In Experiment 1, students experienced either direct instruction (DI) or productive failure (PF), wherein they first generated a quantitative index for variance before receiving direct instruction on the concept. Experiment 2 examined if it was necessary for students to generate solutions or can these solutions be simply given to the students to study and evaluate. Experiment 3 examined if it was necessary for students to generate solutions before receiving the critical features of the targeted concept, or could the critical features simply be told. In Experiment 1, PF students significantly outperformed DI students on data analysis and conceptual insight items. In Experiment 2, only the effects on conceptual insight and near transfer were significant. In Experiment 3, only the effect on conceptual insight remained significant. These results challenge the claim that that direct instruction alone is the most effective approach for teaching novel concepts to learners.