Mental models theorists (e.g., Johnson-Laird & Byrne, 1991) suggest that syllogism terms are represented in working memory as ‘abstract tokens’. However, the role that working memory sub-systems and prior knowledge play in the representation and processing of such terms is poorly specified. Two experiments are reported in which the representational distinctiveness of syllogism terms was manipulated. In Experiment 1 participants were required to evaluate the logical validity of conclusions for syllogisms whose premises contained visualisable terms (e.g., spotty or hairy), character terms (e.g., friendly or stupid), or nonsense terms (e.g., drenful or furplish). A logic x content interaction was observed, such that the effect of logic was greatest with syllogisms whose premises were visualisable and smallest with syllogisms whose premises contained nonsense terms. In Experiment 2 participants were required to evaluate conclusions for syllogisms containing either phonologically similar terms (e.g., fuds, fods and fids) or dissimilar terms (e.g., harks, paps and fids). Again, a logic x content interaction was observed such that a strong effect of logic arose with the dissimilar content, but not with the similar content. Set within a mental models framework, hypotheses are proposed to explain the importance of phonological and visuo-spatial distinctiveness in syllogistic reasoning.