A core component of scientific inquiry is the ability to evaluate evidence generated from controlled experiments and then to relate that evidence to a hypothesis or theory. The control-of-variables strategy (CVS) is foundational for school science and scientific literacy, but it does not routinely develop without practice or instruction. This meta-analysis summarizes the findings from 72 intervention studies at least partly designed to increase students' CVS skills. By using the method of robust meta-regression for dealing with multiple effect sizes from single studies, and by excluding outliers, we estimated a mean effect size of g = 0.61 (95% CI = 0.53–0.69). Our moderator analyses focused on design features, student characteristics, instruction characteristics, and assessment features. Only two instruction characteristics – the use of cognitive conflict and the use of demonstrations – were significantly related to student achievement. Furthermore, the format of the assessment instrument was identified as a major source of variability between study outcomes. Implications for teaching and learning science process skills and future research are discussed.