Connoisseurship, physics envy and the wages of error

On 29 May 1919, during a solar eclipse on the island of Principe off the west coast of Africa, Arthur Eddington observed light from the Hyades star cluster bent by the Sun’s gravitational field. The results of his calculations, announced a few months later, confirmed the accuracy of Einstein’s prediction. Thus the unchallenged reign of Newton’s theory of gravitation finally came to an end. Einstein and Eddington became famous overnight (Eddington later receiving a knighthood and the Order of Merit). What however happened to Newton? Did we disinter his body and bury him in unhallowed land? Did we withdraw his knighthood (probably awarded for his administrative services at the Royal Mint rather than his brilliance as a scientist), or remove his face from the various effigies that commemorate his great achievement? Of course not; the very idea is preposterous. Because in science – as everyone who has read Popper or Kuhn understands – success is nothing more than the deferral of failure until a better explanation comes along. There are no permanent, definitive descriptions of the universe – although there are some spectacularly useful ones.

But I wish the same clarity could be found in the way the courts treat forensic knowledge in different fields. Specifically, the duties and potential liabilities for error required of art experts, dealers and auction houses occasionally go well beyond the limits of what can reasonably be required of their expertise. So often, the attribution of works of art cannot be confirmed by scientific tests or the examination of documents, and rests on a process termed “connoisseurship”, about which it is impossible to write without incurring disapproval from one side or another. There is a perceived élitism and political incorrectness that restricts open discourse.  The connoisseur is often portrayed as a magus, and the subject is excluded as unsuited to the academic curriculum. Those who do write about it either sound more like a medium at a séance, or seek to persuade us that attribution is a reliable discipline like any other branch of science.

What then is the nature of connoisseurship as a form of knowledge, and how precisely does it differ from other fields? To what special forms of cognitive error is it prone? What does the art historian do to arrive at an attribution when there is no evidence to go on other than what is before our eyes? (This incidentally is much the most common position. Only fakes from a later period can be conclusively detected by laboratory tests of materials. Archival research only occasionally produces evidence that was not already known. In what follows I am not discussing the minority of cases amenable to these techniques, which remain part of physics, chemistry or archival science. I also talk principally about pictures, although connoisseurship or parallel skills can play a role in all the visual arts, as well as in music and literature – although questions of attribution arise less frequently in these fields.)

Most people will understand that the process starts with the experience of looking at a great many accepted works by artists of the same period: that is the foundation of expertise. Some will have heard of the methods of connoisseurs such as Giovanni Morelli, who pioneered the accumulation of photographic evidence of the different ways in which artists painted specific details such as hands, ears or draperies (but as Berenson noted, such evidence is most helpful in distinguishing minor artists). Perhaps it is thought that the connoisseur accumulates (whether documented systematically or merely accumulated in the subconscious memory) these data in a sort of multidimensional space in which each artist’s œuvre occupies a finite volume. The object of art historical research is to define those volumes precisely, the assumption being that they are both well-defined and discrete; and then the task of the connoisseur is to locate a new work in this space, reading off the artist’s name from the coordinates.

That isn’t how connoisseurs think of the process. True, minute observation is essential, and to varying degrees individual experts may find cataloguing specific attributes a useful technique for focusing that observation: but as a process for coming up with a name, I don’t think the method I described even plays much of an unconscious role. Most connoisseurs who deliver a confident attribution are able to do so (if at all) almost instantaneously, and the process of discussion of comparisons with other artists’ styles is provided as an after-thought, an attempt to show the working which is in fact anything but, since there isn’t any working. (There are many occasions when I am puzzled by new works; very few of these turn into confident attributions by a diligent examination of my files. And I doubt if I am unusual in that.)

The coordinate-based approach is ineffective for a number of reasons, among them that the œuvre volumes are fuzzy, not hard-edged, and we usually have too few dimensions: two artists’ œuvres may not overlap in a full analysis in n-space, but projected into (n-1)-space may be indistinguishable.

Now I really will sound pretentious if I try to describe the alternative process which I think many connoisseurs probably believe in (even if few articulate the thought): they have a Platonic idea of what a Rembrandt or a Rubens actually is, and they can (or think they can) spot this spirit when it is present in a new work.

You can immediately see why discourse about connoisseurship is so dangerous. But actually this is a process which goes on in other fields without eyelids being batted. If you play the piano, you will know exactly what I mean: when you sight-read a piece (particularly one in an unfamiliar genre or tricky rhythm) for the first time, you may press the right keys in the right order without it making sense, but later everything clicks into place: you “get it” (i.e. the composer’s intention) in a way that a computer cannot (this arguably is what distinguishes the human intellect from artificial intelligence).

Nor is this process absent in more rigorous fields such as mathematics: when Johann Bernoulli was shown Newton’s anonymous solution to a problem he had set, such was the elegance of the proof that the Swiss mathematician was able to recognise the lion by its claw (“tanquam ex ungue leonem”). That assessment may seem to have much in common with a connoisseur’s attribution based on brushtrokes; but this is to miss the point: beauty may well inspire a mathematician to formulate a theorem (G. H. Hardy’s classic Mathematician’s Apology remains well worth reading), but, when it comes to proof, what matters is whether it is right; a pedestrian proof which follows the rules of mathematical logic is preferable to a demonstration which, however elegant, contains an error.

Physics, like mathematics, also has an objective process for rejection – Popper’s falsifiability, although one based on experiment rather than logic; as Lagrange wryly put the distinction from a mathematician’s point of view (Napoléon called him the “haute pyramide des sciences mathématiques”), “Ces astronomes sont singuliers! ils ne veulent pas croire à une théorie quand elle ne s’accorde pas avec leurs observations.”

Other intellectual disciplines do not have this objectivity. Not only connoisseurship, but fields as disparate as economics and psychiatry simply do not have a process for falsification. Only 3% of the conditions listed in the new edition of the Diagnostic and Statistical Manual of Mental Diseases can be verified by chemical tests in the laboratory. But there is a further difference that distinguishes connoisseurship from economics or psychiatry: apart from cases of collaborative works (more prevalent in large painting studios than say with drawing), there usually is a correct attribution (the veil over which we seek to remove), but this “right answer” may not be a well-defined concept in other disciplines. Unfortunately we can never be sure collectively when we have reached this answer: the lightbulb in a single head is invisible to the rest of us, and indistinguishable from self-delusion (except by inference from that expert’s track record). The difficulty is that if we can’t objectify our tests, we can neither explain why we are right, nor be proved wrong. We are Ramanujan without Hardy. The worst insult a physicist can face is Wolfgang Pauli’s famous phrase: “Das ist nicht nur nicht richtig, es ist nicht einmal falsch!” (“not even wrong”).

Against this backdrop, how are art historians treated when, like the Greek poet, they nod? In an increasingly litigious world, with ever larger sums at stake, it is unsurprising that collectors who find that works they have bought no longer correspond to the catalogue descriptions seek recourse at law.

This isn’t the place to attempt to analyse this complex subject from all angles, or to discuss specific cases, but I do want to highlight the worrying tendency in the French courts to hold auctioneers and their experts to “strict liability”, i.e. to make them responsible for the accuracy of the description even where they have not been negligent in forming their view. For the reasons I have set out I regard this as a category confusion: the nature of connoisseurship is not amenable to such false clarity.

The English courts, currently at least, take a rather more pragmatic line, as intelligently summarised by Lord Justice Nourse in Harlingdon & Leinster v Christopher Hull [1989] EWCA Civ 4:

Even if fakes are put on one side, many old master paintings cannot be safely attributed to a particular member of a group of artists, some of whom may still remain obscure. All this is a matter of common knowledge amongst dealers in the art market and, I would expect, amongst all but the most inexperienced or naive of collectors. It means that almost any attribution to a recognised artist, especially a picture whose provenance is unknown, may be arguable.

Nourse concluded that

the astuteness of lawyers ought to be directed towards facilitating, rather than impeding, the efficient working of the market. … The potential arguability of almost any attribution, being part of the common experience of the contracting parties, is part of the factual background against which the effect if any, of an attribution must be judged.

However, such pragmatism in a case of dealers buying from one another (as in Harlingdon) has its limits, and English courts nevertheless seek to provide some recourse for disgruntled purchasers by imposing a duty on auctioneers to carry out appropriate due diligence to support their attributions. In view of my comments above about the “lightbulb” nature of the best connoisseurial attributions, you will understand why I am sceptical about the relevance of this hurdle. It may keep auctioneers awake at night, but it isn’t going to wake many of the “sleepers” that pass through their hands.