The theory of the Quaternary climate will be incomplete unless it is embedded in a more general theory for the fuller Cenozoic that can accommodate the onset of the ice-age fluctuations. Here we construct a simple mathematical model for the late Cenozoic climatic changes based on the hypothesis that forced and free variations of the concentration of atmospheric greenhouse gases (notably CO2) coupled with changes in the global ocean state and ice mass, under the additional influence or earth-orbital forcing, are primary determinants of the climatic state over this long period. Our goal is to illustrate how a single model governing both very long-term variations and higher frequency oscillatory variations in the Pleistocene can be formulated with relatively few adjustable parameters. Although the details of this model are speculative, and other factors neglected here are undoubtedly of importance, it is hoped that the formalism described can provide a basis for developing a comprehensive theory and systematically extending and improving it. According to our model the major near-100 ka period ice-age oscillations of the Pleistocene were caused by the downdraw of atmospheric CO2 (possibly a result of weathering of rapidly uplifted topography) to low enough levels for the ‘slow climatic system’, including glacial ice and the deep ocean state, to become unstable.