Two research teams published paleoceanographic reconstructions based on the same sediment core from the Iceland Basin. One paper reported that the Atlantic Meridional Overturning Circulation (AMOC) strengthened during the Bølling-Allerød interstadial, roughly 14,700 years ago. The other concluded that AMOC weakened over exactly the same interval. Both teams had counted the same species of planktonic foraminifera—Neogloboquadrina pachyderma sinistral—and both had used its oxygen isotope ratio and abundance as a proxy for ocean circulation. But they had not, it turned out, counted the same shells. The difference came down to a choice of sieve mesh: 63 micrometres versus 150 micrometres. The discrepancy highlights how a seemingly minor procedural choice can invert a climate proxy.

A Single Sieve Mesh Altered Two Climate Archives

Paleoceanographers reconstruct past ocean conditions by extracting sediment cores from the seafloor and analysing the microfossils preserved within. Foraminifera, single-celled organisms that build calcium carbonate shells (called tests), are among the most widely used proxies. After the sediment is disaggregated, it is washed through a sieve to isolate the foraminifera from the finer mud. The mesh size determines which size fraction of tests is retained for counting and geochemical analysis.

The two studies in question—let us call them Group A and Group B—both worked on core MD99-2251, raised from a water depth of roughly 1,200 metres in the Iceland Basin. Group A used a 63 µm sieve, a common standard in many labs, and counted N. pachyderma sinistral tests from the >63 µm fraction. Group B used a 150 µm sieve, also a common standard, and counted from the >150 µm fraction. When they plotted their records against depth, the two curves diverged in opposite directions. Group A’s record showed increased abundance of the polar species N. pachyderma sinistral during cold intervals and decreased abundance during warm intervals, consistent with a stronger AMOC during warm times. Group B’s record showed the opposite pattern.

The discrepancy was not subtle. At several depths, the two abundance estimates differed by more than 20 percentage points. Each team, naturally, had interpreted its own record as a faithful reflection of past ocean conditions. Neither had considered that the sieve itself might be selecting for different ecological signals.

Why Grain Size Matters for Paleoceanography

The sieving step is so routine that many method sections describe it in a single sentence: “Sediment was wet-sieved through a 63 µm mesh and the >63 µm fraction was dried and picked for foraminifera.” But that sentence masks a critical choice. Foraminifera tests vary in size from roughly 50 µm to over 500 µm, depending on species, growth conditions, and water depth. A 63 µm sieve retains nearly all adult tests of N. pachyderma sinistral, which typically range from 100 to 300 µm. A 150 µm sieve, by contrast, retains only the larger half of the size distribution.

As a related article on this site noted, “Standard mesh varies between labs: 63, 125, 150 µm.” Each mesh captures a different size fraction, and size is not ecologically neutral. In many planktonic foraminifera, shell size correlates with depth habitat. Smaller individuals of N. pachyderma sinistral tend to live higher in the water column, in the mixed layer, while larger individuals live deeper, near the pycnocline or in intermediate waters. Because the water column is stratified in temperature and nutrient content, the two size groups record different hydrographic conditions.

In the North Atlantic today, the surface mixed layer is warmer and fresher than the underlying water, which is colder and saltier. During the last deglaciation, the vertical structure of the water column shifted repeatedly as ice sheets melted and freshwater pulses altered stratification. A sieve that retains mainly small tests may preferentially sample the surface signal; a sieve that retains large tests may sample a deeper signal. If those two signals changed in opposite directions—for example, if surface waters warmed while intermediate waters cooled—then the two sieve fractions would yield opposite proxy records.

The Conflicting Records from the North Atlantic

The core MD99-2251 was collected during the 1999 IMAGES cruise and has been studied by multiple groups. Its location in the Iceland Basin places it under the influence of the Iceland-Scotland Overflow Water, a key component of the lower limb of AMOC. The site is also sensitive to the position of the Polar Front, which shifts during climate transitions.

Group A, which used the 63 µm sieve, published its results in Paleoceanography and Paleoclimatology in 2023. They reported that the abundance of N. pachyderma sinistral in the >63 µm fraction decreased during the Bølling-Allerød, indicating a retreat of polar waters and a strengthening of AMOC. Group B, using the 150 µm sieve, published in Quaternary Science Reviews in 2024. They found the opposite: N. pachyderma sinistral abundance increased during the Bølling-Allerød, suggesting a southward advance of polar waters and a weakening of AMOC. Both interpretations seemed plausible given the existing literature. The Bølling-Allerød was a period of rapid warming in the North Atlantic region, but the response of AMOC is debated. Some models suggest that meltwater input from the decaying Laurentide Ice Sheet could have weakened overturning, while others propose that the warming itself strengthened the circulation by increasing evaporation and cooling at high latitudes. The two sieve-based records seemed to embody that very debate. Yet the cores were the same. The contradiction could not be explained by different age models or analytical errors in the oxygen isotope measurements. The only systematic difference was the sieve mesh.

Methodological Autopsy: What the Sieve Actually Does

To understand how a sieve can invert a proxy signal, it helps to examine the size distribution of N. pachyderma sinistral tests in core MD99-2251. In a typical sediment sample, the test diameters range from about 80 µm to 400 µm, with a peak around 150–200 µm. A 63 µm sieve retains nearly the entire distribution, including the smaller individuals that are more abundant. A 150 µm sieve cuts off the left tail, removing roughly half the tests by number and skewing the retained assemblage toward larger, deeper-dwelling individuals.

In a 2025 reanalysis, a third group measured the size distribution of N. pachyderma sinistral across the core and found that the mean test size varied systematically with climate. During cold, stadial periods, tests were on average smaller; during warm, interstadial periods, tests were larger. This size shift is consistent with ecological observations: in colder waters, foraminifera grow more slowly and reach smaller adult sizes. But the size shift also means that the fraction retained by a 150 µm sieve changes in composition through time. During cold periods, fewer tests exceed 150 µm, so the >150 µm fraction is a smaller subset of the total population, possibly biased toward the largest, deepest-dwelling individuals. During warm periods, more tests exceed 150 µm, and the subset includes a broader range of depths.

Group A’s 63 µm record, by including the entire size spectrum, captured the abundance of the species as a whole, which decreased during warm intervals as the Polar Front retreated. Group B’s 150 µm record, by contrast, captured mainly the deep-dwelling fraction, which may have increased during warm intervals because deeper waters remained cold even as surface waters warmed. In effect, the two records were not measuring the same variable. One was a surface-water proxy; the other was an intermediate-water proxy.

Reconciling the Divergent AMOC Signals

Once the sieve effect is understood, the two records are not contradictory but complementary. The 63 µm record shows that the surface layer warmed and the Polar Front retreated during the Bølling-Allerød, which is consistent with a northward shift of the subpolar gyre. The 150 µm record shows that the intermediate waters remained cold, which is consistent with continued advection of polar water at depth. Together, they suggest that the vertical structure of the water column became more stratified during the Bølling-Allerød, with a warm, fresh surface layer capping a cold, salty deep layer. Such stratification could have been caused by meltwater input from the Greenland Ice Sheet, which freshened the surface and suppressed vertical mixing.

This interpretation, if correct, has implications for AMOC. Increased stratification tends to reduce deep convection and weaken overturning, consistent with Group B’s inference of a weaker AMOC. But the surface warming and northward shift of the Polar Front are also consistent with a stronger surface flow, which could have compensated for the weaker deep flow. The net effect on AMOC remains uncertain.

The key insight is that a single sieve fraction cannot capture the full complexity of the ocean’s vertical structure. Paleoceanographers are increasingly aware that size-sorting during sieving introduces a bias that must be accounted for. Some labs now routinely analyse multiple size fractions—for example, 63–150 µm, 150–250 µm, and >250 µm—to reconstruct depth-resolved signals. Others use laser diffraction to measure the full grain-size distribution of the sedimentary foraminifera and then statistically deconvolve the mixed signal.

As a related article on this site notes, “A Budget Office’s Sieve Standard Split Two Ocean Sediment Core Chronologies” (read more here)—a reminder that procedural standards can have far-reaching effects on scientific conclusions.

Lessons for Future Climate Reconstructions

The MD99-2251 case is not an isolated incident. A 2022 survey of published paleoceanographic studies found that roughly 40% of papers using foraminifera proxies did not report the sieve mesh size in their method section. Among those that did, the mesh ranged from 63 µm to 250 µm, with no consistent standard across labs or even within labs over time. The problem is compounded by the fact that different species have different size distributions, so the same mesh can select different subsets for different proxies.

One obvious solution is to require authors to report the full size distribution of the foraminifera they counted, not just the mesh cutoff. A study highlighted in a related article used machine learning to automatically classify foraminifera tests by size and species from digital images, which could make such reporting feasible. Another approach is to standardise on a single mesh size across the community—but which one? A 63 µm sieve captures more tests, which improves counting statistics but includes small juveniles that may have different geochemistry. A 150 µm sieve reduces the juvenile bias but discards information about the smaller fraction.

Perhaps the best practice is to analyse multiple size fractions in parallel, as some labs already do. This approach was used in a recent reanalysis of core MD99-2251, which measured N. pachyderma sinistral abundance and oxygen isotopes in three size fractions: 63–150 µm, 150–250 µm, and >250 µm. The results showed that the 63–150 µm fraction tracked surface conditions, the 150–250 µm fraction tracked intermediate conditions, and the >250 µm fraction tracked a deeper signal. The combined record provided a three-dimensional view of the water column that neither single-fraction record could offer.

As the same site’s article “A Single Electron Microscope Service Contract Resolved Two Lab Data Discrepancies” (read more here) illustrates, methodological transparency can resolve apparent contradictions.

A Sieve Is Not a Neutral Tool

The story of the two AMOC records is a reminder that every procedural choice in science carries a bias. A sieve is not a neutral filter; it is a lens that selects which fragments of the past we get to see. The same core, the same species, the same counting protocol—but a different mesh size produced two opposite conclusions. How many other paleoclimate debates are actually artefacts of methodology?

While solutions exist, the episode underscores the need for methodological vigilance. Journals are increasingly requiring authors to deposit raw grain-size data alongside their proxy measurements. Community initiatives like the Paleoceanography Standards Working Group are developing best-practice guidelines for sieving and size reporting. And new technologies—from automated image analysis to micro-CT scanning—are making it possible to characterize the full size distribution of foraminifera without destructive sampling. Yet even with these advances, the deeper epistemological challenge remains. Paleoceanography, like all historical sciences, relies on indirect evidence. The ocean’s memory is stored in the sediment, but we can only read it through the strainers we choose. If we change the strainer, we change the memory. The next time a paleoclimate study reports a surprising result, it may be worth asking not just about the age model or the calibration, but about the sieve.