In 1995 a trade association and a New Hampshire fabric mill started buying women's blazers off the rack, one at a time, the way any other customer would.
The blazers were made by Harve Benard, Ltd., and the labels said 70 percent wool, 20 percent nylon, 10 percent cashmere. The buyers were the Cashmere & Camel Hair Manufacturers Institute and L.W. Packard & Co. They were not shopping. They were collecting samples, and they handed what they bought to two scientists — Professor Kenneth Langley and Dr Franz-Josef Wortmann, both retained by them — who tested the garments separately.
All of that is recorded in a published opinion of the United States Court of Appeals for the First Circuit, which in April 2002 reversed a partial summary judgment that had been entered against the two buyers and sent the case back for further proceedings.1 A reversal is not a finding of liability, and no reported decision from after the case returned to the district court appears in the public record. A suit filed in 1996, decided on appeal six years later, and no reported ending after that: it is a fair introduction to how much of this is visible from outside.
Set the outcome aside. Look at who went looking.
No regulator bought those blazers. A trade body did — one whose members' revenue depends on cashmere commanding a price that ordinary wool does not. It paid retail for garments it did not want, and it paid two experts to take them apart. That is an expensive way to find out what is in a jacket, and somebody was willing to pay it because there was a price gap at the end of the question large enough to justify the bill.
Nobody was ever required to do it. That has not changed since.
Eight countries, 132 garments, one Italian laboratory
In June 2026 the European Commission published the results of a joint market-surveillance campaign run across eight member states.2 Before the figure, the thing that governs it. EU law requires market-surveillance authorities to choose what they test on a risk basis, weighing a product's known record of non-compliance and any complaints received.3 The programme's own published account of its method describes samples collected by national authorities against plans drafted by technical experts and constrained by the available budget.4 So what follows is a hit rate among garments picked because somebody expected them to fail, not a rate for the clothes on the rail.
Of 132 items tested at an accredited laboratory in Italy, 49 — 37 per cent — did not correctly reflect the materials used.
The Commission describes three kinds of failure: the fibres correctly named but the stated percentages wrong; a different and cheaper fibre present than the one declared; or the fibre declared or named incorrectly. It has not published how the 49 divide across those three, so nobody outside the laboratory can say what share were gross substitutions and what share were a blend running four points off. It has also not published the analytical uncertainty of the testing, or how the Regulation's 3 per cent manufacturing tolerance was applied.5 Two of those three categories cannot be artefacts of a marginal measurement: a cheaper fibre is either present or it is not, and a wrong fibre name is not a rounding error. At least some of the 49 are not close calls. How many, nobody outside can say.
Enforcement followed, in overlapping actions rather than a tally: sales of 18 products were stopped, two manufacturers were told to take corrective measures, two products were relabelled, and 41 entries went into the EU's market-surveillance information system.2 Against the other categories in the same campaign cycle, textiles did comparatively well — gadgets tested for hazardous substances failed at 49.7 per cent, and radio-controlled toys, checked against the radio-equipment rules, at 60.2 per cent.6 This is not a story about textiles being uniquely bad, and the eight authorities did what the law asks of them.
It is a story about what that number leaves out. Because there is a second gap in the same garment. It is also large, it is entirely lawful, and there is no percentage for it anywhere.
What the wool market published that week
Every week the United States Department of Agriculture publishes a wool price table. Here is the one for the week ending 17 July 2026, quoting Australian wool in US dollars per pound delivered to Charleston, South Carolina.7 These are raw-fibre input prices — what a spinner pays for material, not what anyone pays for a jumper. And the spread across grades is not a spread between good wool and bad: thirty-micron wool is coarser, not defective, and goes into carpets, upholstery and heavy outerwear, where fineness is not what the buyer wants.
THE LADDER — Australian wool, one trading week, by grade
17 micron ....... USD 8.35 / lb
18 .............. 8.10
19 .............. 7.45
20 .............. 6.96
21 .............. 6.85
22 .............. 6.77
23 .............. 6.56
25 .............. 4.29
26 .............. 3.80
28 .............. 2.82
30 .............. 2.40
32 .............. 2.15
>>> Every one of these is "100% wool" on a label. <<<
Twelve grades. The price falls at every step, from USD 8.35 a pound at the top to USD 2.15 at the bottom — near enough four times. The label has one word for all twelve.
Now set that ladder beside the substitution the Commission was actually testing for. Using the 17 July 2026 USDA wool report and the 16 July 2026 Cotlook A Index, 21-micron wool from the table, at 6.85 a pound, ran about seven and a half times the price of cotton at 90.65 US cents a pound.7,8 The two are not a like-for-like comparison: the cotton figure is an index of cotton quotations, while the wool figure is Australian wool delivered to a US port. The gap is therefore an order of magnitude rather than a measurement.
Put that on a garment. A 400-gram jumper — an assumed weight, not a measured average — carries about £3.89 more in raw fibre value as wool than as cotton.9
Here is the shape of it. One of those gaps is unlawful if the tag says wool. It is detectable, it sits in a dataset, and eight member states tested for it. The other breaks no rule at all, and there is no dataset, because there is nothing to look for. A garment made from the bottom of that ladder and a garment made from the top carry the same four characters on the tag.
The 37 per cent is real, but it measures only the label failures the campaign tested. It is the only part of the problem that has a number. Everything true about the missing measurement is equally true of every garment in that sample that passed.
There is a fair objection here, and it is the one a careful reader forms unaided. Is the composition label even supposed to convey grade? A nutrition panel does not rate flavour. A hallmark says nothing about whether the ring is well made. Every disclosure has a scope, and criticising an instrument for failing at a job it never had is a category error rather than a finding.
That objection is right about the instrument and beside the point about the reader. The Regulation is a naming instrument. Its work is to make "wool" mean the same thing in Lisbon and in Vilnius, and it does that work. The gap sits somewhere else: between a disclosure's designed scope and its received meaning, with nothing about the composition line's appearance marking where that scope ends.
Part of the space does get filled, voluntarily — and the clearest instance is the price table above. Twelve grades, twelve prices, published weekly and free by a government agency, entirely outside the composition regime. Grade information turns up where somebody profits by supplying it, and it carries no verification duty either.
One boundary needs stating plainly. The claim here is about the tag: it does not distinguish the two ends of that ladder, and a shopper cannot use it to tell which one she is holding. Whether retail prices track micron grade under a constant "100% wool" label is a different question, and this report did not investigate it and does not answer it.
"Weigh to an accuracy of 0,0002 g"
Regulation (EU) No 1007/2011 is not vague about measurement. It is extraordinarily precise about it.
The word "accuracy" appears eight times. Twice it is the duty: the manufacturer, or the importer where the manufacturer sits outside the Union, "shall ensure the supply of the label or marking and the accuracy of the information contained therein."5 The other six times it is laboratory metrology. Weigh to an accuracy of 0,0002 g — a ten-thousandth of a gram. Give the individual results and the arithmetic mean, each to an accuracy of 0,1 percentage point of total fibre mass. Report the result obtained, the method used and its degree of accuracy.
The instrument knows exactly how to find out what is in a garment, in forensic detail, across a long annex of dissolution methods.
And Article 19(1) attaches every word of that method to one thing: "the checks referred to in Article 18." Article 18 is a single sentence, and it is the whole of the article — market-surveillance authorities shall carry out checks on the conformity of the fibre composition of textile products with the supplied information. The only actor in the Regulation directed to check anything is the state, if and when it chooses.
The person who prints the number gets Article 15(1): ensure accuracy. No method, no evidential standard, no records. The shop that sells you the coat gets Article 15(3), which asks only that the product bear the appropriate labelling — a duty discharged by the tag existing.
Then there is Annex II. The phrase "technical file" appears five times in the Regulation, and all five are there, in the requirements for an application to add a new fibre name to the official list. The instrument's single use of the word "verifiable" is in the same provision: the characteristics in a proposed definition "shall be verifiable via testing methods to be provided with the technical file along with the experimental results of analyses."5 Annex II then lists what evidence a word must carry — a chemical formula, a melting point, a density, an infrared spectrum, and data on the accuracy and repeatability of the methods used to establish them.
The only thing in the entire Regulation that must be documented, evidenced and verified is a word. The vocabulary gets a technical file. The garment does not.
The Commission's own guidance to businesses on complying with this Regulation runs to some 47,000 characters and does not contain the word "accurate" at all.10
The paragraph that came from 1968
Public pricing exists. TÜBİTAK BUTAL's Materials-Textile Laboratory in Türkiye is accredited to TS EN ISO/IEC 17025 and publishes 2026 rates: 2,400 Turkish lira for identifying and quantifying a two-fibre composition under TS EN ISO 1833 methods, and 3,600 lira for three fibres.24
Another published public-institute tariff comes from Kanpur, effective October 2021.11 Identifying what a fibre is: 500 rupees. Determining the ratio in a two-fibre blend: 800 rupees at the industry rate, plus tax, on a minimum sample of 250 grams — less fibre than goes into the jumper above. An Indian public-institute rate is not a European commercial one; it is a scoped comparison, not a market rate. And the price is not what decides this in any case: no duty to test exists to be priced against.
Something cheaper than testing was on offer, and it was not a price. It was a defence.
The United Kingdom's implementing instrument gives a trader a complete answer to a prosecution on two conditions, and both must hold.12 First, that the offence was due to one of five named causes — a mistake, an accident, the act or default of another, some other cause beyond his control, or reliance on information supplied to him by another person. Second, that the trader took all reasonable precautions and exercised all due diligence. Both, not either. Nor is the first limb free: to rely on the supplier route a defendant must serve notice naming the person he relied on, at least seven clear days before the hearing. You may pass the claim along. You must be willing to say who handed it to you.
But nothing in it says that diligence includes measurement. That is the operative effect, checkable from the text: the cheapest lawful route to a complete statutory defence does not contain a test.
The mechanism sits in the list itself. A laboratory result is not one of the five named causes, so testing cannot by itself carry a trader through the first limb. Relying on what a supplier told you is one of them, printed in the instrument.
Those five causes are section 24 of the Trade Descriptions Act 1968. A legislation.gov.uk full-text search returns more than twenty instruments using the second limb's wording.13 The five-cause paragraph itself had already been used once before, in the Business Protection from Misleading Marketing Regulations 2008 — the same five causes, the same order, the same words.14 Only the cross-reference changes: regulation 6 in the Business Protection from Misleading Marketing Regulations becomes regulation 5 in the Textile Products (Labelling and Fibre Composition) Regulations.
The paragraph came out of the statute book, not out of any decision about cloth.
Brussels, 2014
There is a standard way the European Union makes a product-law promise checkable. It has a name — conformity assessment — and a toolkit of modules. The lightest of them, Module A, is the closest thing in that toolkit to what textiles has: pure self-declaration, the manufacturer declaring on his sole responsibility, with no third party involved anywhere. Even that one requires the declaration to rest on technical documentation that, wherever applicable, contains test reports, and requires the manufacturer to keep it available to national authorities for ten years.15
Textile fibre composition sits inside the same body of law as toys. The EU's market-surveillance regulation lists the Textile Regulation as item 40 of the 70 instruments it classifies as Union harmonisation legislation.3 The Textile Regulation's own opening words take Article 114 of the EU treaty as their basis — internal-market harmonisation, the same family as the toy directive.5,16 Recital 14 of the Textile Regulation adopts one half of the framework, the market-surveillance half, the half that says someone may check afterwards.
It adopts none of the other half. No conformity assessment, no technical documentation, no declaration of conformity, no notified body, no module, no retention duty. It received the self-declaration without the file.
That same market-surveillance regulation carries a second list: Union harmonisation legislation without provisions on penalties. Textiles occupies list position 17 in that second list of 19, compared with the seventy instruments in the first.3
There may be good reasons for this, and they belong on the page. The framework decision is a drafting template rather than a rulebook for manufacturers, and it permits legislation to depart from its principles where the specificities of a sector make that appropriate.15 Its modules engage only where legislation requires conformity assessment in the first place, and the choice among them is expressly calibrated to "the nature of the risks entailed by the product," with a direction to avoid disproportionate burden. Toys, low-voltage equipment and construction products can injure people. A mislabelled jumper costs money. A ten-year documentation duty on every garment placed on the EU market would be a large burden for a harm that is financial and reversible, and proportionality is a real principle rather than an excuse.
That is a fair answer. It is also an answer nobody has ever had to give, because the Regulation states no reason, and the question does not appear to have been asked.
It was not asked in 2014, when the Commission formally reviewed this Regulation. That report named the mechanism in its own text: "in the absence of sufficient checks and controls on the fibre composition of textile products, there is a risk of consumers buying products labelled as containing expensive fibres (e.g. cashmere, silk) which are actually made of less expensive fibres."17 A footnote recorded that in one member state about 35 per cent of tested products had incorrect fibre content — with no member state named, no sample size and no sampling method, which is why that figure cannot be set beside the 2026 result as any kind of trend. The report also identified the constraint honestly: market surveillance and testing activities "affected by limited availability of resources, both human and financial, resulting in a reduced number of samples that can be tested."
Then it concluded that "no major gaps, inconsistencies or administrative burden have been detected that would require amendment of the Regulation." Its three suggestions were a guidance document, a better process for handling new fibre-name applications, and a request for standardisation work. Its enforcement recommendation was that member states be encouraged to increase their participation in joint enforcement action, pooling resources and expertise.
And that recommendation was carried out. The eight-country exercise of 2026 is precisely what pooled joint enforcement looks like. This is the strongest thing that can be said for the regime, and it is true: the Commission diagnosed a resourcing problem, prescribed a resourcing remedy within its powers, and the remedy happened.
It also shows the limit exactly. Pooling raises the number of garments the state can test. It creates no duty on any manufacturer, importer or retailer to test anything. After 2026, as before it, nobody in the chain owes a measurement, and the population nobody tested is unchanged in kind. Changing that would have meant amending the Regulation, which the same document declined to do.
In June 1959 a rule took effect in the United States requiring every manufacturer of a textile fibre product to keep records showing the fibre percentages, and to preserve them for at least three years.18 It applies irrespective of whether any guarantee has been given or received. The rule states its own purpose: to establish a traceable line of continuity from raw material through processing to finished product. It requires nobody to test anything. What it requires is that the number rest on a documented basis somebody can later demand. That was fifty-two years before the European instrument was written, and the European instrument attaches no duty to document the basis of the declaration at all. Its only two uses of the words "record" and "retain" are bench instructions in a laboratory annex, telling a chemist to write down a titration result and to keep the fibres in the flask.
None of which would have made the labels true. That is the awkward and necessary fact, and it comes from the same campaign. Products carrying the full apparatus — conformity assessment, CE marking, notified bodies, a technical file retained for ten years, of the kind the Toy Safety Directive imposes16 — did not do better. Radio-controlled toys, checked against the radio-equipment rules, failed at 60.2 per cent; the gadgets tested for hazardous substances failed at 49.7 per cent.6 Both worse than textiles. In that campaign, conformity-assessment apparatus did not guarantee accuracy.
What it produces is something narrower and more useful. When a toy fails, there is a technical file, a declaration and a named person who signed it. When a composition line fails, there is a tag. The apparatus does not make the number true. It makes the number answerable — and the shopper is reading it as though it already were.
The people who needed it most stopped reading it
In 2018 a machine changed what people could see. When near-infrared sorting was introduced on the Dutch market, deviations between what the scanner read and what labels claimed became apparent to the sorters standing at the line. Not to regulators, not to shoppers. The Dutch Parliament asked the government to look into it, and the Ministry of Infrastructure & Water Management commissioned a study of 10,901 post-consumer garments.19
Of those, 7,749 still had a legible label. A further 295 were set aside because their stated percentages did not add to 100 within a 5 per cent margin — which the researchers attributed to possible data-entry error on their own side. That left 7,454 garments, sorted by near-infrared scanner rather than tested by wet chemistry, with a label counted as accurate when its stated composition fell inside the band the scanner assigned. The bands are wide: one of them runs from 45 to 94 per cent cotton. On that sample — Dutch, discarded, machine-read, and judged against those bands — 59 per cent of labels came out accurate, so 41 per cent did not.
Near-infrared analysis has its own error structure. The published calibration error for cotton content runs around 3.3 per cent, and rises when a garment contains fibre types the model was not trained on.20 That is a property of the technique, not a discount on the finding — and the study's own caveat points the other way, noting that because it sorted into bands rather than measuring exact percentages, the true deviations might be larger than the ones it recorded.
That 41 per cent also describes a different population from the Commission's, measured by a different method, at a different point in a garment's life. The two figures are not versions of each other and cannot be averaged. Worn clothes shed, too: cotton can lose between 1 and 20 per cent of its fibre over 50 washes, so a blend's ratio drifts physically through use, and a label can become wrong without anyone having said anything untrue. That confound applies here and not to the Commission's newly purchased garments.
This investigation continues below.
Want the next one in your inbox?
One finding inside the study settles the question of error direction that the rest of this report can only circle: do the label errors plausibly favour the seller, or run against the seller's own interest?
Inside one fraction of it — the garments the scanner read as roughly 65 per cent cotton — 31 per cent carried labels claiming 95 per cent cotton or more. The researchers said deviations of roughly 30 percentage points or more could not be explained by fibre loss alone, and judged intentional exaggeration plausible.
Set against the preceding 31 per cent that overstated cotton, 29 per cent in that same 65-per-cent-cotton fraction understated their own cotton content or overstated their polyester — errors running against the seller's own interest, for which the researchers judged intentional inaccuracy improbable. Both patterns sit in one dataset, at almost the same frequency. A regime that measures nothing cannot tell them apart, and neither can anybody else.
What the study establishes without needing any rate at all is what the recyclers said. Most of them do not rely on composition claims on labels to determine what their feedstock is made of. Labels are often missing or illegible by the time garments reach them; checking them manually would raise processing costs significantly; and those who do check find deviations. So they built their own instruments — near-infrared sorting, and, when in doubt, light and burning tests.
An industry was offered a legally mandated, printed, free statement of exactly the data its margins depend on. It priced that number at zero and paid to replace it.
And then the complication, which is too good to leave out. The recyclers' stated barrier is not accuracy at all: it is the label itself. Of all 10,901 garments, 55 per cent carried a tag made of polyester — a different material from the garment it was sewn into. Nearly a quarter had no label left. And in 33 per cent of that same full sample, the incompatible tag was the only thing standing between the garment and recycling, because a polyester label sewn into an otherwise homogeneous cotton garment makes the garment a blend for mechanical recycling. The sewn-in permanence that makes the claim feel official is the same property that makes the object harder to recycle.
Something similar holds at the border. Import duty differs by material — the study says the difference between the import charges applied to cotton and wool can be as much as 10 per cent — so customs has a live fiscal reason to care whether a declaration is honest, and may intercept goods on that basis. But that check, in the study's words, only compares customs declaration forms with the composition claims on garment labels, and does not include any assessment of whether the claims themselves are accurate. Two documents get checked against each other, and the cloth is never opened.
One document, twenty-one paragraphs apart
If labels were wrong at anything like these rates, wouldn't people notice?
A British government review of the enforcement regulations, dated 8 May 2017, answered that in its own opening pages: "As consumers cannot independently verify what fibres the textile products they purchase are composed of, consumer confidence in this sector may be eroded as they would have no way of holding retailer's claims to account." The next sentence extends it up the chain — individual traders further down the supply chain "may also not be in a position to test every garment or batch of garments."21
Twenty-one paragraphs later, the same document records that Citizens Advice received 66 complaints about textile labelling in 2016, 0.01 per cent of the 612,195 complaints it received in total, and concludes that "in general there appears to be high compliance." It adds that the government has no information on the number of fines issued, "if any," and that Trading Standards do not track enforcement activity in this area but estimate it to be low.
That review was not designed to measure compliance — its stated remit was the enforcement provisions, not the objectives of the underlying regulation, and it should be read accordingly. But the circle it draws is instructive. A defect the same document says consumers cannot detect generates few complaints, and the absence of complaints is read as evidence the defect is rare. There is no feedback loop here capable of generating the scepticism that would otherwise discipline the claim, and the government said so first.
The composition line does not need to manufacture trust. It only needs nothing to correct it. Its form does the rest: a number, summing to 100, sewn in permanently, in a shape prescribed by law and identical on every garment in Europe. Every one of those is a property of presentation, and none is evidence that anyone measured anything.
The next instrument is already drafted, and it makes the same move.
The EU Digital Product Passport Registry became operational on 20 July 2026; textile-specific requirements are currently planned through a future ESPR delegated act in late 2027.22,27 Article 9(1) of the current ecodesign regulation already says: "The data in the digital product passport shall be accurate, complete and up to date."23 That is the 2011 construction again — a duty of result, with no method attached.
The article then hands the delegated acts a list of nine things to specify, covering what data goes in, who may read it, who may write it, how it is carried and how it is updated. Within those nine specifications, none expressly requires an independent check that the product data entered in the passport are true.23 Article 9(3)(b) instead directs the passport to facilitate the verification of product compliance by competent national authorities. The state again.
Then there is the border. Article 15 is headed "Customs controls relating to the digital product passport," and under Article 15(2) customs may release a product for free circulation once it has verified — electronically and automatically — that the identifier and the commodity code correspond to the data held in the Digital Product Passport registry.
The paper-to-paper check, written into law, at machine speed.
The textiles delegated act has not been adopted, and it could still attach an evidential duty — a technical file, a retention period, anything of the Module A kind. If it does, this projection fails, and gladly. As of 21 July 2026, no adopted revision proposal appeared on the Commission's review page. The review materials emphasise accurate, intelligible and comparable information without misleading information, and better fibre identification, but do not yet show a garment-level evidential duty; a proposal could change that.26 And the campaign's own final report, if published for textiles, would settle several things this analysis cannot: how many garments sat in each blend tier, how the 49 divide across the three failure types, and how the laboratory handled tolerance and uncertainty.
Those denominators matter more than they look, and the arithmetic does most of the work. A garment declared as one fibre has to be right about one thing. A two-fibre blend has to be right about two. So a failure rate that climbs as garments get more complicated is partly just what compound criteria produce, before anyone has done anything wrong.
The campaign's tiers climb exactly like that: 15 per cent for garments declared as a single natural fibre, 46 per cent for blends of natural fibres, 64 per cent for blends of natural and artificial ones. How much of that climb is more than the arithmetic predicts would be the real finding. It cannot be worked out, because no per-tier sample sizes have been published.
The category figures need the same warning read twice. Scarves failed at 80 per cent — there were five scarves. Activewear failed at 13 per cent on fifteen items, which sits below even the single-fibre tier, and nothing published says why. That is unknown, and it should be left that way.
Four characters on a tag
Nothing here makes the composition label worthless. What it does is draw a boundary around it, and the boundary sits much closer in than most readers assume.
The tag will tell you what the seller declares the thing to be: wool. That is a real question, worth answering, and the mandate is what makes it answerable at all: without a compelled statement in a prescribed format there is no claim to test a garment against, no failure to record, no stopped sale, no campaign. Where somebody's money rides on the answer, verification has appeared — a trade body has run a monitoring programme in cashmere since 1984,25 precisely where the price gap is wide enough to fund the laboratory bill. And a shop genuinely cannot check the coat it is selling you, because the statutory method dissolves the sample. The retailer's presence-only duty is physics, not laziness. The argument here has never been with the shop. It is with the point in the chain where a production batch could be sampled as routine, and no rule requires it.
What the tag will not tell you is which wool. Twelve grades traded in that one week, spread fourfold from top to bottom, every one of them wearing the same four characters. Coarser wool is not defective wool — it goes to different products for different reasons — and whether retail prices track grade is a question this report did not investigate. What it does say is narrower: nobody anywhere in the chain owes anyone a measurement to establish which grade is in your hands.
The same question travels further than clothes. What is this number a number of, and who was required to check it? Put it to a nutrition panel, an SPF rating, a thread count, a carbon figure. It costs nothing to ask, and unlike the number it interrogates, it does not go out of date.
One question closes some of that distance, and it works at a counter or on a customer-service line: does the composition statement rest on a supplier's declaration, or on a test report? A seller who tests answers in a sentence. A seller who does not sends a policy. And a label whose percentages fail to add up to 100 has failed arithmetic before anyone has tested anything, which you can check standing still, with no equipment.