Analogical transfer occurs when someone transfers insights about shared structure from one task to another. Here, we explore this ability in categorizations tasks. Previous work found that people could transfer knowledge between two unidimensional rule-based tasks, but could not for two information-integration tasks. In four experiments, we extended this to between-modality transfer, whilst also better controlling for several incidental features of the tasks such as the number of relevant stimulus dimensions. We found that similar numbers of participants engaged in between-modality analogical transfer, regardless of categorization task. However, the ability to report the relationship between the two tasks did not result in any performance benefits. As analogical reasoning is often touted as a cognitive short-cut, these findings have interesting implications for learning of tasks that have a more fine-grained measure of performance.