Under present-day climatic conditions the air content of ice shows a high sensitivity to the atmospheric pressure and hence to the elevation at the surface of the ice sheet. The observation has been used to infer past ice sheet thickness variations of Antarctic and Greenland. A high-resolution air content profile (more than 1000 measurements) covering approximately the last 200,000 years was obtained along the 2546-m long Vostok ice core. Three analytical techniques were used, leading to consistent results which show large amplitude and rapid air content variations. The Vostok results support thicker/thinner ice in the central part of East Antarctica during warm/cold periods. However, constraints imposed by ice sheet dynamics suggest that the Vostok air content signal cannot be interpreted only in terms of ice sheet thickness variations. Apart from ice thickness changes, the two other potential sources of air content variations are atmospheric pressure and ice porous volume at the air isolation level. Several atmospheric general circulation models have been applied to the last glacial maximum. They show atmospheric pressure changes which can only explain part of the air content variations in the Vostok ice core. On the other hand, the ice porous volume at the depth of air isolation undergoes fairly well-quantified thermal variations, but they are too small to play a dominant role in the Vostok signal. On the basis of new data concerning the present day ice porous volume variations we suggest that a wind influence on ice porous volume at the air isolation level could be a source for the unexplained air content variations of Vostok. Equivalent contributions from elevation, air pressure, and nonthermal porous volume changes could explain the air content drop during the penultimate deglaciation. Wind speed changes by about 7 m s-1 could be the source of the large and rapid air content variations observed during glacial stages. ¿ American Geophysical Union 1994