Craquelure, Cradles and Contaminated Evidence

Craquelure records construction, and it has been tested

A crack pattern is not part of a painting's design. It develops as the object responds to its environment, mainly to changes in relative humidity, along lines determined by the materials the artist chose and the way the layers were assembled. The connoisseurs' claim that craquelure carries attributional information was tested experimentally at the Hamilton Kerr Institute by Spike Bucklow, with a parallel study of crack geometry by A. J. Varley at the Cambridge University Engineering Department.

The design of the survey explains what the results can and cannot support. More than five hundred crack patterns were photographed from areas of sky or flesh pale enough for the cracks to read clearly. The sampled areas were small, twenty to thirty square centimetres, chosen in homogeneous passages so that no stylistic clue was visible and chosen to avoid cracks with identifiable causes such as impacts or panel splits. The paintings came from four traditions distinguishable by materials and methods: fourteenth and fifteenth-century Italian panels, fifteenth and sixteenth-century Flemish panels, seventeenth-century Dutch canvases and eighteenth-century French canvases. Forty photographs, ten from each group, were sorted by thirty people ranging from specialists with decades of experience to sixth-form students with none.

The responses were consistent individually and collectively, and the differences between patterns were statistically significant. The correlation between crack pattern and art-historical category ran at about 75 per cent, patterns could be used to confirm or refute a proposed category with better than 90 per cent success, and pair-wise discrimination between categories reached 97 per cent. Machine methods on the same data performed less well: a neural network classification achieved 61 per cent, discriminant analysis 82 per cent, and pair-wise averaging 94 per cent.

Bucklow's own caveat is the sentence that should travel with those numbers. Artists do not always work consistently with the same materials and methods, so crack patterns alone are unreliable indicators of authorship. What the pattern indexes is the construction of the object, and construction is a property of a workshop and a period more than of a person.

Drying cracks record a mistake, not the passage of time

The distinction separating a competent condition report from a misleading one is between age cracks and drying cracks, and it turns on mechanism rather than appearance.

Drying cracks form when an oil-rich mixture is applied beneath a leaner one. The more oil a layer contains, the more slowly it dries, so the upper lean layer sets first and becomes inflexible while the layer beneath continues to move, and the upper film splits. They tend to be wide, to appear in dark passages where the medium content was highest, and to be accompanied by wrinkling in thickly loaded whites. They can appear within months. They say nothing whatever about age.

The National Gallery of Art's examination of Girl with a Flute is the cleanest published illustration. Magnified examination showed wide drying cracks in the dark areas and wrinkles in the whites, and the painter scraped the surface to level them before applying the final paint, while making revisions to the arm and the flute. The same report notes that areas of wrinkled paint and drying cracks occur in one of Vermeer's own late works, The Guitar Player at Kenwood House, which is the necessary corrective: a paint defect indicts technique, not authorship, and it indicts it only in that passage.

Insist that a condition report distinguish the two and say which passages carry which. A forger can induce a crack network mechanically, but reproducing the pattern proper to a particular ground and support, in a film that has been shrinking and responding to humidity for three hundred years, is a much harder problem, and that is why the crack survey works at all.

Lead soaps: the chemistry that never stopped

In 1996, preparing to clean Rembrandt's Anatomy Lesson of Dr Nicolaes Tulp with her then colleague Jorgen Wadum at the Mauritshuis, the conservator Petria Noble found the surface riddled with pockmarks. Each crater was a few hundred micrometres across, no wider than a full stop, and together they gave the painting a dull, hazy, gritty surface. A museum report of 1978 had noticed them and guessed at gas bubbles percolating out of the paint.

Noble called the physicist Jaap Boon, and the investigation that followed changed the field's understanding of how oil paint ages. The white material in the craters proved to be lead soaps, salts of lead and long-chain fatty acids formed over centuries as free fatty acids from the medium reacted with lead ions from the ground and paint layers, together with newly formed inorganic lead compounds in an oil-derived network. The aggregates grow, crack the upper paint layer and protrude through it. The same degradation was then recognised in fourteen further seventeenth-century paintings in the Mauritshuis alone, and in works from the sixteenth to the twentieth century, including a soap pustule in Goya's Portrait of Andres del Peral in the National Gallery, London. Where soaps form at the ground, the paint can lift and flake away.

For attribution this matters in a limited way. Metal soap formation is slow chemistry running for decades or centuries inside a sealed film, and it cannot be induced convincingly on a short timescale, which makes it one of the few positive indicators of a paint film's age rather than a mere exclusion. It dates the material, imprecisely, and names nobody. It is also now known to be very common, so its presence is unremarkable while its complete absence from an old picture with a lead-based ground is the observation worth pursuing.

Thinning, cradling and transfer: what the nineteenth century deleted

Everything above assumes you are looking at the object the artist made. Frequently you are not.

Nineteenth-century structural practice on panels was brutal by design. Warped panels were planed to a fraction of their original thickness, humidified to relax the warp, and then invariably fitted with a heavy cradle. Giovanni Secco-Suardo's manual of 1866 recommends hot cinders and sand, hot bricks to prolong the process where necessary, and for severely deformed panels longitudinal grooves cut at intervals of one to two centimetres with wooden strips glued in afterwards. The consequences are visible today: Giulio Romano's Birth of Bacchus in the J. Paul Getty Museum, an oil on panel of 127.3 by 79 centimetres, carries splits and the surface deformation known as the washboard effect, caused by a nineteenth-century thinning and a heavy cradle. The shrinkage that followed can be measured directly, by how far the battens now project beyond the sides of the panel, sometimes as much as 0.5 centimetres on a panel only 90 centimetres wide.

Transfer went further still, and was a Paris speciality. The workshop of Jean-Louis Hacquin, who died in 1783, carried out transfers for the French royal collection; his son Francois-Toussaint, 1756 to 1832, transferred many of the paintings taken from Italy during the Napoleonic period, the official report on Raphael's Madonna di Foligno in 1799 to 1800 being the best-documented case. Raphael's Saint Cecilia, taken to Paris in 1798, was transferred from panel to canvas and has since taken on the surface characteristics of a canvas painting. In Italy the practice remained common until the late 1950s for panels with severe woodworm, flaking or warping.

The evidential cost is total. A transferred painting has no panel to date by tree rings, no joins, no bevels, no tool marks, no maker's mark and no guild brand. A thinned and cradled panel has lost the end grain dendrochronology needs and gained a deformation that will be misread as age. The National Gallery of Art's Girl with the Red Hat states the problem plainly: early interventions obscured the original panel structure and blocked dendrochronology outright.

Lining, and the surface you are actually looking at

Canvas paintings were treated with the same confidence. For most of the nineteenth and twentieth centuries a lining was considered a normal component of any treatment, and new linings were often applied indiscriminately, with little regard for the resulting loss of the painting's original surface texture. The adhesives ranged from flour paste and animal glue to wax resin, applied with hand irons, hot tables and vacuum pressure; the vacuum envelope, developed at the Courtauld, was adopted at the National Maritime Museum as a gentler alternative to intense pressure and high heat.

The turning point was institutional. Westby Percival-Prescott, who led the National Maritime Museum's picture conservation department from 1961 to 1983, conceived the Conference on Comparative Lining Techniques, held at Greenwich on 23, 24 and 25 April 1974 with delegates from twenty-four countries. His opening paper, The Lining Cycle, showed how the techniques of a given period conditioned the way its paintings later responded to the linings applied to them. The conference put the necessity of lining itself in question, and Percival-Prescott urged conservators to avoid relining unless strictly necessary.

What lining does to evidence is specific and predictable. Heat and pressure flatten impasto, so the relief a raking-light photograph records may be a fraction of what the artist left. Wax-resin impregnation saturates the paint, alters colour, and is effectively irreversible in the canvas fibres. Tension is reset to whatever the liner chose. And the tacking margins are frequently trimmed away, removing the original priming edge, the tack holes and the cusping a radiograph would otherwise use to establish that the canvas has not been cut down. Arguing from surface texture, format or cusping on a lined painting means arguing about a restorer's decision, quite possibly taken in the 1860s by someone whose name is not in the file.

Reading a contaminated record

The working rule is to establish the treatment history before arguing from the material evidence, and where the file is silent to read the object for it: cradle battens, added strips, a lining canvas, a wax-saturated back, inpainted losses under ultraviolet.

The obverse of the problem is that conservation evidence can itself be forged. The Cologne study of Beltracchi's Campendonks records that he patinated the tacking margins to give them a dirty appearance and sometimes used newer wood which he artificially aged. It also models the discipline required in response: every examined Campendonk used a keyed stretcher with two keys in each corner and the fake used an unusual type, but since stretchings are frequently renewed and original stretchers may have been replaced, that observation can only support a suspicion of forgery and cannot carry it.

Van Meegeren remains the sharpest illustration of what the record can hide. His phenol-formaldehyde additions hardened the paint film and defeated the only age tests then in use, the alcohol swab and the heated needle, and his version of The Procuress at the Courtauld returned no modern pigments when it was analysed in 2009. It was confirmed as his work only when a later analysis found Bakelite. The same object told two different stories to two competent examinations, and neither was careless. The difference was the choice of instrument, and that choice depended on what the examiner believed the object had already been through.