Figurative cave paintings from the Indonesian island of Sulawesi date to at least 35,000 years ago (ka) and hand-stencil art from the same region has a minimum date of 40 ka1. Here we show that similar rock art was created during essentially the same time period on the adjacent island of Borneo. Uranium-series analysis of calcium carbonate deposits that overlie a large reddish-orange figurative painting of an animal at Lubang Jeriji Saléh—a limestone cave in East Kalimantan, Indonesian Borneo—yielded a minimum date of 40 ka, which to our knowledge is currently the oldest date for figurative artwork from anywhere in the world. In addition, two reddish-orange-coloured hand stencils from the same site each yielded a minimum uranium-series date of 37.2 ka, and a third hand stencil of the same hue has a maximum date of 51.8 ka. We also obtained uranium-series determinations for cave art motifs from Lubang Jeriji Saléh and three other East Kalimantan karst caves, which enable us to constrain the chronology of a distinct younger phase of Pleistocene rock art production in this region. Dark-purple hand stencils, some of which are decorated with intricate motifs, date to about 21–20 ka and a rare Pleistocene depiction of a human figure—also coloured dark purple—has a minimum date of 13.6 ka. Our findings show that cave painting appeared in eastern Borneo between 52 and 40 ka and that a new style of parietal art arose during the Last Glacial Maximum. It is now evident that a major Palaeolithic cave art province existed in the eastern extremity of continental Eurasia and in adjacent Wallacea from at least 40 ka until the Last Glacial Maximum, which has implications for understanding how early rock art traditions emerged, developed and spread in Pleistocene Southeast Asia and further afield.
Since the 1990s, thousands of rock art images have been documented in the karst caves of the Sangkulirang–Mangkalihat Peninsula in East Kalimantan, a province in the Indonesian portion of Borneo2,3,4,5,6,7,8,9,10,11 (P.S., unpublished observations) (Fig. 1). This remote and difficult-to-access region contains 4,200 km2 of karst outcrops9,12 that are formed of late Eocene to early Pliocene limestone13. The karst terrain is characterized by densely forested mountain chains and towering cliffs that reach heights of several hundred metres. The Sangkulirang–Mangkalihat Peninsula is adjacent to the edge of the Sunda Shelf—a continental shelf that descends to about 2,500 m in depth—and therefore even during low sea-level stands in the Pleistocene, the karsts were situated essentially at the southeastern tip of Eurasia (Fig. 1). Fifty-two rock art sites have been recorded in eight different karst mountain areas between the Berau and East Kutai districts, spanning a distance of about 100 km. The art is often found in remote, high-level caves that contain little other evidence of human habitation. Few sites in the region have been excavated; the oldest published archaeological remains date to 19,761 ± 87 years before present (bp, taken as ad 1950; an uncalibrated accelerator mass spectrometry 14C date on charcoal)14. Previous uranium-series (U-series) and 14C dating of a cave drapery that overlies a hand stencil at Lubang Jeriji Saléh suggested a minimum date of about 10 ka for this motif15 (Supplementary Information).
On the basis of the superimposition of different styles, the rock art of the Sangkulirang–Mangkalihat Peninsula comprises at least three chronologically distinct phases9. The oldest style is characterized by large in-filled, reddish-orange-coloured paintings of animals—mainly the Bornean banteng (Bos javanicus lowi), a type of wild cattle that is still extant on the island (Extended Data Fig. 1), but also includes what may be now-extinct taxa16 as well as hand stencils produced using pigment of the same distinctive hue (Extended Data Fig. 1). The second phase is dominated by hand stencils that are dark purple (‘mulberry’) in colour, which are often clustered into distinct compositions (Extended Data Fig. 1). Many of these stencils are partly in-filled with painted lines, dashes, dots and small abstract signs that possibly represent tattoos or other marks of social identification, and in some instances hand stencils are linked together by painted mulberry lines that form intricate tree-like motifs, which perhaps symbolize kinship connections. Some older reddish-orange hand stencils appear to have been ‘retouched’ with mulberry paint to create similar in-filled designs and tree-like motifs (Extended Data Fig. 1). This phase also features small, carefully executed mulberry-coloured paintings of anthropomorphs (Extended Data Fig. 2). These elegant, thread-like human figures—henceforth, ‘Datu Saman’ following the established term for this style6—are sometimes shown in small groups, and are usually portrayed with elaborate headdresses and an array of other objects of material culture that includes possible spear throwers. Some figures are depicted in narrative scenes as hunting or pursuing small deer or as engaged in enigmatic social or ritual activities (for example, ‘dancing’; Extended Data Fig. 2). The final rock art phase is characterized by anthropomorphs, boats and geometric designs that are usually executed in black pigments (Extended Data Fig. 1). This rock art style is the only one that has thus far been documented elsewhere in Borneo; it is found at other locations in Indonesia and may be associated with the movement of Asian Neolithic farmers into the region from about 4 ka, or more recently17,18.
To date the earliest beginnings of cave art production in this region of East Kalimantan, and to establish the timing of stylistic changes in the rock art, we undertook a comprehensive programme of U-series dating of calcium carbonate deposits associated with parietal motifs. Over two field seasons, we collected a total of 15 calcium carbonate samples that were associated with 13 motifs at 6 separate cave sites, and which offered the opportunity to provide minimum and/or maximum dates for the images under study. Individual samples were divided into multiple aliquots (between 3 and 7; 65 in total) and the resultant dates are in stratigraphic order, which demonstrates that uranium and thorium are in closed-system conditions (Supplementary Table 1).
The oldest minimum dates are for a large reddish-orange, solid in-fill painting of an animal at Lubang Jeriji Saléh, for which we obtained dates of 40 ka (sample LJS1) and 39.4 ka (sample LJS1A) (Fig. 2, Extended Data Fig. 3, Supplementary Table 2). The image is incomplete and the animal depicted is therefore unclear, but it appears to be a large ungulate that possibly has a spear shaft protruding from its’ flank6. A figurative painting of a banteng—painted in the same style as the motif that we dated—is located nearby. Additionally, we dated samples LJS5 and LJS6, each of which was associated with a separate reddish-orange hand stencil from Lubang Jeriji Saléh; this provided a minimal date of 37.2 ka for these stencils (Fig. 3). One of these hand stencils—from which sample LJS5 was taken—had previously been dated to a minimum of about 10 ka, but was associated with a large porous cave drapery with suspected open system conditions for uranium and thorium15 (Supplementary Information). Our new date is associated with dense flowstone underneath the porous cave drapery (Supplementary Information). Another reddish-orange hand stencil at Lubang Jeriji Saléh (sample LJS2) has a minimum date of 23.6 ka and a maximum date of 51.8 ka (Extended Data Fig. 4). In addition, two maximum dates of 103.3 ka (sample LT1) (Extended Data Fig. 5) and 82.6 ka (sample LK1) (Extended Data Fig. 5) were obtained for a reddish-orange hand stencil at Liang Téwét and an animal painting of the same colour at Liang Karim, respectively. These dates correspond to flowstone layers present on the rock face ‘canvas’ before the art was produced, and thus provide the maximum ages of the images.
The second style of cave art—which, as noted, is characterized by mulberry-coloured paintings— yielded two minimum dates of 16.2 ka (sample LJS4) and 15.7 ka (sample LJS3), both of which were from a single hand stencil from Lubang Jeriji Saléh, as well as a maximum date of 20.9 ka for sample LJS4 (Extended Data Fig. 5). At Liang Banteng, two separate mulberry-coloured hand stencils that feature internal decorations and tree-like motifs with links to other hand stencils were dated to a minimum of 19.7 ka (sample LBT1) (Extended Data Fig. 6) and 17.5 ka (sample LBT2) (Extended Data Fig. 6). At Liang Sara, a mulberry-coloured hand stencil was dated to a minimum of 14.6 ka (sample LSR2) and a nearby Datu Saman figure—also produced using mulberry-coloured pigment—yielded a minimum date of 13.6 ka (sample LSR1) (Fig. 4). This small anthropomorph is depicted wearing a large ornate headdress and brandishing an elongated object, possibly a spear. It is also superimposed over a mulberry-coloured hand stencil. Additional age determinations provided respective minimum dates of 9.3 ka and 0.6 ka for a mulberry-coloured hand stencil (sample LH2) (Extended Data Fig. 7) and an unidentified mulberry-coloured figure (sample LH1) (Extended Data Fig. 8) from Lubang Ham.
In summary, U-series dating shows that the oldest parietal art in eastern Borneo dates to between 51.8 and 40 ka, and that a distinct rock art style appears in the region between 20.9 and 19.7 ka. Concerning the latter, at least some components of the art phase characterized by mulberry-coloured pigment appear at the height of the Last Glacial Maximum; specifically, hand stencils with internal decorations, and tree-like designs that link the stencils together. The single minimum U-series date that we have on a mulberry-coloured Datu Saman motif suggests that explicit portrayals of human figures had emerged by at least 13.6 ka, although they could potentially have arisen at the Last Glacial Maximum. To our knowledge, the large animal painting from Lubang Jeriji Saléh—created at least 40 ka—is the oldest figurative rock art image in the world. It is also one of the earliest-known figurative representations of an animal, being comparable in age with the mammoth-ivory figurines from the Swabian region of Germany19,20. It is also clear that Pleistocene rock art in Sulawesi, where dated motifs (n = 14) span the period between about 40 ka and 27.2–22.9 ka1, is not regionally unique. It is likely that the early parietal art of Sulawesi—which contains hand stencils and paintings of large animals that are stylistically similar to those from Kalimantan (Extended Data Fig. 9)—was introduced from the latter region. Thus, we propose that two major Palaeolithic cave art provinces had emerged at opposite edges of the Eurasian mainland by 40 ka: the renowned Franco-Cantabrian province of western Europe21 and a province in island Southeast Asia (ISEA) that straddled the Wallace Line.
It is also apparent that a Pleistocene rock art sequence—which spans at least 20,000 years—arose in Borneo long after the arrival of modern humans in ‘Sundaland’ (a biogeographical region that encompasses parts of ISEA exposed during times of low sea levels) at around 73–63 ka in Sumatra22 and long after the peopling of Australia (70–60 ka)23. This raises the question of who made the first cave art of Borneo. The oldest anatomically modern human fossil from Borneo—and in ISEA—is the ‘Deep Skull’, a partial Homo sapiens calvarium excavated from the Niah Great Cave (Sarawak, Malaysian Borneo) in the 1950s that has now been dated to about 40 ka (human occupation at Niah Cave dates back to approximately 50 ka)24. The Deep Skull is morphologically closer to modern-day eastern Asians than to Australo-Melanesians25. Therefore, we suggest two possible models of early modern-human migrations into Pleistocene ISEA. (1) The first H. sapiens to enter the region comprised an Australo-Melanesian population that had expanded as far south as Sahul by 70–60 ka but did not produce rock art in ISEA (or, if they did, it has not been discovered or dated); this initial wave was followed by a later migration of an eastern Asian population that arrived about 52–40 ka and produced the earliest rock art in Borneo. (2) Alternatively, the first modern human colonizers in ISEA may have been dispersed into small groups that did not produce rock art, with the latter emerging much later owing to local increases in human population density and flow-on effects on social signalling. Whichever of these models was the case, the absence of evidence for Pleistocene rock art production elsewhere in Borneo reinforces the view that East Kalimantan was an important centre of Palaeolithic cave art development.
The distinct rock art style that appears in Borneo at about 21–20 ka is evidence for a major cultural change in Pleistocene ISEA that has not previously been documented. The emergence of this tradition may reflect a population turnover at the Last Glacial Maximum. Alternatively, it is also possible that the karsts of East Kalimantan—one of the richest biodiversity hotspots in ISEA26—were a highly favourable environment for human populations. If so, increased levels of inter-group contact in the karsts may have driven the development of a rock art style that was focused on creating visual records of emerging systems of social organization and cultural identity, group affiliation and territorial demarcation. The Datu Saman figures are also notably similar to small anthropomorphs depicted in ‘Dynamic Figure’ art from Arnhem Land27 and ‘Gwion Gwion’28 art from the Kimberley (Extended Data Fig. 10), which represent the oldest parietal-art styles from Australia that consistently portray humans; although these styles have long been assumed to be of Pleistocene antiquity, they are not reliably dated. In Franco-Cantabrian cave art, human figures are vastly outnumbered by animal motifs and became common only during the Magdalenian period (16,500–12,000 calibrated years bp)29. As noted, with a single minimum date of 13.6 ka, the Datu Saman figures could originate at the Last Glacial Maximum, or they could represent a later stylistic development. In any case, the rock art of the Sangkulirang–Mangkalihat Peninsula documents a clear shift in the development of parietal art from depicting large animals to consistently representing the human world, in the form of human figures and decorated hand stencils with branch-like designs.
It is now evident that rock art emerges in Borneo at around the same time as the earliest forms of artistic expression appear in Europe in association with the arrival of modern humans (45,000–43,000 calibrated years bp)30. Thus, similar cave art traditions appear to arise near-contemporaneously in the extreme west and extreme east of Eurasia. Whether this is a coincidence, the result of cultural convergence in widely separated regions, large-scale migrations of a distinct Eurasian population or another cause remains unknown.
No statistical methods were used to predetermine sample size. The experiments were not randomized and investigators were not blinded to allocation during experiments and outcome assessment.
A small segment (~100–200 mm2) of each speleothem was removed from the rock art panels using either a battery-operated rotary tool equipped with a diamond saw blade, or a small hammer and chisel. Each speleothem sample was sawn or chiselled in situ so as to produce a continuous microstratigraphic profile extending from the outer surface of the speleothem through the pigment layer and into the underlying rock face. The only exceptions were LBT1, LBT2 and LH1, in which the sample broke at the speleothem–paint boundary (LBT1 and LBT2) or above the pigment layer (LH1). LT1 and LK1 provided only maximum dates. All of the sampled speleothems comprised multiple layers of dense and non-porous calcite in clear association with painted motifs. In the laboratory, the samples were micro-excavated in arbitrary ‘spits’ over the surface of the speleothem, creating a series of aliquots ~1 mm thick. The pigment layer was either visible across the entire length of the micro-excavated subsample or was visible at the rear of the samples (LBT1 and BTT2) or on the rock wall (LH1). In total, we obtained 63 U-series age determinations (Supplementary Table 1).
Speleothem samples collected in this study formed from thin films of water on cave surfaces over a long period of time. When precipitated from saturated solutions, calcium carbonate usually contains small amounts of soluble uranium (238U and 234U), which eventually decay to 230Th. The latter is essentially insoluble in cave waters and will not precipitate with the calcium carbonate. This produces disequilibrium in the decay chain, in which all isotopes in the series are no longer decaying at the same rate. Subsequently, 238U and 234U decay to 230Th until secular equilibrium is reached. Because the decay rates are known, the precise measurement of these isotopes enables calculation of the date of the carbonate formation31.
U–Th dating was carried out using a Nu Plasma multi-collector inductively-coupled plasma mass spectrometer (MC-ICP-MS) in the Radiogenic Isotope Facility at the School of Earth and Environmental Sciences, University of Queensland, following chemical treatment procedures and MC-ICP-MS analytical protocols that have previously been described32,33,34. Powdered sub-samples weighing 3–170 mg were spiked with a mixed 229Th–233U tracer and then completely dissolved in concentrated HNO3. After digestion, each sample was treated with H2O2 to decompose trace amounts of organic matter (if any) and to facilitate complete sample-tracer homogenization. Uranium and thorium were separated using conventional anion-exchange column chemistry, using Bio-Rad AG 1-X8 resin. After stripping off the matrix from the column using double-distilled 7 N HNO3 as eluent, 3 ml of a 2% HNO3 solution mixed with trace amount of HF was used to elute both uranium and thorium into a 3.5-ml pre-cleaned test tube, ready for MC-ICP-MS analyses, without the need for further drying down and re-mixing. After column chemistry, the U–Th mixed solution was injected into the MC-ICP-MS through a DSN-100 desolvation nebulizer system with an uptake rate of around 0.1 ml per minute. The U–Th isotopic ratio measurement was performed on the MC-ICP-MS using a detector configuration to enable simultaneous measurements of both uranium and thorium34,35. The activity ratios of 230Th/238U and 234U/238U of the samples were calculated using the previously published decay constants36. U–Th dates were calculated using the Isoplot Ex 3.75 Program37.
It is common for secondary calcium carbonate to be contaminated by detrital materials such as wind-blown or waterborne sediments, a process that can lead to U-series dates that are erroneously older than the true age of the sample. This is due to the pre-existing 230Th present in the detrital components, which is somehow analogous to the radiocarbon marine reservoir effect. As the detrital 230Th cannot be physically separated from the radiogenic 230Th for measurement, its contribution to the calculated 230Th age of the sample is often corrected for by using an assumed activity ratio of 230Th/232Th in the detrital component. Given the fact that the detrital component within a cave is often composed of wind-blown or waterborne sediments that chemically approach average continental crust, the mean bulk-Earth or upper continental crustal value of 232Th/238U = 3.8—corresponding to an activity ratio of 230Th/232Th of 0.825—and an arbitrarily assigned uncertainty of 50% have been commonly assumed for detrital or 230Th corrections32. In this regard, the degree of detrital contamination may be reflected by the measured activity ratio of 230Th/232Th in a sample, with a higher value (such as >20) indicating a relatively small or insignificant effect on the calculated age and a lower value (<20) indicating that the correction on the age will be significant31. Because 232Th in the sample is largely present in the detrital fraction and plays no part in the decay chain of uranium, the detrital 230Th in a sample with a measured activity ratio of 230Th/232Th > 20 would make up only <0.825/20 = ~4.1% of the total 230Th in the sample.
Sometimes the assumed activity ratio of 230Th/232Th of 0.825 (±50%) for the detrital component may not cover all situations. If the actual activity ratio of 230Th/232Th in the detrital component significantly deviates from this assumed range, the detrital correction scheme may introduce significant bias—especially to samples with an activity ratio of 230Th/232Th < 20. In such situations, the activity ratio of 230Th/232Th in the detrital component can be obtained through direct measurement of sediments associated with speleothems22,38,39, or computed using isochron methods or stratigraphical constraints40. In our case, the speleothem layers associated with painted rock art are very thin and therefore are not suitable for extracting sufficient sediments for direct measurement. On the other hand, these speleothems have extremely low growth rates that are much slower than typical stalagmites used previously22 and their growths were often episodic, which suggests that the previously published least-squares approach22 is also not appropriate. Considering the above limitations, we used a limiting stratigraphical constraint method. For instance, using the assumed activity ratio of 230Th/232Th of 0.825 (±50%) for the detrital component, the corrected ages of all five sub-samples of LJS2 are in stratigraphic order (Supplementary Table 1), which indicates three episodic growth phases (at 50.3, 26.1 and 16.8–12.1 ka). However, if we increase the assumed activity ratio of 230Th/232Th for the detrital component to 1.8, then the corrected age for the stratigraphically younger sample LJS2.2 will be reversed and become older than that for the stratigraphically older sample (see ‘Corrected Age-II’ in Supplementary Table 1). In this regard, it is reasonable to argue that the activity ratio of 230Th/232Th in the detrital component for this cave site should be < 1.8 for the ages to be in stratigraphic order, with the assumed activity ratio of 230Th/232Th of 0.825 (±50%) being more reasonable. This is because—using the assumed activity ratio of 230Th/232Th of 0.825 (±50%) for detrital correction—the corrected dates of LJS2.1, LJS2.2 and LJS2.3 are 12.6 ± 0.5, 14.9 ± 0.4 and 16.8 ± 1.1 ka, respectively, which are stratigraphically more coherent than other correction schemes (see ‘Corrected Age-I’ in Supplementary Table 1). We are therefore confident that we used the optimal approach for detrital correction. Regardless, the choice of the correction schemes has almost no effect on the interpretation of the critical samples used to constrain the ages of the rock art. For instance, using the assumed activity ratio of 230Th/232Th of 0.825 (±50%) for detrital correction, the corrected dates of LJS1.3, LJS1A.3 and LJS2.5 are 40.9 ± 0.8, 39.9 ± 0.6 and 50.3 ± 1.6, respectively. Using the activity ratio of 230Th/232Th of 1.8 ± 50% for detrital correction, the corrected dates of LJS1.3, LJS1A.3 and LJS2.5 are 39.1 ± 1.8, 38.8 ± 1.2 and 47.2 ± 3.1, respectively, which are indistinguishable within their respective age uncertainties from the former.
Synchrotron powder diffraction
Small, thin spall flakes of pigment were collected for complementary analyses during sampling for the dating programme. Very small chips of rock art paint (micro-spall) were crushed into homogenized powders manually using an agate mortar and pestle (P1, P2, P3 and P4, Supplementary Information). Sample P1 consisted of mulberry-coloured pigment from the hand stencil located to the left of dating samples LJS1 and LJS1A. P2 consisted of reddish-orange pigment from the animal painting associated with dating samples LJS1 and LJS1A, and P3 consisted of reddish-orange pigment directly related to the dating samples. P4 consisted of mulberry-coloured pigment associated with dating sample LJS4. Once powdered, the rock art paints were placed into 0.3-mm-diameter borosilicate capillaries and mounted on the beam line. Diffraction data were collected at the Australian Synchrotron at a wavelength of 0.77412(3) Å, calibrated using a NIST SRM 660b, from 5-85° 2Theta, with a Mythen microstrip detector with an inherent step size of 0.002°, using two detector positions and a collection time of 5 min per position. Samples were rotated at around 1 Hz during data collection to ensure good powder averaging. Phase identifications were undertaken using Panalytical Highscore with the ICDD PDF4 database.
Synchrotron X-ray fluorescence microscopy
Small, thin spall flakes of pigment were collected for complementary analyses during sampling. We scanned small flakes of rock art paint that adhered to the limestone panel surfaces (samples P1, P2, P3 and P4) at the X-Ray fluorescence microscopy beamline of the Australian Synchrotron using the Maia 384C detector array with incident excitation beam energy of 18.5 keV41. The energy resolution of the detector is 275 eV at Mn Kα. The >2-mm-thin chips of paint adhered to limestone were mounted on standard magnetic arms using protective films and adhesive tapes that are invisible to X-rays (chiefly Ultralene). The rock art pigments were positioned at an optimal working distance of 10 mm from the detector. Scans were collected at 2-μm spatial resolution with 1 ms and 2 ms dwell time per pixel for the cross-sections and surfaces, respectively. Scans were collected with full-spectrum X-ray fluorescence data deconvoluted into spatially resolved elemental ‘heat’ maps using the dynamic analysis method implemented in the GeoPIXE software suite42.
Scanning electron microscopy
Field emission scanning electron microscopy (JSM-7100F) was used to image surface morphology and the spatial distribution of chemistry was investigated using electron dispersive X-ray spectroscopy (a JED-2300F EDX) undertaken in both spot assay and element-mapping modes. Samples P1 and P2 were platinum-coated for conductivity.
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
The authors declare that all the data supporting the findings of this study are available within the paper and its Supplementary Information.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The fieldwork was authorized by I. M. Geria, the director of the National Centre for Archaeology in Jakarta (Arkenas) and B. Sancoyo and I. M. Kusumajaya, the director and former director of the Balai Pelestarian Cagar Budaya Kalimantan Timur. We further acknowledge the Indonesian State Ministry of Research and Technology for facilitating the research. We thank Griffith University for additional project support. Field assistants included Tewét, Bombé, Amril, Joang, Satriadi, M. Hendra, Stepanus, Satriadi, Sugianor, Heldi, Aidil, Joel, Ghojen, Budiansyah and Firman. Technical laboratory assistance involved A. Nguyen and Y. Feng. We thank S. O’Connor for critical feedback on the manuscript. This research was supported by grants from the Australian Research Council to M.A. (DE140100254 & FT170100025). Part of this work was carried out on the powder diffraction and X-ray fluorescence beamlines at the Australian synchrotron, which is part of ANSTO.
Nature thanks M. Bar-Matthews and R. Dennell for their contribution to the peer review of this work.