Tegiszhol

What it is for

Exposure — how much of a claim rests on dates nobody can check

The register on its own is a list. The use is this.

For any chronological claim resting on radiocarbon evidence, what share of that evidence could nobody check? That question is answerable today, with no new laboratory work, by anyone holding the CSV.

It corrects nothing. It tells you where correction would matter — which is the difference between “our chronology might be wrong somehow” and “this attribution rests on N dates carrying a documented up-to-millennium error mechanism and no test for it.”

Three things you can do with it

1. Re-run your chronology without the unscreenable dates. If the conclusion survives, it is considerably stronger than it was — you have shown it does not depend on unverifiable evidence. If it does not survive, you have found something. This needs no new measurement, only the list.

2. Aim new isotopic work where it buys the most. The site table below ranks collections by how many uncheckable determinations sit in each. Measuring δ¹³C/δ¹⁵N on one collection can resolve a dozen dates at once — where the material still exists.

3. Check a single date before citing it. If a date is doing real work in an argument, it is worth knowing whether anyone could have screened it.

The result that prompted this page

The Sintashta → Petrovka → Alakul sequence turns on distinctions of roughly two centuries, and is usually described as beyond radiocarbon’s resolution.

In this corpus, dropping the uncheckable dates attributed to Петровка moves that group’s median by +195 years.

Read that carefully, because it is easy to overstate. Only three determinations are removed, so the shift is fragile and could be an artefact of small numbers. It is not evidence that the Petrovka chronology is wrong. It is evidence that the question “does this sequence survive removing the dates nobody can check?” is worth asking of a properly calibrated model — because the magnitude of the possible movement is the same as the magnitude of the distinction being argued over.

That is the whole argument of this project in one row of a table.

Exposure by stated attribution

Labels are exactly as each source prints them — not harmonised, not translated. That is deliberate, and it produces an instructive artefact: Afanasievo (7 dates, 100% exposure) and Afanasievo culture (24 dates, 0% exposure) are the same culture written two ways by different authors. Their exposure is opposite. Harmonising the labels is real work that has not been done here, and until it is, these rows describe publications, not cultures.

“Median if dropped” is the group’s median age recomputed with the unscreenable dates removed, and the change from the full median. Small n after dropping is marked; treat those as questions, not results.

Exposure by site

Ranked by how many uncheckable determinations sit in each collection — the practical ordering for anyone planning new measurement.

What this does not say

No date here is claimed to be wrong. Reservoir offsets are real, documented at over a millennium on this material, and unpredictable — which is exactly why an unscreened date cannot be assumed either sound or faulty. Most of these dates are probably fine. Nobody can currently demonstrate which.

Nor is this a criticism of the excavators. Much of this material was dated decades before isotopic screening was routine, by people doing careful work with the methods of their time. The register exists to aim new effort, not to second-guess old effort.