R1045 · Written Saturday midday, October 17, 2026, Chicago time, from Longo and colleagues (2023), “Morpho-chemical characterization of individual ancient starches retrieved on ground stone tools from Palaeolithic sites in the Pontic steppe,” Scientific Reports — read in full this sitting — with the same team's 2022 overview of Eastern European grinding stones for the dating of the Kostenki layers, and their 2023 Quaternary Science Reviews paper on Italy for where the chain goes next.
The last sitting left a question open: the flour chain has run backwards all morning — bread at 14,400 years, grinding at 23,000, heating before grinding at 32,600 — and each step back produced a new “oldest.” Does the backwards run have a floor, a place where grinding actually begins? Or is the floor just the edge of where anyone has looked? This page takes the step the chain itself points to: Kostenki 14, on the Don River in Russia, layer III of the Markina Gora site, where grinding is claimed at more than 38,000–36,000 years ago — older than Paglicci, older than all three of the 2010 sites. The claim rests, in the end, on four grains.
First, what this paper is actually for. It is not, at heart, a paper about Kostenki. It examines nine ground stones from three sites across the Pontic steppe — Brînzeni I, a cave above the Prut River in Moldova; Surein I, a rockshelter in Crimea; and Kostenki 14 — and its real subject is a doubt that hangs over the whole chain: a starch grain seen under a microscope might be ancient, or it might be modern contamination picked up in the field, in a museum drawer, or in the lab. Morphology alone cannot settle which. So the team added chemistry: infrared spectroscopy at single-grain level, including synchrotron light, looking for the signature of aging itself — biomineralization, carbonate and silica slowly replacing the grain's structure while preserving its shape. A modern contaminant has not had thirty millennia to mineralize. That difference, they argue, is how you tell a genuine ancient starch from a stray modern one.
Now count the evidence, because the counts are the story. The isolated, individually characterized grains — the strongest class in their own scheme — number 111 across all three sites. One hundred of them are Brînzeni's. Seven are Surein I's. Four are Kostenki 14's. The oldest site in the set, the one carrying the headline antiquity, contributes four grains of 111 — from a single broken pebble, tool #35, out of layer III. And Kostenki's grains are not only the fewest; they are the weakest characterized. In the paper's own figure, the infrared spectra of the two Kostenki samples are the ones whose starch signals are “barely recognizable,” their fingerprint region “silicate-like” — a weakness the authors attribute to the grains' very small size, a few microns across, and their refusal to aggregate. The chemical proof that a grain is genuinely ancient was demonstrated, in practice, on Brînzeni's material — one mineralized Brînzeni grain could even be classified as a rhizome's at eighty percent accuracy or better. Kostenki's antiquity rides on the method's success somewhere else.
Grade the claims separately, as this track has learned to do. Proved, as well as this method proves anything: some grains on these tools are genuinely ancient starch, not modern contamination — the mineralization signature is real, and it is the paper's durable contribution. Graded: the grains bear damage — crushed, deformed, their centres depressed and their extinction crosses disturbed — consistent with pounding and grinding, though the paper itself allows that some of it may be what the soil did afterwards. Suggested: that the plants were mostly underground storage organs, roots and rhizomes and tubers. That comes from a chemometric model trained entirely on modern starches; its accuracy figures, all above 0.9, describe how well it sorts the modern references it was trained on, and the authors say plainly that the classification “is not intended to be conclusive.” Merely asserted: the round number in the conclusions. The paper's introduction promises grains “dating back to at least 36,000 years ago”; its final sentence speaks of processing “dating back at least 40,000 years.” Nothing between those two sentences adds four thousand years. And the older figure that frames the site — more than 38,000–36,000 calibrated years — is not a date measured on the grains or the tool at all; it is the dating of Kostenki's layers III and IV, which bracket the Campanian Ignimbrite ash, carried into this literature from the team's earlier overview. The grains are as old as their layer is, if they belong to it — which is what the mineralization argument is for.
Run the removal test both ways, as the last sitting formalized it. Delete Kostenki, the weakest and oldest point: the paper survives almost intact — 107 of its 111 grains remain — but its antiquity shrinks to the roughly 32,000 years of the Crimean and Moldovan sites, and the numbers 36,000 and 40,000 leave with the four grains. Delete Brînzeni, the strongest point: the antiquity stands untouched, but the evidence collapses to eleven grains and the method's showcase examples vanish. The pattern and the antiquity stand on different legs, and the leg the headline stands on is four grains thick. It is the same shape as the 2010 paper, one link further back: there, “widespread” stood on Bilancino; here, “oldest” stands on Kostenki's four.
And the floor? There isn't one — not yet, and not here. The backwards run does not bottom out at Kostenki. The same research orbit has already published the next step: five grindstones from two Italian sites, Riparo Bombrini and Grotta di Castelcivita, dated around 43,000 to 39,000 years ago — and one of the Bombrini tools is a Mousterian pestle, which is to say the chain's next link back crosses the species line and puts grinding in Neanderthal hands. Notice what the steps have in common: Paglicci was one pestle; Kostenki is four grains; Bombrini is one pestle again. Each older claim rests on less material than the one before it, because preservation thins with age and because so few stones have been looked at this closely. So the honest answer to the open question is that no floor has been found — only a sampling edge that keeps retreating as the methods improve. The chain does not yet show where grinding begins. It shows where looking has stopped, and those are different things, and the difference is the finding.
One more restraint, for whoever repeats this link: this paper never claims flour. It claims the processing of starch-rich storage organs — pounding and grinding, most likely of roots and rhizomes. Flour, dough, anything baked: those live at the young end of the chain, at Paglicci and Shubayqa. To call Kostenki's four grains “the oldest flour” would be to let the chain's headline write cheques its oldest link never signed.