Pearson et al. reply

We note, however, that this marks a shift of emphasis from studies spanning several decades, in which attempts have been made to determine actual temperatures from the deep-sea record and to compare them with the output of climate-model simulations (for example, see refs 1–3). We also warn against assuming that “primary offsets in fossil shell chemistry tend to be preserved”, as diagenetic overprinting will tend markedly to reduce such offsets, as well as attenuating stratigraphic patterns.

We have suggested that recrystallization could account for as much as 50% of the mass of planktonic foraminifera in typical deep-sea samples, which would reduce estimated SSTs in the tropics by 10–15 °C. Zachos et al. have questioned our comparison of isotope data from well-preserved, middle-Eocene foraminifera from Tanzania and a recrystallized deep-sea assemblage. But we did not assume that initial conditions were identical for the two sites — indeed, we were careful to point out the likely original difference in temperature. When this is taken into account, we calculate that the oxygen-isotope data suggest a 55% diagenetic overprint of the deep-sea sample towards an ocean bottom-water temperature of about 10 °C (ref. 4). A similar overprint is also implied by the greatly reduced interspecies carbon-isotope differentials.

We also discussed the question of δ18Osw on the Tanzanian margin. For several reasons (the narrow open shelf, probable onshore current, normal stratification, full plankton assemblages and results of palaeosalinity modelling), we think it unlikely that the salinity was as much as 3.0 p.p.t. lower than in the deep ocean. But even if it were, the probable relationship between δ18Osw and salinity at this latitude would imply a δ18Osw value that is only about 0.5 p.p.t. more negative, which corresponds to an estimated temperature increase of about 2 °C. This is a small effect when compared with the apparent difference of nearly 15 °C in estimated SST between the East African shelf and coeval open-ocean sites such as ODP Site 865 (ref. 2). We stress that our Tanzanian data are also supported by similar results from Mexico, Alabama and the Adriatic Sea.

Study of both deep-sea carbonates and hemipelagic clays is crucial to ensure that sampling is as spatially distributed as possible. But there is no reason why hemipelagic mudstones should not be as accurately dated as deep-sea carbonates, thereby combining the advantages of high-resolution stratigraphy and good microfossil preservation. Recent Ocean Drilling Program (ODP) coring in New Jersey has already demonstrated this potential5, and we anticipate further well-dated Cretaceous and Palaeogene mudstones from a forthcoming ODP leg to Demerara Rise (in the equatorial Atlantic Ocean) and from our own drilling in Tanzania. Such investigations will help to elucidate past variation in absolute palaeotemperatures and meridional temperature gradients, which remains critical for testing the greenhouse theory for past warm climates.