How Forgeries Are Detected

What Art Authentication Measures

A painting is a stack of materials with a history. Ground, paint, varnish, nails, canvas or panel, later fills, later retouch: each layer can belong to a different decade. Authentication asks which layers match the claim, and which layers arrived later. The American Institute for Conservation (AIC) treats examination as the basis for every later action, and tells members to declare age, origin or authenticity only on sound evidence. The same code flags authentication, appraisal and dealing as activities with a high risk of conflict of interest. A laboratory report is evidence. It is not a certificate that a dealer can stamp on a crate.

Three questions travel together and must stay distinct. Attribution asks who made the work. Dating asks when the materials were produced and assembled. Authenticity, in the forensic sense, asks whether the object matches the story attached to it. A panel felled in 1610 can carry a 1612 painting or a 2012 fake on old wood. The tree-ring date answers the first fact. It does not answer the other two.

Connoisseurship still does the first sort. A specialist who has looked at hundreds of works by one painter will see a wrong hand faster than a spectrometer will. Giovanni Morelli, in the late nineteenth century, taught students to compare ears and drapery folds as if they were anatomical specimens. Bernard Berenson and Max Friedländer built catalogues on that trained eye. The method fails when the object is a pastiche, when the specialist has a financial stake, or when the question is chemical. The laboratory then takes over the negative proof: this layer cannot be from 1650 because titanium dioxide in pigment form did not exist.

Law-enforcement work sits one step further out. The FBI Art Crime Team, formed in 2004, investigates theft, fraud, looting and trafficking. The Bureau states that the team has recovered more than 20,000 items valued at over one billion US dollars. Agents commission laboratories. They do not replace museum scientists.

Materials and Chemistry: Anachronism as Negative Proof

The strongest laboratory result in art authentication is a terminus post quem: a date after which a material became available. If a paint layer offered as fifteenth-century contains a pigment first manufactured in 1828, that layer cannot be fifteenth-century. The chemist does not need to name the forger. The calendar does the work. The same logic applies to binders and to optical brighteners in paper. The table below is a working chronology for the pigments that most often end a false claim. Dates are first manufacture or first documented artists' use, not the year a colour became cheap.

A pigment date is a floor. It does not identify a hand. The Corot viridian result is the warning row: handbook chronologies lag behind a single well-connected colourman. The opposite error is more common. An analyst maps titanium with X-ray fluorescence, then treats every titanium signal as "modern fake." Titanium can sit in a 1950s restorer's fill on an otherwise period picture. Stratigraphy decides whether the anachronism is in the original paint or in a later campaign.

Pigments, particles and why the compound matters

Elemental maps are the first screen. Handheld or scanning X-ray fluorescence (XRF) records calcium, iron, lead, mercury, cobalt, zinc, titanium and the rest, point by point or as a raster. A titanium map that follows the shape of losses in a lead map is a restoration map. A titanium map that sits inside an intact, continuous paint layer is a problem for any pre-1916 claim. Nicolas Eastaugh's 2008 analysis of a painting offered as a Heinrich Campendonk of 1914 or 1915 made that second pattern public: titanium white was not a commercial artists' pigment at the claimed date. The chemistry ended the date. The market history of that case belongs elsewhere.

XRF names elements. It does not name compounds. Cobalt can be smalt (a potassium glass used from the late Middle Ages) or Thénard's cobalt blue. Lead can be lead white, lead-tin yellow, red lead or a later repair. Raman spectroscopy and Fourier-transform infrared spectroscopy (FTIR) take the next step: they identify crystal forms and organic functional groups. Scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS), used on a mounted cross-section, shows particle size, layer order and whether a "period" pigment sits under or over a modern fill. Joyce Plesters built that cross-section habit at the National Gallery after she joined the Scientific Department in 1949. Conservators still work from her premise: you cannot clean what you have not identified.

Natural and synthetic ultramarine illustrate the need for morphology. Both are sodium aluminium silicate with sulphur. The synthetic form, awarded to Guimet in 1828, lacks the calcite and pyrite companions of ground lapis lazuli and tends toward small, rounded, even particles. A "Titian" whose blue is synthetic ultramarine throughout the original modelling is a post-1826 object, whatever the stretcher stamps say. A "Titian" whose blue is lapis, with later synthetic patches in the losses, is a period picture with a nineteenth-century campaign of repair. The sample location is the argument.

Binders, grounds and the support as a clock

Pigment is the headline. Binder and support are the quieter clocks. Until the twentieth century, European easel painting used drying oils, egg tempera, animal glue and, in some traditions, wax. Acrylic dispersions, alkyds and phenol-formaldehyde resins are later. Pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS) can name those polymers in a sample the size of a pinhead. Paul Coremans, working with Wiebo Froentjes and Martin de Wild after 1945, found phenol-formaldehyde in paint films offered as Vermeer and de Hooch. A seventeenth-century painter had no access to that resin. The published science is the detection. The recipe is not the subject of this page.

Grounds and canvases carry their own dates. A linen support can be radiocarbon-dated; the date is the harvest of the flax, plus storage, plus reuse. A wooden panel can be dated by dendrochronology; the date is the last measured growth ring, plus missing sapwood, plus seasoning. Both methods date the support. Neither dates the brushstroke. The dating a painting (pending) guide covers sample size, calibration curves and the post-1950s "bomb peak" in carbon-14. An old board is a supply. It is not a proof.

How a Laboratory Examination Proceeds

A competent examination is a sequence of tests chosen for the claim. The object arrives with a stated artist and date, plus documents (invoices, exhibition labels, photographs, previous reports). The examiner tests the claim against the object. AIC's Guidelines for Practice require a reason before any test that might change the object, and prior consent before a sample comes off. The written record names the examiner, the date, the methods, the sample sites and the interpretation. A verbal "looks right" from a dealer is not that record.

Cost and risk shape the order. Non-invasive imaging and surface microscopy come first because they cost less in material and they map where a sample would be wasted. Provenance research runs in parallel: a letter with a postcode that did not exist at the letter's date is a documentary anachronism, as useful as a pigment one. The Getty kouros file, discussed below, shows both kinds of failure at once. Scientific work on the marble could not close the question. The paper trail did not hold.

The output is a report with error bars. "Consistent with a seventeenth-century Northern Netherlandish palette" is a real finding. "Painted by Vermeer" is not a finding a spectrometer can issue. Catalogue raisonné committees, artist foundations and named specialists still make the authorship call, and many of those bodies have stopped issuing opinions after lawsuits. The laboratory's job is to say which materials are present, which layers are later, and which claimed dates the materials forbid.

From the surface to the structure

The first hour is looking. Raking light shows deformations, transfers and a crack network that may or may not run with the weave. Ultraviolet light shows later varnish and many later retouches as a different fluorescence. A stereomicroscope shows whether a signature sits in the original paint or in a later film, and whether dirt in the cracks is a settled deposit or a packed slurry. None of this dates the object. All of it tells the scientist where not to sample.

Technical imaging comes next. Infrared reflectography can reveal carbon underdrawing: a searching line, a change of pose, a pentimento. A forger who copies a finished reproduction has no reason to invent that search. X-radiography reads lead-rich paint, nails, joins, later patches and, in some cases, an earlier composition on the same support. The National Gallery's work on Raphael's Madonna of the Pinks, published around the 2004 acquisition, used infrared evidence of changes "in vastly more detail than a forger would mimic," in Ashok Roy's phrasing, together with pigments that fit Raphael's period. Imaging methods, wavelengths and camera geometry belong with the sibling article on infrared and X-ray imaging (pending). The authentication point is narrower: imaging tests whether the making process matches workshop practice.

XRF mapping then screens for elemental anachronism without a sample. A conservator who already suspects a nineteenth-century fake of a fifteenth-century design will look for zinc, chromium, cobalt in the modern sense, and titanium. A conservator who suspects a genuine picture with a heavy restoration will look for those same elements in the shape of losses. The map decides the next sample, if a sample is justified.

Sampling, ethics and how to read a result

A sample is a loss. AIC says to take the minimum required, to record the site, and to retain the material where possible. Cross-sections come from edges, existing losses or the reverse, not from a face that still reads as original. The mounted sample shows the order of layers: ground, paint, varnish, overpaint. SEM-EDS and Raman then name the particles in each layer. A modern pigment in the fourth layer up, over a discoloured varnish, is restoration. The same pigment in the first paint layer, under an aged varnish, is the object.

Interpretation is the step that fails in public. "We found titanium" is not a sentence that can stand alone. The report must say where, in which layer, in which compound (anatase or rutile), and whether the surrounding stratigraphy is original. The National Gallery's long investigation of The Virgin and Child with an Angel (NG3927) is the published lesson. The picture entered the collection in 1924 with the Mond Bequest as a Francesco Francia of 1490. In 1954 a second version appeared (now Carnegie Museum of Art, Pittsburgh). Trustees ordered technical work and judged the London picture a forgery. In the 1990s new technical work reversed that judgement, and the Gallery published NG3927 as an early Francia in 1998. In 2010, for the exhibition Close Examination: Fakes, Mistakes and Discoveries, Ashok Roy and Giorgia Mancini re-examined the picture and restored the 1950s conclusion on different grounds: a nineteenth-century Italian fake, with lead-tin-antimony yellow among the telling materials. The object did not change. The questions and the methods did.

Imaging as a First Pass

Infrared and X-ray images have become the public face of "science vs fake" because they look like revelations. They are maps. Carbon underdrawing appears in infrared if the underdrawing is carbon-rich and the upper paint is transparent in that waveband. Many later painters used iron-gall inks or non-carbon sketches that infrared will not read. Lead white in the upper layers can mask what lies below. None of those silences prove a forgery. They prove that the method did not see.

Used with a known corpus, imaging still earns its place. Workshop replicas share cartoons; the infrared lines match. Autograph works show changes of mind. Copies after a print show a confident outline with no search. Coremans's X-radiographs of the disputed "Vermeers" showed a mismatch between surface cracks and ground cracks, and modelling in the faces that did not match the lead-white structure of accepted Vermeer heads. That is process evidence.

Readers who want wavelengths, InGaAs cameras, and how conservators read a reflectogram should use the infrared reflectography (pending) guide. The authentication rule on this page is short: image first, sample second, and never treat a blank infrared plate as a positive identification.

Dating Wood, Canvas and Paint: What the Clock Cannot Say

Dendrochronology matches the ring-width sequence in a panel to a regional master chronology. Baltic oak, the staple of many Netherlandish panels, has dense reference data. The last measured ring is a date. If sapwood is absent, as it often is after a panel-maker squared the board, the felling date is later by an estimated number of sapwood rings (published Baltic estimates run from about six to about nineteen years). Seasoning and transport add more time. A Jordaens project that combined dendrochronology with art history (published 2021) added a minimum of two years after the earliest possible felling before treating the support as ready for a painter. The result is a terminus post quem for the panel, not a sitting date for the artist.

Radiocarbon dating measures carbon-14 in organic material: wood, canvas, binder, sometimes a carbon-containing pigment. Accelerator mass spectrometry has cut the sample size, so museums now consent more often. The date is the death of the organism that fixed the carbon. A forger who buys a seventeenth-century canvas and paints on it in 1938 will obtain a seventeenth-century radiocarbon date for the cloth. A restorer who consolidates a medieval panel with a modern resin can pull a binder date toward the present. Nuclear weapons tests from the 1950s raised atmospheric carbon-14 (the bomb peak); organic material grown after that rise can be distinguished from pre-bomb carbon, so a "Renaissance" oil with a post-bomb binder is a modern object. The dating a painting (pending) article is the place for calibration, wiggle-matching and sample ethics. The authentication limit is the old-wood problem: the clock dates the tree or the flax, not the day a brush touched it.

Thermoluminescence (TL) dates the last heating of ceramic. It is a tool for pottery, porcelain and bronze casting cores, not for easel paintings. Oxford Authentication, a commercial laboratory that specialises in TL, published UK prices of £200 plus VAT per pottery object and £300 plus VAT per porcelain object, with US figures of 400 and 500 dollars, results in three to four weeks. Those numbers show the cost of a single specialised test on a fired object. They do not transfer to a painting.

From the Connoisseur's Eye to the Museum Laboratory

Authentication as a profession is older than the spectrometer. Morelli's morphological method and the catalogue raisonné as a legal-adjacent document both tried to make looking reproducible. They produced a literature of "school of" and "workshop of" that still structures museum labels. They also produced famous errors, because a skilled pastiche aims at the same morphological habits the connoisseur was trained to trust.

Museum laboratories grew from conservation crises. The National Gallery in London appointed the physicist F. I. G. Rawlins as Scientific Advisor in the early 1930s; the Scientific Department dates its founding to 1934. The brief was objective data on methods and environment. After the wartime storage of the collection in Manod quarry, a public cleaning controversy led to the Weaver Committee. The Gallery hired the chemist Anthony Werner in 1948 and set up a chemical laboratory. Joyce Plesters joined in 1949 and made paint cross-sections a standard tool. The National Gallery Technical Bulletin has published that work since 1977. Nature, reviewing the department in 2010, noted a fifteenth-century group portrait bought in 1923 that chemical analysis exposed as a forgery because it contained pigments unused before the nineteenth century. The same review recorded the Corot viridian surprise and Roy's defence of the Madonna of the Pinks.

French museum science has a long institutional memory. The Centre for Research and Restoration of the Museums of France (C2RMF) took its present form in 1995, merging research and conservation services that had operated since the 1930s, in premises under the Louvre. Its particle accelerator, AGLAE (Accélérateur Grand Louvre d'Analyse Élémentaire), installed for heritage use from 1988, is the only accelerator dedicated to cultural materials. NewAGLAE, from 2017, added more sensitive detectors. PIXE and related ion-beam methods give elemental data, including trace elements, without a sample. Access is for public-collection research, not for a private owner who wants a certificate.

The Getty Conservation Institute in Los Angeles and the Metropolitan Museum's Department of Scientific Research in New York run the same pairing of imaging, spectroscopy and micro-sampling on collection objects. Published Met studies show the routine: stereomicroscopy, X-radiography, infrared, scanning XRF, then a few samples from edges. Authentication of a private "find" is not the job advertised on those pages. The job is to understand how a collection object was made and how it has changed.

Independent pigment historians and commercial labs fill the gap the museums leave. Eastaugh's Art Discovery work on the Campendonk titanium white is the model of a negative proof that the market could not ignore. In the Clementine Hunter investigation, five laboratories, including Orion Analytical, documented artificial ageing residues (compacted dirt, paper-towel fibres) that did not match authentic Hunter surfaces. The finding is material. The trial narrative is not the subject of this page.

Who Runs the Work, and Why the Setting Matters

The people who master detection are the people who see enough real objects to know a period palette from a plausible one. Plesters learned that difference on National Gallery cross-sections. Roy learned it on a corpus with documented provenances. Coremans learned it under court instruction, with studio materials to compare against the paintings. C2RMF scientists learn it on several hundred objects a year. That volume is the method. A single spectacular test on a single disputed canvas is a weaker design.

Setting matters because incentives differ. A museum laboratory answers to a collection and to a conservation code. It can publish a fake in its own holdings, as the National Gallery did with NG3927, without selling the object the next week. A commercial authentication service answers to a paying client who may want a sale. AIC's warning about conflict of interest is aimed at that pressure. A catalogue raisonné committee answers to a scholarly reputation and, after a generation of lawsuits, to legal risk; several artist foundations have closed their authentication boards rather than underwrite opinions. The FBI Art Crime Team answers to a criminal statute. Each of those rooms can use the same Raman spectrum. They will write different last sentences.

For method, read museum technical studies and the National Gallery Technical Bulletin. For ethics, read AIC and Icon. For a private object, an independent lab can report materials with a written scope that stops there. Browse museums when you want the institutions that staff these laboratories. Do not expect the laboratory door to open as a public authentication desk.

How Detection Fails

Every method on the bench has a known miss. The misses cluster in two families: the support is old and the paint is not, and the test answers a different question from the one the client asked. Contamination (a restorer's pigment, a consolidant) belongs to the first family if you misread the layer. The Corot viridian case is a supply fact mistaken for guilt: the handbook date was late, the artist was early. Detection fails in public when a report collapses those families into a yes or no.

The Getty kouros is the sculpture that made the second family famous. The J. Paul Getty Museum acquired the marble youth in 1985. Norman Herz's isotopic work pointed to Thasian dolomitic marble. Stanley Margolis argued that de-dolomitization of the surface (magnesium leached, a calcite crust left) could occur only over centuries, and so could not be faked. Later work showed that the surface effect can be induced. Stylistic objections never went away. Documentary provenance did not hold. The museum's label came to read "Greek, about 530 B.C., or modern forgery." Timothy Potts, as director, said the sculpture left display in 2018 because of the dispute. Marble source is not carving date. A surface crust is not a calendar. The tests were real. They did not prove the claim.

Old supports, later paint, and restoration that looks like guilt

Reuse is normal workshop practice. Painters cut down panels, turn canvases, paint over unsold pictures. A forger who understands that history will buy an old support on purpose. Dendrochronology and radiocarbon then "confirm" the period. The examiner who stops at the support has been used. The paint, the ground, the binder and the craquelure relationship still have to match. Coremans's mismatch between paint cracks and ground cracks is the classic catch. A titanium-white fill in a genuine loss is the classic false catch.

Restoration is a common source of anachronistic pigment. Nineteenth-century restorers used zinc white, chrome yellow and synthetic ultramarine without any intent to deceive. Twentieth-century restorers used titanium white because it hides. A scanning XRF map that lights up titanium along old tears is a treatment history. The same map, misread as a single date for the whole picture, becomes a false conviction. Cross-sections exist to prevent that error. If the institution will not permit a sample, the report must say that the elemental anachronism is unlocated in depth.

What a test cannot prove

A laboratory cannot name a painter. Period materials are shared across a city and a generation of followers. Vermeer's lead-white and ultramarine were also the materials of his neighbours. A palette consistent with 1660 Delft is consistent with a 1660 pupil and a 1660 master. Authorship is a historical judgement that uses the laboratory as one witness.

A laboratory cannot prove a negative with the same force as a positive anachronism. "We did not find titanium" means the sampled or mapped areas did not show titanium. It does not mean no modern material exists in an unsampled fold of the reverse. "The infrared is blank" means this waveband did not record carbon underdrawing. It does not mean the painter drew nothing.

A laboratory cannot authenticate a work from a photograph. Commercial services that offer AI scores from a JPEG are answering a different question: statistical resemblance of a surface pattern. They do not replace stratigraphy. They do not date a binder.

A laboratory cannot convert a materials report into a market value. Valuation is a separate professional act, with its own standards and its own conflicts. This article does not advise anyone to buy, sell or hold an object on the basis of a test.

What Scientific Authentication Costs Today

Prices move with city, instrument time and the depth of the written report. The figures below are published tariffs and advertised ranges from 2024 to 2026, not quotes, and not a recommendation to spend. A signal-only log costs less than a narrative report. A report written for court costs more.

Per-sample materials analysis at a small specialist lab can sit around 150 US dollars for FTIR or Raman, 150 for SEM-EDS, and 200 for GC-MS (History Echoes Analytical Services, published menu). That is the cost of naming a compound in a sample the client has already taken. It is not the cost of an authentication. Thermoluminescence on a ceramic, as noted, is in the low hundreds of pounds or dollars per object at Oxford Authentication. A mid-level file that adds provenance reading and a decision on whether science is warranted is advertised by some private practices in the range of 1,500 to 5,000 dollars. Full forensic programmes that combine imaging, sampling, specialist consultation and a formal report are advertised from about 5,000 dollars and, on high-stakes objects, well above 15,000.

Museum laboratories do not sell this service to the public. Their equipment exists for the collection. Independent conservators who are AIC or Icon members may examine an object and still refuse to issue an "authentication" because their code treats that word as a conflict trap. Catalogue raisonné inclusion, where a committee still sits, can involve a fee, a waiting list, or a closed door.

The cost that does not appear on a tariff is the cost of a wrong question. Dating the panel when the dispute is the paint wastes the fee. Sampling a restoration and calling the picture modern wastes the object. The cheaper first pass remains the one AIC describes: a documented examination, imaging, and a written reason for every later test.

FAQ

These questions come from the same confusions that sit on the search page: cost, certificates, whether a lab can name a painter, and whether an old panel is enough. The answers stay inside the detection science of this article. They do not value an object, and they do not tell you how to age a surface. If you own a work and want a materials report, start with a documented examination and imaging, then a written reason for any sample. If you are reading for study, use the cases named above (National Gallery NG3927, Corot viridian, Getty kouros, Coremans's binder findings) as checkable publications rather than as folklore.

Can a laboratory prove who painted a picture?

No. A laboratory can prove that a layer contains materials available at a given date, that a support was felled or harvested in a given window, and that a later campaign sits on top of an earlier one. Authorship is a judgement that compares those facts with a body of accepted works, documents and connoisseurship. Period-plausible materials are shared. The National Gallery's Francia file shows even a well-equipped lab changing its mind as questions and methods change.

What does titanium white in an Old Master painting prove?

Titanium dioxide as a commercial white pigment dates from 1916, with artists' oil colour around 1921. In a paint layer offered as pre-1916, that compound is a terminus post quem: the layer is twentieth-century or later. The first task is to decide whether the titanium sits in original paint or in a restorer's fill. XRF finds the element. Cross-sections and Raman find the layer and the crystal form (anatase or rutile). Eastaugh's Campendonk analysis is the published case of titanium white ending a claimed pre-commercial date.

Why can an old panel still carry a modern fake?

Dendrochronology and radiocarbon date the wood or the flax, not the painting event. Panel-makers trimmed sapwood, painters reused boards, and a later hand can paint on a period canvas. The old-wood problem is the reason a support date is a floor for the support and nothing more. Binder analysis, pigment stratigraphy and the relationship between surface craquelure and ground craquelure still have to be done. See dating a painting (pending) for the chronometric detail.

How much does it cost to get a painting authenticated?

A single instrumental analysis can cost about 150 to 200 dollars per sample. A ceramic TL test is in the low hundreds of pounds or dollars. A structured private examination with a written report is advertised from the low thousands; a full forensic file from about 5,000 dollars upward. Museums do not authenticate private pictures as a public service. A certificate of authenticity printed from a template is not this process and is not a laboratory result.

Is a certificate of authenticity the same as authentication?

No. A certificate of authenticity (COA) is a document someone issues. It can be a living artist's statement, a gallery invoice, a foundation stamp, or a piece of paper with no standing. Authentication in the sense of this article is an argument from examination, documents and, where justified, analysis. AIC members are told to declare authenticity only on sound evidence. Search interest in "free COA" is a different market. It does not substitute for a technical report.

Should imaging come before a sample?

Yes. Infrared reflectography and X-radiography, with raking light and ultraviolet, map restorations, underdrawing and structure without a loss of original material. They also show where a sample would be redundant or harmful. AIC requires a justification before any test that may change the object. The sibling guide to infrared reflectography (pending) covers the imaging methods. Sampling follows if a pigment or binder question remains that imaging cannot answer.

What if the laboratory and the connoisseur disagree?

Record both, and name the disagreement. The Getty kouros is the case where science, style and documents never converged. The National Gallery's NG3927 is the case where the laboratory disagreed with itself across decades as methods improved. A pigment anachronism in original paint outweighs a stylistic "feeling" that the picture is right. A clean materials report does not outweigh a specialist who can show that the drawing, the handling and the workshop practice do not match the named artist. The report should state which question each party answered.