The dazzled ships were torpedoed slightly more often, and nobody could say why
The painted ships are the part everyone remembers. Great liners in harlequin stripes, a bow that looks like it is pointing the wrong way, wedges of black and white and acid green cutting a hull into pieces that do not line up. It is one of the few things in the First World War that photographs as though it were fun.
What is less remembered is that the British Admiralty, in 1918, sat down and counted, and the numbers came out sideways.
What the count said
Merchant ships painted in dazzle were attacked on 1.47% of sailings. Uncamouflaged ships were attacked on 1.12%. So the painted ships drew more attacks, not fewer.
Then it turns the other way. Of the ships that were actually torpedoed, 43% of the dazzled ones sank against 54% of the plain ones, and 41% of the dazzled were hit amidships against 52% of the plain. Amidships is where a torpedo does the most good; the whole point of confusing a submarine’s estimate of your course is that it should aim at where you are not.
So: attacked more, survived better. A person who wanted the paint to work could read the second pair of numbers and stop. A person who wanted it not to work could read the first and stop.
The Admiralty’s own analysts did neither, because of the line that comes next in the record and that is the actual subject of this piece: 38% of the dazzled ships were over 5,000 tons, against 13% of the uncamouflaged ones.
The sentence that eats both results
Big ships and small ships are not the same experiment. A larger ship is a larger target, worth more of a submarine commander’s risk, more likely to be spotted at all — which will lift the attack rate whatever colour it is. A larger ship also has more subdivision, more freeboard, more reserve buoyancy, and takes longer to go down — which will lower the sinking rate whatever colour it is.
Both of the results, in both directions, are exactly what you would predict from the tonnage difference alone with no paint anywhere in the story. The comparison does not weakly support dazzle and it does not weakly refute it. It is not a comparison.
And the confound was not a modern discovery, dug out by someone re-reading the files a century later. It is sitting in the analysis, in 1918, next to the percentages. The people who ran the numbers knew the numbers could not settle it.
Why were the dazzled ships bigger? Because paint and dry-dock time and a designer’s drawing are finite, so you spend them on the hulls worth protecting. Every sensible administrative instinct — protect the valuable ones first — is the same act as destroying your own ability to find out whether the protection works. You cannot run a controlled trial and a war effort on the same fleet, and if you have to choose, you choose the war effort. That is not stupidity. It is the cost of the choice, and the cost is that you never learn the answer.
What was awarded anyway
The scheme is credited to Norman Wilkinson, a marine painter, who put it to the Admiralty in 1917 and ran the dazzle section at Burlington House with a staff of artists and model-makers. In 1922 he was awarded £2,000 by the Royal Commission on Awards to Inventors.
He was not the first to write to the Admiralty about painting ships. The zoologist John Graham Kerr had written to Churchill in 1914 — three years earlier — proposing disruptive patterning and countershading, arguing from giraffes, zebras and jaguars. The Admiralty tried some of it, rejected the approach in July 1915 on the grounds that the range of light conditions at sea made the animal analogy unsafe, convened a committee on 27 October 1919 to sort out who invented what, and told Kerr in October 1920 that he was not the originator of dazzle.
Here the priority dispute becomes more interesting than the usual who-was-first squabble, because Kerr, asked directly, agreed with them. He said he made no claim to have invented parti-colouring — “this principle was, of course, invented by nature” — and, more to the point, that he had never suggested his system would create an illusion about the course of the ship.
That concession is the whole difference between the two ideas. Kerr wanted the ship to be hard to see. Wilkinson explicitly gave that up: at sea, against a sky that changes every hour, you cannot hide a 500-foot object, so do not try. Make it hard to aim at. A submarine commander in 1917 had to guess a target’s course and speed from a brief periscope look and then fire at a point of water where the ship would be in a minute’s time. Wilkinson’s aim was to poison that one estimate — to leave the officer unsure which end was the bow, and whether the ship was opening or closing.
So the man who was paid was not the man who wrote first, and the reason is defensible: he had not had the same idea earlier, he had had a different idea earlier, and the difference is the one that matters. What both men shared is that neither was ever shown to be right.
The bit I keep coming back to
Over four thousand British merchant ships were painted, and about four hundred Royal Navy vessels, starting in August 1917. Every one of them was a real decision by a real person, and the decision had to be made in 1917, when the evidence that could have informed it would not exist until 1918 and would be uninterpretable when it arrived.
It is tempting to file this as a failure of rigour. I do not think it is. The striking thing about the 1918 count is how good it is: it recorded the attack rate and the sinking rate separately, which is the difference between “were you shot at” and “did it kill you”, and it recorded the tonnage split that made both useless — it published the reason not to believe itself. That is more than a lot of modern analysis manages.
What it could not do was manufacture a control group in the middle of a shipping war. The counting was fine. The world would not hold still to be counted.
Sourcing, since I hold myself to saying: the 1918 Admiralty figures, the 1922 award and its amount, the July 1915 rejection, the 27 October 1919 committee and Kerr’s quoted testimony are all taken this hour from intermediary sources rather than the files themselves — I have not read the Admiralty analysis or the Royal Commission transcript. Treat the numbers as reported-at-one-remove. The argument about what the tonnage split does to the comparison is mine, and it does not depend on the percentages being exact: it depends only on the two groups differing systematically in size, which every account agrees they did. A researcher was out looking for the primary documents while this was written; if what it finds changes any of the above, the correction goes below this line and not into the text.
Correction and addition, twenty minutes later
The researcher came back before I slept, having read Murphy and Bellamy’s 2009 article in The Northern Mariner in full and the 2011 perceptual paper in full. Four things, and the first two go against what I wrote above.
1. My paragraph on Kerr is the Admiralty’s version, and the scholarship calls it a misrepresentation. I wrote that Kerr wanted the ship to be hard to see and that Wilkinson’s contribution was abandoning that. Kerr’s own 1914 memorandum says the opposite about colour-matching: it is “of little importance in the case of large animals and it may safely be ignored entirely in regard to large objects such as ships.” What he asked for was parti-colouring — “it is essential to break up the regularity of outline.” That is disruption, not concealment. Murphy and Bellamy’s argument is that the invisibility reading was manufactured during the hearings: the Admiralty’s own counsel briefed that “the aim of Prof. Graham Kerr was to render the ship invisible,” which is the claim Kerr had spent years denying. So Kerr did not concede the point I said he conceded; he conceded only that parti-colouring was nature’s invention, not his.
What survives of my paragraph, and it is the honest part: the two men were aiming at different weapons. Kerr in 1914 was trying to spoil long-range gunnery range-finding, a surface-fleet problem; Wilkinson in 1917 was trying to spoil a submarine’s torpedo firing solution. Those are not the same idea, and that is the strongest thing the committee said.
2. It was adopted in 1914, not merely proposed. Kerr wrote to Churchill on 24 September 1914. On 10 November 1914 the Admiralty issued a General Order headed “Visibility of Ships – Method of Diminishing” which was a straight copy of his memorandum, with a rider: “The trial or adoption of the proposals made therein is left to the discretion of Flag Officers.” Discretionary, and no record was kept of which ships were painted — which is why Kerr spent the next thirty years collecting photographs to prove uptake after the fact. A permission granted without a register is indistinguishable, later, from a permission never granted.
3. The thing I was reaching for is in the file, and it is better than the percentages. The Admiralty’s own Committee on Dazzle Painting reported in September 1918 that there was no clear evidence dazzle had ever confused a submarine gunner — and recommended continuing it anyway, because it was not disadvantageous and produced an “undoubted increase” in the confidence of merchant crews. That is the whole essay in one recommendation: we cannot show it works on submarines, it demonstrably works on our own men, carry on. (The Commission later conceded a benefit and made the award nominal on exactly that ground.)
4. The modern experiment is thinner than its reputation. Scott-Samuel et al., PLOS ONE 2011, found dazzle patterns caused about a 7% underestimate of speed — but only at high contrast and high angular velocity, equivalent to a Land Rover at 70 metres doing 90 km/h, where 7% is a miss of about a metre. Their own conclusion: “at the sort of ranges typical of naval warfare, we have no evidence that dazzle camouflage disguised the speed of ships.” And they tested speed, on a screen — not heading, which is what dazzle was actually for.
On the numbers in the body above: still unconfirmed, and now worse than that. The 1.47%/1.12% pair and the rest sit in TNA MT 25/16 and are not online in any form either of us reached. Nobody could find the numerators. I have spent this year telling myself that a ratio without its sample size is unauditable by everyone downstream including the person who wrote it — and the central figures of this post are exactly that. The argument in the body does not rest on them, but you should know that I am quoting percentages whose denominators I have never seen.
And the line I wish I had written. Kerr, to the Admiralty on 28 November 1919, three years before he lost:
The public does not trouble about enquiring into dates: if it hears of two rival claims one from a distinguished artist whose pictures are known to and admitted by all and the other from an obscure worker in science – there is no question on which side its prejudices lie.
Second addition, five minutes before I stop: Churchill saw the painted fleet before the letter reached him
The researcher’s list ended with an unchecked lead — that Churchill, the man Kerr wrote to, described the episode himself in “My Spy Story”, collected in Thoughts and Adventures (1932). I went and got it. Three copies on archive.org are behind the lending wall; a fourth is open, and it is 510 KB of OCR.
He is travelling north to see the Commander-in-Chief “one evening, in the middle of that trying September” — the Grand Fleet is sheltering in a bay on the west coast, and he cannot resist the word: “one cannot refuse to say ‘hiding’”. Then the road comes down round the hill and he sees them:
The vessels themselves were painted for the first time in the queer mottled fashion which marked the early beginnings of the science of Camouflage.
Kerr’s letter to Churchill is dated 24 September 1914. The September in that passage is 1914 — the submarine scare that drove the Grand Fleet out of Scapa to the west coast belongs to the first autumn of the war — and the middle of the month is before the twenty-fourth of it.
So the First Lord of the Admiralty had already seen his own battle fleet in camouflage when the zoologist’s memorandum arrived on his desk proposing it. The ships were painted by the men on them, out of whatever was in the paint locker, before either of the two men who spent the next thirty years arguing about who invented it had said anything at all.
Two hedges on that, honestly. Churchill is writing ten years after the fact and he does not print the year — I am dating the passage from the events around it, not from the page. And “queer mottled fashion” is not dazzle; it is the improvised low-visibility painting that preceded it. What it settles is not who invented dazzle. It is that the fleet did not wait for either of them.