Narrative cave art in Indonesia by 51,200 years ago

To demonstrate the efficiency and reliability of this technique, we re-dated what was previously the oldest known surviving pictorial narrative, a rock art scene at Leang Bulu’ Sipong 4, which we had already dated to a minimum of 43.9 ka. At this cave site, a 4.5-m-wide panel on the rear wall comprises several figurative paintings of human-like figures interacting with Sulawesi warty pigs and dwarf bovids (anoas, Bubalus sp.) (Fig. 2). The former are depicted with material culture objects (spears and/or ropes) and some display what can be construed as attributes of non-human animals. These figures are interpreted as representations of therianthropes (composite human–animal beings). This enigmatic scene may represent a hunting narrative, while the prominent portrayal of therianthropic figures implies that the artwork reflects imaginative storytelling (for example, a myth).

We originally dated a total of four coralloid speleothems overlying animal figures on this panel to a minimum of 35.1 ka (35.7 ± 0.6 ka; sample BSP4.2), 43.9 ka (44.4 ± 0.5 ka; BSP4.3), 40.9 ka (41.1 ± 0.2 ka; BSP4.4) and 41 ka (41.3 ± 0.4 ka; BSP4.5). Using our LA-U-series method, those same speleothems and the associated artwork are now dated to a minimum of 27.6 ka (28.3 ± 0.6 ka; sample BSP4.2), 39.6 ka (43.2 ± 3.6 ka; sample BSP4.3), 39.5 ka (40.4 ± 0.9 ka; sample BSP4.4) and 48 ka (51.2 ± 2.2 ka; sample BSP4.5) (Fig. 2). Our LA-U-series approach provides either similar ages within error or older ages when compared with previous dates for the same samples (Fig. 2). The only exception is for sample BSP4.2, for which the LA-U-series data for the calcium carbonate deposits closer to the pigment layer are younger. We attribute this discrepancy to our selective avoidance of areas within the sample showing clear alteration (Methods, Extended Data Fig. 4 and Extended Data Table 1).

The rock art scene at Leang Bulu’ Sipong 4 can now be demonstrated to be at least 4,040 years older at 48 ka. On the basis of previous results, it appears that our mechanical microsampling approach either fortuitously avoided areas of diagenesis or that the averaging of microexcavated arbitrary layers rendered any localized digenesis inconsequential to the overall age calculation for these layers. Again the exception is for sample BSP4.2, for which the older minimum age for solution data near the pigment layer could be attributed to localized diagenesis. Our mapping data also show clear evidence of diagenesis occurring at the surface of samples and sometimes near tiny cavities within them. The latter areas would be impossible to avoid when microexcavating arbitrary spits for solution-based U-series methods and could result in erroneous age determinations. Using the LA-U-series mapping approach, these areas of localized digenesis can easily be avoided (that is, not integrated) when calculating U-series ages from map data. Notably, for example, the ²³⁰Th/²³²Th activity ratios for the laser-ablation regions of integration (ROIs) (Extended Data Table 1) are substantially higher than solution data for the same samples owing to the selective avoidance of areas with higher detrital content.