Abstract. The visual-impedance hypothesis postulates that relational expressions which elicit visual images without a spatial component impede reasoning (Knauff and Johnson-Laird, in press). The goal of the present article is to summarize some experimental findings that support this hypothesis. Previous studies yielded four sorts of relations: (1) visuo-spatial relations, such as "abovebelow", that are easy to envisage visually and spatially, (2) visual relations, such as "cleaner-dirtier" that are easy to envisage visually but hard to envisage spatially, (3) spatial relations, such as "ancestor of-descendant of", that are hard to envisage visually but easy to envisage spatially and (4) control relations, such as "better-worse", that are hard to envisage either visually or spatially. Two behavioral studies showed that visual relations slow down reasoning in comparison with control relations, whereas visuo-spatial and spatial relations yield inferences comparable to those of control relations. The results of an fMRI study showed that in the absence of any correlated visual input (problems were presented acoustically via headphones) reasoning about all four sorts of relations evoked activity in the left middle temporal gyrus, in the right superior parietal cortex, and bilaterally in the precuneus. However, only the visual relations also activated areas of the primary visual cortex corresponding to Brodmann's area 18 (V2). The findings corroborate the theory that individuals rely on mental models for deductive reasoning, and that visual imagery irrelevant to reasoning impedes the process.