The oldest art is where the limestone drips

Claim (verbatim)

For a century the earliest figurative art was European by default. Then uranium-series dating of the calcite crusts growing over cave paintings in the Maros-Pangkep karst of Sulawesi and in Borneo (Aubert and colleagues, Nature, 2014 and 2018; Brumm and colleagues, Science Advances, 2021) pushed the oldest known figurative images - a warty pig at Leang Tedongnge, a wild bovid at Lubang Jeriji Saleh, hand stencils - back to roughly 40,000-45,500 years, in Island Southeast Asia, not Europe. The dates are real, but they are minimum ages of a very particular kind: they date the calcite that happens to have grown over a painting, so 'the oldest art' is discoverable only where datable carbonate forms and is sampled. That makes the location and age of the 'earliest' art partly a map of geology and fieldwork, not of where image-making began. The disciplined test asks whether the record behaves like a sampling process: if the earliest date recovered from a region climbs as more panels are dated there, the 'oldest art' frontier is sampling-bounded, and the deeper origin lies in perishable and undatable media everywhere. The missing knowledge is when figurative image-making truly began; its apparent frontier is a countable function of sampling effort.

Prediction clause (verbatim)

Prediction: across the uranium-series-dated Pleistocene rock-art corpus, the oldest minimum age recovered per site (or per region) will correlate positively and strongly (rank correlation >=0.5) with the number of dated calcite samples taken there - the signature of an extreme-value/sampling process rather than a true origin point - and no region with only a handful of dated samples will define the global maximum (primary clause: the >=0.5 sample-count-versus-oldest-age rank correlation). Disambiguation: only U-series minimum ages on overlying/underlying calcite are pooled; radiocarbon on pigment and stylistic dates are excluded so the taphonomic mechanism is isolated. Coverage guard: if fewer than 8 sites have published sample counts alongside their extreme ages the correlation voids for insufficient sampling rather than resolving.

Kill-dataset (verbatim)

Kill: the compiled U-series rock-art dataset - Aubert et al. (2014, Nature, Sulawesi; 2018, Nature, Borneo), Brumm et al. (2021, Science Advances, Leang Tedongnge), and Hoffmann et al. (2018, Science, Iberia) - with each site's number of dated calcite samples and its oldest recovered minimum age; the operation is the rank correlation between per-site sample count and per-site extreme age.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

The mechanism is a published caveat: U-series dates on cave-art calcite are explicitly minimum ages, sampling- and preservation-dependent (Aubert et al. 2014, 2018; Brumm et al. 2021; Hoffmann et al. 2018 for Iberia), and methodological critiques stress that finer/more sampling tends to yield older minima (the laser-ablation redating of Sulawesi to ~67.8 ka is a live illustration). But no located analysis computes the conjecture's statistic - a rank correlation (>=0.5) between the per-site number of dated calcite samples and the per-site oldest recovered minimum age across the pooled corpus, the fingerprint of an extreme-value/sampling process. The caveat is stated; the cross-site correlation is unrun.

• M. Aubert et al., 'Pleistocene cave art from Sulawesi, Indonesia,' Nature 514 (2014); and M. Aubert et al., 'Palaeolithic cave art in Borneo,' Nature 564 (2018)
• A. Brumm et al., 'Oldest cave art found in Sulawesi,' Science Advances 7(3) (2021)
• D.L. Hoffmann et al., 'U-Th dating of carbonate crusts reveals Neandertal origin of Iberian cave art,' Science 359 (2018), together with the subsequent methodological critiques of U-series rock-art minimum ages

Predictions

No prediction registered yet.