Leonardo da Vinci

Ask how many paintings Leonardo made and you will be given a number between fourteen and twenty, and the person giving it will usually not tell you which pictures they have counted. The honest answer is that there is a secure core of roughly a dozen panels that no serious scholar disputes, and a contested ring around it of perhaps eight more, and that the ring is where all the money and all the argument live.

The secure core is small enough to list. The Annunciation and the unfinished Adoration of the Magi in the Uffizi. Ginevra de' Benci in Washington. The unfinished Saint Jerome in the Vatican. The two versions of the Virgin of the Rocks, in Paris and in London. The Last Supper on the refectory wall of Santa Maria delle Grazie in Milan. The Lady with an Ermine in Krakow. La Belle Ferronniere in the Louvre. The Mona Lisa in the Louvre. The Virgin and Child with Saint Anne in the Louvre. The Saint John the Baptist in the Louvre. Even here there are qualifications: Leonardo's contribution to Verrocchio's Baptism of Christ in the Uffizi is a passage in another man's picture, and the Adoration and the Saint Jerome are abandoned rather than finished.

The ring is where the disagreement is. The Portrait of a Musician in the Ambrosiana in Milan is given wholly to Leonardo by some and split between Leonardo and a pupil by others. The Madonna Litta in the Hermitage is given by a substantial body of opinion to Giovanni Antonio Boltraffio working from a Leonardo cartoon. The Benois Madonna, also in the Hermitage, is accepted by most and doubted by a few. The Madonna of the Yarnwinder exists in two principal versions, the Buccleuch and the Lansdowne, and the question is not whether Leonardo designed the composition but how much of either panel he painted. The Bacchus in the Louvre began life as a Saint John and is now generally given to a follower on a Leonardo design. The Salvator Mundi is the extreme case. La Bella Principessa is accepted by one prominent scholar and by almost nobody else.

What makes the Leonardo situation different from the Rembrandt situation is that there is no institution whose job it is to adjudicate. The Rembrandt Research Project ran from 1968 to 2014 and published six volumes; whatever its critics said, there was a body, it had a method, and it put numbers on the record that other people could argue with. Leonardo has never had one. The functional equivalents are the great monographs, the exhibition catalogues of the major retrospectives, and the wall labels of five museums. When those disagree, and they do, nothing resolves them.

Names matter here, because the positions are personal and published rather than institutional. Martin Kemp, emeritus professor at Oxford and the author of the standard English monograph, is the most permissive of the senior authorities and has argued for both the Salvator Mundi and La Bella Principessa. Carmen C. Bambach of the Metropolitan Museum of Art, whose four-volume study appeared from Yale in 2019, is among the most restrictive and gives the bulk of the Salvator Mundi's execution to Boltraffio. Frank Zoellner, whose Taschen catalogue is the most widely used single-volume reference, catalogues the Salvator Mundi as Leonardo and workshop with a question mark. Pietro C. Marani, the Milanese specialist who has worked on the Last Supper conservation and on the Leonardeschi, occupies a middle position. Luke Syson, who curated the National Gallery's 2011 exhibition, made the curatorial decision that put the Salvator Mundi into the canon in the first place. Dianne Dwyer Modestini, who restored the panel, is not a disinterested party to that argument and has never claimed to be, but she is the person who has looked at the surface longest and hardest and her published account is the primary technical document.

Leonardo was born on 15 April 1452 at or near Vinci, in the Tuscan hills west of Florence, the illegitimate son of Ser Piero da Vinci, a notary, and a woman named Caterina about whom almost nothing certain is known. Illegitimacy barred him from the notarial profession his father and grandfather followed and from the university; it did not bar him from a craft. Around 1466 or 1467 he entered the Florentine workshop of Andrea del Verrocchio, a goldsmith by training who ran the most technically various shop in the city: sculpture in bronze and marble, painting, metalwork, festival apparatus, engineering.

The first documentary trace of Leonardo as a professional is his entry in the register of the Compagnia di San Luca, the Florentine painters' confraternity, in 1472, when he was twenty. He remained in or around Verrocchio's shop for several years after that, which is normal and is also the origin of the attribution problem that has followed him for five centuries. A Renaissance bottega was a firm. Its output was signed with the master's name and made by whoever was competent to make it.

The picture where this becomes visible is Verrocchio's Baptism of Christ in the Uffizi, tempera and oil on panel, 177 by 151 centimetres, inventory 1890 number 8358, dated by the gallery to about 1470 to 1475. It came from the Vallombrosan monastery of San Salvi in Florence, moved to Santa Verdiana in 1730, entered the Florentine Galleries in 1810 and has been in the Uffizi since 1919. The kneeling angel at the left is given to Leonardo, and the Uffizi's own current position goes further than the traditional one: it states that studies now consider Leonardo's interventions more extensive, including the riparian landscape, the golden light and the figure of Christ itself.

The gallery adds a detail that is worth more than most connoisseurship. It considers it likely that a third and older painter also worked on the panel, on the evidence that some passages are less developed than others, specifically the hands of God and the dove of the Holy Spirit at the top. That is what a fifteenth-century picture actually is: a panel with three hands on it, one of them a boy who would later be Leonardo, all of it sold under one name.

The tradition that Verrocchio gave up painting in shock at the pupil's superiority comes from Giorgio Vasari, writing in 1550 and again in 1568, and Vasari is a tertiary source of famous unreliability who wrote to a thesis. The Uffizi cites him as narration and then sets out the technical position separately, which is the right way to handle him. The medium change is evidence. The anecdote is a story about a story.

What matters for everything that follows is the principle the Baptism establishes. In the practice Leonardo was trained in, one panel could carry two hands without dishonesty and without a label, because the workshop, not the individual, was the unit of production. Leonardo went on to run such a workshop himself, in Milan, for the better part of two decades. Every attribution argument about a Leonardo picture is downstream of that fact.

Two further documents from the Florentine years belong in any honest account. On 9 April 1476 an anonymous denunciation to the Officers of the Night, the Florentine magistracy for sexual offences, accused Leonardo and three others of sodomy with a seventeen-year-old model named Jacopo Saltarelli. The case was dismissed, apparently because one of the accused was connected to the Medici. It is recorded here because it is a document, and because the suppression of it in nineteenth- and early twentieth-century accounts is part of how the biography was shaped. In 1478 he received his first independent commission, for an altarpiece in the chapel of San Bernardo in the Palazzo della Signoria, and did not deliver it; in 1481 the monks of San Donato a Scopeto commissioned the Adoration of the Magi, and he abandoned it unfinished when he left for Milan. The pattern of non-delivery starts at the beginning.

Leonardo went to Milan around 1482 and stayed, with interruptions, until the French took the city in 1499. He returned in 1506 and remained until 1513. Those two Milanese periods produced the workshop, and the workshop produced the problem.

Giovanni Antonio Boltraffio, born about 1467 into a Milanese family of some standing, is the only member of the shop with a substantial documented independent career and a securely identified body of signed work. He is therefore the name that specialists reach for when a picture is too good for a routine assistant and not quite right for Leonardo. Bambach's position on the Salvator Mundi is precisely that: the execution is Boltraffio's, with Leonardo's intervention in places. The Madonna Litta in the Hermitage attracts the same argument for the same reason.

Gian Giacomo Caprotti, called Salai, entered the household in 1490 at the age of ten. Leonardo's own note on him is one of the few pieces of characterisation of a workshop member in the artist's hand anywhere in the Renaissance: thief, liar, obstinate, glutton. Salai stayed for a quarter of a century, went to France, returned to Milan, and died in 1524. The inventory drawn up after his death is reported to list paintings, among them one described as a Joconda. Two cautions belong on that sentence, and the second is rarely made. The first is that even if the entry says what it is said to say, it establishes that Salai owned a picture of that description, not that it was the Louvre panel; copies were already circulating. The second is that this record has not seen the document, and neither, in most cases, have the writers who cite it. It is one of the most repeated pieces of evidence in the literature and it needs a primary archival citation before anything is built on it. Salai is also one of the two names most often proposed for the Prado version of the Mona Lisa, the other being Melzi, with Spanish pupils of Leonardo proposed by others; the Prado does not settle it.

Francesco Melzi, from a noble Milanese family, joined around 1506 or 1507 and is the one member of the circle who mattered more after Leonardo's death than during his life. Leonardo's will of 23 April 1519 left Melzi the books, the instruments and, decisively, the papers. Melzi took them to the Villa Melzi at Vaprio d'Adda, kept them for half a century, and compiled from them the compilation later known as the Codex Urbinas, the basis of every printed edition of the Treatise on Painting. When Melzi died in 1570 his heir Orazio had no interest in the material, and the dispersal that followed is the reason the manuscripts are now scattered across Milan, Paris, London, Madrid, Turin and Windsor.

Around these three stand the painters the Milanese literature calls the Leonardeschi: Marco d'Oggiono, Cesare da Sesto, Andrea Solario, Giampietrino, Bernardino Luini. They produced, between them, the great mass of pictures that look like Leonardo, derive from Leonardo's designs and are not by him. They are the reason multiple versions exist of almost every late Leonardo composition, and they are the reason a great many of those versions have at some point been sold as autograph.

The vocabulary the trade uses for this is not neutral and repays attention. Leonardo asserts the hand. Leonardo and workshop asserts the design and the supervision and concedes the execution, wholly or partly. Workshop of Leonardo concedes the hand entirely and keeps the shop. Follower of and after concede everything except the composition. The distance between the first and the second of those, on a picture in a saleroom, has been worth hundreds of millions of dollars exactly once, and worth a great deal every other time it has arisen.

The clearest demonstration that Leonardo's problem is structural rather than accidental is that he painted the same altarpiece twice, that both versions survive, that both are accepted, and that they hang eight hundred kilometres apart.

The commission was placed in 1483. Leonardo and the brothers Evangelista and Ambrogio de Predis contracted with the newly formed confraternity of the Immaculate Conception of the Virgin Mary to supply the painted parts of a large altarpiece for the confraternity's chapel in San Francesco Grande, the Franciscan house that was the largest church in Milan after the cathedral. The altarpiece included a wooden sculpture of the Virgin, probably set above the painting. The contract set a price and a date; neither held. The National Gallery's own account is blunt about what followed: a dispute over money led Leonardo to sell the first version elsewhere, the confraternity and the painters eventually came to terms, and the picture was not completed to the confraternity's satisfaction until 1508, twenty-five years after it was ordered.

The doctrinal setting is not decoration and explains the commission. The Immaculate Conception was contested: the Franciscans supported it, the Dominicans opposed it, and Sixtus IV had adopted the feast for the Western church only in 1476, less than a decade before the contract. The subject was so new that there was no settled way of picturing it. A confraternity existed precisely to promote the doctrine, and it hired a painter to invent an image for something that did not yet have one.

The earlier version, on panel transferred to canvas, is in the Louvre with the inventory number INV 777, dated by the National Gallery to about 1483 to 1485. The later, on panel, is in the National Gallery in London as NG1093, dated there to about 1491 or 1492 to 1499 and 1506 to 1508, which is to say painted in two campaigns separated by seven years. The differences are not subtle once you know to look: the angel in the Paris version points at the infant Baptist and looks out of the picture, and in the London version does neither; the London Baptist carries a gilded reed cross; the London handling of rock and foliage is harder and cooler.

The technical work the National Gallery has published on NG1093 established something no amount of connoisseurship could have, and it came in two stages fifteen years apart. In 2005 infrared reflectography found, under the paint, the underdrawing of a completely different composition: a Virgin in another pose, abandoned. Around the 2019 display the Gallery went back with macro X-ray fluorescence scanning and hyperspectral imaging and recovered more of it, including the first design for the angel and the Christ Child, both set higher in the field, with the angel facing outward and holding the child in a much tighter embrace. The XRF worked because the drawing material contains zinc, so a map of the zinc distribution across the panel draws the abandoned composition out from under the landscape. That is direct evidence that the London panel is not a copy of the Paris one but an independent working-out that changed course and then, in the end, came back to something almost identical to the Louvre design.

The same campaign found something smaller and more human. High-resolution imaging in raking light shows a finger or palm print in the priming layer, visible through the thin paint on the Virgin's left cheek. The Gallery does not claim to know whose hand it was, and says the most likely explanation is somebody patting the priming down to make an even surface. That is the correct way to report a fingerprint on a Renaissance panel, and it is worth setting beside the much stronger claims made for fingerprint evidence in attribution disputes elsewhere.

The version-safety lesson is exact. There are two Virgin of the Rocks. Any statement about "the Virgin of the Rocks" that does not say which one is worthless, and a photograph of one labelled as the other is a factual error of the kind this publication exists to prevent.

The Virgin of the Rocks is the best-documented case of Leonardo's method of construction, so it is the right place to say what these pictures are made of intellectually as well as materially. Nothing here is hidden and nothing here is speculative; it is the National Gallery's own account of its own panel.

The setting is a wilderness of rock over still water, the pinnacles and fissures driving vertically against the flat of the water. It is fertile: plants grow strongly beneath the Baptist and the Virgin, and sprout from the rock above. But the flowers are not flowers. They do not resemble any real species; they are invented hybrids of different plants. The Gallery's reading of those two facts together is that the scene may be set in the earliest moments of creation, which is what the opening verses of Genesis describe, the earth made out of the watery deep. That is a proposal about a picture made for a confraternity dedicated to the Immaculate Conception, which is a doctrine about a person conceived outside the sequence of original sin. A world before the first sin, painted with plants that do not exist yet, for a chapel dedicated to a woman held to have been conceived before the fall. The elements fit and the Gallery says so; nobody claims a document in which Leonardo states it.

The optical machinery is documented in his own hand. He worked out from his studies in optics that colour is perceived differently with distance, that mountains read blue and pale when far off, and he softened their edges to imitate what vision actually does; the technique is now called aerial perspective. He blocked out most of the sky, hemmed the figures with overhangs and dark hollows, and lit them as if from a source outside the picture, so that faces and bodies emerge from gloom. And he built the transitions in the flesh with the smoky grading that takes its name from the verb to turn to smoke. In notes for the treatise on painting he was compiling he remarks that even ordinary faces look beautiful at dusk, when the light is low. That is a painter recording, as a technical observation, the thing his pictures do.

The reason this matters to an attribution page is that these are the qualities a workshop hand can approximate and a copyist cannot invent. Nobody has ever mistaken a Giampietrino for a Leonardo on the strength of the optics. The arguments about the marginal pictures are almost never about whether the composition is Leonardo's. They are about whether the surface in front of you was built by the man who worked the optics out.

The Salvator Mundi is a half-length Christ in frontal view, the right hand raised in blessing, the left holding a transparent orb, on a walnut panel of roughly sixty-five by forty-five centimetres. Iconotheca's own articles on authentication and on sleepers both use it, and neither of them needs repeating here. What follows is the object's record, with the documented separated from the reported, because almost every popular account of this picture runs the two together.

The 2005 lot, verbatim. The restorer's own published account reproduces the catalogue line, misspelling and all: lot 664, After Leonardo da Vinci, Italian, 1452-1519, Christ Salvador Mundi, oil on cradled panel, 26 by 18 and a half inches. Estimate USD 1,200 to 1,800. St Charles Gallery, New Orleans, April 2005. It sold below its low estimate. The two figures in circulation resolve exactly once the fees are separated: USD 1,000 at the hammer plus USD 175, which is precisely seventeen and a half per cent, gives the USD 1,175 that is usually quoted. A third figure of USD 10,000 circulates and reconciles with nothing; this record has found no source for it and does not print it. The buyers were Robert Simon, a New York Old Master dealer, and Alexander Parish; Warren Adelson joined the partnership afterwards. By 2011 the owner of record was an entity called R. W. Chandler, which is a materially important provenance fact, because the lender to the National Gallery was a company and not the three named dealers.

What comes before 2005 is thinner than the catalogues suggest. The first secure link in the chain is 1900, when Sir J. C. Robinson sold a Salvator Mundi to Sir Francis Cook for GBP 120; Cook's own 1913 catalogue called it a free copy after Boltraffio. It sold at Sotheby's in London on 25 June 1958, as Boltraffio, for GBP 45, to the Kuntz family; Minnie Stanfill Kuntz left it in 1987 to her nephew Basil Clovis Hendry senior, and it was sold from his estate in 2005. Everything earlier is proposed rather than established: a 1650 etching by Wenceslaus Hollar inscribed with Leonardo's name, an entry in the 1638 to 1641 inventory of the Marquess of Hamilton, entries in the inventories of Charles I in 1651 and Charles II in 1660, a lot in a 1763 Buckingham House sale. Those documents exist. What is not established is that any of them describes this panel. Charles Hope's judgement is flat: there is little reason to suppose that the Salvator Mundi was ever in the royal collection, and no good reason to suppose that Hollar copied the picture in England.

The restoration, in the restorer's own words. Dianne Dwyer Modestini began work in 2005 and continued for years; her account and a published timeline disagree about whether it was finished in 2008 or continued into 2010, and further retouching is recorded around the 2011 exhibition. She describes the state it arrived in: the head and the curls on the proper right grotesquely repainted, the blessing hand perfectly preserved, the background repainted a muddy brown in the nineteenth century, and, on the original background, very little remaining, having for the most part perished when the painting was brutally scraped down. Old damage had been covered with gesso fill and painted over.

And then the sentence that matters more than any other in this dossier. Modestini writes that she also suppressed a pentimento of the blessing thumb that had been exposed during the radical scraping of the background. Read that carefully. The double thumb is not a sceptic's invention: a second, earlier position of the thumb was physically visible on the surface when the picture came to her, and she painted it out. That has two consequences and both of them cut. First, a pentimento is the standard argument for autograph status, because a copyist working from a finished picture does not revise. Second, the argument now rests on a feature that the present surface no longer shows, because the restorer covered it. The authors of the 2019 Oxford book on the picture discuss the decision candidly, along with the repainting of the background black. A reviewer of that book puts it plainly: masking an original thumb and repainting the background black.

Two things are therefore true at once, and any account that suppresses either is misleading. The picture assessed by scholars in 2008 and exhibited in 2011 is in part a reconstruction. And there is no honest way to judge the quality of a passage that has been reconstructed. Hope makes the same point from the other side: an attribution resting on the least damaged parts of a picture in this condition, he writes, can hardly be taken seriously.

Christie's provenance for lot 9B is the longest printed chain ever attached to a Leonardo sold at auction, and it repays reading as a document about how provenance is written rather than as a record of where the picture was.

It opens with two possiblys: possibly commissioned after 1500 by Louis XII of France and Anne of Brittany after the conquest of Milan and Genoa, and possibly by descent to Henrietta Maria of France, by whom possibly brought to England in 1625 on her marriage to Charles I. It then reaches solid documents that are solid about something else. The Commonwealth Sale entry of 23 October 1651, a peece of Christ done by Leonardo at 30-00-00, presented as part of the Sixth Dividend to Captain John Stone; the return of works to Charles II at the Restoration in 1660; probable inheritance to James II; probable removal by Catherine Sedley, Countess of Dorchester, or by John Sheffield, first Duke of Buckingham and Normanby; the Prestage sale in London on 24 February 1763, lot 53, L. Da. Vinci A head of our Saviour, at two pounds ten.

Every one of those entries exists. Not one of them has been shown to describe this panel. Charles Hope's verdict on the whole royal section is that there is little reason to suppose the Salvator Mundi was ever in the royal collection. A reader should be able to see the difference between a documented inventory line and a documented identification, and a provenance printed as a single continuous paragraph makes that difference invisible.

From 1900 the chain is real. Sir John Charles Robinson, who owned it as a Bernardino Luini, sold it that year to Sir Francis Cook. It descended through three baronets at Doughty House, Richmond, catalogued as a free copy after Boltraffio and later as Milanese School, to Sir Francis Cook, fourth baronet, and sold in his sale at Sotheby's in London on 25 June 1958, lot 40, as Boltraffio, at GBP 45 to Kuntz. Note in passing that the nineteenth-century owner's attribution was Luini, which is the name Matthew Landrus proposed again in 2018.

And then the catalogue reads, in full: Private collection, United States. Robert Simon, New York. Private sale; Sotheby's, New York. Acquired from the above by the present owner. That is the entire printed account of 1958 to 2017. The New Orleans auction of April 2005, the lot number, the estimate, the price, the consortium, the eleven years of restoration and the 2011 exhibition do not appear. Nor does the name of Alexander Parish, or Warren Adelson, or the holding company R. W. Chandler, or Yves Bouvier, or Dmitry Rybolovlev.

None of that is a breach of any rule. Houses routinely compress recent private ownership, and a consignor may decline to be named. It is recorded here because it is the single clearest demonstration on this page of why a publication like this one exists. The most heavily documented provenance ever printed for a Leonardo devotes several hundred words to the possible movements of a possibly identical picture in the seventeenth century and four lines to the period in which everything that determined its price actually happened.

The single most consequential event in the Salvator Mundi's modern history was not the sale. It was the exhibition.

In January 2008 Robert Simon took the panel to the Metropolitan Museum of Art, where three infrared images were made, of the head, the blessing hand, and the mantle, stole and orb. That was an imaging session. The Metropolitan has never endorsed the attribution, then or since; Warren Adelson tried to sell the picture to the museum in 2009, and five days after the 2017 sale the museum's then director Thomas Campbell publicly questioned both the price and the standard of the conservation.

Later in 2008 the picture was shown in London to five Leonardo specialists: Carmen Bambach, David Alan Brown, Maria Teresa Fiorio, Martin Kemp and Pietro Marani. Two corrections to the version of this story that circulates. Nicholas Penny saw the picture separately, in New York, before he became director of the National Gallery; and Luke Syson, who would curate the exhibition, is not named among the five. No vote was taken and no collective statement was issued, and the result, on two independent accounts, was the same: two supported the attribution, two declined to commit, and one, Bambach, rejected it. Ben Lewis records it as two yeses, one no and two no comments, and adds that the majority agreed the face had been so badly damaged that they could no longer tell who had painted it. Hope's summary of that arithmetic is one sentence long: no consensus, one might think.

In July 2011 Robert Simon publicised a full Leonardo attribution, claiming general acceptance by scholars including Bambach, Marani, Fiorio, Brown and Kemp. Bambach had rejected it. That is the gap between what happened in the room and what was said about it afterwards, and it is the gap on which everything downstream was built.

The exhibition itself, Leonardo da Vinci: Painter at the Court of Milan, ran from 9 November 2011 to 5 February 2012 under Luke Syson, and showed the Salvator Mundi as one of nine paintings attributed to Leonardo. The Gallery's press release said the picture would be presented as the work of Leonardo. Syson's catalogue entry described it as an autograph work and did not mention other scholars' suggestions of assistant involvement. No reservation about the attribution was expressed anywhere in the catalogue. The absence of a qualifier was a decision, not an oversight.

What that decision did is the general lesson. A loan exhibition at a national museum is not formally an authentication and every museum will say so. In the market it functions as one. The picture entered the exhibition as a recently rediscovered panel with a restoration history and a dealer consortium behind it, and left as a Leonardo that had hung in Trafalgar Square. Every subsequent valuation rests on that.

The sale question is where the record is sharpest and least comfortable. In August 2011 Robert Simon told The Art Newspaper that the painting was not on the market and that he hoped it would remain publicly accessible in a museum; the Gallery said it had accepted the loan after confirming that the work was not about to appear for sale. In the same month it was being offered at USD 200 million. Ben Lewis has reported that the dealer Michael Franses attempted to sell it on Simon's behalf in 2009 and 2010, assembling two dossiers, one on restoration and one on provenance. The Dallas Museum of Art asked to borrow it with a view to purchase in July 2012 and made a formal offer, which the owners declined. In January 2020 the Gallery's position was that the owners told us that the painting was not presently for sale. Note the words that are doing the work in both institutional statements: about to appear for sale, and presently for sale. Neither excludes private marketing before or after the show.

The named dissent is worth setting out precisely, because press coverage tends to reduce it to a vague murmur, and because several of the dissenters are the same people who were in the room in 2008. Bambach placed most of the original surface with Boltraffio as early as 2012, allowing Leonardo the blessing hand, portions of the sleeve, the left hand and the orb, and in her 2019 Yale volumes described the picture as a Boltraffio painted under Leonardo's close supervision around 1505 to 1510. At the first international conference on the picture, held in Leipzig from 13 to 15 October 2022 and organised by Frank Zoellner, Fiorio attributed the work largely to Boltraffio and Marani said that Leonardo was involved but that the work was executed largely by assistants. Zoellner argued there for rethinking the concept of authorship itself. Matthew Landrus proposed in 2018 that Bernardino Luini might have worked on it before Leonardo left Milan in 1513; the percentage figures widely attached to his name are stronger than anything in his published article and should not be quoted. Charles Hope and Jacques Franck reject it outright. Against that stand Kemp, Simon, Modestini and the National Gallery's 2011 catalogue. Of the five scholars who saw it in 2008, three have since published positions that give the execution substantially to the workshop.

Before the auction there were two private sales, four days of history that produced years of litigation. In April 2013 Sotheby's private sales department approached the three owners on behalf of a client whose identity they were not told. The picture went to Yves Bouvier, the Geneva freeport operator and dealer, at a figure reported at USD 75 to 80 million to the sellers; Bouvier's total outlay, once Sotheby's commission is added, was about USD 83 million. He resold it to Dmitry Rybolovlev for USD 127.5 million, reportedly the following day. The markup pleaded in the New York complaint is USD 44,945,000, and that court figure should be preferred to the rounder numbers derived by subtraction in secondary accounts.

Two separate proceedings came out of that and they are constantly confused. The dispute between Rybolovlev and Bouvier, which ran through Monaco, Singapore and Geneva, ended in a confidential global settlement in December 2023; the Monaco charges had been dismissed in December 2019 and the Geneva investigation was closed. The action against Sotheby's did not settle. Accent Delight International against Sotheby's, number 1:18-cv-09011 in the Southern District of New York before Judge Jesse Furman, went to trial from 9 to 30 January 2024, and on 30 January the jury found for Sotheby's on all counts. Any account that says Rybolovlev settled with Sotheby's is wrong.

Christie's sold the picture in New York on 15 November 2017 as lot 9B, with an estimate given in the region of USD 100 million. The hammer price was USD 400 million. The price including the buyer's premium was USD 450,312,500. Those are two different numbers and the difference between them, USD 50,312,500, is Christie's commission, not part of what the bidder bid. The arithmetic closes exactly against the house's published 2017 schedule of twenty-five per cent to USD 250,000, twenty per cent to USD 4 million and twelve and a half per cent above: 62,500 plus 750,000 plus 49,500,000. It is worth knowing that the top tier is twelve and a half and not twelve, because the wrong rate is a common error and produces a total nearly USD 2 million short.

The placement was a decision about buyers. An early sixteenth-century panel was offered in a Post-War and Contemporary evening sale rather than in an Old Master sale, after a pre-sale tour through Hong Kong, San Francisco and London. Christie's engaged the advertising agency Droga5, which built the campaign around the phrase the last da Vinci, with graphics that recalled the marketing of a 2006 feature film. The 160-page catalogue asserted an unusually uniform scholarly consensus that the painting is an autograph work. Scott Reyburn has made the sharpest observation about that document: Christie's standard graded-attribution terms, the vocabulary of attributed to, studio of and follower of that the house applies to every other Old Master it sells, were absent from the catalogue.

The lot carried a third-party irrevocable bid, meaning a floor had been agreed and the risk laid off before the sale; the guarantor has been reported as the Taiwanese collector Pierre Chen. An irrevocable bid changes what a saleroom result means, because part of the demand in the room was contractual. The bidding ran between eighteen and twenty minutes, sources differ; it opened at either USD 70 million or 75 million, sources differ again; past USD 200 million it came down to two telephones held by Francois de Poortere of Old Master Paintings and Alex Rotter of Post-War and Contemporary, increments were cut from 10 million to 5 and then to 2, and it ended on a jump from USD 370 million to 400 million. The underbidder is not documented. Any name attached to that position is speculation and this record does not carry one.

The buyer is documented in one respect and unresolved in another. On 6 December 2017 the New York Times identified the bidder as Prince Bader bin Abdullah bin Mohammed bin Farhan al-Saud, who had lodged a USD 100 million deposit in order to bid and who was appointed Saudi Arabia's first minister of culture the following June. On 7 December the Wall Street Journal and the Washington Post reported that the actual buyer was Crown Prince Mohammed bin Salman, acting through a proxy, on unnamed sources. On 8 December Louvre Abu Dhabi confirmed that it had acquired the work. Those two statements have never been publicly reconciled. A debut at Louvre Abu Dhabi was announced for 18 September 2018 and postponed by a tweet on 3 September with no reason given. The picture has not been publicly shown since 15 November 2017.

The result stands as the highest price ever paid for a work of art at auction. It should be read alongside the fact that the same object had changed hands privately twice within the preceding five years at very much lower figures, that it had sold at Sotheby's in 1958 as a Boltraffio for GBP 45, and that it has not been seen in public since. A price is a fact about a transaction. It is not a fact about a picture.

The Louvre's Leonardo retrospective, mounted for the five-hundredth anniversary of his death and running from 24 October 2019 to 24 February 2020, was the largest gathering of his work ever attempted. The Salvator Mundi was not in it. It was absent from the press preview and from the inauguration, and a month before the opening the loan had not been approved.

What makes the absence more than a loan negotiation that failed is the book. The museum prepared a forty-six-page study, Leonard de Vinci: Le Salvator Mundi, by the chief curator of paintings Vincent Delieuvin with Myriam Eveno and Elisabeth Ravaud of the Centre de recherche et de restauration des musees de France, published by Hazan at eight euros, in December 2019. It appeared briefly in the Louvre bookshop, a small number were sold, and it was withdrawn. When journalists asked, the Louvre first said that no such publication exists. It later described the book as a project in case the Louvre got the chance to present the painting. That first denial is on the record and the museum's own book contradicts it.

The book's conclusions are the reverse of what is usually reported. Jean-Luc Martinez, then the Louvre's president, writes in the preface that the results of the historical and scientific study allow us to confirm the attribution of the work to Leonardo da Vinci. The C2RMF text states that the examination seems to us to demonstrate that the painting was indeed executed by Leonardo. Delieuvin concludes that the elements examined invite us to privilege the idea of a work that is entirely autograph. Dianne Modestini obtained a physical copy by chance and shared scans with journalists; the first photograph of the book was published in November 2022.

Against that stands Antoine Vitkine's documentary The Savior for Sale, screened on France 5 on 13 April 2021, which reported, on two anonymous government officials, that the Louvre's examination concluded Leonardo made only a contribution and that a Leonardo and studio label was indicated; that Saudi Arabia demanded the picture hang beside the Mona Lisa and be presented as wholly Leonardo; and that President Macron declined those conditions. The Louvre would not comment on the film and said only that its scientific content, in publications and exhibitions, is entirely independent.

Both accounts of the same laboratory study cannot be right, and this is the sharpest documented-against-reported divergence in the whole subject. The printed book says entirely autograph. The documentary's officials say a contribution only. There is a further twist that belongs beside them: Delieuvin's chapter in the exhibition catalogue that actually was published in 2019 attributes no version of the Salvator Mundi solely to Leonardo. The same curator, in the same year, wrote the withheld booklet and the published chapter.

Set the institutional positions side by side and the picture is genuinely strange. The National Gallery exhibited it as Leonardo in 2011 with no qualifier. The Prado, in the catalogue of its 2021 exhibition on the copy of the Mona Lisa curated by Ana Gonzalez Mozo, placed it not among works by Leonardo but in a second list of attributed works, workshop, or works authorised and supervised by Leonardo, reproduced the Ganay version full page and did not illustrate the USD 450 million painting at all. The Louvre prepared a study asserting autograph status and withdrew it. Three of Europe's principal collections have taken three different positions, one of which was taken and then unpublished.

The object itself is unlocated in the sense that matters to a record like this one. There is no institution that will confirm it holds the picture, no published conservation report since the restoration, and no exhibition. The claim that it was moved to the yacht Serene comes from a single journalist, Kenny Schachter, writing in June 2019 on two anonymous principals, and has been publicly doubted. In August 2024 the BBC reported, through the Princeton professor Bernard Haykel, that it has been in storage in Geneva since the purchase and that the Crown Prince intends it as the anchor object of a Riyadh museum not yet built. In November 2024 AGBI reported two sources naming a museum inside the King Salman Park development in Riyadh, and quoted a ministry of culture official saying it will be at the King Salman Park heritage centre, which they have already started building; the same article states that Saudi Arabia has still not confirmed that it is in possession of the painting, and a park official asked whether it would be displayed said only that they were not sure about that. Hartwig Fischer, formerly of the British Museum, has been appointed founding director of a Museum of World Cultures in that complex.

None of that is a holding statement. For provenance purposes the correct entry after 15 November 2017 is private collection, location not disclosed, and any record that says more than that is guessing.

The most important technical discovery in Leonardo studies this century was made on a picture that is not by Leonardo.

The Prado holds a version of the Mona Lisa, inventory number P000504, long treated as one of the many later copies and displayed against a black background. Conservation work carried out at the museum in 2011 and 2012 removed the black overpaint and revealed a landscape underneath. Infrared reflectography of the panel then showed something no one expected: the underdrawing does not merely reproduce the finished Louvre picture. It reproduces the Louvre picture's successive changes of mind, the pentimenti, stage by stage. Where Leonardo altered the position of a hand or the line of the veil, the Prado panel's underdrawing records the same alteration.

A copyist working from a finished painting cannot do that, because the finished painting does not show him the abandoned first thoughts. The only way to produce that sequence is to work alongside the original as it is being made. The Prado picture is therefore a studio copy, executed in Leonardo's workshop at the same time as the Louvre panel, by a pupil sitting beside him. It is the closest thing that exists to a photograph of Leonardo at work, and it fixes the Mona Lisa's execution to a workshop with assistants present.

It also does something for the wider argument that is easy to miss. It demonstrates, with physical evidence rather than connoisseurship, that Leonardo's workshop routinely produced high-quality versions of his compositions while he was making them. That is the exact production context in which a picture like the Salvator Mundi, or the Madonna Litta, or either Madonna of the Yarnwinder, has to be assessed. The existence of a very good early version is not evidence of forgery and it is not evidence of autograph status. It is evidence of a workshop.

There is no single test that identifies a Leonardo. There is a cluster of habits, and the usefulness of each is different.

Left-handed hatching. Leonardo was left-handed. In his drawings the parallel hatching characteristically runs from upper left to lower right in the direction natural to a left hand, the reverse of the usual right-handed stroke. On a drawing this is genuinely diagnostic, subject to the obvious caution that it can be imitated by anyone who knows about it, and everyone has known about it for a century.

The mirror script. The notebooks are written right to left in mirror writing. This is a fact about the manuscripts rather than an authentication test, but it matters for the study of them: the pages are read in a mirror, and transcription errors in the older editions are common.

Sfumato built in glazes. The characteristic transition without line is achieved by many extremely thin translucent layers rather than by blending wet paint. Analysis of the Louvre pictures has shown glaze layers of a few micrometres, applied in numbers, with lead-based driers and a high proportion of medium. This is expensive, slow, and hard to fake convincingly because it is a sequence of decisions rather than an appearance.

Manipulation of the paint with the hand. Fingerprints and palm-smudging in the paint surface of several accepted works record him working the wet layers with his hands. Fingerprint evidence has been pushed further than it can bear in attribution disputes: a print in a paint layer establishes that somebody touched it, and the comparison of one smudged partial print to another is not the forensic science that its use in the press implies.

Underdrawing and transfer. Infrared reflectography of accepted works shows free, searching underdrawing with changes of mind. Spolvero, the pricked-and-pounced transfer of a cartoon, indicates a design being repeated rather than invented; it is normal workshop practice and its presence points towards a version rather than the primary object. The combination that argues hardest for autograph status is a free underdrawing plus pentimenti in the paint layers plus a technique that no assistant is doing elsewhere.

Support. Poplar for the Florentine panels, walnut for several of the Milanese ones. Walnut is a marker for the Milan workshop rather than for Leonardo personally, which is exactly why it is cited both for and against the Salvator Mundi. Dendrochronology on walnut is far less informative than on Baltic oak, so the support gives a region and a milieu, not a date.

If the paintings are the smallest corpus of any major master, the drawings are among the largest and the most important. They are also the part of the estate whose history can actually be traced.

Leonardo left his papers to Francesco Melzi in the will of 23 April 1519. Melzi held them until his death in 1570. Under his heir Orazio the collection was broken up: material left the Villa Melzi, some of it acquired by the sculptor Pompeo Leoni, who worked at the Spanish court and who mounted large numbers of the sheets into albums. Leoni's albums are the single most consequential intervention in the history of this material, because he trimmed and rearranged what he mounted, and because his albums are the units in which the drawings subsequently travelled. One of them is the volume that came to England and is now the core of the group in the Royal Collection at Windsor, several hundred sheets covering anatomy, hydraulics, botany, architecture, warfare, caricature, drapery, portrait studies and the late deluge drawings. The Royal Collection is candid that the precise route by which the volume entered royal ownership in the seventeenth century is not documented.

The two chalk and ink drapery studies at Windsor, RCIN 912524 and 912525, for the raised sleeve and the crossed stole of a Salvator Mundi, are the strongest evidence that Leonardo designed the composition. At the Leipzig conference in 2022 they were described as the only elements of the disputed panel that can be traced directly back to Leonardo. Christie's cited Kenneth Clark's entry under number 12524, in both the 1935 and the 1968-69 editions, in the printed literature for the 2017 lot. So the picture's strongest documentary support is an entry in a drawings catalogue.

What that support proves is narrower than it is usually made to sound. It proves that Leonardo designed the drapery. It cannot establish which of the surviving panels is the autograph execution of that design, because a workshop assistant working from Leonardo's own cartoon would reproduce exactly the same drapery. That is not a hypothetical. Infrared examination of the de Ganay version, exhibited at Maastricht in 2026, revealed a perforated preparatory cardboard used to transfer the drawing by pouncing, which is direct physical evidence that at least one version of this composition was produced mechanically from a workshop pattern. The Windsor sheets are neutral between the competing positions, and the distinction between designing a picture and painting one is the whole argument in a sentence.

The drawings market has its own lesson about how thinly this material was once held. Head of a bear, the silverpoint that made GBP 8,857,500 in 2021, was owned by Sir Thomas Lawrence and then by the dealer Samuel Woodburn, and sold at Christie's between 4 and 8 June 1860 as part of lot 1039, with a second drawing now in Edinburgh, for two pounds five shillings the pair. It then went through Colnaghi to Captain Norman Robert Colville by 1936, stayed in his family trust, and was bought privately through the London dealer Johnny van Haeften in 2008 by the consignor of the 2021 sale. Nothing about the object changed in a hundred and sixty years. Everything about what it was worth did.

The anatomical sheets are the part of the corpus with the strongest claim to being scientifically ahead of their time, and they are also the clearest case of work that had no influence because it was not published. The dissections were real, conducted on human cadavers in Florence, Milan and Rome, and Leonardo's own note records the number of bodies. The drawings of the heart, of the foetus in the womb, of the muscles of the shoulder and of the vessels are better than anything printed for two centuries after them. They sat in a private album. Vesalius published in 1543 and changed anatomy. Leonardo did not publish and changed nothing until the sheets began to be reproduced in the nineteenth century.

The manuscripts are the largest single body of material Leonardo left and the least understood by the general reader, who tends to imagine notebooks and to be surprised by the scale. Something over seven thousand pages survive. Scholars who have worked on the dispersal estimate that this is a minority of what existed, with figures of a quarter to a third commonly given; the estimate is an estimate and should be cited as one.

The principal survivals are these. The Codex Atlanticus in the Biblioteca Ambrosiana in Milan, the largest single collection, assembled by Pompeo Leoni out of sheets he cut and remounted, which is why its arrangement bears no relation to Leonardo's. The Codex Arundel in the British Library. The Codex Trivulzianus in Milan. The manuscripts lettered A to M at the Institut de France in Paris, which were taken to France after the Napoleonic seizures in Milan and which suffered a second theft in the 1840s at the hands of Guglielmo Libri, a mathematician and manuscript thief who removed leaves and whole volumes and sold them abroad; some material returned, some did not. The Codex on the Flight of Birds in Turin. The two Madrid codices, which had been in the Biblioteca Nacional in Madrid throughout and were catalogued under a wrong shelfmark, and were rediscovered in 1965: the most important twentieth-century addition to the corpus, and a lesson about what is actually lost as against what is merely mis-shelved.

The Codex Leicester is the one that belongs in a market record. It is a single gathering of eighteen double-sided sheets, seventy-two pages, written in Florence and Milan in the early sixteenth century, and its subject is water: currents, erosion, obstruction, the movement of rivers, and by extension the fossils found in mountains and the light on the dark part of the moon, which Leonardo correctly explained as sunlight reflected from the earth's oceans. It was acquired in Italy by Thomas Coke, later Earl of Leicester, in the early eighteenth century and remained at Holkham Hall. It was sold at Christie's in London in 1980 to Armand Hammer, who renamed it the Codex Hammer, and again at Christie's in New York in 1994 to Bill Gates, who restored the older name. It is the only major Leonardo manuscript in private hands and the only one that has repeatedly changed hands at public auction, which is why the prices in this record attach to a book rather than to a painting.

The forgery history of this artist has an unusual shape, because the earliest problem was not forgery at all. It was authorised copying.

The sixteenth century. The workshop and the Leonardeschi produced versions of every major late composition. Dozens of Mona Lisas survive from the sixteenth and seventeenth centuries. Nearly forty versions of the Madonna of the Yarnwinder are recorded. For the Salvator Mundi the count usually given is at least thirty copies and variants by pupils and followers, of which two are held to be lifetime. It is worth naming who is counting: the figure of at least thirty comes from Robert Simon, who is the dealer who bought the panel in 2005 and drove the Leonardo attribution, and is therefore an interested party. For the Mona Lisa and the Yarnwinder this record could find no named counting authority at all, which is itself a finding about how loosely these numbers circulate.

The Leda is the extreme case, because no autograph survives at all. Leonardo worked on two compositions between about 1503 and 1510, a kneeling Leda known from three preparatory sheets at Chatsworth, in Rotterdam and at Windsor, and a standing Leda of about 1508 with the swan, two sets of infant twins and broken eggshells. The lost original's last documented sighting is Cassiano dal Pozzo, who recorded it at Fontainebleau in 1625 and noted that it was done on three long panels which had split apart, which is very probably how it was lost. Raphael drew the standing composition independently. What survives is a set of painted versions given to other hands: the Wilton House picture to Cesare da Sesto, the Uffizi or Spiridon Leda possibly to Francesco Melzi, the Borghese version to Il Sodoma, further examples by Giampietrino and others. Every painted Leda in the market is therefore a version of something nobody has seen for four hundred years. None of these was made to deceive. All of them have, at some point in five centuries, been offered as originals.

The nineteenth century and the Flora affair. In 1909 Wilhelm von Bode, the most powerful museum director in Germany, bought a wax bust of Flora at a London auction for 185,000 Goldmark for the Kaiser Friedrich Museum in Berlin, as a Leonardo. Within a year the family of the English sculptor Richard Cockle Lucas had come forward to say that Lucas had made it in 1846 after a painting he had been asked to copy. His son Albert Durer Lucas swore a statutory declaration that he had helped, described the method, and said his father had stuffed the hollow interior with debris. Museum staff opened the back of the bust. Inside were a wood fragment, newspaper and other nineteenth-century material, including a letter datable to the 1840s, exactly as the son had described before it was opened. The row was enormous: within two years more than seven hundred and thirty articles had appeared in the German and English press and in France, Italy, Austria and Denmark. Bode did not concede, and displayed the Flora beside other Lucas waxes, which demonstrated the sculptor's habit of making sculpture after historical masterpieces and so undercut his own case.

The question was closed scientifically in 2021. A team led by Ina Reiche published radiocarbon dating of the wax in Scientific Reports. The bust is made largely of spermaceti, whale wax, which required a marine reservoir correction; the authors calibrated on a mixed curve of fifteen per cent atmospheric to eighty-five per cent marine and obtained a combined calibrated range of 1712 to 1950 for the wax samples and 1667 to 1950 for the terrestrial materials inside. Their conclusion is that the bust was not produced during the Renaissance and therefore cannot be attributed to Leonardo, and that its chemical composition matches two documented Lucas reliefs of 1848 to 1850. The Bode Museum now labels it as English, nineteenth century.

The historiographical point is the one worth keeping. The radiocarbon did not discover anything the 1910 investigation had missed. The nineteenth-century newspaper was inside the bust in 1910 and the son's account was on the record. What the 2021 paper defeated was not an absence of evidence but a director's refusal to accept the evidence already in his hands.

Hahn against Duveen. In 1920 Andree Lardoux Hahn, an American, offered a version of La Belle Ferronniere to the Kansas City Art Institute as the autograph Leonardo. Joseph Duveen, the most powerful dealer of the age, told the New York World that it was a copy, working from a photograph and without having seen the picture. Hahn sued. In 1923 the two paintings were set side by side at the Louvre and examined by a panel that included Bernard Berenson and Roger Fry and the directors of several national museums. The case came to trial in New York in 1929 before a jury deliberately empanelled without expertise in Leonardo or in connoisseurship; after about fourteen hours it split nine to three and could not agree. A retrial was ordered and Duveen settled for USD 60,000 plus costs, which were substantial. Harry Hahn published an attack on the whole apparatus of expertise in 1946. The picture is now catalogued as by a follower of Leonardo, probably before 1750, and it sold in that form at Sotheby's in New York on 28 January 2010, lot 181, against an estimate of USD 300,000 to 500,000, for a reported USD 1.5 million. Read that result carefully: the market paid roughly three times the low estimate for a demoted picture whose catalogue note recites its own litigation. Notoriety is a priced provenance attribute.

Hahn against Duveen is the founding case of the modern authentication liability problem, and it is the reason experts and foundations in the United States are so reluctant to put opinions in writing. It is also the reason a page like this one can name who holds which position but cannot expect any of them to sign a certificate.

La Bella Principessa. A profile portrait of a young woman in coloured chalks and ink on vellum sold at Christie's in New York on 30 January 1998, sale 8812, lot 402, as Young Girl in Profile in Renaissance Dress, catalogued German School, early nineteenth century. It was consigned by Jeanne Marchig of Geneva against an estimate of USD 12,000 to 16,000 and bought by the dealer Kate Ganz at a hammer of USD 19,000, USD 21,850 with premium. Ganz sold it to Peter Silverman in 2007 for a reportedly similar sum. In that year Martin Kemp accepted it as Leonardo, identified the sitter as Bianca Sforza, and argued with Pascal Cotte that the sheet had been cut from a vellum Sforziad in Warsaw, presented to Galeazzo Sanseverino on his marriage to Bianca in 1496, on the evidence of stitch holes in the margin.

The field did not follow, and the objections are specific rather than atmospheric. Carmen Bambach and Everett Fahy reject it. Klaus Albrecht Schroder, director of the Albertina, said that nobody is convinced it is a Leonardo. David Ekserdjian suspects a counterfeit. Pietro Marani made the most technical objection: imitators have emulated Leonardo's left-handedness for centuries, and he points to monotonous detail, the pigments used, the firmness of touch and the absence of craquelure. Nicholas Penny kept it out of the National Gallery's 2011 exhibition with the flattest possible statement, that they had not asked to borrow it. Two structural facts sit behind all of that: the sheet has no provenance before the twentieth century, which is extraordinary for a Leonardo, and vellum, which Leonardo used only once, is easily obtained in old sheets by a forger.

A fingerprint claimed by Peter Paul Biro as highly comparable to one on the Saint Jerome was advanced in support and has collapsed; forensic examiners found it insufficiently detailed and Biro himself conceded he might have been mistaken. In 2015 the convicted British forger Shaun Greenhalgh claimed in his memoir, reported by Waldemar Januszczak in The Sunday Times, that he drew it in 1978, modelling the sitter on a supermarket checkout assistant in Bolton, on vellum taken from a 1587 land deed and mounted on wood from a Victorian school desk. Kemp calls that hugely unlikely and cites chalk analysis pointing to a date before the seventeenth century. Greenhalgh's account also contains errors about his own age at the time and about the model's name. Note the interests on both sides: Januszczak's own imprint published the book he reported on, and Kemp's standing is committed to the attribution. The picture has not returned to public auction. Press valuations of GBP 100 million and upwards are not transactions. The USD 19,000 hammer of 1998 remains the only market fact about this object.

The Isleworth Mona Lisa. A canvas sold out of Montacute House and acquired by Hugh Blaker in 1913, named after his studio in west London, presented ever since as an earlier autograph version of the Mona Lisa and promoted since 2010 by a Geneva-based foundation. Kemp, who is permissive on other pictures, rejects it, and his sentence on the supporting cast is the one to keep: the roll-call of significant contemporary Leonardo specialists who openly and unequivocally supported the attribution in public was, he wrote, precisely zero. He gives it to the Leonardeschi and criticises the routine treatment of the hair against the Louvre picture's characteristic rivulet pattern. Zoellner excludes it. Alessandro Vezzosi and Carlo Pedretti, both present at the 2012 unveiling, made encouraging but noncommittal remarks that have since been reported as endorsement; they did not endorse Leonardo. The support side claims image-processing matches and a carbon date consistent with 1503 to 1506, and Jean-Pierre Isbouts has claimed that twenty-four of twenty-seven recognised Leonardo scholars agree; that claim and Kemp's precisely zero cannot both be true, and this record prints both rather than choosing.

The support argument matters more than it looks. Leonardo painted on panel. Defenders cite his experimentalism and his notes on preparing canvas. The rebuttal that holds is Jehane Ragai's: the accepted Leonardo works now on canvas, the Benois Madonna, the Louvre Virgin of the Rocks and the Madonna of the Yarnwinder, were transferred from panel. They were not painted on canvas. There is no accepted Leonardo painted on canvas for the Isleworth picture to sit beside.

The theft of 1911. Vincenzo Peruggia, a Louvre employee who had helped build the picture's glass case, entered the museum at about seven in the morning on Monday 21 August 1911, the weekly closing day, in a white workman's smock, lifted the panel off its four iron pegs in the Salon Carre, stripped the case and frame, wrapped it in the smock and, finding a door locked, was let out by a plumber who took him for staff. He kept it in a Paris apartment for two years, then took it to Florence and approached the dealer Alfredo Geri under an alias, expecting a reward. He was arrested on 12 December 1913. The Italian court credited his patriotic motive and gave him a year and fifteen days, reduced on appeal to the seven months he had served.

The story that a confidence man named Eduardo de Valfierno commissioned the theft in order to sell six forged copies to collectors who each believed they had the original rests on a single article published in the Saturday Evening Post in 1932 by the journalist Karl Decker, who said he had heard it from Valfierno in 1913 and had withheld it until after his death. There is no external corroboration. The six copies have never been produced or traced. The story is repeated as fact in serious books and it should not be.

The Leonardo market is the smallest of any artist of comparable fame, because almost nothing is available and almost nothing ever will be. Everything in the accepted painting corpus except one panel is in a public collection, and the collections that hold them are the Louvre, the National Gallery, the Uffizi, the Vatican, the National Gallery of Art in Washington, the Hermitage, the Ambrosiana and the National Museum in Krakow. None of those is a seller.

That leaves four kinds of transaction with real evidence behind them.

The painting. No picture catalogued as an autograph Leonardo other than the Salvator Mundi has been sold at public auction in the modern era, and it is the only Leonardo painting remaining in private hands. Christie's New York, 15 November 2017, lot 9B: hammer USD 400 million, USD 450,312,500 with premium. The frequently printed claim that it is the only Leonardo ever sold at auction is self-refuting, because the same panel sold at Sotheby's in London on 25 June 1958 as lot 40, catalogued Boltraffio, for GBP 45, and again in New Orleans in 2005 as an after-Leonardo. The claim has to be worded around the picture's own earlier sales under other names.

The drawings. Two sheets define the drawing market. Horse and Rider, a silverpoint of 120 by 78 millimetres connected with the Adoration of the Magi, sold at Christie's in London on 10 July 2001 against an estimate of GBP 3.5 million for GBP 8,143,750 including premium, about 2.3 times the low estimate and an auction record for any Old Master drawing at the time. Only the fee-inclusive figure is published; the hammer is not, and this record does not reconstruct it. Head of a bear, seven centimetres square, sold at Christie's in London on 8 July 2021 as lot 20 of The Exceptional Sale for GBP 8,857,500 including premium, against an estimate of GBP 8 to 12 million, which means it sold below its low estimate and still set a record. The bear result is the more informative of the two, because it prices a securely attributed, physically tiny, extremely well provenanced object with no attribution risk at all. That is what the market pays when the question of authenticity is closed, and it is roughly one fiftieth of what it paid when the question was open.

The manuscript. The Codex Leicester sold at Christie's in London on 12 December 1980 to Armand Hammer, at a figure reported as GBP 2.2 million and quoted in American accounts as USD 5,126,000, which is the same sum at the 1980 rate and not a second sale. It sold again at Christie's in New York on 11 November 1994, consigned by the Armand Hammer Museum at UCLA, to Bill Gates, at USD 30,802,500 including premium. That result held the auction record for any book or manuscript for more than a quarter of a century. It has since been passed twice: by a first printing of the United States Constitution at USD 43.2 million in November 2021, and, for manuscripts specifically, by the Codex Sassoon at USD 38.1 million with premium in May 2023. Anyone still citing the Codex Leicester as the record is a decade out of date.

The institutional purchases. Ginevra de' Benci was bought by the National Gallery of Art in Washington from the Princely House of Liechtenstein in 1967, in a private transaction, and is the only Leonardo painting in the Americas. The Lady with an Ermine passed to the Polish state in 2016 as part of the purchase of the entire Czartoryski collection for EUR 100 million, a figure far below any market assessment of the collection and structured as a state acquisition of a national patrimony rather than as a sale of individual objects. Both are examples of the rule that a museum acquisition price is frequently not a market price. The Mona Lisa cannot be bought at any figure: it is inalienable French national property, and the insurance valuation attached to it for its American tour in the early 1960s is a curiosity of record-keeping rather than a valuation of a thing that could change hands.

Leonardo listed his own books. The lists are in the notebooks, one of about forty titles from the 1490s and a later one running to well over a hundred, and they are the best evidence anyone has for the contents of a Renaissance artist's mind because they are the artist's own inventory rather than a posthumous one. What they contain is instructive in what it lacks as much as in what it holds: mathematics, surgery and anatomy, natural history, cosmography, agriculture, military engineering, Italian poetry, a Bible, Aesop, Ovid, Livy, Pliny. He read in the vernacular, he taught himself what Latin he had late and imperfectly, and he described himself in a famous phrase as a man without letters.

The relationship with Luca Pacioli is documented and consequential. Pacioli, the Franciscan mathematician, was in Milan at the Sforza court in the 1490s; Leonardo drew the polyhedra for Pacioli's treatise on proportion, De divina proportione, and the two left Milan together in 1499. Those drawings are the only illustrations Leonardo made for another man's printed book. The programme of that treatise is Platonising and theological, and Leonardo executed its plates. That is a documented intellectual association and it is as far as the evidence goes.

The Vitruvian figure belongs in the same frame and is routinely misread. The drawing in the Gallerie dell'Accademia in Venice is an illustration of a passage in Vitruvius on the proportions of the human body and its inscription in the circle and the square; it belongs to a Milanese circle of architects and theorists working on that passage in the 1480s and 1490s, and Leonardo's version is the one that solved the geometric problem by shifting the centres. The proposition it embodies, that the human body's proportions repeat the proportions of the world, was ordinary Renaissance doctrine and was held to be a theological statement about the order of creation as much as an architectural one. Reading it as a private emblem of hidden knowledge puts a nineteenth-century idea into a fifteenth-century sheet.

The religious commissions were religious commissions. The Virgin of the Rocks was ordered by a confraternity devoted to the Immaculate Conception, for a chapel, under a contract that specified the subject; the doctrinal argument about the Immaculate Conception was live and Franciscan and Dominican theologians were fighting over it, which is exactly why a confraternity existed to promote it. The Last Supper is a refectory painting, put where Dominican friars ate, representing the meal they were themselves re-enacting, at the moment of the announcement of the betrayal. The Salvator Mundi is a devotional image of Christ as ruler of the world, an established northern type imported into Italy, made for private prayer. In each case the programme is documented, orthodox and completely public. There is no evidence in the record of a concealed scheme in any of them, and the mass of popular writing that asserts one has produced no document, no inventory entry and no contemporary reading.

Where his mind actually went somewhere unusual is visible without any of that. He dissected human bodies at a time when few did, and wrote down that he had. He argued from the fossils in the Italian hills to a history of the earth that had nothing to do with a single flood, in the Codex Leicester, in plain language, two and a half centuries before geology said the same thing. He explained the ashen light on the unlit part of the moon as sunlight reflected from the earth's water. He observed that the sun does not move. He designed weapons for whoever was paying, wrote that he would not publish a method of staying under water because of the wickedness of men, and left the anatomical work unprinted. The intellectual programme is not hidden. It is in the notebooks, and it went nowhere in his lifetime because he did not publish it.

An artist page that pretends the arguments are settled is worth nothing. These are the questions a specialist would still put, and none of them has a consensus answer.

The Salvator Mundi's authorship remains genuinely unresolved and the physical evidence that would move it is not available, because the picture is not available. So long as no institution holds it, no further technical examination can be published and no scholar can revise a position on new data. The 2017 price has therefore frozen the argument rather than settling it.

The relationship of the two Virgin of the Rocks to the 1483 contract and to the litigation is better understood than it was but is not closed; the question of what the confraternity actually received and when still turns on documents that are read differently by different scholars.

The Madonna Litta, the Portrait of a Musician and the two Madonnas of the Yarnwinder each sit on the line between Leonardo and a named pupil, and each is likely to move at least once more in the next generation of scholarship. So is the Bacchus.

La Bella Principessa will either enter the corpus or leave the conversation, and the mechanism by which that happens will be interesting in itself, because it will happen without an authenticating body.

Above all, the Leonardo field has no procedure. Every other artist at this level has, or has had, some body with a published method: a research project, a committee, a catalogue raisonne with editions and revisions. Leonardo has monographs and museum labels. That has worked tolerably while the corpus was static and everything was in public collections. It did not work when a picture came out of a regional saleroom with four hundred million dollars attached to it.