125,000-Year-Old Arabian Stone Points Carry Traces of Poison

HUNTERS in eastern Africa have traditionally tipped their arrows with a tree toxin that can stop the heart. Chemical traces linked to the arrow poison tree have now turned up on stone points left beside a vanished lake in southern Saudi Arabia about 125,000 years ago. Some of the same points carry the fractures of impact and the resin of a mounting, according to an archaeological field study in Science led by Chris Clarkson of Griffith University and Frans van Buchem of King Abdullah University of Science and Technology (KAUST). The residues cannot, however, say how any poison was made, how strong it was or whether it ever brought an animal down.

The points come from ‘Uruq Aljelibah, a site on the southwestern margin of the Mundafan basin, in what the KAUST press release calls the sands of the Empty Quarter. During the Last Interglacial, a warm and humid stretch of prehistory, the basin held a lake. From a satellite elevation model, the researchers reconstruct a body of water of about 340 square kilometers and up to about 29 meters deep, full enough to spill through an outlet on its northern side. Freshwater mollusks and tiny crustaceans in the sediments point to lasting, low-salinity water, a lake and not a brief flood.

Sand from the shoreline and lake beds, dated by a luminescence method that measures how long grains have been buried, gave ages between about 119,000 and 127,000 years. Two hippopotamus teeth dated by other techniques agreed, within their uncertainties. Wet spells of this kind recurred in the region: a 2020 analysis of a stalagmite from a cave in Yemen counted 21 of them over about 1.1 million years, with grassland developing during them.

Along the old shore the researchers recorded more than 500 fossils at more than 130 spots. Hippopotamus accounts for 132 bone fragments and an extinct long-horned buffalo for 12, with a few each from camel, Arabian oryx and gazelle and more than 100 pieces of ostrich eggshell. From a plot of 100 by 125 meters they also collected 5,511 stone artifacts larger than two centimeters, more than 98% of them chert. Nearly all the cores, the lumps of stone that flakes are struck from, had been worked by the Levallois method, in which a knapper shapes the core in advance so that a flake of a planned form comes off. In one cluster, 24 flakes fit back together, a sign that the stone was worked on the spot.

Poison Tree Residues on Mounted Points

Of 89 artifacts examined for residues, 75 kept some, and 44 kept traces consistent with hafting, meaning attachment to a handle or shaft. The points made the strongest case. Plant fibers and resin sat toward their bases, where a mounting would go. Several carried compounds linked to the arrow poison tree, Acokanthera schimperi, including markers related to cardiac glycosides, the family of heart-acting toxins to which the tree’s poison belongs. On at least one point with impact damage, the toxic plant compounds occurred alongside animal fats and collagen.

The clearest example has large fractures of the kind linked to impact, damage on both faces, resin and plant fiber at its base and animal fat and collagen toward its tip. Those traces support use as a mounted stone weapon tip, the authors write, though the breaks cannot say whether it was thrust or thrown.

Contamination is the standing worry with residues this old. Chemical results were screened against blank samples, any compound that also appeared in the controls was excluded, and a subset of residues went to two laboratories, which both recovered the most abundant compounds. The authors still word the conclusion with care. Because the toxic compounds sit on purpose-made weapon forms, in some cases with mounting residues, impact damage and animal traces, they call deliberate use in hunting gear more plausible than incidental contact. The paper’s title says poisoned hunting weapons; its text leaves the preparation and potency of any poison unresolved.

If that reading holds, it reaches a long way back. A 2026 study of arrow tips from Umhlatuzana Rock Shelter in South Africa found toxic plant alkaloids on stone from a level dated to 60,000 years ago, which its authors presented as the first direct evidence of that poison on Pleistocene weapon tips. A 2024 analysis of a 7,000-year-old bone quiver from Kruger Cave, also in South Africa, identified a recipe drawn from several plants and lent support to reading a 24,000-year-old wooden stick from Border Cave as a poison applicator. The Mundafan shoreline is far older than any of them, the candidate plant is a different one, and the Science paper’s text does not rank the find against the others. The release ventures only that it is understood to be among the first such evidence recorded in Saudi Arabia. The arrow poison tree belongs to a genus otherwise restricted to Africa, and it is the one species that also grows in Yemen.

Palm Leaves and Plant Fiber on the Lakeshore

Most of the residues are plant matter of a humbler kind. Denticulates, tools with a notched, saw-like edge, were the artifacts most strongly associated with plant fibers and palm, and their wear is consistent with sawing or shredding. Leaf fragments stuck to two artifacts keep their cell structure well enough to be compared with modern palm leaves. Tools retouched on one face carried compounds associated with wood, rhizomes (underground stems) and other plants. The authors read all this as people making or mending bindings, cordage, matting and hafts as well as getting food.

“This site gives us a rare and richly detailed view of how people lived in a landscape that looks nothing like Arabia today,” Clarkson said in the release.

On the ground itself, the paper is more guarded than the release, which describes animal remains, tools and plant material found together in a single, dated context and calls the stone flint where the paper says chert. Erosion has exposed the artifacts on an open surface, so the site is no sealed living floor, and the authors treat it as a slightly time-blurred record of repeated activity. Fossils did fall more often than chance would predict in the patches densest with points, denticulates and Levallois flakes, and more than 50 bones of one young hippopotamus lay with artifacts within about 1.5 square meters. Neither that closeness nor the ages of the animals demonstrates hunting, the authors caution. The behavioral case rests on that spatial pattern together with independent evidence of weapon use and animal residues on the tools.

Who Made the Stone Points Is Still Unknown

No human fossils have been recovered at ‘Uruq Aljelibah. Fossils place Homo sapiens in northern Arabia by about 85,000 to 95,000 years ago, and footprints in the Nefud Desert date to the Last Interglacial, but the identity of the Mundafan toolmakers is an inference. The site does not pick out a migration route or a source population either.

For anyone picturing Arabia’s green phases mainly as a corridor between Africa and Asia, the site adds a different use of the land: a lakeshore able to sustain people for meaningful periods, where they made tools, mounted weapons, worked fiber and came back. The lake did not last. A second high stand at Mundafan dates to about 79,000 years ago, a separate wet pulse in a longer dry sequence.

Laurence Hapiot, head of heritage research at KAUST and a co-author, said in the release that building heritage science in Saudi Arabia “will enable us to turn the smallest traces into the biggest stories.” The traces at Mundafan leave work for the next round: experiments that could tell a thrust point from a thrown one, and a human fossil from the site.

FAQ