Differences between the ages of the Bolling onset and mwp-1a in the original data sets An age discrepancy exists between the onset of the Bolling and mwp-1a in both 14C and 230Th chronologies. In the following, all reported 14C ages are conventional 14C ages before 1950 (5568 yr half life, corrected for fractionation). Errors are given at the 2 sigma level for both 14C and 230Th ages. All 230Th ages are expressed as years before 1950. The relative sea levels (RSL) are the depth of Acropora palmata recovery corrected for a constant uplift of 0.34 m/kyr (230Th ages are used for the uplift correction). Mwp-1a is bracketed in Barbados between samples from cores 9 and 12 (main text Table 2, Figure S1). Mwp-1a is dated between about 12235 and 11635 14C yr BP (using a 365 yr reservoir age) [Fairbanks et al., 2005], i.e. younger than the onset of the Bolling in lake and deep-sea sediments (12700 14C yr BP) [Bard et al., 1987]. It is dated between about 14100 and 13650 230Th yr BP, which again is younger than the onset of the Bolling in Greenland and Cariaco Basin (~ 14650 cal yr BP) [Rasmussen et al., 2006; Lea et al., 2003]. The Barbados data (Figure S1) narrowly constrain the sea-level rise with respect to both 14C and 230Th chronologies (note that dating uncertainties are considered at the 2 sigma level). The 14C chronology is subject to greater uncertainty than the 230Th chronology, and this contrast is amplified when considering corrections for variable reservoir ages that affect 14C ages. Clearly, significant improvements to the dating of mwp-1a should be sought by means of the 230Th technique. The Sunda sea-level record [Hanebuth et al., 2003] is entirely based on 14C ages. As a consequence, it does not have the necessary precision around mwp-1a to discuss shifts in the order of a few centuries. This situation is exacerbated when discussing its chronology in "calendar years" after calibration [Kienast et al., 2003]. In Kienast et al. [2003], the calibrated 14C-based chronology is presented with a shaded zone (their Figure 3a) that represents dating uncertainties only at the 1 sigma level. This would need to be doubled to be comparable with the 2 sigma interval commonly represented for the corals (e.g., auxiliary Figure S1), in which case the Sunda record would be compatible with a sea- level jump anywhere between 13.5 and 15.5 ka BP. In addition, we consider that the chronologies for South China Sea cores are in general not sufficiently understood, given that apparently arbitrary reservoir-age corrections are applied that range between 1300 years [Kienast et al., 2003] and 400 years [Kienast et al., 2001], to achieve matching chronologies for the different cores. Apart from the fact that the very high value of 1300 years would imply almost pure intermediate water, it is even more challenging to consider a 1000- yr gradient of 14C reservoir age within such a small basin as the South China Sea. Consequently, we infer that the 230Th-based chronology for mwp-1a derived from the Barbados coral record remains the most accurate for comparison with other records of climate change. Additional References Bard, E., M. Arnold, P. Maurice, J. Duprat, J. Moyes, and J.-C. Duplessy (1987), Retreat velocity of the North Atlantic polar front during the last deglaciation determined by 14C accelerator mass spectrometry, Nature, 328, 791-794. Kienast, M., S. Steinke, K. Stattegger, and S. E.Calvert (2001), Synchronous tropical South China Sea SST change and Greenland warming during deglaciation, Science, 291, 2132-2134.