The fragments in question, often described in the literature as “debitage,” are the crumbs of sarsen — a hard, silicified sandstone — left behind or displaced during the monument’s construction and later disturbance. While the towering trilithons have attracted centuries of attention, these palm-sized chips have long been regarded as geoarchaeological afterthoughts. That changed when previous researchers applied geochemical provenancing to the fragments, measuring their elemental compositions and comparing them with known sarsen sources across southern Britain. That earlier work concluded that most of the stones, including the massive sarsen megaliths themselves, could be traced to the Marlborough Downs, roughly 25 kilometers north of Stonehenge, while some fragments pointed to more distant or exotic origins.

Michelaki and her colleagues — including David Barham, Michael P. Gorton, William C. Mahaney, Susanne Aufreiter, and Ronald G. V. Hancock — did not collect new samples. Instead, they took the previously published datasets and subjected them to a different kind of analysis. The crux of their critique concerns normalization, a widely used technique in geochemistry in which the concentration of each element is divided by the concentration of a reference element — in this case zirconium, or Zr. Normalization is intended to strip away noise caused by dilution effects, such as varying amounts of quartz in a rock, allowing researchers to compare the underlying chemistry of samples on a level playing field.

The technique has deep roots. Archaeometric studies of pottery, sediments, and metals have long relied on normalizing elements — aluminum for estuarine sediments, scandium for ceramic pastes, zirconium for silica-rich rocks — to counteract the enormous spreads in raw elemental concentrations that obscure meaningful comparisons. In the Stonehenge work, Zr normalization formed the backbone of the provenance assignments: when fragment ratios matched the ratios of a candidate source, the fragment was deemed to originate there. On the surface, the logic is sound and elegant.

But the new reanalysis shows that the elegance comes at a cost. When the same data are plotted without normalization, using raw elemental concentrations and a visual, geochemically informed graphical approach that explicitly acknowledges analytical variability, a different picture emerges. Similarities that appeared compelling under Zr-normalized ratios are not always supported across multiple elements. In some cases, normalization generated apparent agreements between fragments and sources that dissolve the moment other elements are considered. In other cases, genuine coarse geochemical groupings visible in the raw data simply vanish when everything is divided by zirconium.

The problem, the authors explain, is mathematical as much as geological. Dividing one element by another collapses two independent pieces of information into a single ratio, and if the normalizing element itself varies substantially across samples — as Zr can in silcretes, where zircon-bearing grains may be unevenly distributed — the ratio can swing wildly for reasons that have nothing to do with shared provenance. Two chemically unrelated rocks can end up with similar ratios, while two chemically related rocks can be driven apart. The result is a classification scheme that may obscure real compositional differences and invent apparent ones.

So what happens to the Stonehenge fragments when the crutches are removed? Using their non-normalized approach, Michelaki’s team found that only some of the fragments can be tentatively associated with previously characterized sarsen sources. Many others remain stubbornly unsourced, or appear to represent geochemistries that have not yet been documented anywhere in Britain’s known sarsen outcrops. The authors are careful with their language — they use “grouping” rather than “group” to stress that these chemical clusters are far more dispersed than the tight compositional families typical of a single geological source. But the implication is striking: the geochemical variability of the silcrete materials at Stonehenge may be considerably greater than the earlier analyses suggested.

Far from undermining the broader narrative of Stonehenge’s construction, this increased variability actually aligns with a growing body of evidence that the monument is a composite of materials drawn from across Britain. The most dramatic example is the Altar Stone, whose recent sourcing to the Orcadian Basin in northern Scotland — hundreds of kilometers from Wiltshire — stunned the archaeological community and demonstrated that long-distance stone transport, whether by human effort or glacial action, was part of the monument’s biography. The bluestones from the Preseli Hills of Wales tell a similar story. Against this backdrop, sarsen fragments with unassigned or unexpected chemistries are less an anomaly than a further hint that the builders drew on a wider and more varied portfolio of stone than once assumed.

The study is also notable for what it declines to do. Rather than offering new definitive provenance assignments, the authors frame their contribution as a methodological intervention: a reflexive evaluation of how analytical choices shape archaeological interpretation. The Stonehenge case becomes a cautionary tale for provenance studies in general. Archaeometry has a long history of these debates, from the Olmec controversy, where competing statistical treatments of neutron activation data produced radically different pictures of ceramic exchange in ancient Mesoamerica, to ongoing discussions about how to define chemical reference groups in pottery. The lesson, repeatedly learned and just as repeatedly forgotten, is that data do not interpret themselves — and that a technique convenient enough to become standard practice deserves periodic scrutiny.

The timing of the intervention is significant. The original provenancing of the sarsens, published in Science Advances in 2020, was widely celebrated as solving one of Stonehenge’s enduring mysteries, and subsequent work on the fragments extended that framework. A parallel debate has been playing out in the pages of the journal Archaeometry over the sourcing of sarsen stone 58, with the present team contributing geochemical analyses of that monolith and engaging in exchanges with other researchers over methodology. The new paper effectively widens the conversation: if normalization can obscure differences and manufacture agreements in the fragment data, then provenance claims built on those ratios — for the fragments and potentially for the megaliths themselves — warrant reexamination under multiple analytical lenses.

None of this means the Marlborough Downs connection is wrong. The 2020 study’s core insight, that the great sarsens share a distinctive chemistry matching a specific outcrop called West Woods, was based on a robust correspondence and remains the most parsimonious explanation for the megaliths. But the new work insists that confidence should be calibrated to the evidence. Some fragment-to-source matches may hold up under non-normalized scrutiny; others may not; and a substantial portion of the material may simply await the characterization of sarsen sources that have not yet been sampled. Britain’s sarsen distribution is patchy and imperfectly mapped, scattered in a belt from Dorset through Wiltshire to Kent and beyond, and glacial processes during the last British-Irish Ice Sheet’s advance and retreat have redistributed stone across the landscape in ways that complicate simple local-versus-exotic dichotomies.

For the wider field, the stakes extend well beyond one monument. Provenance studies underpin claims about trade networks, migration, territoriality, and social organization from the Neolithic to the historic period, and they increasingly feature in heritage policy debates about stone conservation and sourcing. When a single statistical transformation can tip the balance between “local stone” and “stone from nowhere we know,” the responsibility on practitioners is considerable. The authors’ recommendation is not to abandon normalization outright — it remains a legitimate and often useful tool — but to apply it reflexively, testing whether conclusions survive when the data are viewed in their raw form, element by element.

In the end, the study is a portrait of science working as it should. Two teams, using the same numbers, reach different conclusions, and the disagreement itself becomes a discovery: the stone fragments of Stonehenge are more chemically diverse, and their journeys more complicated, than anyone had asserted. For a monument that has resisted definitive explanation for a thousand years of inquiry, that may be fitting. The researchers, who note their work was enabled by “curiosity, stubbornness, and retirement,” have reminded the field that at Stonehenge, as in geochemistry, certainty should always be earned — and re-earned — rather than assumed.