Granular materials consisting of assemblies of dry and macroscopic particles of size typically larger than 100 μm, are of widespread use in many industrial applications. In civil engineering, chemical or food industries, numerous processes are designed to transport, store or mix together solid powders. Interestingly, this class of material displays original physical properties and remains a challenge to fundamental understanding. Among those properties, one can cite, as a response to stress, the occurrence of strain localisation and vault effects, also the existence of intermittent flows (avalanches) and a non-Newtonian rheology, furthermore, as a response to shearing and shaking, one observes violent size segregation phenomena. All this phenomenology is original and has no equivalent in the physics of the solid and the liquid state. Contrasting with atomic systems, for which equilibrium statistical mechanics and linear response theory provide a basis for the passage from a microscopic to a macroscopic viewpoint, here the particles are so large, that fluctuations of thermal origin are irrelevant and all classical methods of statistical mechanics fail. Furthermore, a strong disorder is present due to the irregular shape of the grains and to the fact that the contact forces between the grains are non-linear and dissipative. It is clear that such a complexity prevents from following standard routes in order to derive macroscopic behaviours. There were some attempts to provide a theoretical description adapted from solid state mechanics or from hydrodynamics, but until now, they seem to be inappropriate to describe in a unified way the observed behaviours. In this paper, we present series of experiments where the complexity of the “sand pile” problem is reduced by using simple granular model systems.