The dog backs up. Not bolting, not panicking, just a steady backward walk with his chin tucked, and the collar slides over his ears and off. Then he is loose on the sidewalk with a car coming and you are holding an empty loop of webbing.
This happens to greyhounds and lurchers, to whippets, to salukis, and to any dog whose skull is narrower than his neck. It also happens to anxious dogs who have learned that backing out is how you end a walk you did not want to take. For those dogs the equipment question is not about pulling at all. It is about whether the thing stays on.
The short answer
| Martingale collar | Harness | |
|---|---|---|
| Stops the dog backing out | Yes | Yes, if fitted properly |
| Keeps force off the throat | No | Yes |
| Reduces pulling | No | A front clip does, against a flat collar |
| Best for | Narrow heads, sighthounds, dogs that slip collars | Pullers, flat-faced dogs, collapsing trachea |
| Fails when | Fitted too tight, or too loose | Fitted loose, or the dog is a committed puller |
If your dog has slipped a collar, or has a head narrower than his neck, a martingale is the correct piece of equipment. If your dog pulls, a harness is the better choice and the martingale is not a pulling tool. Everything after this is about why the split falls there.
What a martingale is, and the fit that ruins it
Two loops of the same material. The large loop goes around the neck and adjusts like any collar. The small loop is the control, and the leash clips to a ring that the small loop runs through, so pulling straightens the loop and takes up its slack.
When the dog pulls, the small loop closes and the large loop narrows with it. When he stops, both open back up. Because the small loop only has so much slack to give, the collar stops closing at a fixed point. Hence the name limited-slip collar, and hence the different treatment from a choke chain.
The 2025 Cavalli and Protopopova review of walking equipment puts the problem in one line, in its definition of the device:
If kept loose on the neck, it cannot constrict more than a flat collar would when pulled; if kept tight, it would function as a choke collar rather than a martingale.
There is a correct band of fit, and it is narrow.
Set the large loop loose, and the small loop still gives its full travel when the dog pulls. That length does not change. Loose enough that the shut size still measures wider than his head, and it never presses at all: it rides low, it rotates, and a dog backs out of it. You have bought a safety device and disabled it.
Set the large loop snug at the neck, and the same travel now bites. The collar closes several sizes below his neck measurement. You have built a choke collar with its stopping point set too small.
The measurement that actually decides
Everything hinges on the circumference you end up with when the loop is fully shut, so measure that.
Run a soft tape around the widest part of his head, across the cheekbones and just in front of the ears. That number is the one the collar has to beat.
Now measure his neck where the collar actually sits, which is high, just behind the ears, not down at the widest part. On a sighthound those two numbers sit close together, and that is why these collars are cut the way they are.
Then pull the small loop shut by hand, off the dog, and measure around the opening it leaves, the same way. The shut opening has to be smaller than the head measurement, or the collar goes straight over his ears at the moment you need it to hold. And it has to stay above the neck-at-the-top measurement, or the strap is biting whenever he pulls.
When you cannot get both, the control loop is travelling further than that dog can afford, and no adjustment of the large loop fixes it. You need a collar with a shorter control loop, which is why sighthound owners buy from makers who ask for the head measurement as well as the neck.
What the strain gauge found
In 2024, Johnson and Wynne put a strain gauge between the handler and the leash and walked twenty-three shelter dogs for five minutes on each of four setups: a martingale, a front-connection harness, a polymer prong-style collar and a standard metal prong collar.
The dogs pulled hardest on the martingale. Significantly harder than on any of the other three, with the harness, the polymer collar and the metal prong collar all indistinguishable from each other.
The obvious reading is that the martingale hurt and the dogs were trying to get away from it. The data does not support that. The researchers checked for stress signals and did not find them raised on the martingale. Their suggestion ran the other way: the dogs may have pulled harder on it precisely because it was the most comfortable of the four, since a harness loads the shoulders and chest and a prong collar applies pain. Their word was “possible”, and it is a hypothesis, not a finding.
One caveat the authors raise themselves, and it matters. Every dog was walked in the same order, martingale first, and dogs are most likely to pull early in a walk, when they are most excited. Some of that gap could be first-walk enthusiasm rather than the collar. They also warn that shelter dogs on five-minute walks are not the same as your dog on your street.
What the study does establish is narrower and still useful. A martingale did not restrain those dogs’ forward drive at all.
The result that turns the whole thing
A 2021 study by Shih and colleagues tethered fifty-two shelter dogs to a wall-mounted tension meter, once on a neck collar and once on a back-connection harness, in a counterbalanced design.
The dogs pulled harder and for longer in the back-connection harness than on the collar — but only when a food treat was the reward. With a toy, the difference disappeared. And no stress signals differed between the two.
Set the two side by side. One found a back-clip harness increased pulling against a collar. The other found a front-clip harness significantly decreased it against a collar. Neither is wrong and they are not in conflict. The obvious difference is where the leash attaches, though the setups differ too, one dog tethered and lunging at a visible treat against one being walked, so treat the clip position as the likeliest explanation and not a proved one.
Anyone who thinks of a harness as the gentle option has to sit with this. Against a flat collar the front clip did cut the total pull, significantly and by a large margin. What it did not do is turn a puller into a dog that walks on a loose leash. The gentleness is somewhere else: the force comes off the throat and spreads across a chest and shoulders built to take load.
Choosing by what you are protecting
Not by which device wins, because none of them solves pulling.
Protect the throat. A dog that pulls hard puts load on a trachea, and some dogs are structurally less able to take it: flat-faced breeds, toy breeds, dogs with a collapsing trachea, older dogs with a cough. For those dogs the harness is the correct answer, and a martingale is the wrong one no matter how well it is fitted.
Protect against the escape. A dog that backs out of collars, or whose head is narrower than his neck, needs something that cannot pass over his skull. A properly fitted harness also prevents escape, but a determined dog can back out of a loose one, and a harness tight enough to stop that often rubs.
The 2025 review is blunt about the pulling case: for dogs that pull, non-tightening front-clip harnesses appear to offer the best balance between discomfort and prevention of pulling, while tightening harnesses, martingale collars and head collars “can pose greater discomfort” and should be used with caution. That places the martingale in the “specialist tool” column rather than the “upgrade on a flat collar” column.
When a martingale is the wrong tool
-
He pulls and has a normal-shaped head. The martingale adds nothing, and it puts a tightening band on his throat for no benefit.
-
He is flat-faced or has a collapsing trachea. Pressure on that airway is the specific thing to avoid.
-
He has learned to lean. A tightening collar teaches some dogs to lean into it, and you end up with a dog who pulls and a throat taking load.
-
You want to leave it on all day. Don’t. A collar that can tighten is a collar that can catch on something, and a dog shut in a crate wearing one is genuinely at risk. Walking equipment comes off indoors.
If you need both
A dog that slips collars and pulls is the awkward combination, and it is common. Run a harness for the walk and a martingale as back-up.
Fit the harness properly and clip the main leash to it. Then put the correctly fitted martingale on and give it its own second leash — unless he is on the list above, a flat-faced or toy dog whose throat should not take load at all, in which case get the harness fit right and skip the collar. If the harness fails, or the dog backs out of it, the collar is what is still holding him. Without that second lead you are carrying a spare collar, not a back-up.
Do not run one clip through both rings. A single clip joining the two drags the collar every time the harness comes under tension, and then both sit wrong.
The part equipment cannot do
None of this equipment teaches a dog to walk on a loose leash. The collar keeps a bad moment from becoming an injury while you do the actual work.
The work is dull and it works. Stop when the leash goes tight and wait for the dog to come back and give you slack. Reward the slack, every time at first. Change direction when he forges ahead. Spend five minutes on a quiet street before you try the squirrel route, and give it a few weeks before you decide it is not working.
If that has not moved things after a few weeks, a head halter gives you control over the head that no collar or harness can. It is also easy to misuse: introduce it with food over a couple of weeks, keep the leash loose, never jerk it, and have someone show you the fit.
Back to the dog on the sidewalk
The reason a martingale exists is that one moment: the dog backing up, the collar sliding, the hand suddenly holding nothing.
Fit it so the shut loop cannot reach his ears, put it high on the neck, take it off when you get home, and accept that it will not slow him down on the way there. If you need him to slow down, that takes different equipment and several weeks of a different kind of work. Buying the collar that stops the escape and expecting it to stop the pulling as well is where most people end up disappointed, and the collar is not the part that failed.