(The review first appeared in the Frankfurter Allgemeine Zeitung, May 3, 2006)

Joseph Heller, in Picture this (1988, p. 59): “Rembrandt did some fifty-two self-portraits that have come down to us, and several of these Rembrandts are not by him. It is hard to conceive of self-portraits not executed by the subject, but here they are.”

Ernst van de Wetering, in The Self-Portraits, vol. 4 (2005, p. 89) of A Corpus of Rembrandt Paintings(begun 1968): “There is one category of non-authentic self-portraits whose existence became clear to us only when we had virtually completed this volume of the Corpus. This category of ‘self-portraits’ of Rembrandt was executed by workshop assistants or pupils … The discovery imposes drastic limitations on the effectiveness of the methods, presumed to be objective, that we had initially hoped could introduce order into the group of paintings long taken to be Rembrandt’s self-portraits … Indeed, the fact that the usefulness of these ‘objective’ methods of authentication was undermined by the discovery of the category of ‘self-portraits’ painted by other workshop members is due entirely to our effort to apply those methods as consistently as possible.”

How is it possible that a plain, essential fact, obvious to an attentive non-art historian like Joseph Heller, could have remained invisible to a famous team of Rembrandt specialists until it undermined their methods of authentication? At the end of this review I will answer that question, but first …

In 1968 a group of Dutch art historians submitted to the Netherlands Organization for Scientific and Scholarly Research (NWO, then ZWO) a successful grant proposal for a large-scale study of the paintings of Rembrandt van Rijn. The group called itself the Rembrandt Research Project (RRP) and its publication A Corpus of Rembrandt Paintings. The object of the proposal was a scholarly question of some significance: “What statements can be made concerning the authenticity and autograph nature of the paintings attributed to Rembrandt on the basis of research concerning material structure and stylistic properties, iconological interpretation and documentary facts.” In support of this aim, all documents related to the issue were to be reproduced.[1] The project was to be completed within ten years.

In the event, none of these aims was achieved. Because the scientific evidence in the Corpus concerning the “material structure” of Rembrandt’s paintings consisted almost entirely, where it existed at all, of research conducted by a large assortment of third parties, it was incomplete, mutually incompatible and inconclusive. Iconology played almost no part when it came to the determination of authenticity. The relevant documents were not reproduced or used as evidence for authenticity, and ten years after 1968 the RRP had yet to publish the first of a series of volumes. Nor did the RRP make good on its other stated aims:[2] no uniform criteria for responsible attributions were defined, no group of core works singled out as reference points, no controllable vocabulary established. Instead, the RRP went straight ahead with attributions of its own, in the same strain of stylistic and qualitative analysis that connoisseurs had always employed and that had gotten its predecessors into the very quandary from which the RRP wished to escape.

Yet, when vol. 1 of A Corpus appeared in 1982 (vol. 2 came out in 1986, vol. 3 in 1989), it was greeted with widespread enthusiasm. The idea had taken hold, abetted by the publicity surrounding the publication, that the RRP had succeeded in founding a new connoisseurship, more firmly based on system and science than previous methods. The field seemed inclined to accept that the RRP had made good on the proud ambition proclaimed in 1969 by the project leader, Josua Bruyn. The aim of the Project, he said at a high-power symposium in Chicago, was to frame “a precise definition of our observations and of the standards by which we interpret them. Only thus will our opinions become rational judgments.”[3]

Fast forward to 2005. On p. vi of vol. 4, we read a disclaimer, drafted with the help of two lawyers, stating that “the opinions expressed in this volume (IV), and the previously published volumes I-III in the Series A Corpus of Rembrandt Paintings, should be understood as ‘opinions’ that are meant for academic use only.” And more in this vein. Not only has the RRP toned down its claims for the judgements in vol. 4, it has also disavowed the ideology and degraded the judgements previously published in vols. 1-3.

• In the early 1990s the RRP underwent a change of leadership. Josua Bruyn and three of the other founding members of the group resigned en bloc because they were unable to agree with the fifth man, Ernst van de Wetering, who now took over the Project on his own. Vol. 4 reflects his new way of doing things. The main changes:
• The ideal of connoisseurship by committee, which was intended to reduce the subjectivity of attributions, has been abandoned.
• The much criticized grading system of the RRP has been discarded. In vols. 1-3 A stood for genuine Rembrandt, C for definitely not by Rembrandt, while B meant that the RRP could not make up its mind. (This was the case for only 12 of the 280 paintings with main entries.) The entries in vol. 4 are given the Roman-Arabic numbers IV 1 to IV 29.
• The organization of the Corpus has been restyled. Vols. 1-3 covered chronological periods from 1625 to 1642. Vol. 4 adds only the small number of self-portraits that had not yet been published in the earlier volumes, while offering reappraisals of some already catalogued. Since the RRP intends to deal with all of the remaining 200-odd candidate Rembrandts in a single volume 5, whatever system resided in the division of the volumes has been undone.
• The composition of the Corpus has been transformed. Vols. 1-3 contain entries on 280 paintings with brief prefatory essays and appendixes. Vol. 4 consists mainly of lengthy texts by Van de Wetering (more than 270 pages), Marieke de Winkel on costume, Karin Groen on painting ground and Jaap van der Veen on documentation, followed by entries on 29 paintings. The ratio of essays to entries has increased by a factor of 20. The essays are unfortunately buried in monumental Corpusvolumes costing 1000 euros, where they will be inaccessible to most art historians.
• Van de Wetering pays more attention than did Bruyn et al. to questions of condition and is more receptive to the possibility of collaboration or later intervention. This is very much to the good, and brings the Corpus more into line with current thinking about studio practice and the history of restoration.
• High-quality color reproductions have been added, which contribute greatly to the functionality of the book. Previous volumes were nearly exclusively in black and white, as much a matter of principled mistrust of color reproductions, I assume, as of cost.

What brought Ernst van de Wetering to execute these changes? “Reviewing the three past decades,” he writes in the preface to vol. 4, “it is evident that this project – as with every project attempting to chart a complex phenomenon – is not only a path to resolving the problems involved, but also a learning process.”

From whom, we may ask, has Van de Wetering learned? The answer lies in the footnotes. His prefatory essay contains 47 references to publications that have appeared since 1982. Of these, 29 are to his own writings and another 14 to publications by junior members or associates of the RRP working under him. Only four refer to outside authors. None of these are reviewers of a preceding volume or scholarly critics of the Project. Van de Wetering’s lessons, it seems, come from himself.

Van de Wetering’s lessons to himself are not all bad. A felicitous change is that the terms in which attributions are discussed in vol. 4 are richer, more personal and less categorical than in vols. 1-3. Van de Wetering introduces a form of probability logic named after the British preacher and mathematician Thomas Bayes (1702-1761), who, according to Van de Wetering, “observed that our beliefs are not all-or-nothing convictions based on simple yes-or-no answers to decisive questions, but rather that there are degrees of belief, that one arrives at a conclusion through inductive reasoning using arguments of varying probability” (p. 109). This is a welcome contribution to the theory of connoisseurship, albeit more as a description of what connoisseurship is not rather than what it is. The entries in the earlier Corpus volumes, with their categorical judgements, are distinctly non-Bayesian. They often give one the uncomfortable feeling that the observations and arguments were massaged to match the conclusion. This was demonstratively the case with regard to judgements concerning the authenticity of the signatures on paintings published in vols. 1-3. The handwriting experts who were called in, twenty years into the Project, were visibly shaken to find that the RRP had approved the authenticity of nearly all the signatures on paintings which they accepted, and disputed the authenticity of the signatures on paintings they rejected. “These statistically improbable high scores … are explainable if in a number of cases the researchers had allowed themselves to be guided more by the convincing authenticity of the painting than by the handwriting evidence.”[4] Equally non-Bayesian is the ratio of “simple yes-or-no answers” (95% in vols. 1-3, categories A and C) to “degrees of belief” (the 5% of category B). By rights the ratio should be closer to the reverse.

Van de Wetering’s discussions of authenticity in vol. 4 are more open to the contradictory evidence that nearly always attends Rembrandt problems. It cannot be said, though, that he has banned the massaging of evidence. When Corpus number C56, a painting in Berlin of Rembrandt in a feathered cap, was a rejected Rembrandt, in 1986 (vol. 2), it was seen not only as uncharacteristic and unusual for Rembrandt, but also suffering from “cramped placing of the figure,” displaying “clumsy heaviness in the appearance of the figure,” with a cap and feather “not really effective in creating depth,” in the chain a “hurried manner of painting [that] is far from effective”, a color scheme with “no parallel in any of Rembrandt’s works from these years.” The upper part of the face was “weak [in] execution, … flat and patchy.” The “transitions from the face to the hair are noticeably weak,” while “the somewhat primitive bravura of the brushstroke … does not always help to create clarity in the shape of the head or an effect of depth” and so on (vol. 2, pp. 671-72). Now that Van de Wetering has re-attributed the painting to “Rembrandt and studio,” it has become a “brilliant, broadly painted self-portrait, … perhaps … a demonstration of Rembrandt’s mastery of the ‘rough manner'” (p. 603) and we read of its “brilliantly applied brushstrokes that are left emphatically visible” (p. 216).[5] At the least, Van de Wetering owes us an explanation of what happened to all those detailed criticisms of virtually every aspect of the painting, in the X-rays as well as the surface, from vol. 2. Analysis of that kind – close observation and quality judgements coupled to an attribution – was the central contribution of the RRP to art history. We will not learn the lessons that need to be learned from the case of the Corpus until a proper critique is written of how that detailed, highly negative judgement from 1986 relates to the lyrical one of 2005. Indeed, not only the credibility of the Corpus but of connoisseurship itself is at stake here.

No Rembrandt specialist can consult the Corpus without comparing the opinions of the RRP with his or her own. In my book of 1984, Rembrandt, his Life, his Paintings: A New Biography, I arrived at many of the same conclusions as those in vol. 4 of the Corpus, something that is not always acknowledged there. Concerning Corpusno. IV 11, in the Kunsthistorisches Museum in Vienna, Van de Wetering writes, “The authenticity of this painting was never doubted in the literature until 1986, when Tümpel attributed it to an anonymous follower of Rembrandt.” That is inaccurate. I expressed doubt in my book of 1984, p. 380.

Of the 29 entries in vol. 4, I shared the same opinion concerning Rembrandt’s authorship of 26. With regard to the other three paintings, in the Museo Thyssen-Bornemisza (IV 2), Kassel (IV 9) and Aix-en-Provence (IV 16), whose authorship I doubted in 1984, I am now convinced by the arguments of Van de Wetering and others that they were painted by Rembrandt. Of the seven non-self-portraits in the “Addenda et corrigenda” to vol. 4 that were doubted or rejected in vols. 1-3 and are now accepted by Van de Wetering, I published six as Rembrandts in 1984 (figs. 12-14, 49, 211-12) and expressed a more positive judgement on the seventh (Bredius 633; p. 380) than the RRP. Should one be positively impressed that in the 1980s Van de Wetering put his name to wrongheaded deattributions and has now reversed them without acknowledging that other colleagues did not misguide themselves in the first place?

To return to the question posed above: how could Van de Wetering seriously say that only after years of pursuing a wrong course did he discover an error that was plain for all to see? How could he have imagined that an unstable category such as self-portraiture could provide a basis for “‘objective’ methods of authentication”? The answer is that Van de Wetering is mistaken. His claim that this phenomenon was a “discovery” of the Rembrandt Research Project in the early 2000s is demonstrably inaccurate. In vol. 2 of the Corpus, published in 1986, the portrait of Rembrandt in Berlin discussed above, previously called a self-portrait, was attributed by the RRP to Flinck (C 56). The painting had been examined in 1968 by Ernst van de Wetering and Simon Levie, who presumably drafted the entry on it in the Corpus but in any case co-signed it as authors. That entry includes a forthright acknowledgment of the existence of “self-portraits” painted by other workshop members. Recognition of this category is there said to be “in line with thoughts expressed earlier by [Horst] Gerson [Rembrandt Paintings, 1968, p. 66]: “… since we know portraits of Rembrandt by Flinck, we must surely reckon it possible [a misquotation; Gerson wrote, more strongly, ‘plausible’] that other students painted his likeness.” In vol. 4, Van de Wetering footnotes Gerson (with the wrong page number, 62), but downplays his contribution, as if Gerson had not put his finger on a phenomenon Van de Wetering now claims to have discovered. Van de Wetering does not refer at all in this discussion to Corpus entry C56, with its matter-of-fact elaboration on Gerson’s findings. In other words, the justification Van de Wetering brings forward for the “drastic limitations on the effectiveness of the methods” applied by the RRP to the self-portraits in vol. 4 is based on an inaccurate claim. This heavy methodological and historical statement in Self-portraits misrepresents the record.

The 690 pages of this major publication contain an unimaginably vast wealth of information, much of it new, on “the group of paintings long taken to be Rembrandt’s self-portraits” even while deconstructing that category. Many passages and entries are excellent, indispensible contributions to our knowledge of Rembrandt. Yet – and as a glutton for information, I do not say this readily – I find the Corpus of Rembrandt Paintings oppressively massive. There is no way for a researcher, let alone a reader, to digest it without writing his own précis and making his own indexes as he goes along. The minimal and highly selective indexes to the 3200 pages of vols. 1-4 are pitiably insufficient. Who can possibly remember what is in them? Only future students of the same material will have the motivation to plow through these books for the nuggets they need.[6]

Finding out what is not in them is even more difficult. I cannot be certain, but as far as I could tell this 690-page volume on Rembrandt’s self-portraits lacks mention of a fact that is highly relevant and perhaps essential to our understanding of the subject. When Rembrandt was fifteen years old, at the very time he left school for studio, a remarkable print was published that I believe was germane to his specialty in self-portraiture. It was a reproduction by Andries Jacobsz. Stock of a self-portrait of Rembrandt’s great forebear and model Lucas van Leyden, painted when Lucas was fifteen. The caption says of the painting that it showed the “incomparable painter and engraver, when he was fifteen years old, depicted from a portrait of himself by his own hand.” We learn further that Lucas died in 1533. In 1633 Rembrandt etched a self-portrait that comes close in pose to that of Lucas. It was the first of his self-portraits to be signed Rembrandt only, another emulation of Lucas. Who can doubt that the impressionable and ambitious young Leidener Rembrandt, when he began portraying himself by his own hand, was not inspired by his admired model?

Actually, it would seem that Ernst van de Wetering would doubt it. The omission of the print by Andries Stock might be perfectly intentional. Van de Wetering is staunchly opposed to personal interpretations of self-portraiture before the nineteenth century. He takes the fact that Rembrandt’s self-portraits were ‘commodities’ that were sold to collectors for proof that they were devoid of self-reflection, indeed of any personal meaning.[7] His favored interpretation is that self-portraits came into being to satisfy the desire not of artists but of collectors to have images of artists. He cites print series like the Iconography of Anthony van Dyck and the Medici collection of painted self-portraits: “One would have to conclude that Rembrandt’s activity in this field, taken as a whole, should be seen as the result of ‘external pressure’ rather than the ‘internal pressure’ presumed by [Perry] Chapman.”[8] Van de Wetering is not impressed by the fact that Rembrandt created more self-portraits – far more – than any other artist of the early modern period. Nor does he take account of the great diversity of guises Rembrandt assumes in his self-portraits, images for which no “external pressure” can be held responsible and which would not fit into a gallery of self-portraits. His insistence that no personal or psychological meaning can be attached to self-portraiture in the seventeenth century, that no speculation about individual identity is admissible, is too categorical and too limiting. In any case, it is at odds with the Bayesian approach Van de Wetering claims to apply. For all it has to offer, vol. 4 of the Corpus of Rembrandt Paintings is a letdown when it comes to understanding the relation between Rembrandt Harmensz. van Rijn and his self-portraits.

Gary Schwartz
Maarssen, The Netherlands