The aim of this paper is to show that with the aid of a qualitative model of ionic bonding, including polarizability, many crystal structures, mainly of halides and chalcogenides, can be explained or even predicted. Polarization of O 2− and F − ions may be neglected unless these ions have very small, or small and highly charged cation neighbours. The polarizability of a few large cations is assumed also to play a rôle. The cation polyhedron may have a shape different from that expected for the ionic model as a result of nonionic bonding types; the theoretical shapes can be used empirically in the structure prediction. Section I deals with the background of this ionic picture. Section II deals with the sharing of corners and/or edges and/or faces of cation polyhedra. Pictures of various types of sharing polyhedra are given, with octahedra as an example. An explanation is offered for Wadsley's rule of maximal edge-sharing in his shear-structures, which seemed to violate Pauling's third rule. In Section III a classification of crystal structures is given in plots of cation coordination versus stoichiometry, expressed as the ratio nr. anions/nr. cations, for simple compounds and complex anions. The average cation coordination is plotted against stoichiometry if one is interested in the changes of structure under high pressure. Section IV deals with the description of crystal structures by space-filling polyhedra, which are an aid towards finding relationships between structures and towards predicting new ones. Simple structures with close-packed anions or cations are explained by the ionic model using these SFP in Section V. In Section VI some structure predictions (e.g., of Sc 2S 3, Li 2CuO 2, SrCuO 2, Sr 2CuO 3, and SrCu 2O 2) are discussed, and some results of recent structure determinations carried out in the author's laboratory (of the above-mentioned SrCuO-compounds, β-BaPdO 2, MgAl 2S 4, FeAl 2S 4 and MgGa 2S 4) are given.