Five studies were conducted to investigate the relationship between how people communicate about social events and how representations of these events are stored in memory. It was hypothesized that more distant events in memory would be described with more abstract linguistic predicates, and recent events with more concrete language. The 1st study supported this hypothesis. The 2nd and 3rd experiments demonstrated that abstract predicates used as prompts elicit memories that are significantly more removed in time than concrete predicates. Two final experiments showed that these outcomes are not merely a function of the type of semantic cue but an interaction between memory and preferential predicate use. The findings illustrate a link between memory and communicative behavior of a type that has not been previously studied. The results are discussed in terms of a recent, well-supported model of 2 separate fast-learning and slow-learning memory systems.