The echo that did not choose the wall

Claim (verbatim)

Iegor Reznikoff and Michel Dauvois proposed in 1988 that Palaeolithic painters placed images where caves resonate - that the density of red marks in Arcy-sur-Cure and Le Portel tracks points of strong echo and low-frequency resonance, as if sound and image were sited together. The idea is seductive and has been extended by instrumented acoustic surveys, notably Bruno Fazenda and colleagues' 2017 measurements (Journal of the Acoustical Society of America) and the wider 'Songs of the Caves' project. This lane treats the claim as a replication problem, not a revelation. Painted locations are non-random for many mundane reasons - flat paintable panels, wide chambers, accessible niches - and those same features shift acoustics, so a raw correlation between paint and resonance can be an artefact of geometry. The disciplined test controls for where painting is physically easy and asks whether any acoustic preference survives. The missing knowledge is whether sound guided the painters at all; it is decidable only if the acoustic signal beats a spatial null that already knows the cave's shape.

Prediction clause (verbatim)

Prediction: in the Fazenda et al. acoustic-survey dataset, the difference in the resonance/reverberation metric between painted locations and matched unpainted control points, after a spatial-permutation null that preserves chamber geometry and wall paintability, will have an effect size below 0.2 standard deviations and will fail to exceed 95 percent of the geometry-preserving permutations (primary clause: the permutation-corrected effect falls short of the 95th-percentile null). Disambiguation: control points must be drawn from paintable surfaces within the same chamber, not from unpaintable voids, so the test isolates acoustic selection from access. Coverage guard: caves with fewer than 20 measured painted points are reported individually and excluded from the pooled effect so a single richly-measured site cannot carry the verdict.

Kill-dataset (verbatim)

Kill (not yet built, from real sources): the Fazenda et al. (2017, J. Acoustical Society of America) painted-versus-control acoustic measurements and the associated 'Songs of the Caves'/Acoustics and Landscape survey data (La Garma, Cantabrian caves), plus Reznikoff & Dauvois' Arcy/Le Portel resonance points - compute the painted-vs-control acoustic contrast under a geometry-preserving spatial permutation and read its effect size and percentile against the null.

Novelty / leakage triage

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

The paint-meets-resonance correlation is published - Reznikoff & Dauvois (1988) mapped red marks against resonant points at Arcy-sur-Cure and Le Portel, and Fazenda et al. (2017, JASA) instrumented painted-versus-control acoustics across Cantabrian caves. But the conjecture's discriminator is a spatial-permutation null that preserves chamber geometry and wall paintability, isolating acoustic selection from the mundane fact that paintable panels also change acoustics. No located analysis runs that geometry-preserving permutation with the <0.2 SD effect-size / 95th-percentile test; the raw association is published, the confound-controlled arithmetic is not.

• I. Reznikoff & M. Dauvois, 'La dimension sonore des grottes ornees,' Bulletin de la Societe Prehistorique Francaise 85 (1988)
• B. Fazenda et al., 'Cave acoustics in prehistory: exploring the association of Palaeolithic visual motifs and acoustic response,' Journal of the Acoustical Society of America 142(3) (2017)

Predictions

No prediction registered yet.