Art forgery is older than the modern art market. The Roman elite collected copies of Greek statues, medieval pilgrims bought relics of doubtful provenance, and the nineteenth century saw the rise of industrial-scale imitation of fashionable old masters. What has changed is the sophistication of detection. Modern forgery analysis combines art historical judgement with an expanding toolkit of chemical, physical and computational techniques, and it reveals forgeries that would, as recently as half a century ago, have passed for authentic with confidence.

The starting point for most authentication is connoisseurship: the trained eye of a specialist with deep knowledge of a particular artist's style. Connoisseurs look at brushwork, composition, the treatment of particular subjects, and the way figures occupy space. A skilled copy may reproduce surface appearance but frequently fails to capture the habits of thought that made the original distinctive. Connoisseurship, however, has historically been vulnerable to confirmation bias: once a specialist has declared a work authentic, subsequent examiners often find reasons to agree, and reputations can invest heavily in particular attributions.

Material analysis provides a more objective layer. Pigments, binders, canvases and wooden supports all change over time in measurable ways, and many of them were introduced or discontinued at known dates. A painting purportedly from the seventeenth century that contains a pigment first synthesised in 1810 is almost certainly a forgery, regardless of its visual quality. Titanium white, for example, only entered artistic use in the twentieth century; its detection in a supposedly earlier work is a strong signal. A variety of spectroscopic methods, some of which can be applied non-destructively by shining light onto a small area of the painting, reveal the chemical composition of pigments without requiring samples.

X-ray imaging can reveal layers beneath the surface. Many artists reworked their paintings, and the pattern of under-drawings and earlier compositions is characteristic of how a particular artist worked. Forgers who copy a finished image rarely produce convincing under-layers, because they do not know what should be there. Art conservator Dr. Henri Valdes, who has worked on Dutch paintings at the Rijksmuseum, has described X-ray analysis as 'the forgery's most dangerous witness', because even sophisticated forgers often fail to anticipate it.

Dendrochronology - the analysis of tree-ring patterns in wooden panels - can date the wood on which a painting was executed. The technique cannot prove that a panel was painted in any particular decade, but it can establish an earliest possible date and a likely geographic origin of the wood. A panel felled in the 1670s cannot carry a genuine 1620 painting, though it might carry a later copy of one.

Stable isotope analysis and radiocarbon dating extend the physical toolkit. Radiocarbon can date organic materials such as canvas fibres and wood directly, with useful precision over the past few thousand years. This allows paintings on material purportedly from long-ago periods to be tested. Stable isotopes can sometimes trace materials to geographic regions, an occasional help in establishing provenance.

The computational side has grown rapidly. Digital analysis of brushstrokes, pioneered by several research groups in the 2000s, looks for quantitative signatures of individual artists: the characteristic length, pressure and direction of brushmarks. Machine-learning models trained on many authenticated works can sometimes detect subtle deviations in a suspect painting. Dr. Valdes has been cautious about over-reliance on such tools, warning that they can be fooled by copies made specifically to pass them. His preference is for methods that integrate computational output with material analysis and traditional expertise rather than substituting one for the other.

Despite this battery of techniques, forgeries continue to appear, and some become famous only after long periods in major collections. The Dutch painter Han van Meegeren sold forged Vermeers during the 1940s, including one to the German politician Hermann Goering; only after the war, when van Meegeren confessed to avoid a collaboration charge, did the scale of his operation become clear. More recently, the Knoedler Gallery in New York closed in 2011 after selling a series of forged works attributed to twentieth-century American artists; the forger, working in a garage in Queens, produced paintings that fooled leading specialists and sold for millions before the fraud was exposed.

Dr. Valdes has argued that the lesson of these cases is not that forgery will always succeed, but that it is an evolving practice that requires detection to keep pace. Each new technique eventually becomes known to forgers, who adjust accordingly. The process is therefore open-ended. Paintings that are confidently authentic today may be questioned tomorrow, and paintings once dismissed as forgeries may be rehabilitated on the basis of new information. For the careful collector, authentication is less a verdict than a provisional report.