The 2026 study's proposal of 1.25–3 million people at the Aquiry peak (AD 100–300) is highly model-sensitive: it rests on an assumed simultaneous-use fraction and roughly 300 people per average earthwork, and the paper's own one-family-per-house alternative yields about 250,000–600,000 (315,000–760,000 with canopy correction).

The claim grades the estimate's architecture, and the fivefold sensitivity is the paper's own (AMZ-E002); credibility here is about the model structure, not the ancient population.

The source paper explicitly documents that its demographic estimate fluctuates by a factor of roughly five simply by toggling the assumed occupancy rate per mound. This proves that the widely cited "millions" figure is a highly sensitive output of a model, not a direct measurement of biological carrying capacity or skeletal remains. The strongest opposing consideration is the authors' own argument that a single-family reading is "very unlikely" given known workforce requirements for earthwork maintenance, meaning the lower bound may be excessively skeptical.

The population estimate’s sensitivity is laid out explicitly by the authors: assuming 300 people per earthwork and 25–60% simultaneous use yields 1.25–3 million, whereas a single-family-per-house reading produces roughly 250,000–600,000 (315,000–760,000 with canopy correction). The case correctly reports both ranges and the dependence on structural assumptions, making the claim about model-sensitivity factually established. The strongest objection is that the paper itself argues the high-occupancy assumption is more plausible, but that does not alter the truth of the sensitivity description.

AMZ-E002 quotes the paper's own architecture: 25–60 percent simultaneous use times 300 people per earthwork yields 1.25–3 million; the single-family reading of about 25 m house foundations yields 250,000–600,000, or 315,000–760,000 with canopy correction. The claim grades the estimate's sensitivity, not the ancient population, and the fivefold span is in the source. Strongest opposing consideration: the authors label the single-family reading very unlikely and argue that 300 matches later village sizes and workforce needs, so the span is an honest error bar rather than a refutation of the higher path.

The fivefold sensitivity is not an external critique but the paper's own alternative reading, verified verbatim: the single-family-per-house assumption yields 250,000–600,000 (315,000–760,000 with the penetration correction) against the 1.25–3 million headline. The claim grades the estimate's architecture, and the architecture is exactly as stated — occupants-per-earthwork and the simultaneous-use fraction do the work. The strongest opposing consideration is that the authors label the single-family reading "very unlikely" and defend roughly 300 people per earthwork from village sizes and workforce requirements, so the sensitivity range may overstate true uncertainty. But that defense is itself argument from analogy rather than measurement, so the claim stands as written. Established, high confidence.

The sensitivity is not an outside criticism but the paper's own arithmetic, quoted verbatim in AMZ-E002: 300 people per average earthwork times an assumed 25–60% simultaneous-use fraction gives 1.25–3 million, while the single-family house reading gives 250,000–600,000 (315,000–760,000 with canopy correction). That a stated assumption swings the output roughly fivefold is a fact about the model, and the claim asserts only that. The strongest opposing consideration is that the authors label the single-family reading "very unlikely" and defend the 300-person figure from later village sizes and labour requirements — so the lower bound is an error bar, not a competing estimate. Either way, this is a model output, not a census.

The sensitivity structure is the paper's own; the claim grades the estimate's architecture, not the ancient population.

The claim grades the estimate's architecture, and the roughly fivefold sensitivity span is the paper's own (AMZ-E002).

The sensitivity structure is stated in the paper itself, verified verbatim: "Only if the very unlikely estimation that each approximately 25-m long house foundation (mound) belonged to a single family were true would the population be about 250,000–600,000 people…". This claim grades the estimate's architecture, not the ancient population.

The claim accurately extracts the sensitivity analysis published in the 2026 Nature study itself, which explicitly notes that the 1.25 to 3 million population figure relies on an assumed 25–60% contemporaneity factor and 300 occupants per earthwork, whereas a single-family alternative model yields 250,000 to 600,000 occupants (or 315,000 to 760,000 adjusting for canopy occlusion). Stating that the headline demographic estimate is model-sensitive and spans a fivefold range is an objective reflection of the authors' own reported bounds. The strongest opposing consideration is that the study authors explicitly characterize the single-family alternative as "very unlikely" based on the large size (ca. 25 meters) of the house mounds.

The fivefold span is the paper's own sensitivity analysis, quoted verbatim: 1.25-3 million under the assumed simultaneous-use fraction and ~300 people per earthwork, versus 250,000-600,000 (315,000-760,000 with canopy correction) under the single-family alternative. Strongest opposing consideration: the authors label the single-family reading "very unlikely," so the claim honestly reports an error bar but does not by itself refute the headline — the low bound is one no proponent need hold.

AMZ-E002 gives the assumed simultaneous-use range, the 300-person conversion, the 1.25–3 million result, and the authors' much lower single-family alternative. AMZ-E003 confirms that the modeled peak is AD 100–300. The strongest opposing consideration is that the authors call the single-family interpretation very unlikely, but its presence still directly demonstrates strong sensitivity to occupancy assumptions.

The paper’s own population section (AMZ-E002) states the 1.25–3 million headline under simultaneous-use and ~300-person assumptions, and the single-family alternative of ~250,000–600,000 (315,000–760,000 with canopy correction). The claim accurately reports model sensitivity; the authors’ “very unlikely” label on the low end does not erase the stated span.

AMZ-E002 quotes both branches of the paper's own arithmetic: 1.25–3 million on a 25–60% simultaneous-use fraction and 300 people per earthwork, versus 250,000–600,000 (315,000–760,000 with canopy correction) under one family per house mound. The roughly fivefold span across the authors' own stated assumptions is exactly what the claim asserts, so the claim is true as a matter of the paper's text. The only opposing consideration is emphasis rather than fact: the authors label the single-family reading "very unlikely," so citing the low branch as the honest error bar is an editorial choice — but the model-sensitivity assertion itself is not in doubt.

AMZ-E002 quotes both the 1.25–3 million derivation (25–60 percent simultaneous use, 300 people per earthwork) and the single-family alternative of 250,000–600,000, or 315,000–760,000 with canopy correction, from the paper itself. The sensitivity is the authors' own; the emphasis on it is ours.

E002 (strong) quotes both the 1.25–3 million derivation and the paper's own single-family alternative of 250,000–600,000 (315,000–760,000 with canopy correction). The sensitivity is the authors'; the emphasis is this case's.

E002 (strong) quotes the paper's own headline derivation and its single-family alternative yielding roughly a fifth of the headline.

AMZ-E002 (strong, supports) quotes the paper's own derivation — 25–60% simultaneous use, 300 people per earthwork — and its single-family alternative of 250,000–600,000 (315,000–760,000 with canopy correction). The authors label the alternative very unlikely; the sensitivity is theirs, the emphasis the case's.

The population model in Pärssinen et al. (2026) is explicitly sensitive to baseline assumptions, requiring an assumed simultaneous-use rate and 300 occupants per earthwork; the paper's own one-family-per-house alternative drops the demographic estimate to 250,000–600,000 (or 315,000–760,000 corrected). This roughly fivefold spread is documented verbatim in the paper. Strongest opposing consideration: The authors explicitly designate the single-family alternative as "very unlikely" based on the large physical footprint (25 m) of the house mounds.

The paper's own sensitivity analysis, quoted verbatim (AMZ-E002), yields roughly a fivefold span between the headline and the single-family alternative the authors call "very unlikely." The strongest opposing consideration is minimal: the claim reports the model sensitivity the authors themselves disclose, so it is about as secure as a meta-observation about a weeks-old paper can be — though no peer-reviewed rebuttal of the model exists yet either way.

AMZ-E002 directly gives the assumed 25–60% simultaneous-use fraction, 300 people per earthwork, headline range of 1.25–3 million, and single-family alternative of 250,000–600,000 or 315,000–760,000 after correction. These explicit alternatives establish substantial model sensitivity. The strongest opposing consideration is that the authors themselves call the single-family interpretation very unlikely, so it is a sensitivity bound rather than an equally preferred estimate.

The paper’s population section states both the 1.25–3 million path (simultaneous-use fraction × ~300 people per earthwork) and the single-family alternative of ~250,000–600,000 (315,000–760,000 with canopy correction), labeling the latter very unlikely (AMZ-E002). The claim is exactly about that model-sensitivity; authors’ preference for the higher path is noted and does not erase the span.

This is a methodological claim about the paper's own arithmetic, and AMZ-E002 quotes both the headline derivation (25–60% simultaneity, 300 people per earthwork) and the paper's single-family alternative yielding 250,000–760,000. The roughly fivefold span is therefore demonstrated from the source's own text, not imputed. The honest counterweight is that the authors label the single-family reading "very unlikely," so the emphasis on the low branch is editorial; what the evidence establishes is model-sensitivity, not that the low figure is correct.

Pärssinen et al. (2026) explicitly present their population model's sensitivity, noting that an alternative one-family-per-house parameter drops the peak demographic estimate from 1.25–3 million down to 250,000–600,000 (or 315,000–760,000 with canopy correction). The strongest opposing consideration is that the authors themselves dismiss the single-family parameter as very unlikely based on house mound dimensions.

The claim is about the paper's own stated sensitivity, and both poles of the fivefold span are quoted verbatim from the fetched text — the headline model and the authors' own 'very unlikely' single-family alternative. No outside knowledge is needed; the opposing consideration is only that the case's framing emphasis is editorial, which the record itself flags.

AMZ-E002 explicitly gives the simultaneous-use fraction, the 300-person assumption, the 1.25–3 million estimate, and the much lower single-family alternatives with and without canopy correction. The strongest opposing consideration is that the authors call the single-family interpretation very unlikely. Their preference does not remove the demonstrated dependence of the population result on occupancy assumptions.

The same paper’s population section (E002) explicitly supplies both the 1.25–3 M headline (simultaneous-use fraction × 300 persons) and the single-family alternative of ~250–760 k. Strongest opposing consideration is the authors’ dismissal of the lower figure as ‘very unlikely’, yet the sensitivity span is theirs and must be reported.

E002 quotes both calculations verbatim. A 25–60% simultaneous-use fraction (4,165–10,000 earthworks) times an assumed 300 people per earthwork yields 1.25–3 million. The single-family reading yields 250,000–600,000, or 315,000–760,000 with the canopy correction. Even within the preferred model the result spans 2.4-fold on an assumed use fraction and scales linearly with an assumed per-earthwork figure, which justifies 'highly model-sensitive'. The strongest opposing consideration is that the authors call the single-family reading 'very unlikely', so treating it as the lower error bar imports an alternative they reject. The claim nonetheless reports it accurately as the paper's own alternative, and E002's limitation concedes the emphasis is the ledger's.

AMZ-E002 quotes the headline derivation: 25–60% of earthworks in use and 300 people each, giving 1.25–3 million. It also quotes the paper's single-family alternative: 250,000–600,000, or 315,000–760,000 corrected. The fivefold span is the paper's own arithmetic. The authors call the low reading 'very unlikely'; the claim asserts sensitivity, not that the low figure is right.