Recent research highlights the potential benefits of practice without feedback on learner's strategy knowledge. However, most prior work has been conducted in one-on-one settings with short retention intervals. We compared the effects of mathematics practice with and without correct-answer feedback on immediate and 1-week delayed performance in a classroom setting. In a randomized experiment, 243 second- and third-grade children received strategy instruction and then practiced solving relevant problems in small groups within their classroom. During practice, children received immediate feedback (after each problem), summative feedback (after all the problems in the set), or no feedback. During the practice task, immediate feedback led to the best performance. However, practice "without" feedback led to higher levels of mastery on the 1-week knowledge retention test. Thus, instructional practices that seem less effective at first can, for some children, ultimately result in more desirable learning.