Recently it has been claimed that as early as 1420 some European artists constructed their paintings by optically projecting images onto their supports (canvas, oak panel, etc.) and then tracing or painting over these projections. Because projected images obey the laws of perspective, a powerful test of this claim centers on analyzing the geometric accuracy of key Renaissance paintings. This paper investigates new techniques for analyzing the perspective accuracy of paintings. Notably, we focus on a portion of a painting central to the debate of the theory: the chandelier in Jan van Eyck’s Portrait of Arnolfini and his wife. Despite the high level of visual realism of the painting, the technique proposed here highlights large geometric inaccuracies that are very hard to explain as arising from the optical projection route. The contribution of this paper is two fold: i) we present a projective geometry-based technique for detecting and measuring geometric inaccuracies in paintings, and ii) we demonstrate that in the Arnolfini portrait the source of those inaccuracies lies in the imaging process, as opposed to the manufacturing of the actual chandelier. The results presented in this paper cast serious doubts on the validity of the claim that optical tools were employed in painting the Arnolfini portrait.