based approaches are associated with fear, avoidance and defensive aggression, and you do not want that attached to your children.
Give the instinct somewhere to go
Suppressing the sequence does not remove it. A herding dog with nowhere to put watching, chasing and controlling movement will find the moving thing you did not choose.
Play a chase game you start and stop, with a flirt pole or a long toy on a rope, for 3 minutes twice a day. Your dog gets the stalk and the chase, on a toy, with you.
Teach a get it and a stop, in that order. Send your dog after the toy, pause the toy, and pay them for stopping. Control of movement is the reward.
Add scent work, which uses the nose rather than the eyes: scatter feeding, a hidden toy, a food puzzle. It leaves a dog settled rather than revved.
Ask for short bits of thinking work daily. Herding breeds tire out on problems faster than on distance.
Practise settle on a mat in short repeatable sets below the point where your dog cannot hold it. Relaxation is trained as a behaviour in its own right.
Teach the interruption
Learn your dog's run-up: the head drop, the hard stare, the wide arc. Call your dog at the stare, not at the nip.
Pay generously for a turn back to you while the children are still moving. That is the whole skill.
Send your dog to a mat or a bed when the garden speeds up, and pay them there. Give the arousal a destination.
Work below the level that pushes your dog past their limit. If your dog cannot take food, they are too close to the running, so add distance.
Reward-based methods are what the veterinary behaviour guidance recommends, and survey work associates them with higher reported responsiveness and fewer reported problem behaviours.
Safety note
Get help now if your dog has broken skin on a child. The same applies to nipping when nobody is moving, or to growling, freezing or a lunge at the face. Separate your dog from the children completely, book a veterinary examination, and speak to a qualified behaviour professional. Sudden behaviour change warrants a veterinary appointment, and pain frequently presents as a change in behaviour rather than as obvious lameness. A child is never the person who practises this with the dog.
Today's version, in one line: 3 minutes of flirt pole with you, and a gate across the garden door when the running starts.