The bone that will not hold its count

Claim (verbatim)

Alexander Marshack's The Roots of Civilization (1972) read the engraved marks on Aurignacian and later bones and plaques - the Abri Blanchard bone, the Grotte du Tai plaque - as lunar notations, sequences of observations accumulated over months. Francesco d'Errico's microscopic and statistical re-analyses in the 1990s re-read the same objects, arguing from tool-track morphology that the marks were made in episodes and techniques inconsistent with a running lunar tally. The lasting lesson is not which man was right but that the dispute reached down below interpretation to the data itself: when analysts disagree about whether a plaque is a calendar, they frequently also disagree about how many marks it bears and how they group. That instability is itself a measurable property of the corpus, and it is the honest object of study here. If Palaeolithic 'notation' objects carried a stable, readable count, independent microscopic re-readings would converge on the same N; if the marks are ambiguous accretions over-fitted by each reader, the reported counts and groupings will scatter. The missing knowledge - what, if anything, was being recorded - is bounded from below by this: you cannot decipher a tally whose length is analyst-dependent.

Prediction clause (verbatim)

Prediction: assembling every independently published mark-count and grouping for a fixed set of contested Palaeolithic notation objects (Abri Blanchard, Grotte du Tai, Ishango, Lebombo, the Tossal de la Roca plaque), the ratio of the highest to the lowest reported total mark-count will reach or exceed 1.3 for at least half of the objects that have been re-analysed by two or more independent teams, and at least one object will show two incompatible groupings (primary clause: the >=1.3 max/min count ratio in >=50 percent of multiply-analysed objects). Coverage guard: objects with only a single published analysis are excluded from the ratio test and reported separately, so the census measures instability, not scarcity.

Kill-dataset (verbatim)

Kill (not yet built): a census table compiled from the primary publications - Marshack (1972) and his later papers, d'Errico "Microscopic and statistical criteria for the identification of prehistoric systems of notation" (Cambridge Archaeological Journal, 1995) and d'Errico & Cacho (1994), plus de Heinzelin on Ishango - recording each analyst's total mark-count and grouping per object, then computing the max/min count ratio and grouping-incompatibility per object.

Novelty / leakage triage

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

The instance-level disagreements are published: Marshack (1972) read the Abri Blanchard bone and Grotte du Tai plaque as accumulating lunar tallies, and d'Errico's microscopic re-analyses (CAJ 1995; d'Errico & Cacho 1994 on Tossal de la Roca) explicitly re-count and re-group the marks against him, so the phenomenon of analyst-dependent counts is documented object by object. But the conjecture's operation is a census across a fixed object set - every published total scored, with the max/min count ratio (>=1.3 on >=50% of multiply-analysed objects) and grouping-incompatibility computed across the set. That cross-object ratio statistic was not located as compiled; this is precisely the census-vs-instance distinction.

• A. Marshack, The Roots of Civilization (New York: McGraw-Hill, 1972)
• F. d'Errico, 'Microscopic and statistical criteria for the identification of prehistoric systems of notation,' Cambridge Archaeological Journal 5(1) (1995)
• F. d'Errico & C. Cacho, 'Notation versus decoration in the Upper Palaeolithic: a case-study from Tossal de la Roca,' Journal of Archaeological Science 21 (1994)
• J. de Heinzelin, 'Ishango,' Scientific American 206(6) (1962), for the Ishango member of the census set

Predictions

No prediction registered yet.