The paint that gave up on hiding

A ship at sea cannot be hidden. There is no background to disappear into: the hull is a hard vertical edge against a horizon, lit from one side, trailing smoke and a wake. Every serious attempt to make one invisible in the First World War ran into the same wall — a paint scheme that works at dawn is a silhouette at noon, and one that works against grey water is a target against a bright sky.

Dazzle starts by conceding that. The marine painter Norman Wilkinson proposed it to the Admiralty in 1917, and the proposal was, roughly: stop trying. Paint the ship in enormous interrupted stripes and wedges of contrasting colour — black, white, blue, green — sweeping across the hull at angles that have nothing to do with the shape underneath. In some light and at close range the ship ends up easier to spot, not harder. That was accepted as a price, not sought — the design gave up on hiding and went after something else. (Note the date: this is 1917, applied to naval ships from that August. Dazzle is the last eighteen months of a fifty-two-month war, not its whole character.)

The thing it was actually aimed at

The point was a piece of arithmetic happening in someone else’s head.

A submarine of that era did not shoot at where a ship was. Torpedoes ran straight and did not home on anything, so the commander had to shoot at where the ship was going to be — which meant estimating, through a periscope, in a glance short enough not to be spotted, three things: the target’s range, its speed, and above all its course, the angle at which it was crossing his line of sight. Get the angle wrong by enough and the torpedo passes behind the stern or ahead of the bow.

Almost all of the visual evidence for course comes from the ship’s own geometry: where the bow is relative to the funnels, how foreshortened the hull looks, which way the superstructure leans away from you. Dazzle attacks exactly that evidence. False bow waves painted on the hull. Stripes that carry the eye off at an angle the ship is not travelling. Panels that break the hull into what looks like two smaller vessels, or that make the stern read as a bow.

It is not camouflage. It is closer to a lie told in the specific dialect the person on the other end is trying to read. The ship is perfectly visible; what is hidden is one number.

Nobody could prove it worked

Here is the part I find hardest to let go of. The programme was enormous — over 4,000 British merchant ships and some 400 naval vessels, a figure that traces to Murphy and Bellamy’s 2009 paper in The Northern Mariner, which is the one scholarly link in a chain where everything else quotes Wikipedia quoting them. A fact-check I ran on this piece could not open that PDF, so take the number as attributed, not as verified. And after all that, the wartime statistics could not say whether it worked.

They could not say because everything changed at once. Convoying came in around the same time and was, on any reading, the enormous variable. Ships zigzagged. Escorts multiplied. Dazzled ships were disproportionately the ones sailing independently, which is to say the ones already in more danger. Losses fell, but a fall in losses is a fact about the whole system, and the paint is one term in it that no one had isolated. Analyses run during and after the war came out mixed and are still argued over.

So the interesting move is what they did instead. Rather than chase the outcome through the noise, the assessors tested the estimate: scale models on a turntable, viewed through a real periscope by observers who were asked to call the course, and then scored against the truth. That measures the one thing dazzle claims to do, in isolation, with a control — and it is available on a Tuesday afternoon in a room in London, which sinking statistics are not.

And then the honest ending, which I did not have when I started writing and which a fact-check handed me. A 2024 reanalysis by Meese and Strong at Aston went back to the primary observer data from a 1919 American study of exactly this kind and found that most of the misjudgement was not the paint. Observers pull a ship’s apparent heading toward the horizon whether it is dazzled or not — a “horizon effect” worth something like 19–23 degrees, against roughly 7 degrees attributable to dazzle itself. The lab that was measuring the mechanism was substantially measuring itself, and the people who read the result were camouflage experts.

So keep the idea, but keep the correction welded to it. When the outcome you care about is buried under confounders you will never untangle, you can often still measure the mechanism you claimed. If your defence works by making an attacker guess wrong, go and measure the guesses. A weaker measurement of the actual claim beats a stronger measurement of something adjacent that the world will not let you attribute.

There is a coda about credit. The zoologist John Graham Kerr had written to Winston Churchill, then First Lord of the Admiralty, back in 1914, arguing from animal coloration — and what he proposed was not dazzle. Disruptive patterning and countershading are techniques for not being seen; Wilkinson’s case was about not being aimed at. They ended up in a priority dispute that took two separate rulings: the Admiralty rejected Kerr’s claim in October 1920, and the Royal Commission on Awards to Inventors gave Wilkinson £2,000 in 1922 and the credit with it. Two bodies, two years, one distinction doing all the work — the difference between hiding a thing and corrupting the measurement of it.