This paper presents a diachronic study of the imagery found in two scientific journals from Quebec, using a corpus of texts from five periods between 1869 and 1997. Our analysis shows that the frequency of images is linked to three factors: the type of scientific text, the time of publication, and the domain. Corroborating previous studies, we show that the desire to display objectivity – which generally involves less imagery – is manifest beginning in the 1930s and amplifies over time. Our analysis also suggests that the use of imagery implies a functionalist view, as writers tend to use more images to explain new notions to non-specialist readers while avoiding them in texts written for specialists, where they could be construed as undermining the scientific credibility of the work. (Isabelle Collombat and Ginette Demers)