The ability to design unconfounded experiments and make valid inferences from their outcomes is an essential skill in scientific reasoning. The present study addressed an important issue in scientific reasoning and cognitive development: how children acquire a domain-general processing strategy (Control of Variables Strategy or CVS) and generalize it across various contexts. Seven- to 10-year-olds (N = 87) designed and evaluated experiments and made inferences from the experimental outcomes. When provided with explicit training within domains, combined with probe questions, children were able to learn and transfer the basic strategy for designing unconfounded experiments. Providing probes without direct instruction, however, did not improve children's ability to design unconfounded experiments and make valid inferences. Direct instruction on CVS not only improved the use of CVS, but also facilitated conceptual change in the domain because the application of CVS led to unconfounded, informative tests of domain-specific concepts. With age, children increasingly improved their ability to transfer learned strategies to remote situations. A trial-by-trial assessment of children's strategy use also allowed the examination of the source, rate, path, and breadth of strategy change.