Retrieval practice is a powerful way to promote long-term retention and meaningful learning. However, students do not frequently practice retrieval on their own, and when they do, they have difficulty evaluating the correctness of their responses and making effective study choices. To address these problems, we have developed a guided retrieval practice program that uses an automated scoring algorithm, called QuickScore, to evaluate responses during retrieval practice and make study choices based on student performance. In Experiments 1A and 1B, students learned human anatomy materials in either repeated retrieval or repeated study conditions. Repeated retrieval in the computer-based program produced large gains in retention on a delayed test. In Experiment 2, we examined the accuracy of QuickScore's scoring relative to students' self-scoring of their own responses. Students exhibited a dramatic bias to give partial or full credit to completely incorrect responses, while QuickScore was far less likely to score incorrect responses as correct. These results support the efficacy of computer guided retrieval practice for promoting long-term learning.