The contributions of the Rembrandt Research Project to the study of Rembrandt paintings are countless and invaluable. In particular, the insistence of Ernst van de Wetering that the physical study of paintings be integrated into the practice of connoisseurship has changed the face of the field. However, inconsistencies in the six volumes of its Corpus of Rembrandt paintings leave us in uncertainty concerning its reconstruction of Rembrandt’s oeuvre. Schwartz puts his finger on a possible re-attribution that should be forthcoming, but isn’t.
It was fifty years ago this year, in 1968, that the Dutch government funding agency ZWO (now NWO) honored the grant application of an art-history initiative known as the Rembrandt Research Project (RRP). The stated aim of the project was to review the right to Rembrandtness of the 624 paintings that had been published as authentic works in the catalogues of Abraham Bredius in the 1930s. This proving impractical, the most unlikely candidates were at a given moment left out of consideration. The five members of the RRP took twenty-one years to publish three volumes of their Corpus of Rembrandt paintings, covering the first half of Rembrandt’s career as a painter, up to the Nightwatch of 1642. Of the 330-odd paintings from this period that had been accepted by Bredius, the RRP catalogued 286. They accepted Rembrandt’s authorship of 146 paintings, rejected a full 122, and admitted doubt concerning only twelve.
That was the situation in 1989. In 1991 three of the four living members of the RRP withdrew from the group and ceded control to the youngest member, Ernst van de Wetering. By the time he rounded off the project in 2015, in vol. 6 of the Corpus, he had re-attributed to Rembrandt 44 of the rejects in vols. 1-3 and added some twenty more that had not been included in those volumes, coming to a grand total of about 340. This development looks a lot like the cycle of decline and rise that has characterized Rembrandt catalogues since the first one, by John Smith, in 1836. Without the accompanying legend, here is a graph of this cycle. (For the source, see my article “Rembrandt scholarship after the age of connoisseurship” [1995].)
The low mark at the right is Christian Tümpel’s listing of 1986, which doubts Rembrandt’s authorship of all but 265 paintings. Van de Wetering’s 340 would rise to a level near the second bar from the right, marking the 350 paintings – not all the same ones as van de Wetering’s – that I accepted in my book of 1984.
Ernst van de Wetering sees his contribution not as part of a secular trend but as its end. When he was questioned in the press about the relative value of his “Complete survey” of Rembrandt paintings (vol. 6 of the Corpus) versus its predecessors, he said this:
In the art world judgments concerning paintings are considered opinions, as most of them indeed are. I however do not deal in opinions. My judgments are based on scientifically buttressed reasoning. There is no question that sooner or later people will realize that I am right. (“Mijn gelijk komt vanzelf.”) I’m not very worried about that.
In addition to scientific buttressing, van de Wetering also claims to be backed up by the form of philosophic logic known as Bayesian probability. To my mind, there is a contradiction at work here, between an approach that aims to gauge probabilities (Bayes) and one that asserts certainties (van de Wetering). But van de Wetering means this seriously. In one entry in vol. 6 he applies what he thinks of as the Bayesian method so circumstantially that he opens the possibility of using it on one’s own. That is what I wish to do here.
Rembrandt, Self-portrait, ca. 1630
Panel, 22.2 x 16.6 cm
New York, Metropolitan Museum of Art
This is a painting that was dismissed by the RRP in vol. 1 as an imitation of a Rembrandt self-portrait. In vol. 6 van de Wetering reverses that judgment on the following grounds, quoted verbatim:
- Like almost all of Rembrandt’s paintings on panel, this painting is painted on an oak panel.
- The panel measures 22.6 x 16.6 cm, [about the same size as] a number of early works by Rembrandt and from his studio.
- According to dendrochronological data the panel comes from a tree that was felled earlier than 1612 […]
- As with c. 14 other (mainly small) paintings from Rembrandt’s Leiden period, this panel was used twice.
- The underlying painting […] belongs to a type that occurs several times in Rembrandt’s early work.
- The painting shows a likeness of Rembrandt’s face, which is significant, given that Rembrandt from early on painted, drew and etched his face in front of the mirror.
- It belongs to a common type, the small-scale self-portrait which he first produced – probably for art-lovers – in painted form.
- The sitter casts a shadow below right on a cursorily indicated wall surface, as is found frequently with Rembrandt.
- [Concerns a feature of the X-radiograph that is typical of Rembrandt.]
- The type of undulating contours with occasional kinks of the figure in this painting is found frequently with Rembrandt.
- As in the painting under discussion, Rembrandt’s brushwork can vary in the degree of its precision or casualness […]
- Rembrandt often mixed painting and drawing with the brush in variable ways. One also finds this feature in this painting.
- The painting does not have the character of a copy and does not reflect any known prototype.
- Neutron activation autoradiography shows that the painting must once have been signed RHL (in monogram), i.e. in the way Rembrandt signed his works during the period in which this painting fits stylistically.
Taken together, they [these considerations] constitute a mutually reinforcing, coherent web of arguments converging with high probability on the conclusion that we are dealing with a work by Rembrandt – indeed, with a probability so high that, in the context of what is possible to say about historical objects, it can be taken for a certainty.
On the face of it, this is as transparent as connoisseurship can get. In order to accept it at face value, however, we have to ignore, as van de Wetering does, the judgment that he signed off on in 1982, in vol. 1 of the Corpus:
Are these observations mistaken? Irrelevant? So beside the point that they do not even have to be mentioned, let alone countered? In 2005, van de Wetering published as the sole author vol. 4 of the Corpus, on the self-portraits. There he seconded the judgment in vol. 1, dating the painting in the Metropolitan Museum “well after 1630” (p. 653). There is no reference in vol. 6 to this previous opinion of the author’s.
Because of the great authority claimed by and accorded to the Corpus, I feel we are entitled to be given answers to certain questions. Why between 2005 and 2015 did van de Wetering change his mind? Why does he not discuss the specific deficits in the painting he saw in it in 1982 and 2005? What effect would it have on the probability of this painting being by Rembrandt if these judgments were taken up in the Bayesian equation?
Those questions I leave aside for the moment in order to ask another.
In the 1969 edition of Abraham Bredius, Rembrandt: the complete edition of the paintings, revised by Horst Gerson, the painting in the Metropolitan Museum of Art is illustrated on this double spread showing four Rembrandt self-portraits of about 1630 that all, it must be said, look rather different from each other. The treatment of this constellation by the RRP is a study in inconsistency. All were catalogued in vol. 1 of the Corpus:
Bredius 8 (upper left) “with hesitation” as an A-number (Paintings by Rembrandt)
Bredius 9 (upper right) and 10 (lower left) as C-numbers (Paintings Rembrandt’s authorship of which cannot be accepted), and
Bredius 11 (lower right) as a B (Paintings Rembrandt’s authorship of which cannot be positively accepted or rejected).
The entire RRP alphabet.
In vol. 4, van de Wetering revised the RRP judgment of Bredius 11 in Stockholm, overriding the reservations in vol. 1 and assigning the painting to Rembrandt. In vol. 6, he did the same for Bredius 10 in New York, as we have seen. But what about Bredius 9? That painting, which from the time of its emergence in 1857 until 1982 was always accepted as a Rembrandt self-portrait, is referred to only once after vol. 1, in a “Table of Dendrochronological Data,” where the felling date of the tree from which the panel was sawn is dated between 1572 and 1581.
By all appearances Rembrandt, Self-portrait, signed and dated RL 1630
Panel, 49 x 39 cm
Private collection
Rembrandt, Self-portrait, ca. 1626-30
Etching, 7.1 x 5.9 cm
Amsterdam, Rijksmuseum
Now what if we interrogate Bredius 9 by the same set of criteria applied by van de Wetering to Bredius 10?
It would come out like this, with the differences between Bredius 10 and 9 in italics:
- Like almost all of Rembrandt’s paintings on panel, this painting is painted on an oak panel.
- The panel measures 49 x 39 cm, in the mid-range of the dimensions of early Rembrandt self-portraits.
- According to dendrochronological data the panel comes from a tree that was felled earlier than 1581.
- The X-ray shows one large change in the composition of the painting, a property that is usually taken as proof that the work in question is an original and not a copy.
- One indistinct area of the X-ray seems to belong to a prior use of the panel, as we often find in the early Rembrandt.
- The painting shows a likeness of Rembrandt’s face, which is significant, given that Rembrandt from early on painted, drew and etched his face in front of the mirror.
- It belongs to a common type, the small-scale self-portrait which he first produced – probably for art-lovers – in painted form.
- The background is a cursorily indicated wall surface, as is found frequently with Rembrandt.
- Like other early Rembrandt self-portrait paintings, this painting bears a close resemblance to a contemporaneous self-portrait etching, with small variations. [An added criterium: a strong indication not true of Bredius 10.]
- The type of undulating contours with occasional kinks of the figure in this painting is found frequently with Rembrandt.
- As in the painting under discussion, Rembrandt’s brushwork can vary in the degree of its precision or casualness.
- The painting does not have the character of a copy and does not reflect any known painted prototype.
- With the naked eye, one can read the signature and date RL (in monogram) 1630, i.e. in one of the ways Rembrandt signed and dated his works during the period in which this painting fits stylistically.
True, in vol. 1 the RRP came out with a categorical denial that this painting has anything to do with Rembrandt.
To reduce the opinions on the painting of the best prior experts – including one of the initial members of the RRP, J.G. van Gelder – to mere mention of their names in a subordinate clause is an extreme case of – I hesitate to complete the sentence out of fear of being accused once more of RRP-bashing. In my book on Rembrandt of 1984, I accepted the attributions of the RRP and left the painting out on their authority. A colleague who defended Rembrandt’s authorship through thick and thin is Christopher Wright.
When vol. 1 was published in 1982, the RRP was unaware that the panel dates from the sixteenth century. That fact makes it more difficult to think up a hypothesis for its origin if it is not from Rembrandt’s studio or by him.
In sum, if we are to take seriously the arguments advanced by Ernst van de Wetering for the re-attribution to Rembrandt of the self-portrait in the Metropolitan Museum of Art (Bredius 10), discussion must at the least be re-opened concerning Rembrandt’s possible authorship of Bredius 9.
Springer, the Rembrandt Research Project and the Netherlands Institute for Art History (RKD) have performed a valuable service by putting vols. 1-5 of A corpus of Rembrandt paintings online at The Rembrandt Database.
© Gary Schwartz 2018. Published on the Schwartzlist on 8 May 2018.
Previous installments of “The transparent connoisseur”:
1: Free advice to the Van Gogh Museum
2: More Rembrandt core
3: The thirty-million pound question
4: A Berenson scorecard
My critiques of connoisseurship are not attacks on the practice, which is not only one of the great pleasures of art history but also indispensable for the integrity of the discipline. To preserve that integrity, however, I feel it imperative that we subject the general claims and specific application of connoisseurship to sharp questioning. This is done too seldom.
Responses in the Reply box below (these will be viewed by all visitors to the site), or personally to Gary.Schwartz@xs4all.nl are always appreciated and will be answered.
So will donations.
Your donations help defray the costs of the Schwartzlist and encourage Gary Schwartz to write more columns.
Very persuasive analysis. I wonder whether the near full-on posture of the Bredius 9 portrait suggests that it was not a self portrait but a portrait. It’d have been helpful had Rembrandt a scar or a birthmark somewhere on his face, for this would certainly have distinguished a portrait from a mirror-image self portrait. One could argue that the artist’s self portraits suggest that he has a quid in his left cheek (I am thinking especially of Self Portrait with Beret and Turned-Up Collar in the National Gallery of Art, Washington, D.C.) I am not saying that he chewed tobacco, only that if one were looking obsessively for asymmetry, one might imagine this. Interestingly, the perioral bulge in Bredius 9 appears to be on the right side, suggesting that only one of the images, Self Portrait With Beret and Turned-Up Collar, or Bredius 9 is reversed, and we know that to be the former.
Regarding the use of mirror to paint a self portrait, I must take issue with Harvard neurobiologist Margaret S. Livingstone’s 2004 observations as reported by Sandra Blakeslee, New York Times, September 16, 2004. Livingstone notes that many of Rembrandt’s self portraits show his eyes to be misaligned, a condition known as divergent strabismus (commonly known as crossed eyes) which leads to stereoblindness, or two-dimensional vision which is supposed to offer a subtle advantage to an artist who must transfer a three-dimensional figure to a two-dimensional surface. This is an interesting bit of pathophysiology, except that it is wrong as applied to Rembrandt and other self-portrait artists. Livingstone overlooks a normal physiologic process known as saccadic suppression whereby the eyes will very briefly and instantaneously suppress vision when they move. This prevents us from an otherwise dizzying display of changing images as our eyes dart this way and that. Look in a mirror. Look at your left eye. Do you see it move? No. Look then at your right eye. Can you see it moving? No. Now look at your cell-phone display with the camera focused on your own face. Look at your left eye. Now you can see a very brief saccadic movement of the iris. This is because the camera has a split-second delay. That split second is longer than the very brief moment of saccadic suppression, so you are seeing the eye as it moved a tiny moment ago.
The point of all of this optometric science is to demonstrate that Rembrandt paints the eyes as he sees them. When he looks in a mirror to his right, for example, he will see a right eye, (It becomes the left eye in the painting.) which diverges slightly more than the left eye in order to see itself. The opposite occurs, of course when the mirror is on his left. And so, a self portrait must show slightly divergent eyes. This is just the way the eyes appear to the artist. No divergent strabismus nor stereoblindness need be presumed.
Now, in the case of Bredius 9, Rembrandt’s eyes appear to be perfectly aligned. So perhaps it was not a copy, but someone else was painting that portrait.
I offer these musings only as a physician and a Rembrandt enthusiast who has no authority of his own, but wishes to learn more about this matter.
Many thanks for these good observations. Although I too disagree with Margaret Livingstone (see Schwartzlist 219), I must say that I see a difference in the direction of the sitter’s gaze in this painting. I think that Ernst van de Wetering suggests somewhere that Rembrandt had a particular identifying physical feature. If I find that reference, I’ll place it here. But since there are such big differences between the appearance of the sitter in the drawings, etchings and paintings that are accepted as self-portraits, I do not think that any particular facial mark can be regarded as proof or disproof that a work is a self-portrait.
My only qualifications for evaluating the efforts of the RRP rest on my graduate training in art history and on a 40-year study of a Rembrandt painting in the Louvre (Br. 431) that was first rejected by the RRP in 1986 (C51), then reinstated in 2015 (A86): traditionally titled “Philosophe en méditation,” it was demoted by Ernst van de Wetering to an “Interior with a Window and a Winding Staircase,” but probably represents a scene inspired by the Book of Tobit (and qualifies therefore as a history painting). It was very easy to reject the 1986 “opinion” in the subsequent article I wrote for the Louvre: all it took was critically reading the arguments, which I believe would not have been accepted in an undergraduate paper. Van de Wetering himself criticized the argumentation in vol. 6. It is regrettable that his discussion of the painter’s intention is limited to his own effort at making the young Rembrandt work like an art theorist, dutifully filling out a training program to become a painter, and this still in 1631-32: it’s all about “kamerlicht”! For a man who “does not deal in opinions” this is a very speculative interpretation for a painting with 2-3 figures in a purely imaginary setting. None of the other Rembrandt scholars I have asked had any problems with the Tobit hypothesis, at least as an iconographic clue. In the particular case of this little masterpiece, both the early and the late RRP have done what the Germans would call a “Bärendienst” to Rembrandt scholarship of the future, while English-speakers would be thinking along the lines of a white elephant.
Very good that you bring this up, Jean-Marie. I have been applauding your campaign on behalf of the “philosopher” in the Louvre from the start, and it’s good that van de Wetering has reversed the judgment in vol. 2, even citing your article although he does not give you specific credit in the entry, as he should have. I agree with you about all these “light practice” criteria van de Wetering has been inventing. I find them speculative and unnecessary.
Really must NOT get into this discussion, much as I am tempted but is that frontal etching really by Rembrandt?? Such tight unimaginative strokes on the torso with a single neat contour line????? As for the frontal supposed SP of R, it’s mushy, shapeless, incompetent . . . but I know it only from these not exactly high res. images. I have spent a lot of time looking at Bernini’s portrait drawings, too many of which have been called SPs, and many of which are copies . . . so similar issues, though it’s much easier to dissect black chalk lines on cream paper than the black and dark brown “soup” of R’s paintings – tracks of brush strokes can only be seen properly in front of the original.]
The early comments about eyesight and misplacement of pupils of eyes VERY helpful to me – I thank the author for contributing. [I’ll download his name and comments later so if I need to credit him, I can do so.]
Ann, no one has ever doubted the etching, as far as I know. But that does not mean there isn’t room for doubt. I agree that the entire question of self-portraiture is up for basic questioning. This time around, though, I was only out to test the RRP’s standards of attribution.