A Closer Look at the Study Linking Roman Lead Pollution to Falling IQs

I. What the Study Actually Found

In January 2025, a study from the Desert Research Institute made headlines with a striking claim: air pollution from Roman-era silver mining may have measurably lowered IQ scores across the entire Roman Empire, roughly 2,000 years before anyone had invented an IQ test. The idea that lead poisoned Rome is not new, but it comes with baggage. A mid-twentieth-century thesis, built by sociologist S. C. Gilfillan and later amplified by public health researcher Jerome Nriagu, blamed lead-lined cookware and lead-sweetened wine for erratic behavior among emperors like Caligula and Nero. Classicists have since faulted that older thesis for thin citation practices and selective reading of ancient sources, and a recent academic review concludes Roman lead's impact has likely been overstated in past scholarship. The new ice-core study does not revive that discredited argument. It makes a narrower, more rigorously quantitative claim: that background air pollution from industrial-scale mining, not cookware or wine, affected the Roman population.

The study, led by hydrologist Joseph R. McConnell and published in the Proceedings of the National Academy of Sciences, is a genuinely inventive piece of science. The research team, which also included historians and archaeologists, drew on three ice cores pulled from Greenland and the Russian Arctic. That may seem like an odd place to look for evidence of Roman pollution, but fine airborne particles do not stay put. Carried by prevailing winds, they can travel thousands of kilometers before settling out of the atmosphere, which is how Mediterranean mining smoke left a trace on the other side of the continent. Snow that fell there, layer upon layer over thousands of years, compressed into something like a frozen ledger of everything that once drifted through Earth's atmosphere.

By measuring the lead in that ledger and using isotope analysis to trace where it came from, the team showed that lead pollution surged during the Pax Romana, the roughly two-century span of relative peace and prosperity from 27 BCE to 180 CE that marked the height of the empire. The likely source was smelting galena, a silver-bearing ore whose processing releases a large volume of lead vapor for every ounce of precious metal recovered, largely to feed the empire's demand for silver coinage.

From there, the researchers took a more ambitious step. Using FLEXPART — an atmospheric transport model of the kind climate scientists use to trace how smoke or volcanic ash spreads through the air — they worked backward from the ice-core measurements to estimate how much lead pollution had circulated over Europe at the time. They tested two scenarios for where that lead originated, one concentrated in a known mining region in what is now southern Spain, the other spread more broadly across the empire, and both produced similar estimates. Altogether, the study puts more than 500 kilotons of lead into the atmosphere over the roughly 200-year height of Roman industrial activity, a figure the authors say rivals a meaningful fraction of the pollution released during the 20th century's leaded-gasoline era.

The final step moves from atmospheric science into human biology. The researchers applied modern epidemiological formulas, the kind used to relate outdoor air lead levels to blood lead levels in children, translating their atmospheric estimate into a biological one: an average childhood blood lead level of about 3.4 micrograms per deciliter, roughly 2.4 above natural background. From there, established research linking blood lead to cognitive outcomes gave the study's headline figure: a population-wide IQ decline of 2.5 to 3 points across the empire. The authors even suggest the resulting immune suppression may have helped set the stage for the Antonine Plague, which killed an estimated 5 to 10 million people between 165 and 180 CE.

The ice-core and atmospheric modeling portion of this work is careful and technically impressive, and it is not the subject of this critique. What this technical note takes a closer look at is the final leap, from a single, empire-wide atmospheric estimate to a single, empire-wide claim about blood chemistry and cognition. That leap rests on assumptions about geography and human biology that the underlying data cannot fully support.

II. The Trouble With Averaging a Continent

The core problem is this: statisticians call it the ecological fallacy, the error of assuming that something true of a group, on average, must be true of every individual within it. If a country's average household income rises, that alone does not tell you whether any particular family got richer — it describes the group, not any one household in it.

The McConnell study builds much of its final claim on exactly this kind of reasoning. It starts with a single reconstructed estimate of atmospheric lead over Europe and treats that number as though it describes the exposure of ordinary people from Hadrian's Wall to the Levant, from Iberia to the Black Sea, across two centuries. But the Roman world was not one place — it was a patchwork of dramatically different landscapes, climates, and ways of life, and airborne lead would not have settled evenly across it.

Consider what actually determined how much lead a given person breathed. Terrain, coastlines, and prevailing winds all shape how airborne particles disperse and settle, so two towns at a similar distance from a smelting site could see quite different exposure depending on local geography. Urban centers like Rome added lead through plumbing, tableware, and cosmetics that most rural populations never had, leaving them more dependent on whatever drifted through the outdoor air — and some cities drew water through lead pipes while others relied on wells or aqueducts with little lead contact at all. A model built entirely around airborne lead cannot capture any of that. Proximity to an actual mine mattered too: someone near the silver operations at Rio Tinto in Spain, or in the mining districts of Roman Britain, almost certainly inhaled more lead than a farmer in Egypt, however similar their empire-wide "average" exposure looks in a model.

This is not just a theoretical concern. A December 2025 review by Simpson and Garvie-Lok in the Journal of Roman Archaeology compiled every published lead measurement from Roman-era skeletal remains and found exactly this kind of scatter. Tooth enamel levels, the most methodologically consistent measure across studies, ranged from 0.03 to 187 micrograms per gram depending on the site. Samples from Rome and Roman London ran higher than samples from more rural regions in present-day Germany, Croatia, and Spain, though the authors caution that differences in preservation make site-to-site comparisons imperfect. Even with that caveat, the variation this note argues for on first principles is visible in the physical evidence that already exists.

To its credit, the study tested two scenarios for where the pollution originated — one concentrated in a known mining region, one more dispersed — and found similar continental averages either way. That's a reasonable check on the sourcing question, but it answers a different one than the question raised here. Both scenarios model where the pollution started, then let the same atmospheric physics carry it across the same continent; neither asks how that pollution then settled unevenly onto cities versus farmland, or onto a child in Londinium versus one in the Nile Delta.

None of this means the atmospheric science is wrong. It means that turning a continent-wide average into a specific, population-wide health outcome requires geographic detail the current model doesn't include. Closing that gap would call for localized spatial regression modeling: statistical tools that let estimated exposure vary by place, accounting for terrain, weather, and distance from known industrial sites, rather than collapsing an empire into one figure.

III. Bodies Are Not Interchangeable

Even a reader willing to accept that continent-wide estimate runs into a second problem the moment it becomes a blood lead level, and then an IQ estimate.

That conversion relies on equations built from modern populations — children in wealthy, industrialized countries, whose bodies absorb and process lead very differently than bodies in the ancient Mediterranean would have. Lead is not absorbed at a fixed, universal rate. Toxicological research cited in the Simpson and Garvie-Lok review points to nutritional status as a key variable: the gut absorbs more lead when the body is deficient in calcium, iron, or vitamin C, because lead competes with those minerals for the same absorption pathways, and fasting or general undernutrition amplifies the effect. Children are especially sensitive, since they absorb more lead per unit of exposure than adults do, and their nutritional status tends to be less stable to begin with.

Roman populations differed from modern reference groups on nearly every one of these fronts. The typical diet leaned heavily on grain, and calcium and iron intake varied by class, region, and season — often falling well below what a well-nourished modern child receives. Intestinal parasites were also widespread across the Roman world, well documented in archaeological studies of ancient latrines and preserved waste from Britain to the eastern Mediterranean, and chronic parasitic infection itself depletes nutrients and alters gut absorption. Almost none of this applied to the 20th- and 21st-century children whose blood lead and IQ data underlie the study's dose-response curves; those cohorts came from well-nourished, industrialized populations with reliable access to iron and calcium and comparatively low rates of deficiency.

This matters because the resulting error would not simply widen the margin around a good estimate — it could point in different directions for different groups. A population with worse baseline calcium and iron status might absorb more lead per unit of exposure than the modern reference population, pushing the true impact higher than the study estimates; a better-fed, less parasite-burdened group might absorb less. This is not purely hypothetical: the skeletal evidence measures a different quantity than childhood blood lead level, so it can't check the study's numbers directly, but it does speak to the same underlying question of variation, and Simpson and Garvie-Lok conclude that measured lead burdens across Roman sites are generally not so uniformly severe as a single population-wide estimate implies. Wealthy Romans and rural subsistence farmers, whose diets and health looked very different, could plausibly have converted identical air exposure into meaningfully different blood lead levels — a problem for applying one modern dose-response curve to all of them alike.

None of this is an argument that lead exposure was harmless in the Roman world. Lead's toxicity, especially to developing brains, is about as well established as anything in environmental health science, and there is good reason to think Roman-era exposure was substantial. A 2.5-to-3-point shift in IQ is genuinely modest at the level of any one person, well within ordinary test-to-test variation, but the study's claim operates at the population level, where small shifts compound: applied across an empire of tens of millions, a few IQ points can plausibly translate into measurable effects on literacy, administration, and economic productivity, which is exactly why the number is worth scrutinizing carefully rather than dismissing as trivial. The point of this note is narrower than a dismissal: the study's specific figures — a 2.4-microgram-per-deciliter enhancement, a 2.5-to-3-point IQ decline — carry a deeper kind of uncertainty than a standard margin of error can capture, and are better read as an illustrative order of magnitude than a precise, verified measurement of what happened inside any real Roman body.

IV. What the Ice Can, and Can't, Tell Us

None of the above should read as a dismissal of the McConnell study, which remains a genuinely impressive piece of science. Reconstructing continent-scale atmospheric conditions from 2,000 years ago, precisely enough to trace individual centuries of Roman industrial activity, is a real achievement with no close precedent. Researchers who work directly with Roman-era skeletal remains have generally called the underlying pollution reconstruction solid, even as they flag the later analytical steps as shakier. Sean Scott, a chemist at Pacific Northwest National Laboratory who has studied lead exposure in Roman skeletal material, has pointed out that stacking one estimate on another — ice core to blood lead level to IQ score — compounds uncertainty at every step. Even McConnell himself has acknowledged that science doesn't yet have a well-quantified way to connect ambient air pollution, the blood lead levels it produces in children, and the long-term health consequences that follow.

That is the most useful way to read this research: a strong answer to one question, and a preliminary, order-of-magnitude answer to a much harder one. Ice cores tell us, with real confidence, that Roman industrial activity polluted the atmosphere across an entire continent for roughly two centuries — by the study's own estimate, at about a third of the cognitive impact of 20th-century industrial pollution at its peak. That alone is a striking finding, since it overturns the assumption that the pre-industrial world was environmentally pristine. What ice cores cannot yet tell us, with the same confidence, is how many IQ points that pollution cost any particular Roman farmer, soldier, or senator, because that answer depends on geographic and biological variables the model hasn't incorporated.

Closing that gap looks solvable rather than fatal, and the fix runs in two directions. Future work could pair the atmospheric reconstruction with localized spatial regression models — statistical tools that let exposure vary by terrain, weather, and distance from known mining sites, instead of one figure for an entire empire. It could also lean more on paleopathological evidence: the lead content measurable in Roman-era teeth and bones at specific sites, combined with diet and parasite-load reconstructions from the same locations. Matching that bottom-up evidence against the top-down atmospheric model would let researchers check, place by place, whether the continental average is a reasonable stand-in for local reality. Simpson and Garvie-Lok make almost exactly this recommendation themselves, calling for research that integrates bioarchaeological and environmental evidence rather than treating them separately.

That caution reaches beyond this one study. Paleoclimatology increasingly uses ice cores, tree rings, peat bogs, and other environmental archives to make claims not just about ancient weather but about ancient human health — linking volcanic eruptions to famine, or pollution records to disease burden, in ways that carry the same appeal as this study's headline number. The gap between a robust continental average and a precise local biological outcome is not unique to Roman lead; it is a structural feature of any research that tries to connect a global environmental record to a local human one. Building spatial and biological uncertainty into these studies from the outset, rather than treating it as an afterthought once the headline number exists, would make the whole growing field more credible, not less ambitious.

Until that work is done, the responsible way to describe this study's headline number is as a plausible, illustrative estimate rather than a settled fact. The gap between that estimate and the messier picture emerging from bone and tooth enamel is itself the lesson: a striking number from a model is not the same thing as a verified measurement, and the two are worth telling apart, here and in the next study like it. Rome really was polluting its own air on an industrial scale long before anyone had a word for the Industrial Revolution. Precisely what that cost any single Roman's mind is a question the ice, and now the bones, have not yet finished answering.