The forgers, and how experts tell a real Rembrandt 

The Rembrandt problem is not mainly about criminals. It starts with the workshop. He taught for decades, and pupils were expected to learn by copying his compositions and absorbing his handling. A good pupil's exercise, sold honestly as a studio product in the seventeenth century, can look very like a master's work three hundred years later once the signature has been sharpened by a hopeful dealer. Add deliberate forgery, add nineteenth century collectors who wanted a Rembrandt on the wall, and the number of paintings bearing his name grew far beyond anything one man could have painted.

Cleaning that list up became a scholarly project in its own right. From 1968 a team of Dutch specialists, working as the Rembrandt Research Project, began examining the accepted paintings one by one and publishing their reasoning. Their conclusions were contested, sometimes reversed, and occasionally painful for the museums involved. The accepted total fell sharply, and some famous pictures were reassigned to pupils and followers.

Three kinds of evidence do the work. The first is connoisseurship, which sounds like taste and is closer to pattern recognition: how the paint is loaded, whether shadows are thin and transparent, whether the light obeys one source everywhere, whether the sitter's anatomy holds up under the costume. A copyist reproduces the effect but usually smooths the decisions.

The second is the support. Many of his panels are oak, and the growth rings in oak can be measured and matched against regional tree ring records, so dendrochronology can establish the earliest possible date the panel could have been used. A panel from a tree felled after 1669 settles the question. Canvas weave can also be mapped and matched to other canvases cut from the same bolt, which sometimes links two paintings to the same studio in the same period.

The third is materials. Pigments have histories. A blue or a yellow invented in the eighteenth or nineteenth century in a supposedly seventeenth century paint layer is decisive. X-ray and infrared imaging add another layer, revealing underdrawing, changes of mind and reused canvases. Rembrandt changed his mind constantly. A forger, working from a finished image, has nothing to change.

None of this is a lie detector, and attribution still moves. What it gives you is a way of looking that does not depend on authority. Ask where the light comes from, whether the paint is doing different jobs in different places, and whether the picture contains a decision rather than a copy of one.

That is the whole of Rembrandt's Shadow: a miller's son who learned to hide most of a canvas, sold that skill to a city, lost the money, and kept the skill until the end.