Instruction often starts with an explanation of a concept or principle before students are presented with problems to be solved. Recent research indicates that reversing this widely used tell-and-practice sequence (T&P) so that exploratory problem-solving precedes the instructional explanation (i.e., PS-I) might be more beneficial. We aimed to replicate this advantage, but we also hypothesized based on previous research that the effectiveness of PS-I would depend on how scaffolding prompts and specific ways of representing the problems are combined. In an in vivo experimental classroom study, 213 ninth graders were randomly allocated to either a T&P or 1 of 4 PS-I conditions (in a 2 × 2 design). In all PS-I conditions, exploratory problem-solving consisted of a comparing and contrasting cases activity. However, we varied whether the students processed grounded or idealized cases (containing or stripped off contextual detail, respectively) and whether the activity was scaffolded by an invention or a self-explanation prompt. We assessed transfer performance immediately after learning and 4 weeks later. The PS-I sequences were not generally more effective than the T&P sequence, the effectiveness was influenced by an interaction of scaffolding prompts and problem representation. Immediately after learning, T&P students were only outperformed by students who learned with grounded cases and self-explanation prompts, by students who learned with grounded cases and invention prompts, and by students who learned with idealized cases and invention prompts; only the latter retained this advantage 4 weeks after learning. We discuss potential reasons and emphasize that PS-I sequences demand careful design.