Dendrochronology: The Floor Under a Panel's Date

The ring dates the timber, and the timber is always older than the picture

Dendrochronology does not date a painting. It dates the outermost tree ring still present on the end grain of a board, and every link between that ring and the first brushstroke has to be estimated: the sapwood the panel maker cut away, the felling, the shipping, the seasoning, the joining, and the painter's own delay in taking the panel off the shelf. Each estimate is a range, and every range pushes the date later. The honest output of a tree-ring report is an earliest possible date, a terminus post quem, and never a date.

The sapwood estimate does most of the work and carries most of the error. For oak from Poland and the southern Baltic, the statistics published by Tomasz Wazny in 1990 give between 9 and 24 sapwood rings within a 90 per cent confidence interval, with a median of 15. A board whose last surviving ring is 1606 therefore comes from a tree felled no earlier than 1615, most probably around 1621, and possibly as late as 1630. A separate metrical study of oaks from northern Poland recorded a minimum of 9 sapwood rings and a maximum of 36, with half the sample falling between 13 and 19.

Seasoning follows. Peter Klein and colleagues proposed in 1987 that seventeenth-century Netherlandish panels went into use after about two years; Jorgen Wadum argued in 1998 that up to five should be allowed. Fifteenth-century practice was slower, with an average seasoning time of ten to fifteen years, against two to eight years for the sixteenth and seventeenth centuries. On the Melbourne triptych by the Master of the Legend of Saint Catherine, a last ring of 1466 gives an earliest felling date of 1475 and a plausible painting date from 1491 onward. That is a spread of twenty-five years read off a single measured ring.

The practical consequence is that dendrochronology is at its strongest when it says no. A panel cannot have been painted before its tree was cut down. It can very easily have been painted eighty years afterwards.

The Baltic timber trade built the reference curve; an Antwerp guild rule then blunted it

Tree-ring dating works only against a dated reference chronology for the correct region, and for Netherlandish panels that region is not the Low Countries. Straight-grained oak was shipped from Poland and the Baltic states to the Low Countries and England from the middle of the fourteenth century, leaving the ports of Gdansk and the city then called Koenigsberg in a standard ten-foot length of about 280 centimetres. Hamburg's wood scientists began dating panels in the 1970s on the assumption that the oak was locally sourced; only in the 1980s, once Polish reference chronologies existed, was the south-eastern Baltic origin recognised.

The scale of that trade is now measurable. A 2024 study of 294 panel paintings by or connected to Jacques Jordaens and Anthony van Dyck dated roughly 80 per cent of the boards examined and found that over 71 per cent came from the Poland and Baltic region, with another 20 per cent from western Central Europe; three paintings combined Baltic planks with planks from the Ardennes, most likely cut from larger panels. The same study records how panel stock thinned over time: about 2 centimetres for a fifteenth-century panel of up to medium size, about 1 centimetre in the early sixteenth century, and 0.4 to 0.5 centimetres in the seventeenth.

Then comes the irony that every specialist should hold in mind. On 13 November 1617 twenty-two Antwerp panel makers, mirror-case makers and softwood frame makers petitioned for a quality law. The resulting ordinance of 11 December 1617 required joined panels of two or more planks to be submitted to the Dean of the trade for inspection, and decreed that they must contain no sapwood, no mould and no worm. Approved panels were branded with the guild mark, and the maker added his own mark on pain of a fine. The rule that protected the picture destroyed the evidence, because sapwood is exactly the tissue that fixes a felling year. A rule against sapwood guarantees that most panels made under it can yield only a floor.

Two portraits of Rembrandt: what the wood settled, and what it could not touch

In 2025 Marta Dominguez-Delmas published a replication of dendrochronological work Peter Klein had carried out in the 1990s on two versions of the same young Rembrandt. One, at the Germanisches Nationalmuseum in Nuremberg, is signed and dated on style to about 1629. The other, at the Mauritshuis in The Hague, is accepted as a studio copy by one of Rembrandt's pupils. Both are single oak boards. Both lack sapwood.

Klein had dated the last ring of the Nuremberg board to 1606 across 210 rings, and the Mauritshuis board to 1602 across 101 rings. He also reported that the Nuremberg board came from the same tree as the panel of the Rijksmuseum's Laughing Young Man, catalogued as circle of Rembrandt, whose board carries 231 rings including 8 of sapwood and ends in 1614. That match is what converts a floor into a window: if the two boards come from one tree, the outermost ring on the Nuremberg board is the heartwood-sapwood boundary, so the felling can be bracketed rather than merely bounded.

The replication confirmed all of it. Remeasured from digital photography against current Baltic chronologies, the Nuremberg series matched a Polish reference, the same-tree match with the Rijksmuseum board returned an outstanding statistical and visual fit, and the felling was placed between 1615 and 1630, giving an earliest possible production date of 1617 and a likely window of 1617 to 1635. The Mauritshuis board, having no sapwood and no same-tree partner, could only be given a terminus post quem: felled after 1610, painted after 1612.

Two things follow. First, the technique is reproducible. The exercise did surface two measurement errors in Klein's original series, a ring omitted around 1405 and the 1427 ring counted twice, which depressed his statistics without moving his dates. Second, and more important for anyone quoting the report, the wood settled nothing about authorship. Both panels are entirely consistent with the period. The studio copy sits on wood as good as the original's.

Where the wood does decide: a board felled after the painter was dead

The Metropolitan Museum's Christ Appearing to His Mother is the textbook case of tree rings carrying an attribution. Acquired by bequest in 1921, it was long held to be an original by Rogier van der Weyden, with the version in the Gemaeldegalerie, Berlin, dismissed as a workshop copy. In 1991 and 1992 Berlin's curator Rainald Grosshans published the underdrawing of the Berlin panel, which showed Rogier's characteristic hook-ended drawing and several changes of composition. Maryan Ainsworth then found that the Metropolitan's panel carried a mechanical underdrawing made without changes, probably taken from a pattern or a tracing, and Grosshans noted that its perspective had been subtly adjusted towards a one-point scheme belonging to the later fifteenth century. In Ainsworth's own account, what made the redating definitive was the dendrochronologist's report, which dated the panel to around 1485, some twenty years after Rogier's death in 1464.

Note the structure of that argument. The tree-ring date did not demote the picture on its own; the underdrawing had already done the art-historical work. What the wood supplied was a hard boundary that no connoisseurial rebuttal could cross.

The mirror case is just as instructive. When the Museo Nacional del Prado defended its Temptation of Saint Anthony in 2016 against the Bosch Research and Conservation Project, which had reassigned the picture to a follower and dated it 1530 to 1540, the museum cited a dendrochronological date of 1464 for the oak panel and now dates the painting 1510 to 1515. A last ring of 1464 is compatible with a picture of about 1510. It is equally compatible with a picture of 1535. The number excludes an earlier date and nothing else, and citing it in support of a specific attribution asks the method to do work it cannot do.

The failure modes to ask about before quoting a report

Access is the first. Tree rings are read on the end grain, which means the edges of the boards, and edges are exactly what frames, added strips, cradles and old restorations cover. In 2021 a painting attributed to Rubens's studio was examined by X-ray computed tomography at a resolution of 37.3 microns precisely because oak strips covered all four edges. The scan revealed a concealed double panelling. The visible back board is deciduous oak with a terminus post quem after 1557, which refuted the previous reading, while the front board, the one actually painted on, is a tropical wood, probably Swietenia, which no European chronology can date at all. The National Gallery of Art's Girl with the Red Hat is a simpler instance of the same problem: early interventions obscured the original panel structure and blocked dendrochronology outright.

Species is the second constraint. The method rests on oak in northern Europe. Italian panels are usually poplar, and the Metropolitan's own survey of painting supports hedges the dendrochronological contribution with the words when possible. A poplar, walnut or tropical support may simply be undatable.

Ring count and reporting are the third and fourth. Short series do not cross-date reliably, and long ones can still fail: the Nuremberg board's 210 rings returned a Student's t value of only 4.49 against the best Hamburg chronology because of those two measurement errors. In the 1980s a t value above 3.5 over an overlap of 100 rings was treated as a possible match; current practice expects above 5, with a percentage of parallel variation above 63 per cent and a convincing visual fit on the plotted curves. A report that offers a felling date without naming the reference chronologies, the overlap, the statistics and the ring-width series is asking to be taken on trust. Dominguez-Delmas's recommendation is the right standard: institutions and collectors commissioning tree-ring work should require the graphs, the datasets and the digital images of the ring sequences, and should keep them, so that the work can be rechecked without touching the painting again.