To assess students’ understanding of distribution aspects and to establish a tighter relationship between informal statistical notions and graphs, we decided to “reverse” this battery task. In the last two teaching experiments, during the fourth lesson, we asked students to invent their own data according to certain characteristics such as “brand A is bad but reliable; brand B is good but unreliable; brand C has about the same spread as brand A, but it is the worst of all brands.” Many students produced a graph similar to the one in Figure 5 (in this case, the variation of C is less than that of A). A sample response was