[1] Our knowledge of the temperature evolution over Greenland during Dansgaard-Oeschger events (DO) is currently qualitatively described through the water isotopic profile. Using two independent paleothermometry methods, one based on air isotopic measurements and the other on the combined measurements of water isotopes (dD and d 18 O), we show a complete and quantitative reconstruction of temperature at the NorthGRIP site over the period 76 to 62 kyr BP (DO 18, 19 and 20). We confirm that the associated warmings are larger than those conventionally depicted by the water isotopes (11°C, 16°C and 11°C for DO 18, 19 and 20). Secondly, we demonstrate that the relationship between temperature and d 18 O varies rapidly during the last glacial period, even over a DO. Finally, our temperature reconstruction over DO 19 agrees well with that predicted from simple climate models linking the DO to iceberg discharges. (2004), A continuous record of temperature evolution over a sequence of Dansgaard-Oeschger events during Marine Isotopic Stage 4 (76 to 62 kyr BP), Geophys. Res. Lett., 31, L22211, doi:10.1029/ 2004GL021193. [2] The last glacial cycle was characterized by millenial scale climate fluctuations that have been documented in the North Atlantic region through numerous marine and continental records [Bond et al., 1997; Genty et al., 2003]. The GRIP and GISP2 Summit ice cores [Dansgaard et al., 1993; Grootes et al., 1993] and the newly NorthGRIP [NorthGRIP Members, 2004] ice core exhibit 25 DO during the last glacial period. These events are characterized by rapid, e.g., in less than 100 years, and large, up to 16°C [Lang et al., 1999] warmings over Greenland. [3] Ice cores have already provided a wealth of information on DO in Greenland through the water isotopes for temperature changes, chemical records for atmospheric circulation and analysis of air bubbles for changes in greenhouse gases concentration. However, temperature reconstruction from water isotopes is subject to large biases mainly due to (i) the seasonality of the precipitation (i.e., periods without precipitation will not have their temperature recorded) [Fawcett et al., 1997] and (ii) changes in the oceanic source of Greenland snow [Boyle, 1997]. Whereas the latter can be estimated from the combined measurement of dD and d 18 O through the deuterium excess parameter (d = dD À 8 * d 18 O), we need additional information to account for the influence of seasonality and possible variation of the vertical atmospheric temperature profile (the isotopic composition of the snow depends on the condensation temperature). [4] One elegant way to overcome those difficulties is to use the isotopic composition of the air trapped in ice, a method based on the thermal diffusion of gases (nitrogen and argon), which allows estimates of the amplitude of rapid temperature increases [Severinghaus and Brook, 1999; Lang et al., 1999; Leuenberger et al., 1999]. This method gives directly access to the local mean surface temperature, and is therefore not affected by seasonality, the vertical temperature profile in the atmosphere, nor the source temperature of the precipitation. It confirms that the conventional temperature to water isotopes relationship, based on the spatial slope a s (the Dd 18 O ice /temperature slope calculated from present-day surface data) underestimates Greenland temperature change [Cuffey et al., 1995]. The temporal slope a t = Dd 18 O ice /DT (at a given site between two different climates) is lower than a s by up to a factor of 2. We have recently refined this approach, using a sophisticated firnification and heat diffusion model [Goujon et al., 2003] to quantify the temperature increase associated with large DO 12 [Landais et al., 2004]. [5] Here, we go beyond the simple quantification of the temperature increase. A detailed set of d 15 N and d 40 Ar measurements over DO 18, 19 and 20 provides us with strong constraints on the complete temperature scenario between 76 and 62 kyr BP. This period roughly corresponds to Marine Isotope Stage 4 with rapid ice sheet growth [Shackleton, 1987]. We have also measured a continuous dD profile in the ice that, combined with the existing d 18 O record, allows us to account for the source temperature effects. This complete study makes possible the precise estimation of the temperature change over a sequence of DO. [6] The d 15 N and d 40 Ar profiles (Figure 1) show a sharp peak corresponding to each warming. Rapid surface warm-GEOPHYSICAL RESEARCH LETTERS, VOL. 31, L22211,