Every Tour Guide Says the Parthenon's Curves Were Built to Trick the Eye. An Oxford Mathematician Measured the 6-Centimeter Bulge and the 18-Millimeter Column Swell and Says the Illusion Is the Illusion.

For two thousand years the story has been repeated in textbooks, on tours of the Acropolis and now by chatbots: the architects of the Parthenon curved its platform upward, swelled its columns in the middle and thickened the ones at the corners so that the temple would look perfectly straight to the human eye. A new study by Alain Goriely, a mathematician at the University of Oxford, concludes that the optical corrections never existed. "The thesis of this paper is that there is insufficient historical evidence and no scientific evidence that the Parthenon's refinements were optical corrections," he writes in Royal Society Open Science, in a paper titled "The Illusion of Illusions."

The refinements themselves are real and were measured in the 19th century by Joseph Hoffer, John Pennethorne and, most influentially, Francis Penrose, whose 1851 survey turned the idea into orthodoxy. The stylobate, the platform on which the columns stand, rises about 6 centimeters toward its center over the temple's 69.5-meter length. Each of the 10.4-meter columns bulges outward by roughly 18 millimeters at its midpoint compared with a straight taper, a feature called entasis. The corner columns are about 4.2 centimeters, or 2%, thicker than their neighbors, and every column leans inward by about 0.4 degrees, so that their axes would meet some 2.2 kilometers above the building.

Goriely's argument is that none of this is big enough to do what it is claimed to do. Drawing on decades of vision research, he notes that a person's ability to detect the curvature of a line corresponds to a visual angle of roughly 120 to 420 arcseconds under realistic conditions, which at a 25-meter viewing distance means a deviation of 1.5 to 5 centimeters. The 6-centimeter rise of the stylobate is at the edge of what anyone could notice, and from the corners of the building it is plainly visible as a curve, the opposite of the claimed effect. The 18-millimeter entasis is far below the threshold from almost any distance. And the supposed illusions being corrected, that a flat line sags in the middle or a straight column looks pinched, have never been demonstrated in any experiment. "Our eyes naturally see straight lines as straight when no surrounding features distort our perception," he writes.

The historical chain is equally thin. The idea traces to the Roman writer Vitruvius, who described Greek building practice about 400 years after the Parthenon was completed and whose accompanying diagram has been lost. Goriely points out that the ancient Greeks had no formal theory of conic sections until Menaechmus, nearly a century after the temple was built, and no science of visual illusions at all, so the notion that Iktinos and Kallikrates designed parabolic corrections for perceptual defects requires them to have possessed knowledge that did not exist. He calls this the "supreme knowledge fallacy," one of five he identifies, alongside the "points-of-view fallacy": most claimed corrections only work from a single spot on the ground, and "it is very hard to imagine that any architect would build a temple with the express idea that its beauty can only be revealed at a few specific locations."

The paper also tests the alternatives. A structural explanation for entasis fails because Pentelic marble crushes at 50 to 100 megapascals, long before a column of these proportions could buckle. A drainage explanation for the curved platform is physically plausible, since a 0.2-degree slope would clear a 1-millimeter film of rainwater in about two hours, but it cannot explain why the architrave above the columns curves as well. Goriely's preferred answer is that the refinements were inherited empirical conventions whose original purpose was already forgotten by the time Vitruvius wrote them down, much like lightning crosses on churches before Franklin.

Goriely also notes that large language models, when asked about the Parthenon, faithfully reproduce the myth. "Rather than critical truth seekers or supreme intelligence, large language models are often just amplifiers of accepted folklore," he writes. His suggestion to readers is simpler: walk around your own city, look hard at the long straight lines of its buildings, and see whether any of them ever appears to sag.