How Load Reaches the Neck

Torque, not weight

Pick up a five-kilogram plate and hold it against your chest. Now hold it at arm’s length. Same plate, same gravity, completely different experience.

Nothing about the weight changed. What changed is the moment arm — the perpendicular distance between the line along which the force acts and the joint resisting it. Torque is force multiplied by that distance, and it is torque, not weight, that muscles have to overcome.

The neck is the joint where this matters most, for a reason peculiar to it: the load it resists most often is the head itself, and the head cannot be set down.

An adult head weighs roughly 4.5 to 5 kg. Held in neutral, its centre of mass sits nearly over the cervical spine, the moment arm is close to zero, and the muscular demand is small. Move it forward by a few centimetres and the moment arm — and therefore the torque — climbs proportionally. Every waking hour, without a break.

This is the whole basis of the tech neck discussion, and it is the part of that discussion that is unambiguously true.

The four resistance profiles

Every method of loading a neck falls into one of four patterns, and they are not interchangeable.

Gravity acting on a mass fixed to the head — a weighted helmet. Torque is proportional to how far the head’s centre of mass has moved from vertical. Zero in neutral, maximal at maximum tilt. This means weighted headgear trains hardest exactly where the cervical spine is least tolerant, which is the argument for keeping the range small and progressing load instead.

Gravity acting on a hanging weight via a strap — a classic neck harness with plates. Similar in principle, but the geometry depends on where the strap runs and where the weight hangs, which gives you more control over where in the range the peak falls.

A cable over a pulley — roughly constant resistance through the range, because the pulley redirects gravity into a straight line of pull that does not change as the head moves.

Elastic — ascending resistance. Lowest at the start, highest at full stretch. Which means a band loads hardest at end range, where the joint tolerates least. It is also why bands plateau: their peak tension arrives at the point you should not be loading hardest.

A hand — accommodating resistance, matching capacity moment to moment. The only profile that adapts to you rather than to geometry. See manual resistance.

Attachment point determines direction

This is the practically important consequence, and it is why the same harness can train four different things.

The neck resists load along the line between the attachment point and the anchor. Move the anchor and you change the movement — not by a little, but categorically.

• Anchor in front, load at the crown or forehead → extension is trained
• Anchor behind → flexion
• Anchor to the side → lateral flexion
• Anchor diagonally → a combined movement, loaded through a diagonal

This is not a detail. Someone who sets up a harness without thinking about anchor geometry is training whatever direction the geometry happened to produce, which is frequently not the one they intended.

It is also the reason the 1896 patent quoted throughout this site bothered to move the attachment to the crown:

to provide, by reason of the connection with the head stall at the top of the head, for the exercising or movement of the head in every possible direction Without danger or liability of disarranging the headstall or harness.

US Patent 559,270, George O. Edwards, granted 28 April 1896.

A crown attachment lets the line of pull come from anywhere without the harness rotating off position. A side attachment locks you into the sagittal plane.

The bridge, as a worked example

Hackenschmidt understood moment arms in 1908 without any vocabulary for them:

The higher the bridge, the easier will the press be found.

George Hackenschmidt, The Way to Live, 1908, p. 60.

He is exactly right, and the reason is moment arms. In a bridge, the load is body weight acting through the cervical spine. Raising the hips brings the body’s centre of mass closer to being directly over the head, shortening the horizontal distance between the load line and the neck — which reduces the extension moment the cervical musculature must resist.

Lower the hips and that distance lengthens. Same body weight, more torque.

This is the mechanism behind every bridge regression on the bridging page, and it was observed empirically by a wrestler writing for wrestlers, 118 years ago.

Isometric versus dynamic

Isometric — force produced, no joint movement. Torque is whatever the opposing force generates, at one fixed angle.

The advantage is control: no range of motion means no travel toward the angular thresholds where joint reaction forces climb. The limitation is specificity — strength gains transfer best to positions near the trained angle, with progressively less carryover further away. Training at one angle trains one angle.

Dynamic — load moves through a range. Trains across angles, at the cost of passing through positions where the resistance profile and the joint’s tolerance may not agree.

The practical resolution used by most published protocols is to do both: isometrics for the directions and populations where control matters most (the rugby protocol used weighted isometric flexion while training extension and lateral flexion dynamically), and dynamic work where range coverage matters.

Why “how much weight” is a poor question

The number on the plate tells you almost nothing about what the neck experienced without knowing the geometry that delivered it.

Ten kilograms on a harness anchored to produce a short moment arm may be less demanding than three kilograms mounted on the head at maximum tilt. Two setups with identical loads and different anchor heights train different things. And the same weight moved quickly generates more force than the same weight moved slowly, because force is mass times acceleration and momentum is not free.

The better questions: where is the load line, how long is the moment arm, and where in the range does the resistance peak. Someone who can answer those three about their setup understands their training. Someone who only knows the number on the plate does not.

What is not known

Load distribution across cervical levels during common exercises is poorly characterised. Which segments take what share during a harness extension, whether one device produces a more even distribution than another, and whether that would matter — these have largely not been studied.

Where this site reasons about such things, it is reasoning from mechanics rather than reporting measurement, and it says so. The measurement that does exist comes from crash research, and it describes failure rather than training. See injury tolerance.