Search for how a dog harness should fit and you will receive, roughly forty times over, the same sentence: you should be able to slide two fingers under the strap. It is repeated with the confidence of scripture and the specificity of a horoscope. It is also, taken alone, close to useless, because it answers the question how tight while quietly skipping the two that actually matter: tight where, and pressing on what.

A harness is not a decorative loop. It is a load-transfer device. Every time your dog moves toward something interesting, the force of that intention travels down a lead, into a piece of leather or webbing, and out into a body. Where that force lands is the entire design problem. Land it correctly and the dog forgets it is wearing anything. Land it badly and you have built a machine for producing a shortened stride, a bald patch under the front leg, and a dog that has quietly concluded that walks are something to be endured.

So before the rule, the anatomy.

There are four structures under a harness that determine whether it is a good one. None of them appear in the average size chart.

Now the rule is worth something, because you know what it is protecting.

Two fingers, flat, under every strap, checked independently. Not one representative spot. Not the strap that happens to be easiest to reach. The neck loop, the chest strap and the belly band are three separate systems and they can each be wrong on their own.

And scale it to the dog. On a very small dog with narrow straps, two fingers is a physical impossibility and one is the honest check. On a large dog with wide straps, three fingers may be closer to correct. The number was always a proxy. What you are really testing is: does this strap stay exactly where I put it when the dog moves, and does it press on nothing that objects.

Two conditions, both required:

It does not move. Take the harness in your hand and try to rotate it. It should resist. A harness that turns on the body is a harness that will walk itself into the armpit within ten minutes.

It does not restrict. Then walk the dog and watch, from behind and from the side. Full stride, free shoulders, no hitch, no shortened reach. A harness can pass the finger test and still be wrong, because the finger test is static and your dog is not.

Start with what they have in common, because it is the part the rest of the market does not.

Neither goes over the head. This is unusual enough to be worth stating twice. The conventional H-harness, as sold almost everywhere, is dropped over the dog's skull and then buckled. For a great many dogs that single motion is the whole problem: a hand approaching the face, fabric across the eyes, and a daily negotiation that never quite resolves. Our H-harness is placed around the body instead. So is the step-in, by definition. Whichever you choose, nothing passes over your dog's head, ever.

The lead goes on the back ring, on both. The D-ring between the shoulder blades is the lead attachment. The smaller ring at the chest holds the hangtag and nothing else: it is not rated for a lead and should never carry one.

Now the difference, which comes down to a single number.

The step-in harness is sized on chest circumference alone. The dog places its front paws into two openings, the harness closes over the back, and you are done. One measurement, one motion, a few seconds. It is the simplest object we make in this category, and simplicity is not a lesser virtue when you are doing something twice a day for a decade.

The H-harness is sized on chest circumference and neck circumference. That second measurement is the entire argument. A harness fitted at two points around the body rather than one distributes pressure more widely, which is precisely what you want under a dog that puts real force into a lead. More adjustment, more contact, less concentrated load. It asks thirty seconds more of your time and gives it back in how the force is spread. The Marrone H-Harness in natural grain leather shows the construction clearly.

Everything above is theory until you put a tape around a dog. These are our published measurements. Note that the H-harness needs two numbers and the step-in needs one, and note that the breed columns are a starting point rather than an answer: a Poodle appears at four different sizes, which is the entire reason we do not size by breed or by weight.

Step-In Harness. Chest circumference, measured at the widest point just behind the front legs, dog standing on all four legs.

H-Harness. Two measurements: neck circumference at the base of the neck where it meets the shoulders, and chest circumference at the widest point behind the front legs.

There is no anti-pull harness in our collection today, and the reason is a specific objection rather than a general one. Our objection is to the mechanism, not the goal.

The harnesses marketed as anti-pull almost all work by the same mechanism: cords or straps that tighten somewhere sensitive, usually the armpit, when the dog pulls. The dog stops pulling. This is presented as training. It is not. It is discomfort applied at the moment of enthusiasm, and it works precisely to the degree that it is unpleasant, which is why the versions padded enough to be humane are also the versions that do not work.

It also teaches nothing. Remove the device and the pulling returns intact, because the dog learned that a particular object hurts, not that walking beside you is worth doing. In some dogs the association goes further and attaches itself to whatever was in view when the pain arrived: another dog, a cyclist, a child. Pain-based tools have a long history of producing reactivity in animals that had none.

So the honest position. A harness can change how pulling feels, and a well-designed one can make a walk easier for both of you. What it cannot do is teach: the lesson comes from you, and we have written about the method in how to teach a dog to walk on a lead. What the harness must do is narrower and non-negotiable: make sure that when your dog pulls, and it will, the force lands on bone.

This is where we have an interest, so take it with the appropriate salt. The mechanism is nonetheless real.

Leather is a fibre structure that responds to heat and pressure. Worn against a body it softens, drapes, and over some weeks takes the specific geometry of the specific dog. A leather harness at month three fits better than it did at week one. It has been learning the shape.

Synthetic webbing is dimensionally stable, which sounds like a virtue and is the problem. It presents the same edge, at the same angle, to the same patch of skin, on every walk, forever. That edge is where chafing begins, and it is why the armpit complaint is disproportionately a webbing complaint.

Two honest caveats. Leather needs maintenance: a few minutes with the leather care set, twice a year. And leather is not the right choice for a dog that swims on every walk. For everything else, it is the material that ends up fitting the dog rather than the size chart.

The correct answer to how a harness should fit is not a number. It is a description of a state: the dog forgets it is wearing it.

That state has conditions. Pressure only on bone. Nothing on the shoulder blade. Nothing in the armpit. Nothing on the throat. Nothing behind the last rib. Stable enough not to travel, loose enough not to bind. And re-checked, because dogs change and straps loosen and coats grow and the fit you set in March is a different fit in August.

Two fingers is where the conversation starts. It was never supposed to be where it ends.

Pressure on bone. Clearance at the shoulder. Two fingers, everywhere, every time.