1. Eleven Cartridges in Breda

Breda has lost Van Goghs before. In March 1886 his mother and his sister Wil gave up the parsonage at Nuenen and moved seventy kilometres west, and they had the pictures he had left behind crated up and put into store with a local remover and carpenter, Adrianus Schrauwen. Vincent wrote to Wil from Arles in June 1888 asking whether anything might be got out of the junk of his still lying in an attic in Breda. Nobody went. The family moved on to Leiden, the crate stayed where it was, and when they finally enquired in 1903 Schrauwen told them he had sold the contents the previous year to a junk dealer named Johannes Couvreur — at least forty paintings, one guilder the lot, about fifty cents. Some of them came back through the trade over the next thirty years. Most did not.

Eleven small flat dark cartridges fanned across a pine table beside a brown-bowled briar pipe lying on a folded blue cloth, painted in thick impasto with a wedge of yellow light across the boards
The eleven cartridges as they came out of the box, and the pipe, which came out of it with them and is the only object in the whole deposit that anybody has tried to date. It is a briar — root burl of a Mediterranean tree heath, the pipe material of choice from the middle of the nineteenth century on — so it dates nothing.

The eleven cartridges came out of the same trade and very nearly the same streets. When a house-clearance firm on the Terheijdenseweg wound up its business in the spring of 2019, the strongroom inventory turned up a deed box lodged in November 1994 under the name Coenders, never claimed, storage paid twenty-five years forward in a single sum. Inside were eleven magneto-optical cartridges of 640 megabytes apiece, a typed index of ninety-one items, and a pipe. No letter, no receipt, no first name to go with the surname on the lid. Magneto-optical media were sold from 1985 into institutions that needed to keep something for decades and be certain of it: legal deposit, hospital imaging, court exhibits. All eleven read cleanly at the first attempt, which is more than most paper of that age manages.

2. What Is Actually On Them

Ninety-one files, and not one of them is a photograph of a painting. Each is an uncompressed TIFF, sixteen bits to the channel, 4,096 by 3,072 pixels, a little over seventy-five megabytes apiece and just under seven gigabytes together, which is why there are eleven cartridges rather than one. There is no film grain, no page curl, no gutter shadow, no lens falloff at the corners and no varnish. What there is, at every magnification the files will bear, is brushwork: ridges, drag marks, the stiff comma stroke, one loaded bristle skipping across the tooth of a weave that is not there.

Seventy-three of the ninety-one correspond to paintings that exist, out of the roughly 2,100 works and 860 oils he left in barely a decade. The Arles bedroom, the fifteen sunflowers, the Saint-Rémy irises, eleven of the self-portraits he made between 1886 and 1889, a great many wheatfields. The remaining eighteen correspond to nothing recorded anywhere, and all eighteen are landscapes in the double-square format he used at Auvers in July 1890 — which is either the honest residue of a lost summer or the single easiest thing in the whole of art history to counterfeit, depending on who is talking.

He was ill for the last eighteen months of it, and in the asylum at Saint-Paul-de-Mausole from May 1889 to May 1890 by his own arrangement. The argument that a man in that state could not have kept ninety-one items in order is made about once a month and has never had anything behind it. He kept his paint accounts in order throughout. The orders he sent Theo to place with the colourman Tasset et L’Hôte in Paris are itemised tube by tube and colour by colour, and he was still making them out inside the asylum: on 5 September 1889 he wants six large tubes of zinc white, two of emerald green, two of cobalt, two small tubes of carmine, one of vermilion and a large tube of crimson lake.

3. The Colour Argument

Everything said in favour of the collection rests on a fact about the paintings rather than about the files. Van Gogh’s pictures are the wrong colours, and have been for over a century. He painted with lake pigments — dyestuff precipitated onto an inert powder — and the reds among them are fugitive. Carmine lake, made from cochineal, and geranium lake, made from eosin, both lose their colour under light, and both were in his boxes constantly from 1888 onward. It matters more in his case than in most, because he built pictures on complementary pairs — the law of simultaneous contrast that Chevreul published in 1839 out of fifteen years at the Gobelins dye works, and that Delacroix had already put to work in paint. Take the red out of one half of a pair and the other half contrasts with nothing.

A bouquet of roses in a tan jug, the near blooms deep pink and the far blooms bleached white, in thick swirling paint
The Washington Roses split down the middle: pink on one side as file 47 has them, bleached on the other as the canvas has them. The National Gallery says the same thing in its own catalogue entry, without the split.

The consequences are specific and published. The bedroom’s walls were cobalt blue with carmine mixed into it, which is to say violet; the carmine has gone and the walls now read blue. The roses in Washington were pink and are white, so that the complementary contrast with the green they were built against no longer happens. Beneath the top blue of the Getty irises is a faded violet made with geranium lake, and passages that now look like mud were a brilliant orange. Even the yellows are not safe: chrome yellow is lead chromate, and where it was made sulphur-rich it browns as the chromate reduces, which is why parts of the sunflowers are the colour of old varnish and were not.

File 22 shows the bedroom with violet walls. File 47’s roses are pink. File 60’s irises sit under a violet that is not there any more. On the numbers the collection agrees with the chemistry to a degree that no reproduction printed before about 2010 does, and that is the entire case for it.

4. Where the Files Get It Too Right

There is a difficulty, and it is the letters. Van Gogh wrote about nine hundred of them and they are all online, complete, annotated and free, in an edition that took fifteen years at the Van Gogh Museum and the Huygens Institute and appeared in 2009. He described the bedroom to Theo before he painted it, in letter 705, listing the colours in order: pale violet walls, red tiles, fresh butter yellow for the bedstead and the chairs, very bright lemon green for the sheet and the pillows, scarlet red for the blanket, green for the window, orange for the dressing table, blue for the basin, lilac for the doors.

File 22 gives that list back item for item, in the letter’s order, and gives nothing that is not on it. Painters depart from their own plans; the departures are how a canvas is dated, attributed and argued over. Here there are none. The file is not a painting of a room, it is a painting of a paragraph.

A red-bearded man in a blue smock against swirling turquoise, holding a flat dark cartridge level on his open palm
File 09, one of the eleven self-portraits, and the only one in which the hand is holding anything. On the canvases it holds a palette or a pipe; here it holds the object the whole collection is trying to account for.

The second difficulty is narrower and worse. A man painting himself in a looking-glass records the reversal: what is on the right of the picture was on the left of his face. He cut his left ear on 23 December 1888, and the bandage accordingly sits on the right of the canvas he painted the following month. Two of the eleven self-portrait files put it on the left. That is not the mistake of a man at a mirror. It is the mistake of somebody copying a picture that had already done the reversing once, and doing it again.

5. The Dates

Provenance is what the collection has instead of a signature, and the provenance is a date field. The claim is that the files carry their own manufacture in them, and in a sense they do: every one of the ninety-one carries a modification stamp, and every one of the ninety-one reads 1 January 1980. That is not a date, it is a floor. A FAT directory entry packs the whole modification date — year, month and day together — into a single sixteen-bit field counted from an epoch of 1 January 1980, and cannot express anything earlier, so a file claiming to be from 1890 arrives flattened against the bottom of the format along with every other file that has ever had its clock wrong.

There is one file in the set that its defenders never put up on a screen. File 88 is a night sky, and for the upper two-thirds of it a night sky is all it is; then the swirl stops and the rest of the picture is a field of small flat squares of unmixed colour, one loaded stroke to a square. The boundary between the two runs dead level from edge to edge. That is not a thing a hand does. It is a thing an instruction does, and it is the first file anybody who has been shown the set asks to see again.

A swirling night sky of thick spirals and haloed stars above a sleeping village, one dark cypress climbing the left edge, low hills behind
File 55, the Saint-Rémy night sky, swirl for swirl and star for star. The canvas it follows was painted in June 1889 from the east window of his room, before sunrise; the village along the bottom of it was never visible from that window and came out of a sketch made on a hillside.

The dates the collection actually stands on are typed on a sheet of paper in the box. Set that against the instrument the trade really uses. There is a list of 364 works in the hand of Andries Bonger, Theo’s brother-in-law, made about 1890 or 1891 and now manuscript b 3055 V/1962 at the Van Gogh Museum; it is transcribed, published and checkable, and it is how the Washington roses are followed from Theo to Johanna van Gogh-Bonger and out to Paul Gallimard for four hundred francs on 9 June 1891. Sixty-eight of the seventy-three matched files have an entry on that list. The five that do not are the five most reproduced paintings in the collection.

6. The Museum Has Been Wrong Before

Nobody at the Van Gogh Museum has called the collection a fake, and the reason is institutional memory rather than charity. In the 1990s it turned down a landscape for want of a signature — bought in 1908 by the Norwegian industrialist Christian Nicolai Mustad, and shut in an attic for the rest of Mustad’s life after the French ambassador to Sweden, a guest in the house, advised that it probably was not by Van Gogh. The file was reopened in 2011, and on 9 September 2013 the museum announced that Sunset at Montmajour was genuine after all. What decided it was a letter of 5 July 1888 in which he describes a stony heath, small twisted oaks and a ruin on the hill behind.

About a dozen sunflowers crowded in a tan earthenware jug against a pale yellow ground, some heads burning chrome yellow and others scattered among them gone olive and brown
File 31. Half the blooms are painted in the browned state and half in the fresh — one canvas, one afternoon, two chemistries a century apart.

The list is longer than that. The Tokyo sunflowers, bought at Christie’s on 30 March 1987 for a record $39.9 million, spent fifteen years being called a Schuffenecker forgery before the museum’s own research settled them as genuine in 2002. A self-portrait in Oslo was unanimously rejected by scholars for decades and was confirmed authentic in 2020. An institution with that history does not hurry.

It is also not squeamish about objects that are neither paintings nor fakes. Since 2013 the museum has sold Relievos — three-dimensional reproductions made with Fujifilm that copy the front of the painting and the back of it too, and come framed, each one approved by a curator of the museum, at £22,000 apiece. The question of what counts as an authentic Van Gogh object has been a retail question for over a decade, and the museum got there before the forgers did.

7. What the Collection Proves

He died at about half past one in the morning on 29 July 1890, in a room over the Auberge Ravoux at Auvers-sur-Oise, two days after being shot in the chest and five weeks after the roses caught up with him from Saint-Rémy. He was thirty-seven. Theo was there.

One tall dark cypress at the right against a swirling starry sky, a single low farm building with one small lit window below it, a wheatfield running across the whole foreground
Painted from the index rather than from any file — the Coenders sheet gives the last eighteen items as views near Auvers at night. The cypresses are the objection. They are a Mediterranean tree, they belong to the year before at Saint-Rémy, and they are not what stands in a field north of Paris.

Fifteen months before that, in April 1889, he wrote to Theo that paintings fade like flowers, and that this was all the more reason to use the colours boldly and too raw, since time would only soften them too much. Whoever put the eleven cartridges in the deed box had read that sentence — it is the sentence the whole collection is built to answer, and it is the sentence that sinks it.

A file does not soften. Nothing on those cartridges will ever be the wrong colour, and nothing that will never be the wrong colour was painted by him.