This study is inspired by laboratory studies demonstrating that distributing study sessions over time better supports learning and retention than clustering sessions. We compare two implementations of a multi-day inquiry science unit: in the clustered instruction condition, students completed an inquiry unit in five consecutive class periods. In the distributed instruction condition, students completed one activity per week for five weeks. Both conditions resulted in significant and similar gains in understanding and retention overall. Students' self-directed revisits to previously studied materials differed by condition, with students in the clustered condition tending to visit materials studied on previous days. These "distal" revisits explained variance in delayed post-test scores as an interaction effect with condition. Students in the clustered condition who revisited distal materials tended to score higher on the delayed post-test, whereas those in the distributed condition who did so tended to score lower. Our findings illustrate the complexity of realising laboratory findings in classrooms under real-world conditions.