While many view learning as a process of reducing learners' uncertainty, research suggests that instruction that is uncertain can promote learning and transfer better than instruction that is certain. In addition, research on curiosity suggests that uncertainty is a key trigger of curiosity, which in turn can facilitate learning. However, educational research rarely examines the direct effects of uncertainty on curiosity, learning, or transfer. Additionally, research on the effect of curiosity on learning rarely considers state-level curiosity or how curiosity changes over time. In a study with 208 middle school students learning physics, we addressed these gaps. Participants in two conditions completed learning activities where they invented their own equations for physical science phenomena. The Low Uncertainty condition received relevant information on task process before inventing, while information on task process was withheld in the High Uncertainty (HU) condition, which received irrelevant information before inventing. Both conditions learned the physics content equally well, but the HU condition demonstrated greater state-level curiosity and performed better on transfer problems. Moreover, in both conditions, curiosity decreased over time as students gained more information. Surprisingly, curiosity did not predict learning or transfer, which suggests that curiosity was not the mechanism by which uncertainty influenced transfer. This study advances the notion that introducing uncertainty in learning activities can, perhaps counter-intuitively, promote transfer of knowledge across contexts while also rousing learners' curiosity. This work demonstrates a practical way for educators to induce uncertainty, by withholding information about task process. This research also broadens our understanding of how to provoke curiosity in classroom contexts.