This study examined how the testing procedure might affect performances on the water-level task and the verticality task in the course of development, comparing performances on a two-dimensional graphic version (drawings of jars and of a hillside) and on a three-dimensional graphic version (real jars and a physical model of a hill). Results showed that performance improved with age and that scores on the 3D version were better as a whole than on the 2D version. The verticality task was quite difficult for all children, whatever the version. When the verticality task is presented in a blindfolded condition, children performed better than in the normal condition. Major factors (both relevant and irrelevant) that affect performance during development are discussed.