Recent progress in the reconstruction of atmospheric CO2 records from Antarctic ice cores has allowed for the documentation of natural CO2 variations on orbital time scales over the last up to 800,000 years and for the resolution of millennial CO2 variations during the last glacial cycle in unprecedented detail. This has shown that atmospheric CO2 varied within natural bounds of approximately 170–300 ppmv but never reached recent CO2 concentrations caused by anthropogenic CO2 emissions. In addition, the natural atmospheric CO2 concentrations show an extraordinary correlation with Southern Ocean climate changes, pointing to a significant (direct or indirect) influence of climatic and environmental changes in the Southern Ocean region on atmospheric CO2 concentrations.Here, we compile recent ice core and marine sediment records of atmospheric CO2, temperature and environmental changes in the Southern Ocean region, as well as carbon cycle model experiments, in order to quantify the effect of potential Southern Ocean processes on atmospheric CO2 related to these orbital and millennial changes. This shows that physical and biological changes in the SO are able to explain substantial parts of the glacial/interglacial CO2 change, but that none of the single processes is able to explain this change by itself. In particular, changes in the Southern Ocean related to changes in the surface buoyancy flux, which in return is controlled by the waxing and waning of sea ice may favorably explain the high correlation of CO2 and Antarctic temperature on orbital and millennial time scales. In contrast, the changes of the position and strength of the westerly wind field were most likely too small to explain the observed changes in atmospheric CO2 or may even have increased atmospheric CO2 in the glacial. Also iron fertilization of the marine biota in the Southern Ocean contributes to a glacial drawdown of CO2 but turns out to be limited by other factors than the total dust input such as bioavailability of iron or macronutrient supply.