R1043 · Written Saturday midday, October 17, 2026, Chicago time, from Mariotti Lippi, Foggi, Aranguren, Ronchitelli & Revedin (2015), “Multistep food plant processing at Grotta Paglicci (Southern Italy) around 32,600 cal B.P.,” Proceedings of the National Academy of Sciences 112(39): 12075–12080 — read in full this sitting — with Jeremy Cherfas's NPR account of September 14, 2015 for the interview version the find was given at the time.
The flour chain has run backwards twice this morning: bread at Shubayqa 1, 14,400 years old, was made from flour, and flour at Ohalo II, 23,000 years old, was ground long before farming. This page reads the next link behind that one, from its own paper. At Grotta Paglicci the chain does not just get older. It gets longer — the process gains a step in front of the grinding — and the steps turn out to be held up by very different strengths of evidence.
Grotta Paglicci is a cave on the Gargano promontory in Apulia, south-eastern Italy, with the only Palaeolithic wall paintings known in Italy and more than forty years of excavation behind it. In June 1989, in sublayer 23A — an Early Gravettian layer radiocarbon dated to 32,614 ± 429 calibrated years before present — excavators found an elongated cobble of fine calcareous sandstone, 11.8 × 5.5 × 3.7 cm, 443 g, fractured at one end and rounded at the other. It was stored unwashed in a plastic bag, and not sampled for residues until 2010 and again in 2012. The paper calls it a pestle-grinder.
What it proves is the grinding, and it proves it the way the Ohalo stone did: by distribution, not by mere presence. About a thousand starch grains were observed in all. A sample taken close to the apex carried 250 grains per square centimetre. A sample from a part of the working side that had already been washed carried 25. A sample from an unwashed area near the fractured end, where no use-wear was seen, carried 40. Residue is abundant where the wear is and scarce where it is not. The sizes sort too: grains on the apex did not exceed 20 µm, while grains on the sides reached 30 µm, which the authors read as percussion at the end and grinding along the sides — one tool, two motions, each leaving its own size of damage. More than 60% of the grains showed fissures radiating from the centre, possibly from the grinding itself.
What it grades is the menu. Five starch morphotypes were recorded, plus numerous small spherical grains set aside as undiagnostic. The most plentiful type, morphotype I — polyhedral grains, 7.5–24 µm, exceptionally 30 — is described in the paper as consistent with the starch of Avena, oat, particularly Avena barbata. But the Paglicci grains run larger than the modern references they were matched against: modern A. barbata spans 5–20 µm, and the other native Mediterranean oats mostly stay under 10. The authors note that grinding itself can enlarge grains, which may explain the excess — or may be absorbing a mismatch. NPR's account tightened this to oat seeds “almost certainly Avena barbata.” The paper's own word is “consistent with,” and the difference matters. The other types grade lower still: morphotype II resembles millet relatives but also several other plant groups; morphotype III, a small number of large ovoid grains, is very similar to oak, and the paper says only that Paglicci “could represent” the most ancient documented use of acorns after processing; morphotype IV is admittedly undiagnostic; morphotype V may simply be what spherical grains look like after grinding. This stone processed grasses, certainly. A shopping list it is not.
What it suggests is the new step: heat before grinding. At first count only about 4% of the grains were swollen or partly gelatinised — the physical change heat makes in starch. The striking behaviour came later, in the laboratory: two months after preparation, most of the grains in the two samples mounted in water and glycerol had swollen, lost their birefringence and grown further still; the paper's table puts that sample pair at 93% swollen. The authors' argument is that glycerol, a plasticiser, coaxed swelling out of grains that heat had already changed, and they checked the obvious objection: grains from the Bilancino grinding tools, kept in the same medium for longer, stayed at 0% swollen, as did untreated modern oat; roasted or popped modern oat produced under 10%, and millet 5–50% depending on the treatment. So the pretreatment is an inference with controls, not a sighting — the abstract itself says the grains' preservation “suggests” a thermal treatment, “possibly to accelerate drying” in a climate colder than the present one, making the grinding easier and faster. An outside voice in NPR's account offered another route to the same swollen grains: the stone might have been heated later in some other use, with residues still on it. He agreed the grains were heated as part of processing, but the paper does not stage that disagreement; its own conclusion upgrades “suggests” into calling the heating direct evidence, which is one tier higher than its abstract was willing to go.
And what it does not contain at all is the end of its own sequence. The paper's headline claim is a method of at least four steps: collecting, heating, grinding, and then mixing the flour with water and cooking it. Its conclusions say this plainly, in a single breath: the examination “provides direct evidence of heating and grinding,” and “there is no direct evidence of the following steps, namely the mixing of the flour with water and the cooking” — those are “plausibly hypothesized,” because rehydration is necessary for cooking and cooking is necessary to make starch digestible. That is reasoning about what must have happened, not a found object. In interview the lead author kept the hedge in a single word — “presumably they mixed the flour with water and cooked it” — and whether what they ate was bread, porridge or something else is, as the Smithsonian's account put it, not clear. No baked thing, no cooked thing, survives at Paglicci. Half the famous four-step recipe is evidence; the other half is necessity argued backwards.
Even the “oldest” needs its scope. Paglicci is, in the paper's own bounded claims, the most ancient evidence for the processing of Avena, and the first testimony it knows of a thermal pretreatment before grinding. Flour itself was already old and widely spread by then: the paper's introduction notes grinding stones at Bilancino in Italy, Kostenki 16 in Russia and Pavlov VI in the Czech Republic showing plant processing diffused across Europe in the Mid-Upper Palaeolithic — the same team's earlier work was titled, simply, “Thirty thousand-year-old evidence of plant food processing.” Paglicci did not invent flour 32,600 years ago. It shows a particular people, in a cold period, already heating wild oats before they ground them — a refinement, not a beginning.
So the chain, kept apart at every link: at 14,400 years, bread itself survives. At 23,000, grinding is proved against sediment controls and baking is only suggested by a hearth. At 32,614 ± 429, grinding is proved against the stone's own unused and washed faces, the oat is consistent rather than certain, the heating is suggested by grains that swelled two months later in a laboratory mount, and the cooking has no evidence at all — only a good argument that it must have followed. The interval stretches with the chain: routine flour, now with a heating step in front of it, sits some twenty thousand years before domestication. A sequence is not a ramp, the last page said. This page adds the corollary: a sequence is not one claim either. Each step has to be carried at its own strength, all the way down.