Happy or Redundant? Retrieval Practice Enhances Lasting Learning in Non-Interactive Teaching for Low-Quality Explanations

Background: Supporting lasting learning is a central goal in education, particularly in schools where long-term retention is crucial for students' academic success. Generative learning helps construct meaningful mental representations, while retrieval practice supports consolidation and long-term retention. Combining both strategies may foster learning; however, research remains scarce.Aims: We investigated how combining generative activities and retrieval practice affects students’ conceptual knowledge and monitoring accuracy in authentic school settings for immediate and lasting learning. Additionally, we explored associations between generation, retrieval, and explanation quality (i.e., completeness, elaboration, correctness).Sample: The classroom experiment was conducted in authentic physics lessons with N = 344 secondary students.Methods: Students participated in a physics teaching unit about converging lenses and were randomly assigned within each class to one of four combinations of sequential learning activities crossing two factors (non-generation vs. generation; non-retrieval vs. retrieval). Conceptual knowledge and monitoring accuracy were measured immediately and after eight weeks.Results: Conceptual knowledge and monitoring accuracy did not significantly differ between generation, retrieval, or their sequential combination at either test. However, exploratory analyses revealed that retrieval practice supported lasting learning only when prior generative processing was of low quality—that is, when students’ explanations were incomplete, less elaborated, or inaccurate.Conclusions: Combining generation and retrieval does not universally enhance learning. Its effectiveness appears to depend on the quality of prior generative processing. In this sense, retrieval practice may serve a happy role—compensating fragile knowledge—or a redundant one, offering no added benefit when prior understanding is already strong.