Children who observe gesture while learning mathematics perform better than children who do not, when tested immediately after training. How does observing gesture influence learning over time? Children (= 184, ages = 7–10) were instructed with a videotaped lesson on mathematical equivalence and tested immediately after training and 24 hr later. The lesson either included speech and gesture or only speech. Children who saw gesture performed better overall and performance improved after 24 hr. Children who only heard speech did not improve after the delay. The gesture group also showed stronger transfer to different problem types. These findings suggest that gesture enhances learning of abstract concepts and affects how learning is consolidated over time.