The research, led by Victoria E. Taylor of the University of Bergen together with Alison M. Piasecki, Eivind O. Straume and A. Nele Meckler, applied a technique called carbonate clumped isotope thermometry to fossil foraminifera from two of the world’s great ocean basins. Unlike conventional oxygen isotope measurements, which conflate temperature with the isotopic composition of seawater, clumped isotope thermometry exploits the tendency of heavy isotopes of carbon and oxygen to bond with one another in carbonate minerals. The abundance of these carbon-13–oxygen-18 bonds, quantified as the Δ47 value, depends on temperature alone at the time the shell formed. That independence makes it possible to reconstruct deep-sea temperatures directly, without having to guess how salty or isotopically heavy the ancient ocean was.

The team assembled temperature records spanning the Eocene–Oligocene transition from sediment cores recovered by the Integrated Ocean Drilling Program in the eastern equatorial Pacific, at Sites U1333, U1334 and 1218, and from the northwest North Atlantic, at IODP Site U1406. What they found was genuinely surprising. Although the oxygen isotope records from the two basins look remarkably similar, the actual temperatures behind them behaved in completely different ways. The deep Pacific began cooling earlier, with its temperature decline preceding the main phase of the oxygen isotope increase by roughly 100,000 years. The North Atlantic, by contrast, cooled later in the transition, even though its isotope signal shifted at much the same time as the Pacific’s.

This decoupling is the kind of result that forces a rethink of a foundational dataset. For decades, the globally coherent rise in benthic oxygen isotope values at the Eocene–Oligocene transition has been treated as evidence that the deep ocean cooled everywhere in step as the Antarctic ice sheet expanded. The new clumped isotope data reveal that the isotope signal can be misleading. Because oxygen isotope values in foraminiferal calcite depend on both temperature and the isotopic composition of seawater, which itself tracks salinity, two water masses can record identical isotope shifts while experiencing entirely different thermal histories. The authors argue that contrasting salinity gradients, required to maintain similar water densities between the basins, probably masked the divergent temperature behaviour in the traditional records.

Perhaps the most striking finding concerns the direction of the temperature gradient between the basins. Today, the deep North Atlantic is bathed in relatively warm water formed in the northern high latitudes, while the deep Pacific receives colder water sourced from the Southern Ocean. In the early Oligocene, the situation appears to have been reversed: the deep North Atlantic was, on the whole, cooler than the equatorial Pacific. In fact, between about 33.52 and 33.29 million years ago, deep northwest Atlantic temperatures dropped to values approaching those of the modern ocean, levels that, within the limits of the available records, were not reached again until the Pleistocene ice ages tens of millions of years later. That is a remarkable early glimpse of ice-age-like cold in a basin usually associated with warmth.

The explanation the researchers propose is a deep ocean organised in a fundamentally different way from today’s. Rather than a single dominant overturning circulation, the Eocene–Oligocene deep ocean may have hosted two distinct southern-sourced deep-water masses with divergent properties. The compartmentalised Southern Ocean of the time, still partially cut off by the unfinished opening of the Drake Passage and the Tasmanian Gateway, offered greater potential to form cold deep waters in its Atlantic sector than in its Pacific sector. The two basins were therefore effectively decoupled, each receiving deep water whose characteristics depended on where and how it last contacted the surface.

This picture carries a broader implication: the growing Antarctic cryosphere was not merely a passive response to cooling, but an active participant in shaping the ocean’s circulation. The spatially and temporally variable early ice sheet would have altered brine rejection, freshwater fluxes and the buoyancy of surface waters around the Antarctic margin, directly influencing the properties of the deep waters forming in adjacent sectors. The divergent behaviour of the Pacific and Atlantic basins at the transition may thus record the fingerprint of an ice sheet learning, so to speak, how to be an ice sheet, with each phase of its growth leaving a distinct imprint on the water masses it helped create.

The study builds on a growing body of work applying clumped isotope thermometry to the deep past. A 2022 analysis in Science by Meckler and colleagues established the long-term Cenozoic evolution of deep ocean temperature using the same method, and subsequent studies have documented transient deep cooling in the eastern equatorial Pacific at the transition. By extending the approach to the North Atlantic and comparing basins directly, the new work transforms what was a single global narrative into a two-basin story with genuinely different thermal trajectories. The methodological rigour is considerable: the team used long-integration dual-inlet mass spectrometry, community standardisation protocols and careful error propagation, and their Atlantic temperatures reproduce independently published measurements from nearby Expedition 342 sites.

Why does this matter beyond the Eocene? The Eocene–Oligocene transition is one of the best-studied natural experiments in how a greenhouse world tips into an icehouse one, and it is frequently used to benchmark climate models and to frame expectations about future change. If the deep ocean’s response to that tipping was spatially heterogeneous rather than uniform, then reconstructions and models that assume a globally coherent deep ocean may need revision. It also matters for how scientists interpret the oxygen isotope archives that underpin so much of Cenozoic climate science: the same isotope shift can hide different combinations of cooling and salinity change in different places, and clumped isotope thermometry offers a way to untangle them.

The deep ocean, covering most of the planet and storing the vast majority of its accessible heat, is the flywheel of the climate system. What this study reveals is that during one of Earth’s great climatic reorganisations, that flywheel was not turning as a single machine but as two, with the Pacific and Atlantic spinning along divergent paths even as the ice advanced on Antarctica. The samples and data underpinning the work, provided through the Integrated Ocean Drilling Program and archived on Zenodo, ensure that other researchers can now interrogate this pivotal interval further. As the Antarctic ice sheet assembled itself 34 million years ago, it did not simply chill the world’s oceans in unison; it reorganised them from the bottom up, leaving a signature that is only now coming into focus.

Subject of Research: Deep-sea temperature evolution in the Pacific and Atlantic oceans during the Eocene–Oligocene onset of Antarctic glaciation

Article Title: Divergent deep-sea temperatures in the Pacific and Atlantic during the onset of Antarctic glaciation

Article References: Taylor, V. E., Piasecki, A. M., Straume, E. O., & Meckler, A. N. (2026). Divergent deep-sea temperatures in the Pacific and Atlantic during the onset of Antarctic glaciation. Nature Geoscience. https://doi.org/10.1038/s41561-026-02106-x

Image Credits: AI Generated

DOI: 10.1038/s41561-026-02106-x

Keywords: Eocene–Oligocene transition, Antarctic glaciation, clumped isotope thermometry, benthic foraminifera, deep-ocean circulation, palaeoceanography, oxygen isotopes, North Atlantic Deep Water, Southern Ocean, IODP, paleoclimate, ocean gateways