African Pearl Millet Was Not an Independent Invention

TL;DR

• Patrick Munson argued in 1976 that cultivation reached the western Sahel from outside, and that local people then domesticated a local grass, pearl millet, by applying a borrowed technique.
• He named the route it travelled. Adrar des Iforas, down the Tilemsi Valley, to the Niger bend, then west to Dhar Tichitt. No domesticated cereal had come out of the Tilemsi at the time.
• What turns on this is whether Sahelian agriculture was invented or inherited. The plant is African and nobody disputes it. Whether the practice began there is a separate question.
• In 2011 the world’s earliest domesticated pearl millet was published from the Lower Tilemsi Valley, several centuries older than the millet at Tichitt. The ordering he predicted is the ordering the evidence shows.
• His absolute dates were wrong by roughly eight hundred years, which is what happens to a chronology built before direct dating of grain. Recalibration moves dates; it does not move routes.
• The prediction was printed thirty-five years before the data and cannot be retrofitted. It is a registered prediction in everything but name, it came in, and nobody has scored it.

“Thus the ’lines of communication’ were present.” — Patrick J. Munson, Origins of African Plant Domestication (1976), p. 202

What Munson predicted in 1976 #

Patrick Munson dug Dhar Tichitt, the escarpment settlements in south-central Mauritania, in 1968, and wrote the sequence up as his dissertation three years later (Munson 1971). He went there to test George Peter Murdock’s claim, in Africa: Its Peoples and Their Culture History, that West Africans had invented seed agriculture on their own in the fifth millennium BC, somewhere near the headwaters of the Niger (Murdock 1959: 64–70). The claim was already contested on botanical grounds (Baker 1962), but it had not been tested with a spade. Munson’s sequence did not support it. What he found instead was a population that acquired domesticated animals first and crops much later, and a transition into intensive millet cultivation so compressed that he could not read it as a local invention.

So he asked where the practice came from, worked through the candidates, and rejected three of them. Not an independent invention at Tichitt, because cultivation was already thousands of years old in northeastern Africa and because domesticated cattle and goats had reached Tichitt some five hundred years before the first evidence of planting, and those animals were unambiguously introduced from outside. Not an earlier hearth to the south, because nothing in the savanna dated earlier than Tichitt. Not Ethiopia, because the sites in between are later, not earlier, which is the wrong shape for a westward spread.

What he was left with was contact, and he said what contact would look like. On page 202:

“Near Eastern domesticated crops were demonstrably present in Egypt by 4500 B.C., and Near Eastern domesticated animals were introduced into the central Sahara at a very early date, perhaps prior to 5000 B.C. Thus the ’lines of communication’ were present.”

He was careful about what travelled. Wheat and barley are winter-rainfall crops and the Sahel is a summer-rainfall zone, so the crops themselves could not make the trip. That is not an objection to his model; it is the engine of it. People “confronted with the difficulty of growing the winter-rainfall adapted, Near Eastern crops in this summer-rainfall zone,” he wrote on the same page, “began experimenting with such wild, summer-rainfall adapted, local species as Pennisetum.” The idea transfers and the content is reinvented locally. Kroeber had given that process a name in 1940: the receiving culture accepts the idea and then must “develop a new content,” a mechanism he called “stimulus-diffusion” (Kroeber 1940: 1). Munson uses the term himself, in his own summary on page 205, where he traces the ultimate origin of the Saharan cultivation complex to either “independent seed-crop domestication in the eastern Sahara or stimulus-diffusion from earlier developments in the Near East via Egypt.”

Then he drew the map.

“Pressure in the central Sahara … might well have pushed some of these cultivators southwestward into such southern Saharan highland areas as Aïr and Adrar des Iforas, and from there down such corridors as the Tilemsi Valley into the central valley of the Niger in the vicinity of Timbuktu.” (Munson 1976: 205)

On page 203 he had already laid out the geography that makes the route work: the southwestern tip of the Saharan highlands is the Adrar des Iforas, the great bend of the Niger east of Timbuktu “is connected to this area by the valley of the Tilemsi,” and Timbuktu in turn sits a few hundred kilometres from the southeastern edge of Dhar Tichitt-Oualata. A massif, a valley, a river bend, an escarpment. Four named places in a line.

And he stated the entailment. Moving southward from Tichitt, he wrote on page 204, “the date of initial occurrence will become progressively later.” That is a falsifiable claim about ordering, published in a volume that anyone can pull off a shelf (Harlan, de Wet & Stemler 1976).

Is African pearl millet an independent invention? #

What that model denies is not a small thing.

Pearl millet is one of the standard entries on the short list of places where humanity invented agriculture more than once. The list is how we know domestication is a thing people do under the right pressures rather than a freak event that happened in the Fertile Crescent and then spread. The Near East, China, Mesoamerica, the Andes, New Guinea, and the African Sahel, where a wild grass called Pennisetum glaucum became a crop. Take the Sahel off that list and you have not adjusted a footnote. You have removed one of the independent replications that the whole framework rests on.

Munson’s chapter is an argument for taking it off. Not because the plant came from somewhere else. It did not, and nobody says it did. Pearl millet is African, its wild ancestor is African, and every grain in this story was grown in the Sahel. That was never the question. The question is whether the practice was invented there, or arrived there.

Those are different claims and they pull apart cleanly. A group that already knows what planting is, because their neighbours do it and their animals came from the same direction, faces an ordinary agronomic problem when they reach a summer-rainfall zone: the crops they have seen will not grow here. Solving that problem by taking the local grass into cultivation produces a genuinely African domesticate and is not an independent invention of agriculture. It is an application of an imported one. Munson’s whole case is that the Sahel is the second thing wearing the costume of the first.

That is what is on the line, and it is why the route, rather than the plant, is the evidence that decides it. An invention has no upstream. A transmission does, and it leaves one: a line of places where the practice shows up earlier as you move back toward the source. Munson said the western Sahel would have such a line, said where it would run, and named the valley in the middle of it.

Why the Tilemsi Valley was a strange place to name #

The Tilemsi is a fossil river. It drains the Adrar des Iforas southward across northeastern Mali and meets the Niger near Gao, and it has been dry for a long time. People had been there. Raymond Mauny recorded sites in the valley in 1952, and Andrew Smith excavated Karkarichinkat in the spring of 1972 (Smith 1974). What nobody had was a domesticated cereal out of the ground. The argument about African agricultural origins was being conducted with reference to Tichitt and Dhar Oualata, to the Nile, to Ethiopia, and to Murdock’s hypothetical Niger-headwaters hearth (Amblard 1989).

Munson named the Tilemsi because his model needed a conduit between the highlands and the river, and the map offers exactly one. That is a real inference, not a hedge. He could have written “southward from the Sahara” and stayed safe. He picked a valley.

What was found there in 2011 #

Thirty-five years later, Katie Manning, Ruth Pelling, Tom Higham, Jean-Luc Schwenniger and Dorian Fuller published “4500-year-old domesticated pearl millet (Pennisetum glaucum) from the Tilemsi Valley, Mali” in the Journal of Archaeological Science. The material came from the Lower Tilemsi Valley: charred grain from Karkarichinkat Nord, and chaff of domesticated-type pearl millet used as temper in pottery at the nearby site of Er Negf, which produced an accelerator date of 2579 to 2369 cal BC. It is the earliest domesticated pearl millet anywhere in the world. The people who left it were not casual about the crop: stable isotope work on the same burials puts roughly 85 percent of their diet in C4 plants, while their cattle and small stock grazed largely outside the cultivated zone (Finucane, Manning & Touré 2008). This was a working millet economy, not an experiment.

Set that against Tichitt. A potsherd from Akreijit carrying impressions of domestic Pennisetum dates to 2046 to 1598 cal BC. Along Dhar Néma, millet-tempered pottery gives dates between 1740 and 1510 cal BC (Fuller, MacDonald & Vernet 2007). The most recent synthesis states it flatly: domestic millet was present “between 2500 and 2000 cal. BC in the Tilemsi Valley and Foum el Alba (Mali), spreading rapidly in West Africa and beyond in the first part of the second millennium BC” (McIntosh, Casanova & Gauthier 2026).

Earlier in the Tilemsi, later at Tichitt. That is Munson’s ordering, on Munson’s axis, between the two places he named.

The dates were wrong. The ordering was not. #

Munson put intensive millet cultivation at Tichitt after about 1100 BC. The real figure is most of a millennium earlier. That is a large error and there is no point softening it.

It is also the error you would expect, and it is not the kind that touches his prediction. Munson was dating a sequence by ceramic grain impressions and phase seriation, in an era before anyone could take a single carbonised grain and date the grain itself. Direct accelerator dating of individual seeds and of crop-processing waste in pottery temper is what moved these numbers, and it moved them everywhere, not only at Tichitt. The same technique that broke his chronology built the Tilemsi chronology that vindicates his route.

The distinction is ordinary in archaeology. A stratigraphic sequence and a calibration are different objects. Recalibration of the radiocarbon curve routinely shifts absolute dates across a whole region while leaving the relative ordering of deposits exactly where it was, because ordering comes from the ground and calibration comes from the curve. A diffusion model is a claim about ordering. It says: this place before that place, along this line. It does not say what year, except insofar as the modeller borrows whatever chronology the field has at the time.

Munson borrowed the chronology of 1976 and it was wrong. He built the ordering himself, and it was right. Old claims about African prehistory are often part right and part wrong in ways that only came apart once the dating caught up, and the split rarely falls where their critics expected.

Compare Schliemann at Hisarlik. He identified the mound correctly, and he was badly wrong about which layer was Homeric, digging through the relevant city to reach one a thousand years too early. Nobody scores that as a failed identification. The place was the claim.

A registered prediction, thirty-five years early #

Open science has spent two decades building machinery to manufacture a property that Munson’s chapter has for free.

The point of preregistration is not paperwork. It is that a claim recorded before the data arrives cannot be shaped by the data, and a claim recorded afterward almost always is, without the author noticing. “With hindsight bias,” Nosek, Ebersole, DeHaven and Mellor write, “the observer uses the data to generate an explanation, a postdiction, and simultaneously perceives that they would have anticipated that explanation in advance, a prediction.” Their conclusion is blunt: “Prediction is achieved because selection of tests is not influenced by the observed data, and all conducted tests are knowable” (Nosek et al. 2018).

Munson’s chapter satisfies that standard completely and by accident. It was printed in 1976 by Mouton, in the founding volume on African plant domestication, with a page number. He had no degrees of freedom left. He could not have known that a valley he named for geometric reasons would yield the earliest domesticated pearl millet in the world, because the excavation that produced it lay three and a half decades in the future and the dating method that secured it did not exist. There is no version of this where the prediction was tuned to the result.

That is worth more, evidentially, than a modern paper that recovers the same pattern by fitting a model to data already in hand. We normally have to pay for this property with a time-stamped protocol and a locked analysis plan. Here it is sitting in a 1976 book, unclaimed.

Why nobody has scored it #

Because the two halves of the prediction live in separate literatures, and the people holding one half were not looking for the other.

Manning and colleagues do not cite Munson 1976. His name is absent from the paper’s reference list entirely. They were writing an archaeobotany paper about domestication traits, and their finding, that non-shattering evolved in pearl millet before grain size increased, is genuinely about the mechanics of how a wild grass becomes a crop. Munson’s chapter is a thirty-five-year-old regional synthesis with a superseded chronology, which is precisely the kind of citation that drops out.

The sharper case is the 2026 reanalysis. Susan McIntosh, Emmanuelle Casanova and Caroline Gauthier went back to Munson’s own 1968 excavations at Tichitt, ran new accelerator dates on his curated charcoal, and did residue and petrographic work on his pottery. They cite him. They report the revised Tichitt dates. On the same page, they report that domestic millet was present in the Tilemsi Valley between 2500 and 2000 cal BC. Both halves of his prediction are printed in one paragraph. The words “corridor,” “Adrar,” “Iforas” and “diffusion” do not appear anywhere in the article.

This is what happens to an old paper whose dates went bad. It gets cited for its excavation and quietly dropped for its argument, and after a while nobody rereads the argument to see whether new evidence bears on it. Augustin Holl engaged Munson’s historical model directly in 1985 and faulted it for oversimplifying (Holl 1985). The route itself has not been taken up since. Munson lost on chronology and the field moved on without checking what else was in the chapter.

Does local domestication settle it? #

The Tilemsi material is generally read as a local domestication, and the genetics point the same way: whole-genome modelling puts the origin of the domesticated form in the western Sahara and dates the expansion of cultivated pearl millet to about 4,900 years ago (Burgarella et al. 2018), which Fuller and colleagues gloss as a spread “westward to Mauritania and southward into the savanna south of the Niger River bend” (Fuller et al. 2021). Does that not settle it against Munson?

It does not. Local domestication is not independent invention. The genetics establish that the plant became a crop in the Sahel, which Munson also believed, said in print, and built his model around. He never proposed that pearl millet arrived as a finished domesticate. He proposed that the people who finished it already knew what planting was. Genomes locate the workshop. They are silent on where the craft came from, and they cannot be otherwise, because a grass genome does not record whose idea it was.

Manning’s subtitle points at something harder. Pearl millet took an alternative domestication pathway: non-shattering was fixed by around 2000 BC while grain size kept increasing afterwards (Fuller et al. 2021), which is not the Near Eastern order. If Sahelians were working from an imported practice, why does their sequence not match?

Because trait order was never the thing that could tell you. It records how one grass responded to being harvested and resown. It does not record who first had the idea to resow it. And on Munson’s model there was nothing available to copy in any case: the crops the practice came attached to could not cross the rainfall line, which is the premise the entire argument rests on. People applying a bare technique to an unfamiliar plant have no template to reproduce, so a mismatched sequence is unsurprising and, more to the point, uninformative.

What does discriminate is geography. An invention has no upstream and a transmission does, and that difference surfaces as an ordering of places rather than an ordering of traits. Munson understood which of the two he was staking. He chose the one that could be checked against a map, said where the line would run, and it ran there.

What the corridor still buys #

The corridor remains a live route rather than a historical curiosity, and the evidence that arrived since 1976 has been kind to it in ways Munson could not have arranged.

His mechanism survives intact. The livestock argument he leaned on is untouched: sheep and goats have no wild ancestor in Africa, they were domesticated in southwest Asia, and they reached the Sahel as finished products before any African crop existed. Whatever else was moving, animals were, and they prove the channel was open. The agronomic barrier he identified is real and is now better understood than he could state it, and it explains the shape of what crossed. A summer-rainfall zone cannot take a winter-rainfall crop, so the thing that moves is not the seed but the practice, and it lands on the nearest workable grass. It is the same structure that cultural transmission into Africa shows from the genetic side, and that the linguistic evidence for a Sahelian contact zone traces independently, along the same belt where livestock and millets spread.

The crop’s onward history looks less like an isolate every year. Pearl millet is now directly dated in the lower Indus at 2578 to 2358 cal BC, five centuries earlier than the previous secure figure and effectively contemporaneous with the Tilemsi material, in a paper whose authors reach for a maritime corridor to explain it (Jiménez-Arteaga et al. 2025). They read those grains as the arrival of a crop already domesticated elsewhere, which is how a date that close can sit alongside the Tilemsi material without displacing it. A grass domesticated in the southwestern Sahara was in Pakistan within a few centuries. Long-distance transmission is not the exotic hypothesis in this story. It is the observed behaviour of the plant. This is not the shape of the contact claims that fail every evidentiary test at once. Here the archaeobotany, the direct dates and the genomics agree, and they agree on a route.

None of this requires believing everything Munson believed. His central Saharan evidence rested partly on pollen from the Hoggar that he read as a domesticated Pennisetum, and identifying domestication status from pollen is not something the field would accept now. That plank is gone.

The corridor is not. He named four places in a line, said the earliest evidence would appear at the Saharan end and get later as you moved away, and published it. The valley he picked out of the middle of that line turned out to hold the oldest domesticated pearl millet on earth. He got the calendar wrong and the map right, which is the better half to get right, because the map was the claim and the calendar was borrowed.

What that buys is not a footnote about a forgotten archaeologist. It is a reason to stop treating the Sahel as a settled instance of humanity inventing agriculture twice. The strongest evidence for that reading has always been the plant, and the plant was never in question. The route was, and the route has now been found where a man who thought the practice was imported said it would be.

It has been fifty years. Someone should mark the prediction correct.

Sources #

1. Munson, Patrick J. “Archaeological Data on the Origins of Cultivation in the Southwestern Sahara and Their Implications for West Africa.” In Jack R. Harlan, Jan M. J. de Wet & Ann B. L. Stemler, eds., Origins of African Plant Domestication, pp. 187–209. The Hague: Mouton, 1976.
2. Munson, Patrick J. The Tichitt Tradition: A Late Prehistoric Occupation of the Southwestern Sahara. PhD dissertation, University of Illinois at Urbana-Champaign, 1971.
3. Manning, Katie, Ruth Pelling, Tom Higham, Jean-Luc Schwenniger & Dorian Q. Fuller. “4500-Year-Old Domesticated Pearl Millet (Pennisetum glaucum) from the Tilemsi Valley, Mali: New Insights into an Alternative Cereal Domestication Pathway.” Journal of Archaeological Science 38(2) (2011): 312–322.
4. McIntosh, Susan Keech, Emmanuelle Casanova & Caroline Gauthier. “Revisiting Tichitt: New Dates and Data from the 1968 Excavations by Patrick Munson.” Azania: Archaeological Research in Africa (2026).
5. Fuller, Dorian Q., Kevin MacDonald & Robert Vernet. “Early Domesticated Pearl Millet in Dhar Néma (Mauritania): Evidence of Crop-Processing Waste as Ceramic Temper.” In Fields of Change: Progress in African Archaeobotany, 2007.
6. Finucane, Brian, Katie Manning & Mustapha Touré. “Late Stone Age Subsistence in the Tilemsi Valley, Mali: Stable Isotope Analysis of Human and Animal Remains from the Sites of Karkarichinkat Nord (KN05) and Karkarichinkat Sud (KS05).” Journal of Anthropological Archaeology 27(1) (2008): 82–92.
7. Amblard, Sylvie. “Agricultural Evidence and Its Interpretation on the Dhars Tichitt and Oualata, Southeastern Mauritania.” African Archaeological Review, 1989.
8. Holl, Augustin. “Background to the Ghana Empire: Archaeological Investigations on the Transition to Statehood in the Dhar Tichitt Region (Mauritania).” Journal of Anthropological Archaeology 4 (1985): 73–115.
9. Jiménez-Arteaga, Carlos, et al. “New Evidence Reveals Dispersal of Pearl Millet from West Africa to South Asia by 2500 BCE.” Scientific Reports 15 (2025): 32931.
10. Kroeber, A. L. “Stimulus Diffusion.” American Anthropologist 42(1) (1940): 1–20.
11. Nosek, Brian A., Charles R. Ebersole, Alexander C. DeHaven & David T. Mellor. “The Preregistration Revolution.” PNAS 115(11) (2018): 2600–2606.
12. Harlan, Jack R., Jan M. J. de Wet & Ann B. L. Stemler, eds. Origins of African Plant Domestication. The Hague: Mouton, 1976.
13. Murdock, George Peter. Africa: Its Peoples and Their Culture History. New York: McGraw-Hill, 1959.
14. Baker, H. G. “Comments on the Thesis that There Was a Major Centre of Plant Domestication Near the Headwaters of the River Niger.” The Journal of African History 3(2) (1962): 229–233.
15. Smith, Andrew B. “Preliminary Report of Excavations at Karkarichinkat Nord and Karkarichinkat Sud, Tilemsi Valley, Republic of Mali, Spring 1972.” West African Journal of Archaeology 4 (1974): 33–55.
16. Fuller, Dorian Q., Aleese Barron, Louis Champion, Christian Dupuy, Dominique Commelin, Michel Raimbault & Tim Denham. “Transition From Wild to Domesticated Pearl Millet (Pennisetum glaucum) Revealed in Ceramic Temper at Three Middle Holocene Sites in Northern Mali.” African Archaeological Review 38 (2021): 211–230.
17. Burgarella, Concetta, et al. “A Western Sahara Centre of Domestication Inferred from Pearl Millet Genomes.” Nature Ecology & Evolution 2(9) (2018): 1377–1380.