Rembrandt van Rijn

Rembrandt Harmenszoon van Rijn is the most attributed and the most de-attributed painter in Western art. In 1836 the dealer John Smith catalogued 614 paintings as his. By about 1900, after Wilhelm von Bode, Cornelis Hofstede de Groot and Wilhelm Valentiner had finished adding, the corpus stood at more than seven hundred. In 1969 Horst Gerson cut it to 435. The Rembrandt Research Project, founded in 1968 to settle the question scientifically, took it down to roughly 250. Then in 2014 the last surviving member of that project published a final volume counting 340. Not one of those figures was reached by discovering or losing a painting. They are all the same body of objects, sorted differently.

That is the fact a reader should carry into everything else on this page. There is no number of Rembrandts. There is a number per authority per decade, and the authority that produced the strictest number dissolved itself, and the man who produced the most recent one died in 2021 without a successor. A painting can be a workshop product in 1970, a Rembrandt in 2020, and back in dispute the following year, without a brushstroke changing.

What has changed, and changed decisively, is the evidence. Tree rings in a Baltic oak panel can put two pictures in the same felled log. A synchrotron can find a lead layer under the ground of the Night Watch that nobody knew was there. Infrared can show whether the maker was inventing on the panel or following a design already fixed. The connoisseur's eye has not been replaced, but it now has to argue against instruments, and it loses more often than it used to.

The Rembrandt market prices the attribution line, and the gap between a good period picture and an accepted autograph one is measured in orders of magnitude, not percentages. The Adoration of the Kings carried a 17,000 dollar estimate at Christie's in 2021 as a work by an artist associated with Rembrandt and made 992,000. Two years later Sotheby's offered the identical panel as Rembrandt's own; nobody bid in the room and it went to a guarantor at 13,800,000 dollars. The two houses publicly maintained their opposing catalogue entries.

Supply is the second driver. Very little remains in private hands, and the best material now moves between states and institutions rather than through salerooms: the two greatest Rembrandts to change hands this century, the Soolmans and Coppit wedding portraits, went from the Rothschild family to the Louvre and the Rijksmuseum jointly for 160 million euros without an auction, and the last biblical picture of consequence in private hands was withdrawn from a Sotheby's sale and placed privately. The auction record therefore belongs to a portrait sold in 2009 and has not been beaten since.

Prints and drawings are a separate market with separate rules. An etching has to be placed in a state and in a printing before it can be priced, because plates went on being printed commercially until about 1900 and a fourth-state Three Crosses is a different work from a first. Drawings have moved sharply upward: a chalk study made 17.8 million dollars with premium in February 2026, a record for the medium in his hand.

The corpus is an argument, not a count

John Smith's catalogue of 1836 accepted 614 paintings, many of which he knew only from written descriptions or reproductive prints. Bode's eight-volume corpus of 1897 to 1906 rejected a large number of Smith's and added almost as many new ones, fifty to sixty of them in the final volume alone, mostly oil sketches, unfinished works and boldly executed late paintings. His protege Valentiner added more. Hofstede de Groot, who had assisted Bode, then produced his own catalogue in 1916. By the turn of the twentieth century the corpus had passed seven hundred.

The correction came slowly. Abraham Bredius, Jakob Rosenberg and Kurt Bauch rejected attributions made by Bode and Valentiner, but it was Horst Gerson's 1969 revision of Bredius's 1935 catalogue that made the scale of the problem visible: Gerson accepted 435 and threw out pictures that had been counted among the master's major works for a century, including the National Gallery of Art's Descent from the Cross and its 1650 self-portrait, now catalogued as Rembrandt Workshop and as Portrait of Rembrandt respectively.

Arthur Wheelock, who joined the National Gallery of Art in 1973, described what that felt like from inside a museum. The whole world he had been trained in was being questioned seriously. He spent forty-five years there working through it.

The Rembrandt Research Project, 1968 to 2011

The Rembrandt Research Project was organised in 1968 by Josua Bruyn, Bob Haak, Simon H. Levie, Pieter J. J. van Thiel and Ernst van de Wetering, with initial funding from the Netherlands Organization for the Advancement of Pure Research and support from the University of Amsterdam. Its purpose was to study every painting attributed to Rembrandt inside a scientific framework and to resolve the accumulated uncertainty by method rather than by taste.

In its first three volumes, published in 1982, 1986 and 1989 and organised chronologically under Bruyn's guidance, it sorted paintings into three categories. A meant accepted. B meant do not know. C meant rejected. Those are the plain glosses used by Wheelock, who also identifies the fault in them: the scheme has no place at all for a picture made collaboratively in a workshop, which is a description of a very large part of seventeenth-century production.

The RRP applied standards stricter than Gerson's and the corpus was trimmed again, to around 250 paintings. It also did something no previous authority had done, publishing a great deal of new technical information on supports, grounds, paint build-up and brushwork, and re-reading the contemporary documents. Its findings were criticised in print from the beginning, in the Burlington Magazine in 1983 and afterwards by Egbert Haverkamp-Begemann, Walter Liedtke, Peter Schatborn, Mansfield Kirby Talley and others. The recurring charge was that the project had assumed too narrow a range of style and technique for Rembrandt, and had assumed that his work developed in a logical progression. His etchings and drawings show him altering style and technique radically, sometimes inside a single image.

There was also a premise buried in the first three volumes that the later work abandoned. Bruyn held that Rembrandt did not use assistants the way Rubens and Van Dyck did. His own formulation is quoted in the National Gallery of Art's catalogue: with Rembrandt design and execution were closely bound up, and instead of using workshop procedures that could replace part of the master's hand, he allowed invention and execution to be separated only in the early stages of an assistant's activities. If that is true, then every hand that is not Rembrandt's is a different painter and can be subtracted. If it is false, the arithmetic does not work.

1993, and the letter that ended the categories

In 1993 the project broke. Bruyn, Haak, Levie and van Thiel wrote to the Burlington Magazine, in volume 135 at page 279, to say that they were no longer involved with the Rembrandt Research Project. They stated that it would in future be headed by Ernst van de Wetering, and that he intended to eliminate the A, B and C categorisation from subsequent volumes.

He did. Van de Wetering became chairman in 1993 and directed the research alone thereafter. The remaining volumes were reorganised thematically instead of chronologically: volume IV on the self-portraits in 2005, volume V on the small-scale history paintings in 2010 or 2011 depending on the imprint, and volume VI in 2014. The A/B/C system was formally dropped in volume IV. In its place van de Wetering put a different question, quoted from the introduction to that volume: does the work betray a genesis which would suggest that the maker was also the person who developed the conception of the work, and if so, can it be the work of Rembrandt himself or a pupil or an assistant, or was it executed by several people. That is a question about how a picture came into being rather than about whose hand held the brush, and it is not answerable with three letters.

The board voted to terminate the project by the end of 2011. Volume VI, Rembrandt's Paintings Revisited: A Complete Survey, was presented by van de Wetering at the Rijksmuseum in Amsterdam on 8 October 2014, and its publication marks the end of the Rembrandt Research Project. He was by then its last remaining member.

His count was 340 autograph paintings. Christian Tumpel had counted 280 in 1986. The sixty pictures van de Wetering restored were not discoveries; they were works that earlier scholars had rejected in the course of purifying the oeuvre, and his position was that the arguments for those purifications were sometimes poor. This record does not reconcile the 340 with the roughly 250 that the National Gallery of Art attributed to the RRP in the same year. They are two positions held by different people at the same moment and both are recorded here as such.

Who rules on attribution now: nobody

When the RRP ceased activities no scholar took over the chair. Ernst van de Wetering died on 11 August 2021. For decades a Rembrandt was, in practice, whatever the RRP and then van de Wetering personally would support; that mechanism no longer exists and nothing has replaced it. There is no board, no committee and no certificate.

What exists instead is documentation. The Rembrandt Database, run by the RKD with more than forty partner institutions including the Rijksmuseum, the Mauritshuis, the Metropolitan Museum of Art, the Louvre and the National Gallery in London, publishes the underlying material for roughly seven hundred works and several thousand research documents: technical reports, radiographs, condition records, provenance files. It publishes evidence. It does not adjudicate.

The consequence is that individual museums now make and unmake attributions on their own authority, and they do. The Mauritshuis returned Saul and David to Rembrandt in 2015 after eight years of research and restoration, reversing a rejection that had stood since 1969. The Allentown Art Museum in Pennsylvania restored its Portrait of a Young Woman of 1632 to Rembrandt after a 2018 to 2019 conservation campaign that removed overpaint in the costume and degraded varnish; the RRP had demoted it to workshop around 1970. The Gemaldegalerie in Berlin announced on 31 March 2022 that Landscape with Arched Bridge, written off by the RRP in 1989 and given to Govert Flinck, is by Rembrandt after all, on imaging that did not exist in 1989. Each of those is a museum overruling the closed authority in its own collection, and each of them is defensible on its own terms. There is no longer anywhere to appeal.

The Polish Rider and the Man with the Golden Helmet

These two pictures are the standard illustrations of what a de-attribution costs, and they came out differently.

The Polish Rider, about 1655, oil on canvas, 117 by 135 centimetres, was sold by Zdzislaw Tarnowski to Henry Clay Frick in 1910 and is in the Frick Collection. Bode and Bredius had established the attribution; Alfred von Wurzbach's alternative of Aert de Gelder was generally disregarded. Julius Held's article of 1944 accepted Rembrandt's authorship while rejecting the idea that the rider was Polish, reading him instead as an allegorical Christian knight. In 1984 Josua Bruyn, then of the RRP, tentatively proposed that characteristics of Willem Drost, another Rembrandt pupil, could be seen in it. The RRP's 1998 assessment concluded that a hand other than Rembrandt's was involved and that the picture may have been left unfinished and completed by someone else. Simon Schama in 1999 and van de Wetering in 1997 both argued for Rembrandt. Bruyn's proposal remained a minority view and is now generally rejected. The Frick never changed its label, which reads Rembrandt.

The Man with the Golden Helmet, about 1650, oil on canvas, 67.5 by 50.7 centimetres, went the other way. It was one of Bode's own major discoveries; he bought it for the Berlin museum, and its broad brushwork was, for him, the proof of Rembrandt's stylistic independence. It became one of the most reproduced pictures in Germany. A Dutch commission examining doubtful Rembrandts raised doubts in West Berlin in 1984, and in November 1985 the Berlin curator Jan Kelch announced the de-attribution: important details of style did not match Rembrandt's known work, and the picture was most likely by a pupil working around 1650. Kelch was careful about what he was and was not saying. It is not a fake, he said, and it remains a great masterful work. The Gemaldegalerie now catalogues it as Circle of Rembrandt. Hofstede de Groot had recorded in 1915 the tradition that the sitter was Rembrandt's brother Adriaen, and dismissed it himself: Adriaen was a poor shoemaker in Leiden who died in 1652, while the same model appears in pictures dated 1654.

The pattern is worth naming. The one that survived is a picture whose attribution rests on long acceptance and a museum willing to defend it. The one that fell is a picture that Bode himself brought in, and Bode is the point of failure in a great deal of this material.

The Mauritshuis, which counted its own losses in public

Most museums do not publish their demotions. In 2019 the Mauritshuis did, hanging all eighteen paintings it had ever acquired as Rembrandts together and labelling each one with what had happened to it. Eleven are accepted as autograph or have been returned to him. Five are no longer given to him at all. Two carry a question mark.

The five losses are specific and four of them were bought by one man. The self-portrait of a young Rembrandt acquired by Stadtholder William V in the eighteenth century was shown in 1999 to be a studio copy of a self-portrait in the Germanisches Nationalmuseum in Nuremberg. Abraham Bredius, director from 1889 to 1909, bought Study of an Old Woman, long called Rembrandt's Mother, in 1890; it is now thought to be a copy of a lost original. He bought Travellers at Rest in 1894 signed Rembrandt.f.; the signature proved to be false and the picture is given to an imitator. He bought Minerva in London in 1899; it is probably by an apprentice. A Praying Woman shown as an authentic Rembrandt at the 1898 exhibition for Queen Wilhelmina's coronation was de-attributed soon afterwards. The two unresolved pictures, a Tronie of a Man of about 1630 to 1631 and a Study of an Old Man of 1650, also entered under Bredius and were once identified as the painter's father and his brother; neither the attribution nor the identification of the sitter now holds, and the experts do not agree.

The movement runs both ways. Saul and David was one of the most celebrated Rembrandts until its attribution was rejected in 1969. A seven- to eight-year research and conservation project returned it to him in 2015. The technical findings were startling on their own terms: the canvas is made up of fifteen different pieces, two large original sections plus fragments taken from elsewhere, including an old copy after Van Dyck, and a three-dimensional reconstruction showed how much larger the picture had originally been. Van de Wetering also re-attributed the Tronie of an Old Man from the Bredius bequest in the Corpus revisited, a move the museum's own Quentin Buvelot has publicly hesitated over, calling the execution of the fur trim flat and weak. That is what live disagreement looks like when it is printed rather than smoothed.

The panel, and what tree rings can settle

In the Leiden years of the 1620s Rembrandt worked predominantly on oak panels cut from Baltic timber, the same supply that served Jan Lievens and the rest of the Dutch trade. Baltic oak is the best-dated wood in European art: the reference chronologies run to thousands of measured sequences from panels of the fifteenth to the mid-seventeenth century, so a panel can be given a terminus after which it cannot have been painted, and two boards from the same log can be matched to each other.

That last capacity is the one that decides cases. A dendrochronological replication study published in Humanities and Social Sciences Communications on 19 November 2025, led by Marta Dominguez-Delmas, went back to the two panels at the centre of the Mauritshuis and Nuremberg self-portrait problem: the signed early self-portrait in the Germanisches Nationalmuseum and the version in the Mauritshuis long held to be its equal. Peter Klein had measured both in the 1990s. The replication confirmed his dating conclusions while correcting measurement errors in his ring counts, revising the Nuremberg panel to a felling date of 1615 to 1630 and an earliest plausible use of 1617 to 1635 under updated Polish sapwood statistics, and the Mauritshuis panel to a felling after 1610 and use after 1612 to 1615. It also produced an outstanding visual and statistical match between the Nuremberg panel and the Rijksmuseum's Laughing Young Man: same tree.

The independent art-historical work had already reached the answer. Infrared imaging in 1998 revealed an extensive underdrawing beneath the Mauritshuis picture that matched the Nuremberg version, which establishes Nuremberg as the invention and The Hague as the copy, possibly by Gerrit Dou. That is the general principle, and Betsy Wieseman of the National Gallery of Art states it plainly: in a copy the painter already knows what he is going to do, so imaging rarely shows major changes, whereas an original shows the artist fussing with the position of a head or a hand or deciding a colour is wrong. Underdrawing that does not match the paint above it is invention. Underdrawing that matches exactly is transcription.

The study's own conclusion is about method rather than about Rembrandt: reproducible dendrochronology requires that reference datasets, statistical values and raw tree-ring measurements be reported and kept accessible, so that a result can be checked without touching the painting again. Thirty years passed before anyone checked Klein's numbers.

The ground, the impasto, and the tell

The preparation under the paint is the most diagnostic thing about a Rembrandt because it is the part nobody was imitating. In the early panel work the wood was primed with chalk and animal glue. In the Amsterdam years he used a double ground: a first layer containing red earth pigments, then a second containing lead white. In the 1640s and after he moved to quartz grounds, cheaper, lighter and more flexible than lead white and better suited to very large canvases.

The Night Watch of 1642 turned out to have something under even that. Work published in Science Advances as part of Operation Night Watch, using correlated x-ray fluorescence and ptychographic nano-tomography on a paint sample at the PETRA III synchrotron at DESY in Hamburg, found that Rembrandt impregnated the canvas with a lead-containing substance before the first ground layer went on. No such impregnation has been observed in Rembrandt or in any of his contemporaries. The likeliest reason is practical: the canvas was enormous and destined for a damp outer wall, and ordinary glue sizing would not have held. The work was led by Frederique Broers at the Rijksmuseum with Katrien Keune, Koen Janssens and Florian Meirer, funded through NICAS and the Dutch Research Council.

The finding also explains something that had been visible for years without being understood. Lead soap protrusions, small white globules that break the paint surface, occur in areas of the picture where there is no other lead component to have formed them. The hidden layer is the source.

Above the ground, the late work is built with lead-white impasto so heavy that it reads as relief, applied at times with a palette knife, and the pictures are constructed to resolve only at a distance. The National Gallery in London puts the change bluntly: the earlier pictures have a smooth finish, the later ones are designed to work only from a distance. It was exactly that quality which the romantic connoisseurs of the nineteenth century prized, which is why the demotions of the twentieth century fell hardest on late, broadly painted pictures.

The etchings, and why a state is not a variant

Rembrandt's printed work is the part of the oeuvre where his own decisions survive in sequence. Virtually every plate exists in several states, from minor corrections to genuinely new compositions, and he reworked plates over years. The catalogues disagree on how many prints are his: Gersaint listed 341 in 1751, Bartsch 375 in 1797, Dutuit studied 363, Middleton and Dutuit gave 329, Hind accepted 293. Andre-Charles Coppier rejected roughly 140 of Bartsch's, about a third, as doubtful, forged or wrongly attributed. Bartsch numbers remain the standard reference: an etching is cited as B. followed by a number.

Two plates show what the states are for. The Hundred Guilder Print, about 1649, etching with drypoint, exists in five states before Captain William Baillie reworked the plate in 1775, with perhaps 125 to 225 impressions known. The Three Crosses of 1653 was executed in drypoint, with burin work from the third state on. Rembrandt pulled around sixty impressions from the first three states, then cut the plate down to something close to a new picture. A fourth-state impression is not a later printing of the same image; it is a different work made on the same copper.

The paper is part of the argument too. He began with French papers from Troyes and Angouleme, then used Japanese paper for its warm yellowish tone, fine grey Chinese sheets, coarse buff oatmeal paper, and vellum. Choosing the support was a compositional decision, not a supply accident.

The plates outlived him and kept printing. Clement de Jonghe held seventy-four of them at his death in 1679 and they were auctioned two years later. Claude-Henri Watelet acquired eighty-one in 1767 and reworked several. Pierre-Francois Basan published his Recueil de quatre-vingt-cinq estampes originales in 1786 and it stayed in distribution for over a century; plates were still being commercially printed around 1900. Seventy-nine original coppers were recorded as surviving by the Musee du Petit Palais in 1986, seventy-seven of them now at the North Carolina Museum of Art, with only two lost. An impression on the market therefore has to be placed in a state and in a printing before it can be priced at all.

The workshop, and how the division shows

Rembrandt began taking pupils in Leiden in 1627, aged twenty-one. From 1631 to 1635 he ran Hendrick Uylenburgh's Amsterdam workshop and lodged there, producing an unprecedented number of portraits; in 1634 he married Uylenburgh's cousin Saskia. Roughly fifty names are recorded as his pupils, drawn chiefly from Arnold Houbraken's Groote Schouburgh of 1718 and from RKD documentation. Gerrit Dou, Carel Fabritius, Ferdinand Bol, Govert Flinck, Samuel van Hoogstraten, Nicolaes Maes, Willem Drost and Aert de Gelder are the names that matter most for attribution, because their work is what a rejected Rembrandt usually becomes.

The economics are documented. Houbraken reports Joachim von Sandrart's figure of one hundred guilders a year per pupil, and puts Rembrandt's income from teaching alone at 2,000 to 2,500 guilders a year, a very large sum. Houbraken also describes the physical arrangement: a barn on the Bloemgracht in which the pupils had rooms separated from one another by nothing more than paper or linen partitions, so that each could work from the model without disturbing the others.

Teaching was delegated. Samuel van Hoogstraten, with Rembrandt from about 1642 to 1647, functioned in the last years as a senior pupil with teaching responsibilities, correcting the drawings of Barent Fabritius around 1645 to 1647 in a manner traceable in the sheets themselves. His Inleyding tot de hooge schoole der schilderkonst of 1678 describes the pedagogy he had absorbed: copy paintings to sharpen the eye, correct inside the drawing rather than beside it, and let the master improve a pupil's drawing by sketching on it.

Collaboration is documented on the finished works, not merely inferred. Four etched oriental heads of 1635 after prototypes by Lievens are signed Rembrandt geretuckert, Rembrandt retouched. The Munich Abraham's Sacrifice, a version of the Hermitage painting, is inscribed Rembrandt, verandert. En over geschildert. 1636, changed and painted over. The 1656 inventory itself lists paintings described as geretukeert by Rembrandt, at numbers 25, 27, 28, 33, 120 and 123, and one at number 79 as gemodelt, designed by him. None of those works has been identified. An assistant is generally accepted to have painted the hands in the 1633 Portrait of Johannes Wtenbogaert in the Rijksmuseum, one of his most prestigious commissions. Wheelock is certain that assistants worked on Belshazzar's Feast in the National Gallery, a picture whose sole authorship has never been questioned in print, because the peripheral figures are executed in a range of manners inconsistent with Rembrandt's own.

So the practical test is not whose brush made a passage but whether the picture was conceived by the person who executed it. Underdrawing that departs from the paint, pentimenti, changes of mind, a composition arrived at rather than transferred: those indicate invention. Their absence indicates a copy, whoever signed it. And Rembrandt is documented as signing works made by members of his workshop, which is why so many pictures listed in his own inventory as by his hand appear to be workshop productions.

The insolvency of 1656

In May 1639 Rembrandt bought the house on the Sint Anthonisbreestraat for 13,000 guilders and never came close to paying it off. In the 1650s Amsterdam entered a severe depression, his creditors moved, and in July 1656 he applied successfully for cessio bonorum, a surrender of goods. It is not the modern bankruptcy and it was not disgrace: the debtor ceded his assets to be sold by the municipal authority to satisfy his creditors, who then had no further claim on him, and he avoided prison.

An official of the Amsterdam Chamber of Insolvent Estates, the Desolate Boedelskamer, whose secretary was Frans Bruijningh, spent two days walking the house room by room and completed the inventory on 26 July 1656. It runs to 363 items or groups of items. The house itself was sold at foreclosure by February 1658. His paintings, etching plates and drawings were auctioned in November 1657, and further sales followed.

The restriction that came with insolvency was professional. A bankrupt could not trade as a painter under the rules of the Amsterdam guild. In 1660 Titus and Hendrickje Stoffels set up as art dealers and nominally employed Rembrandt, who received board and lodging; his work was sold through their company. He painted for another nine years inside that arrangement, in a poorer district on the Rozengracht, and the last decade contains a great deal of what is now regarded as his best work.

What was actually in the house

The 1656 inventory is the single best document of what this painter looked at, and it is a document about images rather than about texts. It records more than 130 paintings and about 70 albums of works on paper. The Leiden Collection's summary and the Rembrandt House Museum's study of the inventory agree on its character: ancient sculpture including busts of Roman emperors, Flemish and Italian Renaissance paintings, contemporary Dutch work, Japanese armour and other Asian objects, weapons, textiles and costumes, anatomical models, and natural history and mineral specimens. The graphic collection was the exceptional part, with Lucas van Leyden and Raphael foremost, in quantity and in quality of impression. The November 1657 auction included drawings said to be by Raphael, Mantegna and Giorgione.

What is thin is the library. The collection was unusual in its paintings and truly exceptional in art on paper; it is not the inventory of a reader. This matters for how his subjects should be understood. The evidence that survives shows a man working from pictures, props and models, not from a shelf of learned texts, and any account of his intentions that depends on his having read something has to establish that he did.

One holding is directly traceable into the work. Rembrandt owned an album of Mughal miniatures, and twenty-three drawings survive in which he copied Mughal emperors, princes and courtiers, made around 1656 to 1661. They are his only surviving drawings on expensive Asian paper and the largest group of copies after other works of art in his output. The originals were painted in Mughal imperial workshops between about 1610 and 1655 and reached the Netherlands through the Dutch East India Company at Surat, plausibly through Company men in his own circle. He made them at the moment his collection was being taken from him, and he did not paint from them; they are a painter studying somebody else's system of drawing.

Confession, congregation, and what is not documented

Rembrandt's own religious position is one of the most confidently asserted and least documented things about him, and the honest answer is that the record is thin and partly negative.

What is documented: his father belonged to the Dutch Reformed Church and his mother was Catholic. His children were baptised in Reformed churches and he was buried in the Westerkerk, a Reformed church. In June 1654 the Reformed church council summoned Hendrickje Stoffels three times to answer the charge that she had committed the acts of a whore with Rembrandt the painter; in July she admitted it and was barred from communion. Rembrandt was not summoned. Church discipline reached members, and the fact that the council could act against her and not against him is the strongest single indication that he was not a communicant member of the Reformed congregation.

What is argued rather than documented: that he was a Mennonite. The claim descends from Filippo Baldinucci's life of Rembrandt, written in Florence in the 1680s on the testimony of Bernhardt Keil, a former pupil, and it has been repeated ever since. It is supported by association rather than by any register: his dealer Hendrick Uylenburgh moved in Mennonite circles, and he portrayed the Mennonite preacher Cornelis Claesz Anslo, called to the Amsterdam Waterlander congregation on 14 July 1617, both in an etching of 1641, Bartsch 271, and in a double portrait with Anslo's wife. Portraying a preacher is a commission, not a confession. No membership record has been produced.

The correct finding is therefore a negative one, and it is worth stating plainly rather than filling in. Rembrandt appears in no congregation's membership roll that has been produced by scholarship. He painted and etched more Old Testament subjects than any painter of his century, in a republic whose Reformed churches commissioned no altarpieces at all and whose painters consequently had to sell religious pictures to private buyers rather than to institutions. That is the structural fact behind the subject matter, and it is a market fact as much as a theological one. It also marks the real distance from Catholic Baroque practice: no patron was ordering an image for a chapel, no decree governed what it could show, and the buyer was a household. There is no evidence of a hidden programme in his religious pictures and no reason to look for one.

Menasseh ben Israel, and one inscription that can be traced

Rembrandt lived on the Breestraat in the quarter where Amsterdam's Portuguese Jewish community had settled, and one collaboration is documented rather than inferred. Menasseh ben Israel, 1604 to 1657, founded the first Hebrew printing press in Amsterdam in 1626 and directed a yeshivah. For his Piedra gloriosa of 1655, a book on the stone of Nebuchadnezzar's dream in the Book of Daniel, Rembrandt made four etchings, Bartsch 36. Impressions of all four are in the Rijksmuseum, RP-P-OB-12.298 to 12.301: the image from Nebuchadnezzar's dream struck by the stone, Jacob's ladder, David and Goliath, and Daniel's vision of the four beasts. Those four subjects are the argument of the book, not a decorative selection, and they are the only occasion on which Rembrandt is documented as illustrating another man's theological text.

The second traceable case is Belshazzar's Feast in the National Gallery, about 1635 to 1638, 167.6 by 209.2 centimetres. The gallery's position is that Rembrandt derived the form of the Hebrew inscription from a book by his friend Menasseh ben Israel. The letters are set in vertical columns rather than in the ordinary right-to-left rows, and the reading generally offered is that this is why the Babylonian wise men in the text could not decipher what was in front of them: the writing is legible only if you know how to turn it. One character is mistranscribed, a zayin where a final nun is required, as noted by R. Littman in 1993. This is a specific, checkable transmission from a named book into a named picture, and it is the sort of thing this record wants; it is also as far as the documentation goes.

The traffic in the other direction is a caution. An etching of 1636, Bartsch 269, was catalogued for most of a century as a portrait of Menasseh ben Israel. Cecil Roth proposed the identification in his 1934 biography and withdrew it in his 1970 Encyclopaedia Judaica entry, calling it doubtful. Steven Nadler concluded in 2018 that the scholarly consensus is that it is not a portrait of Menasseh at all. The only secure contemporary likeness of him is an engraving by Salom Italia of 1642. A famous connection between two famous men turned out to rest on nothing more than the assumption that it ought to exist.

What a public dissection was

The Anatomy Lesson of Dr Nicolaes Tulp, oil on canvas, 216.5 by 169.5 centimetres, was Rembrandt's first great Amsterdam commission, painted at twenty-five, and it records an event of 31 January 1632. Understanding it requires knowing what the event was, because it was neither a private medical procedure nor a scene the painter invented.

The Amsterdam Guild of Surgeons, of which Tulp was the official City Anatomist, permitted exactly one public dissection a year, and the body had to be that of an executed criminal. The corpse is Aris Kindt, otherwise Adriaan Adriaanszoon, convicted of armed robbery and hanged the same day. The dissection took place in a theatrical lecture room before an audience of students, colleagues and members of the public who paid to attend and dressed formally to do so; a large open book, possibly Vesalius's De humani corporis fabrica of 1543, lies at the foot of the table.

The occasion was civic and moral, not merely technical. It was rationed to once a year, tied to the scaffold, ticketed and public. The body being opened was a body the city had put to death, and the lesson delivered over it was that the design of the human frame could be read by anyone permitted into the room. The surgeons in the picture are there because they paid to be, each contributing to the commission and the principal figures paying more. It is a group portrait bought by its own sitters, of an annual civic rite, using the corpse of a hanged robber as its subject; anyone who reads it as a study in scientific curiosity has taken half of it.

The self-portraits as a practice

Around a hundred self-portraits have been counted, of which over forty are paintings; modern scholarship gives him more than forty painted self-portraits, thirty-one etchings and a few drawings, and van de Wetering devoted the whole of volume IV of the Corpus in 2005 to them. That is by a wide margin the largest body of self-representation before the nineteenth century, and it invites a reading it cannot always support.

The nineteenth and early twentieth centuries read the series as autobiography, a man photographing his own decline, and collectors bought accordingly: the National Gallery of Art's Rembrandts came largely from the Widener and Mellon collections, formed when Bode's influence was at its height and shaped by a taste for late, broadly painted, personally resonant pictures. Wheelock, who is sympathetic to that tradition, defends it in a qualified form: the career has certainly been over-romanticised, but there are usually germs of truth in the myths.

The counter-position, and the reason volume IV exists, is that a large number of these heads were made as saleable works rather than as private records. The seventeenth-century Dutch market had a category for them, the tronie, a study head sold as a picture in its own right, and a painter who was also his own cheapest and most available model had every commercial reason to produce them. Rembrandt's self-portraits were also demonstration pieces of technique and of expression, made in a workshop where pupils were copying them.

Both readings are held by serious people, and this record does not decide between them. What it will not do is convert either into a statement about what the painter felt. A picture of a face at sixty-three is evidence of a picture of a face at sixty-three.

The nineteenth-century inflation, and who caused it

Between roughly 1870 and 1935 the body of paintings ascribed to Rembrandt approximately doubled while prices for them rose steeply. The two facts are not independent, and the mechanism was a small number of academically trained connoisseurs whose authority a dealer could convert directly into money.

Bode is the central figure. His outlook was formed by the romantic ideal of creative genius; he held that Rembrandt must be studied as a whole and that the coarse, the refined, the elegant and the harsh all belonged to him, which is a principle that admits almost anything. He conceded that not all the works were of equal quality and put the failings down to defects incidental to great original genius, which disposes of the obvious objection by absorbing it. His corpus was published by the Paris dealer Charles Sedelmeyer, and Alfred von Wurzbach criticised him at the time for exactly that association. Valentiner, Bode's protege, introduced a large number of further paintings and later advised Peter A. B. Widener and his son Joseph, whose Rembrandts went to the National Gallery of Art. Hofstede de Groot assisted Bode and then catalogued on his own account. Bredius, who had rejected attributions by Bode and Valentiner and is remembered as a corrective, is also the director who bought four of the five Mauritshuis pictures that were subsequently demoted, including one with a false signature.

Almost none of the paintings involved was a deliberate forgery in the ordinary sense. They were period works by pupils, imitators and copyists, some with signatures added later, absorbed into the oeuvre by men who had a professional interest in a large Rembrandt. That is a more durable problem than a forger, because there is no crime to detect, only a judgement to reverse, and reversing it costs a museum a masterpiece.

What attribution is worth in cash

The Rembrandt market prices the label, not the object, and two recent lots show the size of the gap.

In 2024 an auction house in Thomaston, Maine, found a panel in a client's attic in Camden. Its founder catalogued it as Portrait of a Girl, after Rembrandt, on the strength of his eye and a label on the reverse recording a loan to the Philadelphia Museum of Art in 1970. No technical analysis was done and no outside expert consulted. The estimate was 10,000 to 15,000 dollars. Nine telephone bidders took it to a hammer price of 1,175,000 dollars and a total with premium of 1,410,000. Neither Arthur Wheelock nor Betsy Wieseman would attribute it. Wheelock's assessment was that comparisons are complicated for early Rembrandt portraits precisely because the works one would compare it to are not themselves firmly attributed.

The other case is The Adoration of the Kings. Christie's sold it in 2021 as by an artist associated with Rembrandt, estimate 17,000 dollars, for 992,000. Two years later Sotheby's offered the same picture as by Rembrandt himself. The two houses stood by their opposing catalogue entries in public. In the room nobody bid, and it went to a guarantor at 13,800,000 dollars. Same panel, same paint, three orders of magnitude, and the only thing that changed was the attribution line.

The most recent Rembrandt records show the same structure at the top of the market. A rediscovered early history painting sold at Lempertz in 2014 as Dutch school for 1.5 million euros, was restored over nearly a decade, and came to Sotheby's London on 1 July 2026 as Let the Little Children Come Unto Me, begun by Rembrandt in Leiden around 1627 and finished by another hand some years later. It hammered at 8 million pounds against an 8 to 12 million estimate. In February 2026 a chalk drawing, Young Lion Resting, made 15 million dollars hammer and 17.8 million with premium at Sotheby's New York, a record for a Rembrandt drawing, bought by the Amsterdam dealer Salomon Lilian for a collector.

Against those, the auction record for a painting has stood for seventeen years and belongs to a picture whose sitter nobody could name until 2021.