The concept of productive failure posits that a problem-solving phase prior to explicit instruction is more effective than explicit instruction followed by problem-solving. This prediction was tested with Year 5 primary school students learning about light energy efficiency. Two, fully randomised, controlled experiments were conducted. In the first experiment (N = 64), explicit instruction followed by problem-solving was found to be superior to the reverse order for performance on problems similar to those used during instruction, with no difference on transfer problems. In the second experiment, where element interactivity was increased (N = 71), explicit instruction followed by problem-solving was found to be superior to the reverse order for performance on both similar and transfer problems. The contradictory predictions and results of a productive failure approach and cognitive load theory are discussed using the concept of element interactivity. Specifically, for learning where element interactivity is high, explicit instruction should precede problem-solving.