A bottom-up approach of the nature of science is proposed. Teacher-friendly formulations of criteria of scientificity based on actual classroom practices allow the straightforward derivation of evolutive definitions of kindergarten science from arbitrary criteria subsets, and to distinguish 'scientific' and 'non scientific' practices, even when superficially similar. Suitable formulations of Primacy of experiment, Local reproducibility and Robustness of an experiment are proposed. Three actual implementations of investigation sequences about the making of pancakes with 5-6 y.o. children are analysed using this epistemological tool, and their eligibility as 'scientific' kindergarten practices is shown to depend on the set of criteria elected by the teacher, thus on his own pedagogical objectives. This bottom-up approach of NoS appears to be complementary to the usual top-down approach, better suited to more advanced students.