Human activities have already modified the chemical composition of the natural atmosphere even in very remote regions of the world. The study of chemical parameters stored in solid precipitation and accumulated on polar ice sheets over the last several hundred thousand years provides a unique tool for obtaining information on the composition of the preindustrial atmosphere and its natural variability over the past. This paper deals with the chemistry of polar ice focused on the soluble mineral (Na+, NH4+, K+, Ca++, Mg++, H+, F-, Cl-, NO3-, SO4--, and H2O2) and organic (methanesulfonate (CH3SO3-), formate (HCOO-), acetate (CH3COO-), and formaldehyde (HCHO)) species and their interpretation in terms of past atmospheric composition (aerosols and water soluble gaseous species). We discuss ice core dating, the difficulties connected with trace measurements, and the significance of the ionic composition of snow. We examine temporal (from the last decades back to the last climatic cycle) and spatial (including examples from coastal as well as central areas of Greenland and Antarctica) variations in the ionic budget of the precipitation and evaluate ice core studies in terms of the chemical composition of our past atmosphere. We review (1) how Greenland and Antarctic ice cores that span the last few centuries have provided information on the impact of human activities and (2) how the chemistry of deep ice cores provides information on various past natural phenomena such as climatic variations (glacial-interglacial changes, El Ni¿o), volcanic eruptions, and large boreal forest fires.¿ 1997 American Geophysical Union