Evaluating visual arguments in science

Abstract This paper explores a specialized application of visual argumentation within the domain of the natural sciences and points out a broader conceptual problem regarding the testability of visual arguments. We highlight the methodological problems of Leo Groarke’s Key-Component method that make it, in its current form, unreliable for use to test images’ argumentative role. The main reason is that visual content is ambiguous and underdetermined, especially in a scientific context. Focusing merely on visual features without the inclusion of verbal implicit premises poses a considerable challenge to reconstructing reliable premise — conclusion structures. We posit the necessity of advancing a more sophisticated framework specifically designed to evaluate visual arguments systematically. We argue that images should be considered in a linked verbal — visual argument system in scientific arguments, in which images and other evidence complement each other and jointly support a conclusion because relying solely on visual evidence results in underdetermined inferences. The present paper exemplifies this issue through an illustrative case study focused on images of the Mars Phoenix lander. The goal of the present study is twofold: to scrutinize the findings of visual argumentation in order to extend its scope to the natural sciences, and to suggest methodological changes to the KC method to make it more reliable.