Floating and sinking (FS) is a key topic in science education, both at primary and secondary levels. The interpretation of FS phenomena, however, is challenging due to the difficulty of the scientific concepts and explanatory models involved (e.g., density, buoyancy), along with students' everyday experiences, which conflict with scientific explanations. Consequently, many studies over the last few decades have investigated how FS could be taught effectively to students of different ages while utilizing multiple teaching approaches. This meta-analysis summarizes findings from 69 intervention studies on teaching FS conducted between 1977 and 2021. Over all studies, we estimated a mean effect size of g = 0.85 (95% CI = 0.71, 0.99). This large effect size demonstrates that, although FS is a challenging concept, teaching FS is effective even in elementary school. Moreover, in a moderator analysis, we investigate the effect of intervention characteristics, students' age, as well as study design, and assessment features on the mean study effect size. To analyze the effect of these moderator variables, we use a three-level hierarchical meta-regression model for dealing with multiple effect sizes from single studies. We found two intervention characteristics that explain variance in study effect sizes: longer lasting interventions result in larger effect sizes and interventions where hands-on experiments are applied are more effective than those utilizing virtual experiments. Furthermore, studies with a treatment-control group comparison have significantly smaller effect sizes than studies with a pre-post design. We discuss the implications of our findings regarding the moderator variables for effective teaching of FS and further research on FS.