Superconductors / Bond-valence sum analysis / Oxocuprate superconductors / Transition-metal pnictides / Transition-metal chalcogenides Abstract. A crystal-chemical approach to superconductivity is described that is intended to complement the corresponding physical approach. The former approach takes into account the distinction between the stoichiometric valence ( stoich V) and the structural valence ( struct V) which is represented by the bond-valence sums (BVS). Through calculations of BVS values from crystal-structure data determined at ambient temperature and pressure it has been found that in chalcogenides und pnictides of the transition metals Fe, Co, Ni, Mn, Hf, and Zr the atoms of the potential superconducting units yield values of |BVS| ¼ | struct V| ! 1.11 Â | stoich V|, whereas the atoms of the charge reservoirs have in general values of | struct V| < 1.11 Â | stoich V|. In corresponding compounds which contain the same elements but are not becoming superconducting, nearly all atoms are found to have | struct V| < 1.11 Â | stoich V|. For atoms of oxocuprates that are not becoming superconducting and for atoms of the charge reservoirs of oxocuprates that become superconducting, the relation | struct V| < 1.11 Â | stoich V| seems also to be fulfilled, with the exception of Ba. However, in several oxocuprates the relation | struct V| ¼ 1.11 Â | stoich V| for the atoms that become superconducting units is violated. These violations seem to indicate that in oxocuprates it is the local bond-valence distribution rather than the bond-valence sums that is essential for superconductivity. The present analysis can possibly be used to predict, by a simple consideration of ambient-T, P structures, whether a compound can become an unconventional superconductor at low T, under high P and/or by doping, or not.