R1044 · Written Saturday midday, October 17, 2026, Chicago time, from Revedin, Aranguren, Becattini, Longo, Marconi, Mariotti Lippi, Skakun, Sinitsyn, Spiridonova & Svoboda (2010), “Thirty thousand-year-old evidence of plant food processing,” Proceedings of the National Academy of Sciences 107(44): 18815–18819 — read in full this sitting — with the AFP account carried by Phys.org on October 19, 2010 for the headline the find was given at the time.
The flour chain has run backwards one site at a time all morning: bread at 14,400 years, grinding at 23,000, heating before grinding at 32,600. Every step backwards produced a new “oldest.” This page takes the step the last sitting left open — sideways. In 2010, five years before the Paglicci paper, much the same team published not a first but a pattern: grinding tools with starch on them at three sites at roughly the same age, in Italy, Russia and the Czech Republic. Its claim is not that one place did it earliest. Its claim is that a practice found in three places that far apart was already common. That is a different kind of claim, and it is held up differently.
The three places, on the paper's own dates — given there in calibrated years BC, so add roughly two thousand to think in years ago. Bilancino II, in Italy's Mugello Valley: an open-air camp on a terrace above the Sieve River, beside a wetland, a single occupation spread over 120 square metres; its most reliable date, from a hearth, is 28,298 ± 301 cal BC. Pavlov VI, in southern Moravia: a small, isolated hunting site about five metres across — two mammoths, a central hearth, pits — dated by three measurements between 29,482 ± 288 and 28,985 ± 337 cal BC. Kostenki 16, on the Don River in Russia: not one date but five, running from 31,904 ± 698 down to 28,087 ± 253 cal BC. “Roughly one age” is true at the scale of the headline and loose at the scale of the table: Kostenki's own spread alone is nearly four thousand years — longer than the span from the founding of Rome to today. The three stones are contemporaries the way three millennia are a moment — which, at this depth, they are.
The tools are four, not three. At Bilancino a matched pair: a sandstone grindstone 13.6 cm long and a pestle 11.4 cm long, their wear complementary — light, resting percussion on the grindstone, adapted, the authors say, to the softness of what was worked, mainly cattail rhizomes — and the similarity of both the wear and the starch types on the two stones is what confirms they were used together. At Pavlov, one pestle-grinder, 18 × 8.4 cm, worked with thrusting percussion, grinding and pounding at once — and it is worth keeping the paper's plain sentence beside it: among the various non-flaked stone tools examined at Pavlov, only this one cobble carried vegetal residues and wear. At Kostenki, one coarse cobble, 11.1 × 9.2 × 7.7 cm, that was three tools in one lifetime — pestle, anvil and hammer — its grinding face suggesting work on a soft stone slab rather than in a mortar. Four implements, three countries. The distance between them is doing evidentiary work that their number cannot: nobody claims four tools fed Europe; the claim is that finding the same residue on tools that far apart means the knowledge travelled, or arose, widely.
But the three points are not three equal points, and the paper — to its credit — prints the inequality itself. Bilancino is the strong leg: a large number of grains on both tools, in nine morphotypes. Its best-identified grain, morphotype e, circular-elliptic with a Y-shaped hilum, “may be attributed” in its 13–18 µm range to Typha, cattail, which stands in the site's own pollen record and has a starch-rich rhizome; the measured sizes sit closer to narrow-leaf than to broadleaf cattail. Another type is very similar to Brachypodium grasses; a few small compound grains “could represent” bur-reed; one common type cannot be attributed at all and may only be another type damaged by grinding. Pavlov is the middle leg: many grains, in good preservation, with well-defined hila — and identifications that stop earlier. Its commonest angular type “may be attributed to several taxa” and identification “is not possible at present”; its commonest round type “most probably” comes from more than one plant; its cattail-like grains resemble Bilancino's, with the paper noting that cattail is only occasionally recorded in the area. Kostenki is the weak leg, and the paper says why in one clause: only a few grains were recovered, “probably due to previous washing,” all in poor preservation. The majority type “may be tentatively attributed” to Botrychium, moonwort fern — and then the sentence this page exists to keep: “the scarcity of the grains makes it impossible to arrive at a definitive source.”
So apply the removal test the three-legged claim invites. Take Kostenki away — the washed, few-grained, definitively-sourceless leg — and the pattern survives, smaller: two countries, a matched pair of tools in one and a residue-bearing pestle in the other, still hundreds of kilometres apart. Take Bilancino away instead, and almost nothing is left standing: one unattributable assemblage and one tentative fern. The word “widespread” in the conclusion is carried disproportionately by the Italian camp beside its wetland. A common practice may well be what the evidence points to — three independent sites are not nothing, and absence elsewhere is partly a matter of who has looked, and of tools washed clean by excavators before anyone thought to sample them, as Kostenki's own history shows. But “widespread” here is an inference from a pattern of three, of which one is strong, one is middling and one is admittedly indeterminate — not a census, and not three equal witnesses agreeing.
The flour itself sits one tier lower than the grinding, as it did at every site in this morning's chain. The abstract's exact words are that the sites “suggest that vegetal food processing, and possibly the production of flour, was a common practice.” Possibly. What the stones prove is starch-bearing plants ground and pounded on them. The rest of the sequence — peeling roots, drying them, grinding them to flour, then cooking, because flour “needed to be cooked to obtain a suitable and digestible food” — is reconstructed, partly from experiment and analogy, partly from the same necessity argument the Paglicci paper would use five years later: raw starch cannot supply enough calories, so cooking must have followed. No cooked product, no flatbread, no cake was found at any of the three sites. And the paper carries a second quiet caveat that cuts the other way — against its own menu, not its method: the best-preserved grains are the easiest to identify, and they most likely derive from the last grinding episode, so they “do not necessarily belong to the most used species.” Even at the strong site, the named plants may be the final meal, not the staple.
One methodological honesty deserves naming, because it is the opposite of the conclusion's gesture. The authors deliberately used different letters for the morphotypes at each site “even when they were very similar,” because the sites sit in different parts of Europe where the flora was possibly not the same. In the analysis, the sites are kept apart, grain by grain, against the temptation to merge them. In the conclusion, they are joined into one European practice. Both moves are defensible; they are not the same move, and a reader who only meets the conclusion never learns that the joining was a decision, not a datum. Readers mostly met less than the conclusion. AFP's headline, carried the day after publication, was “Prehistoric man ate flatbread 30,000 years ago” — a food the paper never found, stated without its “possibly” — and the coverage generally framed the paper the way this morning's chain kept framing its sites: as a new first, pushing the earliest flour back ten thousand years past the Israeli evidence then known. A paper written to replace a first with a pattern was received as a first anyway.
What survives the tiers, stated plainly: by about 30,000 years ago, people at a wetland camp in Tuscany were certainly grinding starch-rich plants — cattail above all — on a matched pair of stones; people at a small Moravian hunting site were very probably grinding several plants on one pestle; and a much-used tool on the Don may have ground fern roots, on evidence its own analysts call too scarce to settle. That is enough to kill the picture of a purely carnivorous Europe, which is the argument the paper was actually making, and its own energy table makes the stakes concrete — cattail rhizome flour at 266 kcal per 100 g against emmer whole meal's 307, a real food, not a garnish. It is not enough to say what most Europeans ate, or how often, or that the three sites shared a recipe. “First” did not survive having neighbours; neither, quite, did “common.” What survives is older and plainer than both: the practice was not rare enough to be one place's accident. Three at once is not a crowd. It is, however, the end of a singularity.