This paper examines the retention of team cognition with changes in team membership. Hypotheses are developed from shared cognition and interactive team cognition theories. We report a study of the effects of Short (3-6 weeks) versus Long (10-13 weeks) retention intervals and change (Mixed) versus no change (Intact) in team membership during the interval on shared knowledge, team process, and team performance. The study context was a three-person Unmanned Aerial Vehicle (UAV) simulator. The long retention interval resulted in significantly lower team process scores and, except for the Short-Intact condition, all teams suffered a drop in performance after the break. However, those teams recovered prebreak levels of performance after one UAV mission. The counterintuitive result was that team mixing resulted in significant knowledge and process gains. An exploratory communication analysis indicated that Mixed team communication is longer in duration than Intact team communication, and Long-interval teams communicated more frequently than Short-interval teams. Unlike the Long-interval communication frequency effect, the Mixed team communication duration effect lasted throughout the experiment, suggesting greater interaction experience for Mixed teams. An exploratory mediation analysis indicated that the shared cognition Input-Process-Output framework was a good fit for the Intact team data, but not for the Mixed team data. We conclude that there are team-learning benefits of team mixing and that the interactive team cognition theory accounts better for those benefits than shared cognition theory. (Contains 10 tables and 5 figures.)