the deep ocean, as its mixing time is close to theobserved 800-year lag.Finally, the situation at Termination III dif-fers from the recent anthropogenic CO2in-crease. As recently noted by Kump (38), weshould distinguish between internal influences(such as the deglacial CO2increase) and exter-nal influences (such as the anthropogenic CO2increase) on the climate system. Although therecent CO2increase has clearly been imposedfirst, as a result of anthropogenic activities, itnaturally takes, at Termination III, some timefor CO2to outgas from the ocean once it startsto react to a climate change that is first felt in theatmosphere. The sequence of events during thisTermination is fully consistent with CO2partic-ipating in the latter !4200 years of the warm-ing. The radiative forcing due to CO2may serveas an amplifier of initial orbital forcing, which isthen further amplified by fast atmospheric feed-backs (39)thatarealsoatworkforthepresent-day and future climate.References and Notes1. J. R. Petit et al., Nature 399, 429 (1999).2. A. Neftel, H. Oeschger, T. Staffelbach, B. Stauffer,Nature 331, 609 (1988).3. D. Raynaud et al., Science 259, 926 (1993).4. J. M. Barnola, D. Raynaud, Y. S. Korotkevich, C. Lorius,Nature 329, 408 (1987).5. D. M. Sigman, E. A. Boyle, Nature 407, 859 (2000).6. In 1999, Fischer et al.(30)estimatedthattheincreaseof CO2lagged Vostok temperature by 600 " 400 yearsat the start of the last three Terminations, but the gasage–ice age difference at Vostok may be uncertain by1000 years (1)andthusobscuresthephasingofgasvariations with climate signals borne by the ice.7. The firn is the uppermost part of an ice sheet. It canbe schematically divided into three zones with differ-ent properties concerning the movement of air: theconvective zone in which the air is well mixed, thediffusive zone in which vertical transport is driven bymolecular diffusion, and the nondiffusive zone inwhich air does not migrate vertically, and at thebottom of which the air is trapped (17). This en-trapped air is younger than the surrounding ice,which results in an age difference (#age) between theice and the air bubbles that it contains.8. T. Blunier et al., Geophys. Res. Lett. 24, 2683 (1997).9. E. Monnin et al., Science 291, 112 (2001).10. N. Caillon, J. Jouzel, J. Chappellaz, A. Grachev, unpub-lished data.11. C. Lang, M. Leuenberger, J. Schwander, S. Johnsen,Science 286, 934 (1999).12. M. Leuenberger, C. Lang, J. Schwander, J. Geophys.Res. 104, 22163 (1999).13. J. P. Severinghaus, T. Sowers, E. Brook, R. Alley, M.Bender, Nature 391, 141 (1998).14. J. P. Severinghaus, J. Brook, Science 286, 930 (1999).15. N. Caillon et al., J. Geophys. Res. 106, 31893 (2001).16. The argon peak around 2760 m has no counterpart inthe temperature record published in (1). In Fig. 1B, weplotted the temperature profile that we deducedfrom the new deuterium measurements performedevery 10 cm (between 2700 and 2800 m). During thecooling phase of the interglacial, several abrupt tem-perature fluctuations occurred [especially around235,000 years (2740 m)], which were not revealed bythe temperature profile in (1). Those temperaturevariations could have affected the isotopic composi-tion of argon, making the argon peak at 2760 m.However, the sampling frequency in this depth range(2775 to 2750 m) does not allow access to a $40Arrecord as precise as that which we obtained duringthe Termination (i.e., between 2830 and 2775 m) anddoes not allow a peak-to-peak correlation.17. T. A. Sowers, M. Bender, D. Raynaud, Y. S. Korotkev-ich, J. Geophys. Res. 97, 15683 (1992).18. Dynamic densification firn models accounting for heattransfer are now under development. First results pre-dict that part of the measured argon signal should be aresult of thermal diffusion (40, 41). Those models sug-gest that despite the slow time scale of warming duringthe Termination, there is still a small residual temper-ature gradient left over after thousands of years, be-cause of the low thermal conductivity of the firn. Themodels generate a temperature gradient between sur-face and close-off region of about 3 K, which leads to athermal diffusion signal of about 0.11‰ using mea-sured thermal diffusion coefficients (19, 42). The use ofapreciserecordof$15N, which is more sensitive tothermal diffusion than $40Ar/4, over the Terminationshould be useful to confirm the small thermal diffusionsignal predicted by their models. Indeed, $15Ndatashould have a slightly larger value than $40Ar/4 to beconsistent with the presence of thermal diffusion signal(13). However, a precise record of $15NforTerminationIII is not available (fig. S1).19. J. P. Severinghaus, A. Grachev, B. Luz, N. Caillon,Geochim. Cosmochim. Acta 67, 325 (2003).20. L. Arnaud, J.-M. Barnola, P. Duval, in Physics of IceCore Records, T. Hondoh, Ed. (Hokkaido Univ. Press,Sapporo, Japan, 2000), pp. 285–305.21. C. Lorius et al., Nature 316, 591 (1985).22. The deuterium content of the snow in East Antarctica islinearly related to the surface temperature of the pre-cipitation site. Jouzel and colleagues (43)havereviewedall relevant information focusing on the East AntarcticPlateau where both model and empirical isotope-tem-perature estimates are available. Combining argumentscoming from the isotopic composition of the air bub-bles, from constraints with respect to ice core chronol-ogies, from atmospheric general circulation models, andfrom isotopic general circulation models (see referencesherein), the authors suggest that, unlike for Greenland,the present-day spatial isotope-temperature slope canbe taken as a surrogate of the temporal slope to inter-pret glacial-interglacial isotopic changes at sites such asVostok.23. O. Watanabe et al., Nature, in press.24. J. P. Severinghaus, A. Grachev, M. Battle, Geochem. Geo-phys. Geosyst. 2,Paperno.2000GC000146(2001).25. N. Caillon, thesis, University of Paris 6, France (2001).26. M. Battle et al., Nature 383, 231 (1996).27. The nondiffusive zone is at the bottom of the firn andthus warms several hundred years after the surfacebecause of the slow diffusion of heat through the firn(4). Additionally, the low accumulation rates at Vostokmake the downward transport of firn physical proper-ties rather slow (potentially spanning thousands ofyears). For example, if strong winds during the glacialperiods created wind-packed layers that later impededgas diffusion, thus creating a very thick nondiffusivezone, these layers would take several thousand years tobe