The Power of Varying More Than One Thing at a Time

Background Varying one thing at a time (VOTAT) is considered a universal approach for designing informative experiments. Its core idea is that controlled experiments can be designed by manipulating a single variable. However, we show that this single-manipulation heuristic can produce misleading evidence in contexts involving interrelated variables. We argue that varying more than one thing at a time (MOTAT) can be more informative when learners’ prior beliefs render alternative explanations implausible. Aims To test these claims, we compared single-manipulation to MOTAT experiments in a Bayesian learning simulation and a pre-registered empirical study on children’s learning about water displacement. Sample Participants in the empirical study were 185 children aged 6–9. Methods Learners predicted which of two spheres would displace more water and then observed an animation of the spheres being submerged. In the single-manipulation experiments, the spheres differed either only in size or only in material, leading to a confound of these dimensions with mass. In the MOTAT experiments, both size and material were manipulated such that the larger sphere was lighter. We investigated the formal informativeness of single-manipulation and MOTAT experiments (simulation study) and their effect on children’s learning, as measured via a post- and transfer test (empirical study). Results MOTAT experiments were most informative and led to the highest transfer test performance. Conclusion These findings provide initial evidence that VOTAT's core heuristic has a narrower scope than is commonly assumed. When variables are interrelated, deliberately varying more than one variable may be the better choice.