Abstract Objectives Refreshing, or the act of briefly foregrounding recently presented but now perceptually absent representations, has been identified as a possible source of age differences in working memory and episodic memory. We investigated whether the refreshing deficit contributes to the well-known age-related deficit for retrieving nonsemantic associations, but has no impact on existing semantic associations. Method Younger and older adults judged the relatedness of stimulus word pairs (e.g., pink–blue or pink–cop) after repeating or refreshing one of the words. During a later source recognition memory test, participants determined whether each item recognized as old was presented on the left or right (nonsemantic source memory) and presented in a related or unrelated pair (semantic source memory). The data were analyzed using a hierarchical Bayesian implementation of a multinomial model of multidimensional source memory. Results Neither age group exhibited a refreshing benefit to nonsemantic or semantic source memory parameters. There was a large age difference in nonsemantic source memory, but no age difference in semantic source memory. Discussion The study suggests that the nature of the association is most important to episodic memory performance in older age, irrespective of refreshing, such that source memory is unimpaired for semantically meaningful information.