Palaeolithic of the Armenian Highland · based on Quaternary Science Reviews, the Russian TV channel Nauka and specialist publications · November 2026
In brief. In the small limestone cave of Ararat-1, on the edge of the Ararat Depression, archaeologists found 1,770 stone artefacts dating from roughly 52,000 to 33,000 years ago. Obsidian, the highland’s most valuable raw material, was brought there as finished tools and blanks, sometimes from 180–200 kilometres away, and then resharpened on the spot. Local flint, by contrast, was worked right in the cave, from raw nodules to finished flakes. Two kinds of stone thus reveal two ways of living in the mountains: a portable kit for long journeys and a disposable everyday supply. The study itself is more cautious than the headlines: its authors call the picture a conundrum and explain it as a series of short, repeated visits.
Ararat-1 is a modest limestone cave at 1,034 m above sea level. It lies on the eastern slopes of the Ararat Depression in central Armenia, as Sci.news describes in its account of the research. The depression itself is a tectonic basin of about 1,300 km², formed by the collision of the Eurasian and Arabian plates. To the north and north-west it is bounded by the volcanic complexes of Aragats and the Gegham Highland, and Mount Ararat rises to the south. The climate is semi-arid and the vegetation mostly steppe and semi-desert, while the higher ground of the Urts anticline has patches of temperate woodland (Nora et al., 2025).
The cave is a karstic cavity with a main entrance, a central chamber about 6 m high and 4 m long, and three small passages. The deposits are roughly two metres thick and divide into five sedimentological units. Excavations took place in 2019, 2022 and 2023. The work is led by researchers of the Hebrew University of Jerusalem (Ariel Malinsky-Buller and her PhD students, among them David Nora), together with staff of the Institute of Archaeology and Ethnography of the National Academy of Sciences of Armenia (Boris Gasparyan, Artur Petrosyan) and Ellery Frahm of Yale University. Fieldwork and laboratory analyses were supported by a European Research Council grant for the TransCause project on the dynamics of Pleistocene populations of the Armenian Highland.
The first comprehensive report on the cave’s chronology and environment appeared in 2024 in Quaternary Science Advances. According to Sherriff and colleagues, sediment accumulated between 52,000 and 35,000 years ago, and during a stable phase of the late Pleistocene (Marine Isotope Stage 3, MIS 3) the cave was visited by Middle Palaeolithic people living amid a mosaic of semi-arid shrubland, grassy steppe and patches of temperate woodland. The individual layers are dated as follows: the lowest (SU4) to about 52.4 ± 3.6 ka, SU3 to roughly 45–38 ka, and the upper Palaeolithic horizon SU1.4 to about 38–35 ka (luminescence and radiocarbon dating, Nora et al.).
The upper layers also hold much later traces: pottery of the Middle Chalcolithic (about 4600–4400 BC), the Late Bronze Age (about 1450–1200 BC) and the early Middle Ages (about 610–680 AD). The cave thus served as a shelter across tens of thousands of years, from Palaeolithic hunter-gatherers to people of an era already documented by written sources.
The study centred on 1,770 stone artefacts from five stratigraphic units, most of them dating from roughly 52,000 to 33,000 years ago (the Nauka report on the paper in Quaternary Science Reviews). Of these, 1,245 pieces are obsidian and 525 are other rocks. The size structure of the collection is telling: 1,095 pieces, about 62%, are smaller than 20 millimetres. These are tiny flakes that are easy to miss during excavation, and they proved to be the key to understanding the site.
The authors determined production techniques, shapes, the degree of edge retouching and the origin of the raw material. For obsidian they used a portable X-ray fluorescence analyser (pXRF). Each piece was measured two or three times in different positions, because about 81% of the analysed obsidian finds turned out to be small flakes under 20 mm, a size at the limit of the method. In all, 37 raw-material types were identified: 25 are different obsidian sources and four are varieties of chert (flint); the rest include limestone, chalcedony, sandstone, quartzite, basalt and mafic lava. The authors have published their data and code in an open repository.
A particular methodological novelty is the so-called “module flakes”. These are by-products of tool resharpening, classified in five steps: from the initial trimming of an edge (M0), through retouching and reshaping (M1–M3), to deep “rejuvenation” of the tool (M4). From M0 to M4 the flakes grow larger on average and the number of earlier removal scars on their dorsal side rises, reflecting ever more intensive use and reworking of the same edge.
The Armenian Highland has no shortage of volcanic glass. It is prized precisely for its properties: a perfectly sharp edge and a predictable fracture. So the question is not whether obsidian was available, but where it came from and in what form it was brought. Chemical analysis showed that the material for Ararat-1 arrived from many sources at varying distances (distances are given in a straight line; the actual walking route was longer):
Back in 2022, in the thick of the field season, Ellery Frahm reported a tiny obsidian flake that had been chemically tested right in the field lab: it was struck off while resharpening a tool carried more than 200 kilometres some 45,000 years ago. This gave the research its working metaphor: whole routes can be read from a negligible chip.
Why are the authors convinced that obsidian was brought in already shaped? The cave contains practically no large waste from primary working: no cortex-covered pieces, no core fragments, no massive flakes. Instead there are many tiny chips of the kind produced when a cutting edge is sharpened and repaired. The conclusion is clear: the tool arrived as a blank or finished implement, was refreshed at the camp, and was then carried onward. Valuable raw material was used for as long as possible, the shape of the piece changing with each new sharpening. Specialists call this strategy curation (curated technology), a careful, economical treatment of the tool.
The most unusual consequence is that almost none of the obsidian tools themselves remain in the cave. In layers SU3 and SU4 there are only nine retouched obsidian pieces (scrapers, a denticulate, a Mousterian point) and six of other rocks, together making up less than one percent of the collection. Yet the chips left on the floor testify to dozens of tools that are no longer there: their owners refreshed the edge and moved on. The researchers speak of “ghost tools”, whose existence we learn only from repair debris. Interestingly, some sources appear among module flakes but not among finished tools, while others appear only among tools. Meydan Dağ, for example, is found in tools but left almost no repair waste, meaning such pieces lasted long and were rarely resharpened.
Flint in the same cave looks entirely different. Here there are cores, large flakes, flakes retaining natural cortex, and fragments left after core preparation. This is the standard kit of a place where stone was worked from the very beginning. The flint, then, was obtained nearby and brought almost in its original form, and the needed tools were made right at the camp. For everyday, small tasks this was quite enough: why carry from afar what lies close at hand.
Refitting (reassembling flakes) confirmed this. All the reconstructed reduction sequences concern flint, and the longest, of five flakes, reflects centripetal Levallois flaking, a fairly sophisticated technique. But the core itself was not among the finds: after the blanks were struck off, it was evidently taken away. For obsidian, only one fragment could be assembled, from two module flakes.
There is one more detail: retouched flint tools are, as a rule, noticeably larger than obsidian ones, often twice the size. A small obsidian tool was light, durable and precise; flint tools were made faster and larger but were not meant to be carried.
Here the most interesting part begins, and it is where media retellings simplify the picture. The authors themselves admit that the cave leaves them with a conundrum: in the same layers obsidian speaks of high mobility, while flint, at first glance, suggests a longer stay. They ask outright whether one of the two materials might be an outlier from the general technological strategy (Nora et al., 2025; also summarised briefly in the Sci.news piece).
The standard indicators archaeologists use to tell a base camp from a brief stop (the density of finds per volume of sediment and the share of retouched tools, the so-called WABI approach) give no clear answer here. There are two reasons: too many tiny fragments, which are usually excluded from analysis, and a complicated site-formation history. So the researchers took a different route. They counted the smallest flakes (under 20 mm), separated the finds by raw material, and compared the resharpening waste with the distance to each source.
The result: Ararat-1 is best explained as a series of repeated short visits. Flint can then be read either as a sign of a slightly longer stop within a brief visit, or as a signal of seasonality. An independent geoarchaeological study (Oikonomou and co-authors, repository record) reaches a similar conclusion: human visits, associated with fire and food consumption, were infrequent and brief, and the cave was also regularly used by carnivores and herbivores.
Hence the main image that made the headlines: the cave as a place to rest, repair equipment and prepare for the onward journey. A Hebrew University publication describes it as a hub for resting, repairing gear and planning the next step, while the Jerusalem Post stresses that the study challenges the habit of dividing ancient cultures solely by the static shapes of tools found in their camps.
The paper itself is more precise and more careful. The authors write that Ararat-1 epitomises a transit-oriented mobility strategy, and that the assemblage forms a highly fragmented chain of operations stretched across the landscape: blanks prepared in one place, used in another, resharpened in a third. Raw material arrived from several directions at once, which, in the authors’ view, looks like part of an ordinary nomadic cycle rather than special expeditions for stone.
The cave does not stand alone. The areas from which people drew raw material for Ararat-1, Kalavan-2 and Lusakert-1 overlap markedly. The authors see in this a high-elevation interaction corridor linking sites at roughly 900 to 2,000 metres above sea level. These sites share features: flexible Levallois reduction schemes, selective production of retouched tools, systematic tool maintenance and long-distance obsidian supply. From this one may infer an exchange of information and possibly seasonal gatherings of groups. The choice of routes, the authors suggest, was shaped not only by the shortest distance but by habitual paths, social memory and predictable refuges in a harsh landscape.
The words “may infer” deserve emphasis: this is the authors’ hypothesis, not a direct reconstruction.
Related conclusions are drawn in the companion paper by Frahm and co-authors on the scales of stone transport on the Armenian Highland (see also the publication page). According to it, Middle Palaeolithic groups of the late Pleistocene used overlapping territories, whereas bearers of Upper Palaeolithic technologies at Aghitu-3 cave moved over shorter distances. The authors cautiously allow that differences in mobility may have helped distinct stone industries coexist through social transmission of skills rather than through the direct replacement of one population by another (abstract). This is a debated question, and Ararat-1 supplies only one argument for it.
More than 450 Palaeolithic sites are known in the Caucasus, about 60% of them Middle Palaeolithic (according to a survey by Gasparian and Glauberman cited in Nora et al.). The Armenian Highland stands out among them precisely for its abundance of obsidian: its quality is about equally high at all major sources, so differences in how stone was handled reflect human behaviour rather than properties of the raw material. The authors therefore call the highland a kind of field laboratory for testing ideas about “curated” and “expedient” technologies.
Ararat-1 fits none of the region’s accepted “cultures”. It lacks the characteristic “Yerevan points” by which sites of the Yerevan-1, Angeghakot-1 and Lusakert-1 type are defined, and Levallois blanks are relatively few. But it shares a trait with Yerevan-1, Barozh-12 and Kalavan-2: specific M4 flakes that remove an exhausted, heavily retouched edge. The authors propose seeing the link not in a static list of tool types but in the way a heavily retouched tool was rejuvenated.
The depth of this history on the highland is striking. At Nor Geghi 1 in Kotayk Province, on the outskirts of Yerevan, around 325,000–350,000 years ago people used bifaces and Levallois technology side by side, and, as the paper in Science (2014) showed, this is the earliest known combination of the two technologies outside Africa (Royal Holloway press release). Chemical analysis showed that the inhabitants of Nor Geghi used obsidian from sources up to 120 kilometres away (Sci.news). Even then, obsidian was carried over long distances. At the open-air site of Barozh 12 and at Kalavan-2, retouched obsidian blanks were transported roughly 190 kilometres.
On a site devoted to the long history of the Armenian Highland, this theme deserves separate attention, and no stretching of the evidence is needed. No one has demonstrated, and no one can demonstrate, a direct ethnic continuity between the people of Ararat-1 and the first peoples named in written sources: tens of thousands of years lie between them. But the persistence of landscape and resources can be traced well.
In Ararat-1 itself the upper layers contain an obsidian leaf-shaped point, most likely Chalcolithic. According to Ellery Frahm, its raw material comes from the Geghasar volcano, the same source that supplied the site tens of thousands of years earlier (PanArmenian). At the Yeghegis-1 rockshelter in Vayots Dzor, obsidian made up 92.1% of the stone tools in the Chalcolithic, although the nearest sources lay more than 60 kilometres away on foot; the diversity of sources grew over time, which the researchers link to expanding social networks (Frahm et al., Scientific Reports, 2024). A general overview of Armenia’s archaeological heritage notes that the highland’s wealth of obsidian largely shaped Neolithic culture (Wikipedia); Hatis and Arteni, sources used at Ararat-1, are named there among the key ones.
Between Palaeolithic nomads and the inhabitants of Chalcolithic and Late Bronze Age settlements there is no chain of inheritance, but there is the same field of choice: the same volcanic massifs, the same passes, the same altitudinal belts.
Ararat-1 is interesting not for a sensational find but for a way of reading. The tiny fragments on the cave floor, usually discarded from analysis, made it possible to see what the collection itself lacks: ghost tools, carried tens or even hundreds of kilometres 50,000 years ago, resharpened by the fire and taken onward. Beside them lies local flint, from which everything needed on the spot was made. Two strategies in one layer show people who knew the landscape of the Armenian Highland, valued the best stone and could plan: what to carry and what to take on the spot.
The paper appeared online on 14 July 2025; the Nauka TV retelling was published on 23 July 2026.
Main publication
Related studies on Ararat-1
Popular retellings and reports
Regional context
A source-based overview. All links are clickable and lead to primary sources or scholarly and…
After the Nazis came to power in 1933, “Aryan” origin in Germany became a question…
What exactly the Russian contingent undertook to do in the Lachin Corridor, what happened between…
In brief. Suppose the Armenian authorities deliberately take the Genocide off the agenda. The historical…
By Vigen Avetisyan. Data as of October 29, 2026. In brief. As a scholarly concept,…
By Vigen Avetisyan In brief. The claim that present-day Armenia is “historical Azerbaijan” and that…