Mitigation strategies in practice

Existing decontamination methods are unified, not replaced: each performs the same two moves — isolate the foreign component on age-independent grounds, then validate out-of-sample. Radiocarbon climbs ABA → ABOX-SC → ultrafiltration → hydroxyproline AMS; zircon climbs air abrasion → CA-ID-TIMS, CL-guided dating, ²⁰⁴Pb common-Pb, concordance. Every rung has a stated ceiling; none converts an Axis-B problem into an Axis-A guarantee.

Existing decontamination methods are not replaced but unified: each is the same two moves—isolate the foreign component on age-independent grounds, then validate out-of-sample. Radiocarbon climbs ABA → ABOX-SC → ultrafiltration (>30 kD) → compound-specific hydroxyproline AMS, plus ΔR calibration and Bayesian outlier models; zircon climbs air abrasion → CA-ID-TIMS, CL-guided spot dating, ²⁰⁴Pb common-Pb, and concordance. Every rung has a stated ceiling, and none converts an Axis-B problem into an Axis-A guarantee.

This protocol does not replace the decontamination methods each field already uses; it names the logic they share and adds out-of-sample validation as a standing requirement. Both communities have evolved a ladder of mitigations (Table 4). Each rung removes more of the foreign component on grounds independent of the age being measured, and each has a documented ceiling. Reading them as instances of one strategy is the contribution; the reagents are not new.

Table 4. Existing mitigations as instances of the same two moves: isolate the foreign component on age-independent grounds, then validate out-of-sample. Each method has a stated ceiling.
ClockMethodWhat it isolates / removesCeiling (where it fails)
¹⁴CABA → ABOX-SC (charcoal)acid/base-soluble + oxidisable contaminantsrecalcitrant cross-linked carbon
¹⁴Cultrafiltration of collagen[20]>30 kD degraded gelatin/humicscannot remove <30 kD or cross-linked/glue carbon
¹⁴Ccompound-specific (hydroxyproline) AMS[21,22]a bone-specific amino acid onlycollagen-based glue; needs more material
¹⁴CΔR databases + Marine20/IntCal20[15,16]known reservoir offset (curve)offset transferability is locality-dependent
¹⁴CBayesian age models + outlier detection[24]statistically discordant dates (priors)garbage-in if priors/associations wrong
U–Pbair abrasion[26] → CA-ID-TIMS[25]outer / radiation-damaged Pb-loss domainswhole-grain inheritance (needs single-grain)
U–PbCL/BSE imaging + spot dating (SIMS, LA-ICP-MS)specific growth zones (rim vs core)ambiguous textures ⇒ circular assignment
U–Pb²⁰⁴Pb / Tera–Wasserburg common-Pbnon-radiogenic Pb componentleverage grows for young, low-radiogenic grains
U–Pbconcordance + cross-method (⁴⁰Ar/³⁹Ar, (U–Th)/He, astrochronology)internal + external consilienceno written-record ground truth in deep time

Radiocarbon. The pretreatment ladder runs from acid–base–acid, through wet-oxidation stepped combustion (ABOX-SC) for charcoal, to ultrafiltration of bone collagen, which retains the high-molecular-weight gelatin fraction (>30 kD) and discards low-weight degradation products and humics[20]. [F] Its ceiling is explicit and was reached at Vindija (§9): ultrafiltration cannot remove contaminants below the filter cut-off or those chemically cross-linked to collagen. The next rung, compound-specific dating, isolates a single amino acid (hydroxyproline, ~10% of collagen and rare elsewhere in nature) so that essentially every contaminant except collagen-based glue is excluded, at the cost of requiring more bone[21,22]. [F] Crucially, each rung selects a chemically or molecularly defined fraction—an age-independent criterion—never “the part that looks the right age.” For the reservoir problem, the field maintains regional ΔR databases and the Marine20/IntCal20/SHCal20 calibration curves[15,16]; but reservoir-offset transferability remains locality-dependent, the same caveat our leave-one-out test surfaced quantitatively (§6 unequal gains). A statistical layer—Bayesian age–depth modelling with explicit outlier detection (OxCal and similar)—down-weights discordant dates using stratigraphic priors[24], a model-level realisation of “screen, then report honest uncertainty.” [I]

Zircon. Physical and chemical removal of compromised material progressed from mechanical air abrasion of grain exteriors[26] to chemical abrasion (CA-ID-TIMS), which anneals radiation damage and then partially dissolves the damaged, Pb-loss domains, leaving residual closed-system zircon and effectively eliminating Pb-loss discordance[25]. [F] Imaging-guided microsampling—cathodoluminescence or back-scatter imaging followed by spot dating with SIMS or LA-ICP-MS—targets individual growth zones, supplying the rim-versus-core classifier the protocol depends on. Isotopic screening removes the common-Pb component through measured ²⁰⁴Pb or the Tera–Wasserburg intercept, with the honest caveat (§11) that its leverage grows for young, low-radiogenic grains. Finally, concordance is a built-in self-check, and cross-method agreement (⁴⁰Ar/³⁹Ar sanidine, (U–Th)/He) together with astrochronological tuning supplies the consilience that substitutes for written-record ground truth where none can exist. [I]

Where this protocol sits. Every method above is an instance of the same two moves: (i) isolate or avoid the foreign component on grounds independent of the age one is trying to measure—chemistry, molecular identity, crystallography, stratigraphy—and (ii) validate out-of-sample or by cross-method consilience. The contribution of this paper is not a new reagent but a single accounting that makes those two moves explicit, comparable across chronometers, and auditable per claim through the [F]/[I]/[A] tags. [I] And there is a limit no rung removes: no mitigation converts an Axis-B problem into an Axis-A guarantee. Where the foreign component cannot be characterised—an unknown reservoir, an ambiguous CL texture—the correct output is a wider, explicitly stated uncertainty, not a precise number. [A]