Intelligence
Dazzle Camouflage: The Pattern No One Could Prove Worked
Dazzle camouflage is the only major camouflage system in history that was never asked to hide anything. Between 1917 and 1918 the British and American merchant and naval fleets painted more than five thousand hulls in hard-edged geometry — wedges, false curves, converging stripes — with the explicit aim of being seen and misread. It is also, a century on, the camouflage with the weakest evidence base of any that was ever adopted at scale. That combination is what makes it worth looking at closely: it is a design problem solved by artists, defended by anecdote, and only recently subjected to the kind of measurement that might have settled it in 1918.
USS West Mahomet in port, c. November 1918. The wedges are cut across the bow deliberately: the archive's own caption notes how far the scheme distorts the apparent aspect of the bow. Bureau of Ships collection, US National Archives, 19-N-1733 — public domain.
A zoologist, a marine painter, and a man on the Mersey ferry
The idea did not begin with Norman Wilkinson. In 1914 the Glasgow zoologist John Graham Kerr wrote to Winston Churchill proposing what he called parti-colouring: countershading after Abbott Thayer, plus disruptive coloration borrowed straight from the giraffe, the zebra and the jaguar. A general order went to the fleet on 10 November 1914 and ships including HMS Implacable were painted to it. After Churchill left the Admiralty the Navy reverted to plain grey, telling Kerr in July 1915 that the range of light conditions at sea made any theory built on animal analogy unreliable — rejecting, in one sentence, the whole lineage running from animal coloration through the stalking hunter to every camouflage scheme that followed.
Wilkinson — a marine painter, Illustrated London News illustrator and poster designer, then a lieutenant in the Royal Naval Volunteer Reserve — arrived at his version in April 1917, on a train, by inverting the problem. If a ship could not be made invisible to a periscope, the opposite extreme was the answer: paint her so her form breaks and a submarine officer cannot read her course. A stores vessel, Industry, was painted to his design that May; the Director of Naval Equipment, Captain Clement Greatorex, supplied the name.
The priority fight ran for years and was never settled on the merits. Hugh Murphy and Martin Bellamy's reassessment of Kerr in The Northern Mariner makes the uncomfortable case plainly: Kerr's scheme was the first based on stated principles to receive a practical trial, and it failed in application rather than in theory, largely because nobody was ever given responsibility for applying it properly. Wilkinson had what Kerr did not — Admiralty contacts, a plausible manner, and the willingness to go around the Admiralty to the Minister of Shipping when it stalled him.
A third claimant, a Liverpool clerk named Archibald Phillips, had submitted a dazzle-like proposal in May 1915 after watching ships cross the Mersey on his commute. The Liverpool Biennial's account of Phillips and the 1922 Committee of Enquiry is a useful corrective to the single-inventor story: the hearing found for Wilkinson, who received £2,000, and the chairman dismissed Phillips outright. What the record actually shows is that the core idea was in the air from 1914 onwards and that the award went to the man best positioned to execute it.
Three claimants, one award. Kerr's scheme was the first with stated principles to get a practical trial; Wilkinson got the £2,000.
How a dazzle plan was actually made
The production method is the most under-examined part of dazzle and the most interesting. Refused facilities by the Admiralty, Wilkinson was given studio space at the Royal Academy Schools in Burlington House — the male students had been conscripted, so the rooms were empty and the remaining student body was almost entirely women. The Academy's own account of the dazzle section at Burlington House notes that by September 1917 the Admiralty was using five of the Schools.
An Admiralty memo puts the initial team at five artists commissioned as RNVR lieutenants, three modellers, and eleven women colouring plans; Wilkinson later recalled around twenty women on staff, all recruited from art schools rather than clerical pools. One of them, Eva Mackenzie of Edinburgh School of Art, later married him. The Liverpool Biennial's reconstruction of the section's working method is the fullest published account of who did what.
The test rig was a small theatre in Wilkinson's office. A scale wooden model was cut for each of the thirty-seven merchant vessel types, painted with a candidate scheme, set on a turntable and viewed through a submarine periscope against interchangeable sky backgrounds. Designs were selected for maximum distortion of apparent heading, then transferred to hand-coloured lithograph plans — port and starboard drawn separately, because the two sides were never the same — and issued to dock officers at the ports, Edward Wadsworth among them.
The Americans built the same idea larger and blunter. Their experimental station put the model on an adjustable pillar against a screen of water, and the archive card describing it spells out the test that followed: the observer sights through the periscope and fires a small gun at the model. If he misses, the scheme passes. If he hits, they try another one. That is the whole protocol — and it is the most rigorous thing anyone ran before painting several thousand ships.
What the testing actually taught them
Three findings from that periscope theatre are worth holding onto, because they are design constraints derived from observation rather than from theory. The bridge area mattered most, since that is where a submariner looks to read course. Sloping lines and curves outperformed verticals. And the palette that worked was not the circus one dazzle is remembered for: black and white with blues and greens beat violently contrasting colour, partly because German U-boats fitted colour screens to their periscopes that flattened bright hues.
Each hue was keyed to an official colour chart, and the plans say so on their face. The Ministry of Shipping's Type 11, Design B sheet — starboard copy, drawn at an eighth of an inch to the foot for a hull of 527 feet between perpendiculars, signed by Wilkinson himself — carries a note in the corner warning that the colours printed on the plan are there for measurement only, and that the actual colours follow the chart. Its key names five: two greys, a green, a mauve, and black. Read the lithographs as specification documents rather than artworks; the colours on them are labels, not swatches.
The visual logic: false bow waves and borrowed perspective
Strip the decoration away and dazzle is doing three separable jobs, none of them the background matching and silhouette blending most camouflage attempts. The first is bow–stern ambiguity: shapes that carry across the ends of the hull so the observer cannot tell which end is which, or where the vessel actually stops. Some hulls were built to help — the Royal Navy's 24-class sloops were deliberately given a bow-like stern and a single central mast.
One hull, drawn three times, each carrying only the marks that do its own job. No issued scheme was ever this tidy — every real plan ran all three at once, over the whole hull, and differently on each side. Tap to enlarge.
The second is the false bow wave. A pale curve painted low on the hull reads as the wave a fast ship throws, implying both a speed and a heading the ship does not have. The third, and the one the Americans pushed hardest, is forced perspective: painted lines and texture gradients that converge as if receding, contradicting the real three-dimensional geometry with a pictorial one. American camoufleurs called it reverse perspective. Loyd A. Jones at Eastman Kodak, working on the US programme, concluded that distortion of perspective was the only element of dazzle worth anything.
All three depend on a low, close, monocular viewpoint. Dazzle was tuned to a periscope at under a thousand metres and to a shooter with perhaps thirty seconds to fix a firing solution before his own wake gave him away. Wilkinson's own claim was that an error of eight to ten degrees in the estimated course was enough to make a torpedo miss. That is the number the whole system was built to produce, and it is the number nobody managed to verify at sea.
Dazzle was designed for one observer in one situation: submerged, monocular, close, and short of time.
9° AT 1,000 M ≈ 160 M — more than a hull length.
Where to actually look at the plans
The surviving visual record is better than most camouflage records and worse than it looks. The largest accessible set of American plans is the Dazzle Print Collection at the Fleet Library at RISD: 455 lithograph plans covering port and starboard for ten merchant ship types, plus twenty photographs, donated by Maurice L. Freedman, district camoufleur for the 4th district of the US Shipping Board's Emergency Fleet Corporation. Freedman's job was the unglamorous half of the system — take the plans, hire painters, adjust the design where the hull refused to cooperate. The collection's own notes date the first US ship launched already camouflaged, the Everglade, to 29 July 1918.
Roughly 495 American designs were approved and printed in multiples; incomplete sets survive at the National Archives and at RISD, and the British preserved their plans and wooden models rather better than the Americans did, as Roy Behrens sets out in his survey of surviving dazzle plans and models.
The most instructive thing you can do with these documents is compare a plan against a photograph of the ship it specified. Behrens did exactly that for the SS Matilda Weems, matching RISD's plans to 1918 photographs of both sides, and the painted hull departs from the drawing — noticeably on the port side. Ship painters were working at scale, at speed, on curved plate, from a flat drawing. Treat any dazzle plan as an intention, not a record. The collections worth starting from are gathered on our CAMO RESOURCES page.
The pictures that are not evidence
Edward Wadsworth supervised dazzle application at Bristol and then Liverpool and cut nine woodcuts of dazzled ships. One became Dazzle-ships in Drydock at Liverpool (1919), now at the National Gallery of Canada — three metres tall, commissioned by the Canadian War Memorials Fund, and the last serious Vorticist painting anyone made.
It is also not documentation. The pattern on those bows is Wadsworth's invention, closer to his 1915 abstractions than to anything issued from Burlington House. The same caution applies to the period's other famous dazzle images: contemporary photography was monochrome, which is why dazzle is remembered as black-and-white graphic design when the working palette leaned blue-green. Picasso is supposed to have claimed Cubism invented it. The archive does not support that either.
The evidence problem
The Admiralty analysed shipping losses in 1918 and could not draw a conclusion. Dazzled ships were attacked on 1.47% of sailings against 1.12% for uncamouflaged ones — more conspicuous, which Wilkinson had never disputed was the point. Of ships actually struck by torpedo, 43% of dazzled ships sank against 54% of plain ones, and 41% were hit amidships against 52%. That looks like a result until you notice the confound: 38% of the dazzled ships were over 5,000 tons, against 13% of the uncamouflaged. The comparison is between different fleets, not different paint.
Quarter by quarter the numbers simply swap sides. In the first quarter of 1918, 72% of attacked dazzled ships were sunk or damaged against 62% of non-dazzled; in the second quarter it inverted to 60% against 68%. The Committee of Enquiry on Dazzle Painting, sitting in September 1918, found no case in which dazzle demonstrably changed the outcome of a submerged torpedo attack — while noting that crews on dazzled ships were visibly more confident, and that under some conditions in convoy the paint might increase risk.
The American figures are quoted more often and prove less. Harold Van Buskirk reported 1,256 US ships dazzled between March and November 1918, with 18 camouflaged merchantmen of 2,500 tons and over sunk against 78 uncamouflaged. Without the number of uncamouflaged ships sailing there is no rate to compare, so the celebrated "under 1% lost" figure is a numerator in search of a denominator. Meanwhile the convoy and escort system arrived in mid-1917, alongside zigzag manoeuvring and hydrophones. Everything changed at once, deliberately, in wartime. There was never a control group.
Two counts appear to favour dazzle. The fleets being compared are not comparable, and quarter by quarter the sign flips. Tap to enlarge.
What measurement eventually showed
The first quantitative test was a 1919 MIT bachelor's thesis. Leo Blodgett used the Boston Camouflage District's model theatre to run twelve designs past three sets of observers and recorded course-estimation errors averaging twenty to thirty degrees, with the best designs off by nearly sixty. Against Wilkinson's eight-to-ten-degree threshold that reads as vindication, and it was cited as such for a century.
It does not survive re-analysis. A 2025 reworking of Blodgett's data separates two effects that he had folded together: twist, a systematic bias in perceived heading that the camouflage actually causes, and hysteresis, a general pull of judgements toward the beam that occurs regardless of paint. Blodgett's real twist was about seven degrees. Most of his dramatic error was an artefact of how people judge ship headings at all.
The 2024 psychophysical study by Lovell, Sharman and Meese, "Dazzle camouflage: benefits and problems revealed", reaches the same magnitude from the other direction and adds a sting. Wilkinson's own scheme for RMS Mauretania produced roughly one degree more twist than neutral grey. The design that produced the largest twist was not a historical dazzle pattern at all but a regular circles stimulus — a pure texture gradient, forced perspective with the decoration removed. Worse, dazzle patterns reduced the protective hysteresis that plain grey enjoys, so twist and hysteresis tended to cancel. Their overall verdict is that most WWI dazzle would have delivered little or no net benefit — with a genuine exception for long, fast ships viewed at distance by experienced observers, where a strong twist could plausibly have cost a submariner the shot.
The perceptual claim that does hold up is a narrower one. Scott-Samuel and colleagues showed in 2011 that dazzle-type patterns distort perceived speed — but only two-dimensional ones, zigzags and checks, only at full contrast, and only at high angular velocity, where they slowed apparent speed by about 7%. The authors are explicit that this gives no evidence for any speed effect at naval ranges. Their live application is a vehicle at seventy metres against a rocket-propelled grenade, where a 7% error puts the round about a metre behind the target. An earlier target-capture study found no systematic advantage for dazzle over plain colouring at all.
What dazzle became instead
Five months after the Armistice, on 12 March 1919, the Chelsea Arts Club filled the Royal Albert Hall for its Dazzle Ball, decorated by Wilkinson himself with Cecil King, Burnell Poole and Walter Webster, and attended in disruptive costume. Copycat balls followed in Washington and in Sydney that October. The pattern outlived its function almost immediately and has been recycled ever since — Carlos Cruz-Diez repainted the pilot ship Edmund Gardner in Liverpool in 2014 for 14-18 NOW, and HMS President (1918), an actual WWI dazzle ship, was dazzled again in London.
The sharpest descendant is Adam Harvey's CV Dazzle, begun in 2010, which used hair and makeup to break the Viola–Jones face detection cascade — not recognition, a distinction Harvey has spent fifteen years correcting. What makes it the honest heir to Wilkinson is that Harvey has publicly retired his own early looks: once convolutional networks replaced haarcascades around 2016, the patterns stopped working and he said so on the project page. That is the discipline the 1918 Admiralty never managed.
The one measured effect that holds up is narrow, and nothing about it is naval. The pattern outlived its function within months of the war ending.
Dazzle earned its place in design history legitimately — as a rare case of a state handing a perceptual problem to artists and getting a coherent visual system back, produced at industrial scale by a workforce of women art students whose names mostly went unrecorded. What it never earned is the claim usually attached to it. Ninety years later Optifade would run the argument in the correct order — measure the observer's vision first, derive the pattern from it, publish the reasoning — which is exactly the discipline dazzle was never held to. The honest position after a hundred years of argument is that a handful of dazzle schemes, on the right hulls, at the right distance, probably did shift a submariner's estimate of course by a few degrees — and that the rest of the fleet was painted on faith. Look at the plans as what they are: the most ambitious speculative design programme of the war, and one of the least tested.
