Pictures are the tangible traces of our visual world. They are the materialisation of a classifiable, datable, and localizable object; a vocabulary of expressions that, when examined on a large scale, becomes key to understanding changes in visuality over time and space. This is precisely the goal of Visual Contagions, a project at the University of Geneva that aims to use computational analysis to understand what images circulate and how some of them become iconic. To achieve this goal we have developed a novel multi-scalar analysis that combines computational exploration of visual similarities with spatiotemporal classification. The method is applied to a global and diachronic corpus of illustrated periodicals, using algorithms to compare images extracted from the pages and group them into image-types: Vectors of visually similar images. The components of the image-types are then interconnected using their spatiotemporal properties and examined using a combination of digital and traditional art historical methods. The initial result will lead to a panoramic study of globalisation through images, encompassing both their geography and their structural and conjunctural logics of circulation. This study will help to make clear the nature of the iconic, how it manifests itself and how it circulates. Using the outcome from the distant reading, the identified iconic features will be further explored drawing from traditional historical methods, thus making evident the aesthetic, social and historical reasons for the emergence and diffusion of the iconic. The contribution present the first results of the project, focusing on the analysis of a large global corpus of images (120 countries – 6 millions pictures) from illustrated magazines published between 1890 and 1990, and showcasing what circulated most, when, how, and through what type of media.