A cremation from the Early Holocene of Malawi.

All modern Human societies feel the need to dispose of the dead in specific, often highly ritualised, ways, and this appears to have been the case for much of the history of our species, and potentially other members of the genus Homo. Cremation is an effective way to achieve this, transforming the recognisable remains of a member of the community into a small amount of ash and calcined bone fragments in a relatively short period of time. However, this process also requires the effort and co-ordination to build a funeral pyre of sufficient size, and does not appear to have become common or widespread until the Middle Holocene.

The oldest known assemblage of burned Human remains comes from Lake Mungo in New South Wales and have been dated to about 40 000 years before the present, although these remains appear to have gone through a multi-stage ritual process which included burning, rather than being a single cremation event as the main way to deal with the remains. The oldest evidence of a body being burned on a pyre comes from a site called Xaasaa Na′ (Upward Sun River) in the Alaskan Arctic, where a child of about three years, who was burned about 11 500 years ago. The oldest known burned remains in Africa are about 7500 years old and come from Egypt, but it is unclear whether these were deliberately cremated. The oldest known evidence for deliberate cremation in Africa comes from Njoro River Cave in Kenya, and is only about 3300 years old.

In a paper published in the journal Science Advances on 1 January 2026, a team of scientists led by Jessica Cerezo-Román of the Department of Anthropology at the University of Oklahoma present evidence for a deliberate cremation, from the Hora 1 archaeological site at the foot of Mount Hora in the Mzimba District of Northern Malawi.

Mount Hora is a 110 m high granite inselberg (isolated hill) rising above the Kasitu River Valley, which forms a distinctive regional landmark. The Hora 1 archaeological site is a rock shelter at the base of this hill, opening to the east and incorporating a flat dry area of about 80 m². This site is easily accessed from all directions, and contains no enclosed areas which could act as a natural furnace. No other natural rock shelters could be found in the region. Hora 1 has produced a record of Human occupation stretching over more than 21 000 years, and appears to have been used for mortuary practices between about 16 000 years ago and about 8000 years ago.

HOR-1 site (11°39′S, 33°39′E; 1470 m above mean sea level, white stipples) relative to published excavated Later Stone Age (LSA) sites. Light green in inset outlines the Zambezian Biome. Cluster of sites near Mpunzi (or Mphunzi) Mountain includes six additional sites, three with human remains (Mtuzi, Changoni Bible School, and Chencherere II). Cerezo-Román et al. (2026).

This site was first excavated by Desmond Clark and William Rangeley in the 1950s, who excavated an area of 7-8 m² (Clark and Rangeley did not record measurements of the area they excavated, but did take photographs, which have been used to reconstruct the extent of their work). These excavations produced the almost complete remains of two adult Humans, HOR-1 (or UTC-242), a male directly dated to between 9081 and 8725 years before the present, and Hora 2 (or UTC-243), a female directly dated to between 8172 and 7875 years before the present. Both of these skeletons have yielded ancient DNA, suggesting a relationship to modern hunter-gatherer populations in Southern Africa.

Between 2016 and 2019, the Malawi Ancient Lifeways and Peoples Project carried out further excavations at the site, removing and sieving 7.8 m³ of sediment from two blocks which produced about 46 500 objects, including the almost incomplete skeletons of two male infants (Kahora 1 and Kahora 2), which have been indirectly dated (i.e. dated from material found with them, not directly from the skeletal material) to about 14 000 and about 16 000 years before the present, and which have also yielded ancient DNA, as well as fragmentary remains attributed to at least four additional adults (Hora 4 to Hora 7) and five additional non-adults (Kahora 2 to Kahora 3), as well as numerous animal remains. Examination of these fragmentary remains has suggested that they may have been subjected to complex mortuary practices, including secondary burials (i.e. burials either after being buried once and then excavated, or after having been either left unburied for an extensive period, or having been subjected to ritual behaviours which substantially altered the remains), which may have included ritual token-taking.

Views of HOR-1. (A) Plan view of the HOR-1 site georectified onto drone imagery, showing the open overhang, accessibility, and Malawi Ancient Lifeways and Peoples Project (MALPP) excavation Area I and II. (B) View from the north of the open site access. (C) Plan view of the Malawi Ancient Lifeways and Peoples Project Area I excavations, Kahora 1 and Kahora 2 infant burials, and Hora 3 remains, with the outline of the original 1950 excavations and position of the Hora 2 skeleton by Clark and Rangely, reconstructed using an archival site photograph. Cerezo-Román et al. (2026).

Cerezo-Román et al. report the discovery of a large ash feature at Hora 1, along with an associated set of remains attributed to a single cremated adult, who they designate as Hora 3. Hora 3 is calculated to have been chronologically intermediate between Hora 2 and Kahora 1. As such it appears to form part of a genetically continuous population which lived in the area across the Pleistocene-Holocene boundary, although (unsurprisingly) it was not possible to extract DNA from the cremated remains.

The 'ash feature' forms a large cemented block, dated to about 9500 years before the present, which overlies a Pleistocene sediment sequence which begins 17 000 years ago and ends about 12 000 years ago, and is overlain by younger Early Holocene deposits. At the core of this feature they found a set of Human remains identified as 'Cluster 1), which lie on top of an ash layer 10-15 cm thick, and roughly 2.5 m by 1.5 m in extent. Below this are a series of deeper lenses of consolidated ashes, charcoal, and rubified sediment, together referred to as the 'Lower Ash', which are interpreted as evidence for a series of consecutive burn events at the same location.

The Cluster 1 remains sit on top of a layer of incompletely combusted charcoal, and are in turn overlain by several more stratified lenses of ash and rubified sediment. A second set of remains was identified to the northeast of Cluster 1 and 5-10 cm deeper. These were identified as Cluster 2, but appear to be a part of the same individual which became detached during the cremation event.

Plan view of the exposed ash complex, showing the large spatial extent. Jessica Thompson in Cerezo-Román et al. (2026).

Twenty radiocarbon dates were obtained from the core pyre and larger ash feature, providing a chronology for the site. Dates obtained below the core feature begin around 12 699 years before the present, with the latest coming from 9918 years before the present. Layers of ash above the core feature produced dates ranging from 9540 to 9454 years before the present, while a layer of ash on the eastern excavation wall was dated to between 9538 and 9455 years before the present. A Land Snail bead from a layer immediately above the core combustion event provided a date of between 9537 and 9441 years before the present, and is taken as evidence of this layer being covered over quickly by non-combustion sediments. Another large combustion event was indicated by a stratigraphically higher layer of ash, which produced a date of between 9452 and 9142 years before the present, i.e. less than 400 years after the core event, although this layer produced no Human remains. Above this were further layers of non-combustion deposits dating from between less than 9403 and 9031 years before the present. 

Spatial relationships of dated materials, samples, and human remains. HOR-1 excavations in profile (A) and plan (B) views showing positions of dated materials and adult cremated remains with different degrees of certainty (LH indicates 'likely human' based on size, texture, and preservation, but not morphologically diagnostic). (C) shows phytolith (green) and micromorphology (white boxes) sample positions. Thick blue line shows how the profiles correspond in each view. C1 indicates Cluster 1 and C2 indicates Cluster 2. Jessica Thompson in Cerezo-Román et al. (2026).

Phytoliths are produced by plants as a way of handling silica absorbed with water from the soil. Most plants produce phytoliths to some extent, but Monocotyledons, such as Grasses and Palms, which utilise phytoliths both as structural support and a defence against herbivores (silica phytoliths quickly wear down the teeth of animals which lack specific adaptations to deal with them), produce phytoliths which can often be used to identify the maker to genus or even species level. Phytoliths recovered from the Hora 1 pyre deposits imply that the majority of the ash was generated by the burning of wood, but that non-woody plant fibres were also present. 

Sediments and ash layers of the pyre deposit. (A) view southwest of the feature 50 cm south of Cluster 1 of the remains. In both (A) and (B) black arrows indicate the top of micromorphology sample MALAPP 834 and white arrows show the distal radius from Cluster 1 in situ. Dashed box is the profile in (C), which shows layered ash with flecks of charcoal overlying rubified sediment. Distinct root and termite disturbances are visible crossing the boundaries of the intact ash layers. The brownish (grey when dry) sediment above and below is also ash-rich but more mixed (homogenised) and darker due to a higher content of finely comminuted organic material and microcharcoal. The sediments were dampened with a mist of water prior to taking the photos and image levels were adjusted to emphasise colour differences. Jessica Thompson in Cerezo-Román et al. (2026).

These pyre deposits comprise a series of laminated ash layers extending horizontally for more than a metre, and containing ash, charcoal, rubified sediments, and Human remains. These deposits have been penetrated by insect burrows and/or plant root tunnels in places, presumably before the ash layers became cemented. Examination of the rubified sediments suggests that these contain fragments of baked clay which probably originated from structures built on trees by termites, suggesting that deadwood was collected to build the pyre(s). Also present were carbonised fragments of the Bracket Fungus Ganoderma.

Thin section of sample 834. Sample 834 was collected from the pyre, about 1m away from the Hora 3 remains (A). One thin section was selected from the small block sample (B) with reddish disturbed sediments in the lower part, and bright ashes in the top (C). Termite galleries and shelter tubes are very common in woodland areas, especially on deadwood, which presents ready-for-use fuel and was likely preferentially selected to build the pyre (D); (E) displays the clear boundary between compact laminated ashes (lower half of photomicrograph) and mixed ashy sediment on top, which are mixed with sandy clay soil aggregates. A very coarse sand fragment (feldspar) in the lower right likely represents roof spall; (F) A closer view of wavy laminated ash topped by sandy ash; Wavy layers of intact ashes (grey) and elongate, sickle-shaped clay aggregates (orange, red), possibly resulting from termite shelter tubes on tree branches (G); The clear lower contact of the compacted ash is characterised by a sudden increase of sand in the underlying ashy sediment (H); Phosphatic ashes indicate some mineral alterations and weathering of the ashes (I); Termite channels, fortified by well-sorted sandy clay, run through the lower part of the sample and contain burned organic matter (J); Articulated ash fragment in compacted ash (K); The sediment consists of a mix of fire residues, clay aggregates, and sand (L). Flora Schilt in Cerezo-Román et al. (2026).

The ash deposits generally contain fewer artefacts (such as flakes from toolmaking) than non-ash deposits, probably relating to the fact that these were laid down much more rapidly. However, the Human-remains-containing core layer contains many more such objects, suggesting that they were deliberately included within the pyre, possibly for symbolic reasons.

Unifacial points and cores from square E11-a core pyre ash contexts. (B) and (F) Unmodified points; (D), (E), (G), (H), (I), and (J) broken point fragments (dashed red line shows breaks); (A), (C), (E), and (D) point showing unifacial retouch; (H) point fragment showing basal tang. Red stars indicate point areas with possible residue preservation. Justin Pargeter in Cerezo-Román et al. (2026).

A total of 170 Human bone fragments were recovered from the pyre area. Of these, 112 could be clearly identified as specific parts of the Human skeleton, with the remaining 18 identified as 'likely human' on the basis of their size, texture, and preservation. The majority of these bone fragments (160) were recovered within the ash feature, with 152 found within the core feature. Of the elements which could definitely be identified as Human, 97 could be assigned to a specific element, with 90 of these being long bone (i.e. arm or leg) fragments. Other fragments from the core pyre include partial left and right calcanei, an unsided patella, the lower right articular facet of a lumbar vertebra, and four phalanges. All are consistent with a single individual.

Skeletal inventory and thermal alteration colour changes. Preserved skeletal elements and changes in bone coloration in relation to temperature and fire exposure. Additional fragments not shown in the figure: an unsided patella, the lower right articular facet of a lumbar vertebra, and unidentified long bone fragments. Jessica Cerezo-Román in Cerezo-Román et al. (2026).

Analysis of the Hora 3 skeleton suggests that this individual was a female aged between 18 and 60. The individual is presumed to be an adult on the basis that it appears to be skeletally mature, with the distal femur and radius having fused. The articular margin of the distal femoral epiphysis showed marginal lipping, which may indicate the presence of a mild degenerative disease, giving the upper age estimate. No other potential age-markers were preserved. Since it was impossible to recover DNA from the burned and fragmented remains, sex was estimated using overall small size and gracility and the vertical diameter of both preserved femoral heads. On the basis of the diameter of the femora, the individual is estimated to have been between 145 and 150 cm tall in life (about 5 feet). Based upon the maximum and minimum midshaft diameters of the right humerus and femur, Hora 3 it thought to have been less mobile than most modern Southern African hunter gatherers, and more accustomed to manual labour.

(Top) Right distal humerus with cemented/encrusted termite tubes. (Bottom) Cemented elements that include the left proximal femur, a radius shaft, a tibia shaft, and an unidentifiable long-bone shaft. Cerezo-Román et al. (2026).

About 585 g of material thought to be derived from the skeleton of Hora 3 was recovered. This is notably less than is typically produced by a Human cremation, which is usually more than 1500 g. Notably, no head elements were found within the core pyre area, although a fragment of a sphenoid (the base of the skull) was found within the larger ash feature. Fragments of cranium, and in particular teeth, are usually among the most easily preserved and identified elements in cremations, and it is surprising not to have found any such fragments, given that sediment and ash removed from the site were wet sieved to 1 mm.

All of the skeletal remains show transverse, curved-transverse, and longitudinal cracks, which are typical for the burning of a fleshed corpse, but none show signs of the checking pattern fractures which are often seen when dry remains are burned. Most fragments were grey to black, with some brown, bluish-white, and white hues. Fragments from closest to the torso were typically the most blackened and charred, while more distal elements were grey or white - which indicates more exposure to flame in areas which would have been covered by less flesh. Many elements are white on their external surface, while internal surfaces are grey, which implies these elements were intact when the burning occurred. These colours suggest that the temperature to which the distal elements were exposed exceeded 500°C, while elements closer to the torso typically reached maximum temperatures lower than 500°C. Fourier-transform infrared spectroscopy is sensitive to low temperature heating of bone, which often does not involve recrystallisation, and therefore can provide an accurate palaeothermometer for low temperature cremations. Four white/grey fragments from Hora 3 were analysed in this way, with all producing spectral peaks indicative of heating to over 500°C.

Although the body of Hora 3 shows signs of having been manipulated after the cremation, many of the recovered fragments were recovered together, covered in ash, and in some cases cemented together. This includes a right distal humerus and right proximal radius cemented together along with long bone fragments and a possible metacarpal fragment. This appears to show that intact joints were present at the time of final disposal of the remains. Many of the bone surfaces are covered by calcium carbonate, with none of the recovered Human bones showing more than 50% exposure. Nevertheless, eight of the bones showed stone tool marks consistent with defleshing, while none showed any signs of having been scavenged by carnivores. 

Bone modifications made with stone tools. Bone surface modifications made with stone tools. (A), (B), (C), (D), and (G) are cut marks. (E) is a percussion mark. (F) is an indeterminate mark made with a sharp object. (B) to (G) show the variable bone colours and superficial calcination with carbonisation at the interior. (C) shows the minor 'skin' of calcium carbonate overlying the marks. (G) shows remaining adhering ash. (B) to (G) show two aspects of the fragment, with close-ups of the modifications in the boxed area. Jessica Thompson in Cerezo-Román et al. (2026).

Cerezo-Román et al. reconstruct a sequence of events in which people began using the HOR-1 as a habitation site about 21 000 years ago. The earliest recorded burials at the site occurred 16 000 and 14 000 years ago, both being infants. Small ash features consistent with campfires appear around the end of the Pleistocene, with larger pyres appearing by 10 240 years before the present.

Geology of HOR-1 rock shelter.(A) Pinkish and gray walls of the rock shelter at the site location. (B) Exfoliating bedrock higher above the excavation trench. (C) Example of spheroidal exfoliation of the bedrock near the rock shelter. Flora Schilt in Cerezo-Román et al. (2026).

Some time between 9540 and 9454, a pyre comprising at least 30 kg of deadwood and grass, with some hairy leaves from herbaceous plants, was assembled, something which would have represented a significant investment of time and effort by the local community. The body of a small, probably female, adult individual was then burned on this pyre, probably within a few days of their death. The distribution of the bones, particularly those of the arms and legs, suggests that this individual was placed on this pyre in a flexed or pugilistic position (i.e. with the knees and elbows brought in close to the torso). Cut marks on some of the bones show signs of defleshing, although the preservation of the bones and some joints suggests that the flesh of the body (or some other covering) was present at the time of burning. It is likely that the skull was removed before the burning, since neither scavengers nor Early Holocene hunter gatherers seem likely to have been able to remove all of the fragments which would have been produced by burning a skull, while leaving other elements of the skeleton in situ. The ritual removal of skulls, and other body parts, has been documented in a range of modern African populations, but never at a site of such antiquity. 

Spatial locations of femora and right tibia fragments. Cerezo-Román et al. (2026).

The fire achieved was able to reach temperatures above 500°C, although this doesn't appear to have been consistent, and the lamination of the deposits suggests that the fire-makers continued to add fuel to the pyre for some time. The detachment of Cluster 2 from Cluster 1, and the greater degree of burning seen on Cluster 1, may indicate that the body was manipulated during the cremation process to detach parts of it. Knapped material appears to have been added to the pyre, either at the outset or during the process, possibly in association with other funerary objects. Multiple further fires were lit at the same site over the next few hundred years, although no further cremations appear to have occurred.

Reconstruction of the cremation ritual. Sequence of events leading to the formation of the cremation feature at HOR-1. (A) Site location at an inselberg, a natural monument. (B) A large quantity of wood was collected to construct the pyre, suggesting communal labor. (C) Cutmarks on bone show parts of the body were defleshed. (D) Human remains display black coloration and curved transverse fractures, indicating some moisture in the remains. (E) The pyre and body at Cluster 1 were actively disturbed during burning, creating Cluster 2. (F) High temperatures were maintained by attendees adding additional fuel. (G) Convergent points are uniquely associated with the cremation. (H) Bipolar reduction dominates a lithic assemblage that occurs in higher concentrations with the remains than in the rest of the ash feature. (I) The presence of carbonised Ganoderma and the remnants of termite tunnels indicates the use of deadwood as the primary fuel. (J) The absence of cranial and dental remains suggests these may have been collected and removed. (K) Multiple fires were relit atop the original pyre location within communal memory. Patrick Fahey in Cerezo-Román et al. (2026).

Evidence of cremation among African hunter-gatherer populations is extremely rare in the archaeological record, and has not previously been found south of the Sahara. The oldest previously documented example come from the Nabta Playa site in southern Egypt, where a single individual appears to have been burned between 7800 and 7300 years ago (i.e. at least 1300 years before the earliest date for the Nabta Playa stone circle), and this has been interpreted as a 'burned inhumation' rather than a true cremation, possibly caused by the accidental ignition of material placed within the grave. Burned Human remains, possibly as much as 7000 years old, have also been recovered from an ancient midden near Lake Besaka in Ethiopia, although again this does not represent an in situ pyre like that seen at Hora, and again may not represent an intentional burning. The previous oldest known intentional cremations from Africa are associated with Neolithic Elmenteitan pastoralists in Kenya, about 3300 years ago.

A open-air, pyre cremation such as that seen at Hora requires a significant investment in time and labour by the local population, which may be why the practice is so rare among hunter-gatherer populations. In an enclosed furnace, a Human corpse can be burned in about two hours, but open air cremations require considerably longer, during which time the temperature must be maintained, typically by adding more fuel. Nevertheless, such a cremation would be a notable community event, particularly if, as seems to have been the case at Hora, parts of the body were removed during the process for ritual disposal elsewhere.

While the Hora site only records a single event, which seems to have been highly unusual in nature, it adds to a growing picture of a culturally diverse tropical African hunter-gatherer population in the Early Holocene, displacing an earlier impression that these peoples were likely to have been culturally homogeneous, even over great distances and long periods of time. The people at Hora did not erect megaliths or other architectural features. Nevertheless, they appear to have utilised chosen a site of natural prominence to carry out a significant mortuary ritual, a site which appears to have been utilised repeatedly over an extended period of time (at least 16 000 years), albeit with changing rituals during that time.

Evidence for evolving social cooperation and complexity has been recorded from many ancient hunter-gatherer populations around the world, although until now this has largely been absent from Africa. The Hora Rockshelter in Northern Malawi records a population showing both a diversity of behaviour and the ability to stage large events which would have required the investment of time and effort by many individuals. This appears to have been carried over multiple generations, with repeated fires being lit, including at least one subsequent major pyre event, suggesting a communal memory in which the significance of the location was maintained over many generations.

Cerezo-Román et al. conclude that around 9500 years ago the remains of an adult female were burned on a substantial pyre at the Hora Rocksheter in Northern Malawi. This is the oldest known pyre cremation in Africa, and the oldest known adult pyre cremation in the world. The fire appears to have been one of a sequence of fires at the same location, which persisted for centuries after the cremation event, signifying the importance of the site to the population. This site demonstrates the emergence of complex funerary rights, communal projects involving large numbers of people, and the recognition of certain sites as culturally significant long before the emergence of agriculture and organised food production, challenging previous conceptions about the extent to which community co-operation occurred among ancient hunter-gatherer populations in tropical Africa.

See also...

Evidence of smoke mummification in South China and Southeast Asia more than 10 000 years ago.

Archaeologists classify Human burials in two ways. In primary burials, the deceased is placed into the ground without modification, with-or-without grave goods. In secondary burials, the body is modified in some way before being buried. Burials classified as secondary include cremations, mummified remains, and a variety of embalming and body-modification techniques. Burials left by the ancient pre-Neolithic peoples of South China and Southeast Asia, who lived in the area from the Late Pleistocene till between 5000 and 3500 years ago, appear to have undergone complex mortuary procedures before being buried. Such burials are typically found in tightly flexed, crouching, or squatting positions, often with signs of having been tightly bound. These remains often show signs of scorching on their bones, but this appears to be localised to certain areas, rather than indicating a cremation process has occurred. Others show signs of partial dismemberment or mutilation.

In a paper published in the journal Proceedings of the National Academy of Sciences on 15 September 2025, a team of scientists led by Hsiao-chun Hung of the Department of Archaeology and Natural History at the Australian National University,  and Zhenhua Deng of the Key Laboratory of Archaeological Science and School of Archaeology and Museology at Peking University, present the results of a study in which they analysed 94 pre-Neolithic burials from northern Vietnam and Guangxi Province, China.

Hung and Deng et al. theorised that many burials from ancient South China and Southeast Asia were subjected to long periods of smoke-drying, which effectively mummified the bodies before they were buried. This practice was still carried out in parts of Australia and the New Guinea Highlands within recent history. Burials which Hung and Deng et al. believe were similar in nature are extremely numerous in northern Vietnam and Guanxi Province, with a lesser number of examples from the Philippines, Laos, Thailand, Malaysia, and Indonesia. There are also some possible examples from northern China, Korea, Japan, Australia, and other parts of the world. 

Recorded hunter-gatherer sites with flexed and squatting burials in Southeastern Asia from the Late Pleistocene to the Middle Holocene. (1) Gaomiao, (2) Niupodong, (3) Zhaoguodong, (4) Qihedong*, (5) Daowei 1, (6) Xiaoma, (7) Jiaoziyan, (8) Zengpiyan, (9) Miaoyan, (10) Dayan, (11) Xingyi, (12) Baida, (13) Gexinqiao, (14) Beidaling, (15) Liyuzui, (16) Fengyan, (17) Huangmenyan, (18) Xiankezhou, (19) Baxun, (20) Liyupo, (21) Changtang, (22) Lingwu, (23) Huiyaotian, (24) Qingshan, (25) Dingsishan, (26) Qiujiang, (27) Xijin, (28) Ganzao, (29) Hecun, (30) Jiangbian, (31) Chongtang, (32) Lang Cuom, (33) Mai Da Nguom, (34) Pho Binh Gia, (35) Hang Doi, (36) Dong Thuoc, (37) Hang To 1, (38) Hon Hai Co Tien, (39) Cai Beo, (40) Liyudun, (41) Dong Can, (42) Du Sang, (43) Lang Gao, (44) Hang Cho, (45) Hang Muoi, (46) Xom Trai, (47) Hang Mang Chieng, (48) Hang Con Moong, (49) Hang Diem, (50) Mai Da Dieu, (51) Mai Da Nuoc, (52) Hang Lang Bon, (53) Da But, (54) Con Co Ngua, (55) Banyan Valley Cave, (56) Spirit Cave, (57) Tham Lod, (58) Ban Rai, (59) Doi Pha Kan, (60) Ban Tha Si, (61) Quynh Van, (62) Pha Phen, (63) Yingdun, (64) Bau Du, (65) Krong No, (66) Bubog-1, (67) Ille Cave*, (68) Duyong Cave, (69) Sa’gung, (70) Moh Khiew, (71) Gua Kerbau, (72) Gua Kepah, (73) Gua Kajang, (74) Gua Teluk Kelawar, (75) Gua Gunung Runtuh, (76) Gua Peraling, (77) Gua Cha, (78) Niah Cave, (79) Kimanis, (80) Liang Tebo, (81) Keboboh Cave, (82) Gua Tengkorak, (83) Leang Panninge, (84) Cappalombo1, (85) Gua Harimau, (86) Gua Pawon, (87) Gua Braholo, (88) Song Tritis, (89) Song Terus, (90) Song Keplek, (91) Gua Lawa, (92) Song Gentong, (93) Tron Bon Lei, (94) Ratu Mali 2, (95) Liang Lembudu (*suspected flexed). Hung & Deng et al. (2025).

Hung and Deng et al. recorded the positions in which burials were placed within the ground, as well as signs of burning and cut marks upon bones. X-ray diffraction and Fourier-transform infrared spectroscopy were used to detect prior heating of bones, by looking for characteristic recrystallisation structures. Finally, they compared the ancient remains to records of smoke-drying mummification in Australia and the New Guinea Highlands. 

Zengpiyan Cave, a pre-Neolithic site in the northeast of Guangxi Province, was occupied from approximately 12 000 to 7000 years before the present. At this site 26 individual burials were found, some of which were placed in squatting positions in small pits, while others were placed on their sides in flexed postures. 

Human burials exposed in Zengpiyan Cave (approximately 12 000-7000 years before present), Guangxi, southern China, excavated in 1973. (A) Three human burials exposed in excavation square BT2, viewed from the east; (B) Burial BT2M1 in a hyper-flexed position; (C) The three burials viewed from the south; (D) Burial BT2M5 in a flexed position, though not as tightly flexed as BT2M1. Hung & Deng et al. (2025).

Huiyaotian Shell Mound, by the banks of the River Yongjiang in Nanning City in central Guangxi, has produced 169 individual burials, one of which has been radiocarbon-dated to approximately 9030-8975 years before the present. Hung and Deng et al. examined 60 burials from this site, with 21 being flexed, 17 placed in squatting positions, and 22 too poorly preserved to determine. Twelve of the flexed burials were placed on their sides, three on their backs, and one was lying face down (the original position of the remaining flexed burials was unclear).

The Luyupo Shell Mound is located about 50 km from Huiyaotian in Long'an County and contains a series of burials dated to between approximately 8000 and 6700 years before the present. Many of the burials here were flexed, and covered by large stones, which were placed on the heads, chests, or stomachs of the deceased. 

Examples of Early and Middle Holocene Human burials from southern China. This figure shows six Human burials from Huiyaotian in Nanning City: (A) M14, (B) M19,  and (C) M20, and Liyupo in Long’an County: (D) M23, (E) M24, and (F) M28. Both shell-midden sites arelocated in the Guangxi Zhuang Autonomous Region. All individuals are in flexed positions, with several exhibiting hyper-flexed postures. (A) and (E) are male, (B)–(D) and (F) are female. Hung & Deng et al. (2025).

The Hoabinhian-Da But-Quynh Van Group burials of northern Vietnam comprise a series of sites, such as Mai Da Dieu, Hang Cho, Hang Diem, Hang Mang Chieng, Con Co Ngua, Quynh Van, and Bau Du, which record a material culture similar to that of ancient Guangxi, mostly within cave or shell mound deposits. 

Notable among these is the Con Co Ngua site in Thanh Hoa Province, which was excavated between 1978 and 1980, and again in 2015, has produced a shell midden and 267 burials, dated to between 6700 and 6200 years before the present, and attributed to the pre-Neolithic Dabutian Culture. The nature of the shells here has been interpreted to indicate that the site was on the coast when inhabited, although it is more than 30 km inland today. The majority of the burials here (about 77%) were placed in a squatting posture, while the remainder (about 23%) were flexed and placed on their sides. 

Examples of Early and Middle Holocene flexed burials from northern Vietnam. (A) Burial 80M27 from the Con Co Ngua Shell Midden; (B) Burial M16 from Mai Da Dieu Cave, both located in Thanh Hoa Province. Hung & Deng et al. (2025).

The Gua Harimau (Harimau Cave) site on southern Sumatra is considered to be one of the most significant burial sites in Island Southeast Asia, with a chronological sequence which runs from the late pre-Neolithic into the Iron Age. A total of 33 burials have been uncovered, containing 82 individuals, the majority of which are Late Neolithic or Iron Age, which are mostly either in a supine position (on their backs) or secondary burials. There is a smaller number of pre-Neolithic burials, placed predominantly in flexed positions, two of which have been directly dated, the older to between 5715 and 5587 years before the present, and the younger to between 3819 and 3568 years before the present, making it the youngest known burial of the type in Southeast Asia. 

A major question which needs to be addressed in order to understand these burials is how individuals were placed into the hyper-flexed positions in which they are found? Such skeletons frequently retain all of their original articulation, but are folded into very compact positions, which would have required a degree of contortion unlikely for an intact body.

M26, a tightly flexed human burial from Huiyaotian, Guangxi, southern China. (A) Skeletal position of a middle-aged male, and (B) surface scan of Burial M26. The mostly complete skeleton retains its original articulated position. The pelvis lies on the ground, while the Right and Left femora are nearly vertical, with their distal ends oriented upward. Both knees are strongly flexed, positioning the feet just medial to the hip joints, and the foot bones are largely intact. The shoulders, rib cage, and upper limbs have collapsed downward, but the positions of the humerus, radius, and ulna indicate that the arms were crossed beneath the thighs, with each hand placed under the opposite knee. Overall, the body is folded in an extremely compact manner. Hung & Deng et al. (2025).

When unmodified bodies are buried within sediments, decomposition of the flesh creates voids, which in turn allow skeletal elements to move, leading to varying degrees of disarticulation, particularly of small bones such as the phalanges (bones of the fingers and toes). Examination of hyper-flexed skeletal remains from Guangxi and Vietnam shows that they appear to have been buried with no space between their limbs and torsos. This would appear to imply that there was no soft tissue (with the possible exception of dried skin) left on the bodies when they were buried. 

This combination of extreme flexion and preservation of small bones in their original position has been seen in other parts of the world, such as the Mesolithic burials from the Sado Valley in Portugal, which have also been interpreted as having undergone some port-mortem treatment before burial.

Another question which arose from examination of the skeletons is whether they were treated with fire or smoke before being buried. Some of the bones from Huiyaotian and Liyupo show clear signs of fire exposure, in the form of burned and blackened patches, even though the skeletons are still articulated (something which is not usually the case with cremated remains). Curiously, this burning appears to be quire localised, and restricted to certain bones, notably the frontal bones of the skull, the lower limb bones such as the femur, tibia, and fibula, the upper limb bones in the elbow region including the humerus, ulna, and radius, and the pelvic bones. About 15% of the skeletons from Huiyaotian and Liyupo showed signs of such burning.

This combination of highly flexed yet articulate skeletons with occasional burn marks has also been recorded from sites in northern Vietnam, Sarawak (Malaysia), southern Java (Indonesia), and northern Palawan (Philippines). Notably, a male skeleton from Hang Muoi Cave in Hoa Binh Province, northern Vietnam, which shows intense burning of some bones, also shows cranial features characteristic of Indigenous Australo-Papuan populations, and has been radiocarbon dated to between 14 027 and 13 798 years before the present. Another cranium showing a similar burn pattern was found at Dong Can, close to Hang Muoi Cave, has been interpreted as that of an older male from with a similar date.

Examples of hyper-flexed burials with partially burned bones from southern China and Indonesia. (A) and (B) Burial M35, a young male from Liyupo in Guangxi, with its partially burned skull; (C) and (D) Burial ST1 from Song Terus Cave in Java, showing its partially burned left femur, tibia, and humerus. Hung & Deng et al. (2025).

In the 1950s and 1960s, archaeologists  Tom and Barbara Harrisson excavated a series of burials at the West Mouth of the Niah Caves in Sarawak (Malaysian Borneo). They found over 150 burials, 39 of which they classified as Mesolithic or pre-Neolithic. These burials again were typically flexed or seated, sometimes mutilated, and showed signs of burning, and were dated to between 11 698-11 270 and 8454-8354 years before the present. 

The Song Terus individual from southern Java has been directly dated to 8500 years before the present, and was buried in an articulated but highly flexed position with charring on the bones of the left side, notably the left femur, tibia, and humerus.

At Ille Cave on northern Palawan, a fragmentary Human skeleton dated to the Early Holocene shows burning and cut marks which may indicate disarticulation, which has been directly dated to between 9260 and 9006 years before the present.

This consistent presence of localised burning on certain parts of the skeleton strongly implies that the intention was never full cremation. The presence of charring on areas such as the lower limbs, elbows, and frontal areas of the cranium may suggest that the remains were arranged in a seated or tightly flexed position and exposed to heat in a controlled, or semi-controlled, way. This may have been intended to smoke the remains as a way of preserving them, with burning occurring in areas with less muscle or other soft tissue covering the bone. Logically, if this was the practice, and it was as widespread as it appears to have been, many of the flexed burials which do not show visible charring may show signs of bone-recrystallisation caused by exposure to heat. 

X-ray diffraction spectroscopy is known to be particularly effective at detecting thermal alterations which occur within bone when it is heated to temperatures in excess of 500°C. Hung and Deng et al. selected 20 pre-Neolithic archaeological samples from northern Vietnam, as well as two sub-modern samples from Edo Castle in Japan. Nine of the samples showed signs of having been heated to above 525°C, with some of these having potentially been heated to temperatures greater than 645°C. Three of the samples showed no signs of having been heated. The remaining eight samples showed signs which could be associated with having been heated to about 525°C, but which could be associated with long term diagenetic processes.

Fourier-transform infrared spectroscopy is more sensitive to lower heating of bone, which often does not involve recrystallisation. Because this method does not rely on detecting recrystallisation, it is also less prone to being confused by diagenetic effects. Hung and Deng used Fourier-transform infrared spectroscopy to analyse 69 bone samples from 11 pre-Neolithic sites in southern China, northern Vietnam, and Indonesia, as well as one premodern sample from Edo Castle. Five of the samples produced poor quality results, and were excluded from the study. Of the remaining 64 samples, eight showed no signs of having been heated, 33 showed signs of having been heated to low temperatures, and 23 showed signs of having been heated to high temperatures. Notably, different elements from the same skeletons did not always show the same degree of heating. For example, in skeleton M18 from Huiyaotian, the proximal end of the left ulna showed signs of having been heated to high temperatures, while the middle section of the same bone appeared to have been exposed to only moderate heating. Another skeleton, M1 from Mai Da, showed signs of strong heating to the talus (anklebone), but only moderate heating to a tooth. Such variation within a single skeleton is unlikely to have been caused by diagenetic processes. Notably, areas with minimal soft tissue covering again typically showed signs of greater heating.

Thus, evidence of heating was found in almost all of the bones in the sample, but the majority did not appear to have been heated to great temperatures, suggesting that such exposure was only sporadic and localised to certain parts of the skeleton. Interestingly, while some areas which appeared blackened also seemed to have been heated to high temperatures, this was by not always the case, suggesting that another cause, such as blackening by smoke, might be responsible.

Based upon this, Hung and Deng et al. conclude that the skeletons had undergone a form of heat treatment, prior to burial. During this process, most of the body was exposed to only moderate heat, and only certain parts were directly exposed to fire. Furthermore, much of the blackening observed might not stem from prolonged exposure to heat, but instead from prolonged exposure to smoke. This specialised mortuary practice appears to have been carried out across a wide area of South China and Southeast Asia.

Locations of nine sites with confirmed mummification by smoking identified in this study, along with two uncertain and four relevant sites. Burial samples from 11 sites were analysed, and confirmed that smoke-dried mummification occurred at nine of them: Huiyaotian, Liyupo, Hon Hai Co Tien, Cai Beo, Hang Mang Chieng, Mai Da Dieu, Hang Diem, Con Co Ngua, and Gua Harimau. Two other sites, Gaomiao and Hang Cho, remain uncertain due to their weaker Fourier-transform infrared spectroscopy signals. Beyond the 11 primary study sites in China and Vietnam, the six ad￾ditional sites of Zengpiyan, Hang Muoi, Dong Can, Ille Cave, Niah West Mouth, and Song Terus, although not yet tested by X-ray diffraction or Fourier-transform infrared spectroscopy, present strong evidence of comparable cultural practices. Hung & Deng et al. (2025).

Many prehistoric societies are best known from their burial practices, and this tends to be particularly true of those which modified bodies in some way before burying them. The Chinchorro culture of the Atacama Coast of South America are generally considered to have been the oldest practitioners of mummification. Between around 7000 and about 3700 years ago, they developed a practice in which bodies had their brains and viscera removed, and their limbs reinforced with sticks, before being deliberately desiccated by exposure to the arid climate, and having their faces coated with clay and painted. 

The Ancient Egyptians are probably the most famous example of a culture using mummification, due to the complex preservation methods they used, combined with monumental architecture and often lavish grave goods. Here too, mummification predates the dynastic period first appearing around 4500 years ago, where again a form of desiccation mummification was used in the Nile Valley.

The smoke mummification described by Hung and Deng et al. predates either of these examples by thousands of years, and apparently originating in an area with a humid monsoon climate. These mummies were produced by a hunter-gatherer society living in South China and Southeast Asia from the Late Pleistocene to the Middle Holocene, who predominantly buried articulated remains in caves, rock shelters, and shell middens. Many of these burials are associated with sites which are associated with the Hoabinhian Epipalaeolithic cultural assemblage of Southeast Asia, which includes pebble, core and flake tools, but which lacks blades and bifaces. Other sites in South China and northern Vietnam are associated with a 'para-Neolithic' culture, which had earthenware pottery and ground pebble axes, but which did not appear to practice any form of agriculture or Animal husbandry, and which left no stylistic remains. Unlike later burials from the true Neolithic, none of these burials were associated with any form of grave goods.

In order to try to better understand the mummification practice used in ancient South China and Southeast Asia, Hung and Deng et al. carried out an ethnographic visit to the Dani and Pumo peoples of Indonesian Papua, who live in the Bellem Valley of Jayawijaya Regency. The Dani traditionally made mummies by smoking a body until it turned entirely black, often in a tightly compressed position with the limbs bound to the trunk. This process is presumed to have been carried out soon after death, before rigor mortis set in, although there does appear to have been some flexibility about the process, with some bodies being preserved without binding. These mummified bodies were then stored in special rooms within the houses of the living, and brought out on special occasions. Importantly, the bodies were often bound in a very compact way, as was observed in Southeast Asia, and preserved by what appears to be a similar method of smoke mummification. 

Examples of smoked mummies kept in private households in Papua, Indonesia, photographed in January 2019. (A) A Dani hyper-flexed mummy; (B) A flexed mummy from Pumo Village. Both locations are in Wamena, Jayawijaya Regency, Papua. Hung & Deng et al. (2025).

The Anga (or Kukukuku) people of the Aseki region of Morobe Province in eastern Papua New Guinea, also practice a form of mummification, with revered figures such as warriors, shamans, and village leaders, being placed in a seated position on a frame over a constantly smoking fire, typically for about three months. Once the process is completed, the mummies are placed in a protected ledge or niche on a cliff.

Both of these examples appear to show a mummification practice similar to that used in ancient South China and Southeast Asia. However, in neither case is there any record of bodies being buried, or any indication as to what the ultimate fate of the mummies is.

Based upon these observations, Hung and Deng et al. suggest that many (but not all) of the individuals from South China and Southeast Asia which they examined are likely to have been tightly bound shortly after death, then suspended over low-temperature, smokey, fires for long periods of time. The occasional blackened and burned bones are probably the result of occasional loss of control over the fires during this lengthy process. It had previously been suggested that individuals may have been burned by fires set inside graves at Niah West Mouth and other sites, but Hung and Deng et al. could find no evidence to support this. Comparison to modern examples of similar practices in Papau suggest that this was carried out under a roof, potentially within a house or specially constructed hut. They also feel that it is likely that after being smoked, the body was transferred to a special site, such as a residence, cave, or rock shelter, where they would have remained on display for some time, potentially many years, before finally being buried. 

 Proposed scenario for prehistoric smoke-drying mummification, drawing on both archaeological findings and ethnographic parallels from Papua. The illustration shows the multi-step process of smoke-drying mummification, as ascertained through our study. (A) Preparation for the smoking ritual—corpses were bound to varying degrees, often in a hyper-flexed position; (B) Processing – the corpse could be bound, as shown in (A), or unbound, as shown in (B), based on a modern example in Papua New Guinea, while placed above a low-temperature fire. In the Late Pleistocene–Middle Holocene examples, all individuals were bound; (C) Post-smoking treatment – after the smoking was completed, then the smoked mummy was transferred to a residence, a specially constructed hut, a rock shelter, or a cave. In some cases, however, the mummy decayed due to prolonged exposure to open air. As seen in the example presented here, the head and body became separated; (D) Final burial – the smoked mummy eventually was buried. Most were intact at the time of burial. Some cases had decayed by this time in the process, and those remains were rearranged in the burial pit, sometimes appearing as suspected intentionally dismembered burials in the archaeological record. Hung & Deng et al. (2025).

Some of the individuals found at sites such as Dingsishan, Huiyaotian, and Liyupo in Guangxi, had anatomical anomalies, with parts of the bodies placed out of their natural, anatomical position, such as having the head placed within the thoracic cavity, of the legs in the place of the arms. These have previously been interpreted as evidence of intentional dismemberment, but Hung and Deng et al. suggest an alternative explanation, that these bodies may have been exposed so long that they began to decay, or that they became disarticulated while being transported from their display position to their final burial site. In this scenario, attempts at reconnecting the body may have been hurried, as the ultimate burial was not the purpose of the mummification, but the intervening period when the body was on display being more important.

Burials M6, M7, and M8 at the Huiyaotian site, Guangxi, southern China. This image shows a corner of the excavation area at Huiyaotian, excavated in 2011. The three burials - M6 (female aged 40 to 60 years), M7 (a middle-aged adult female), and M8 (an adult male) - were all found in a squatting position. Hung & Deng et al. (2025).

Having established the practice of mummification by smoking in ancient Southeast Asia, Hung and Deng et al. next consider the presence of cur marks on some skeletons, something which has been reported at many sites in Southeast Asia and Guanxi. These have previously been interpreted as signs of ritual dismemberment, but Hung and Deng et al. note that they are often concentrated around the epiphyses of major limb bones such as the humerus and femur, and suggest that they may represent interventions to enable joint release in bodies which had started to undergo rigor mortis prior to the beginning of the mummification process, which would have enabled tighter binding of the limbs. It is also possible that some may represent incisions made to enable the draining of body fluids during the smoking process.

Diagram showing the percentages of cut marks recorded different joints from theHuiyaotian sample of 19 individuals (excavated in 2016), Guangxi, southern China. The small rectangles indicate additional cut marks on long bones that are considered to be from flesh removal, mostly from the femur and tibia. Hung & Deng et al. (2025).

There are also a smaller group of marks, predominantly on the femora and tibiae, which appear consistent with some form of defleshing, probably carried out after the body had been mummified. Hung and Deng et al. note an ethnographic account of smoke mummification in South Australia, in which relatives sometimes removed small pieces of flesh from the thighs of bodies to offer to sorcerers, in exchange for magical protection or retribution. Such marks have been reported on burials from Huiyaotian in Guangxi, Ille Cave in Palawan, and other sites.

This all suggests a complex mortuary tradition dating back to over 10 000 years ago, involving smoke mummification, delayed burial, and, at least in some cases, ritual interactions with the preserved remains.

This practice appears to have been carried out over a huge geographical area, including South China, Southeast Asia, New Guinea, and Australia. It is likely that in many of these areas, which have predominantly wet tropical climates, smoking would have been the most effective way of preserving bodies. Nevertheless, the large geographical area and long period of time over which these practices persisted suggests that the activity had more significance than the simple preservation of bodies. Hung and Deng et al. note that the Anga people believe that the spirits of the deceased are able to roam freely during the day, but return to their bodies at night, while among the Taramindjeri people of South Australia, mummification has been linked to the possibility of immortality. They suggest that similar beliefs are likely to have been present in ancient South China and Southeast Asia, and that this is probably the cause of the long persistence of smoke mummification as a practice.

The final question that Hung and Deng et al. ask is whether there is direct continuity between the ancient smoke mummification practices of South China and Southeast Asia and similar practices in recent New Guinea and Australia. If there is, then this would represent a cultural practice dating back into the Pleistocene, which tracked the expansion of ancient Homo sapiens through the tropical latitudes of southern Asia, and which might potentially be tracked back to our first emergence from Africa. Alternatively, it might all just be a coincidence. 

Studies of genomic, craniofacial, and archaeological evidence have all suggested that the ancient, pre-Neolithic peoples of South China and Southeast Asia, and the ancestors of the modern indigenous populations of Papua and Australia, were related to the earliest Modern Humans to migrate out of Africa, and that they represent a distinct population from the later Neolithic farmers of the same region, who migrated from further north in Asia during the Holocene, eventually giving rise to the speakers of the modern Austroasiatic, Kra-Dai, and Austronesian language groups, amongst others. 

This more ancient population, sometimes referred to as 'first layer' occupiers, are considered to be ancestral to the modern Negrito peoples of the Andaman Islands, Peninsular Malaysia, and the Philippines, as well as  Indigenous Australians, New Guinea Papuans, and the ancient Jomon people of Japan. All of the mummified remains examined had craniofacial features consistent with belonging to this ancient 'first layer' population.

The two-layer model for Anatomically Modern Human (AMH) population movements across eastern Eurasia.The southern and northern major migration contexts for early Homo sapiens dispersal towards eastern Asia may have been separated by mountain ranges and deserts in central Asia. The southern route followed the coastal rim of the Indian subcontinent and continued onwards into the ancient Pleistocene continent of Sahul by at least 50 000 years ago. This southern migration was associated with a continuation of the Epipalaeolithic Hoabinhian pebble and flake tool complex, presumably a direct descendant of the Middle Palaeolithic industries introduced by the initial modern human inhabitants of this region. The northern route of early Homo sapiens dispersal is less clear, but in this scenario, it involved migration from western Eurasia across Siberia around 45 000 years ago. The descendants of this northern migration domesticated millet and rice after 9000 years ago in the Yellow and Yangtze valleys, and migrated south with their food producing economies to reach northern Mainland Southeast Asia by about 5000 years ago. Hung & Deng et al. (2025).

Given this scenario, it is possible that smoked mummification may have originated even earlier, and have been even more widespread, than currently identified. Several sites with flexed burials and partial burning have been documented at a number of Southeast Asian sites more than 12 000 years old. Examples from South China include Dayan, thought to be between 15 000 and 12 000 years old, Miaoyan, thought to be between 19 000 and 12 000 years old, Fengyan, potentially between 42 000 and 33 000 years old, Zhaoguodong, between 12 000 and 9000 years old, Huangmenyan, between 14 000 and 10 000 years old, while in Vietnam, a Human clavicle from such a burial has been directly dated to between 14 027 and 13 798 years before the present. 

Similar mortuary practices have also been recorded further north in Asia. These include the Initial Jomon Period in Japan, between 10 500 and 7000 years ago, the pre-farming period on Gadokto Island, Korea, between 6600 and 6300 years before the present, as well as some parts of the southern Korean Peninsula. Some sites in northern China have flexed burials, dating to between 13 000 and 10 000 years ago. Finally, the Sakhalin Ainu of the Russian Far East are recorded to have practised mummification by sun-drying as recently as the early 19th century.

Flexed burials have also been recorded from archaeological sites in Australia, such as the Broadbench Burial Ground of Queensland, which was excavated in the 1960s, although this has never been formally dated, instead being considered to be from 'Late Prehistory'.

Thus, there is clearly a biological and cultural continuity between the ancient 'first layer' peoples of South China and Southeast Asia, and the modern indigenous populations of Australia and New Guinea. This creates a compelling (but ultimately unprovable) case for the retention of smoke mummification as a custom among hunter-gatherer populations over a wide area and a vast span of time.

See also...