Across 7,363 recorded archaeological sites, distance from a river is the most important predictor of pre-Columbian occupation, and interfluvial areas were more sparsely populated than areas close to rivers.

Strong result over the largest available site database (AMZ-E010), with survey-effort bias toward rivers acknowledged as partially irreducible.

A massive spatial analysis of over 7,300 sites confirms that proximity to rivers heavily dictates pre-Columbian occupation, rendering the deep interfluves statistically sparse. This finding effectively synthesizes the debate by negating both the "empty basin" and the "uniformly crowded metropolis" theories in favor of a riparian corridor model. The strongest opposing consideration is that archaeological discovery is profoundly biased toward accessible river channels, meaning the observed sparsity in the interfluves could partly be an artifact of survey effort rather than ancient demographics.

McMichael et al. (2025) clearly report that, in models based on 7,363 archaeological locations, river distance is the most important predictor of pre-Columbian occupation and that interfluvial areas were more sparsely populated than river-adjacent ones. This aligns exactly with the claim. The strongest objection is that site databases and thus the model are biased toward accessible riverine areas, but even with that caveat the relative contrast between rivers and interfluves remains well-supported.

AMZ-E010, from 7,363 AmazonArch locations, finds distance from a river the most important predictor at 21.5 percent permutation importance and interfluves more sparsely populated than river-proximate areas. This is the single most discriminating basin-scale pattern in the ledger: it contradicts both a pristine wilderness and a uniformly dense occupation, and it predicts the corridor-heavy world the chronicles describe. Strongest opposing consideration: known-site databases inherit survey-effort bias toward accessible rivers, so part of the gradient is discovery, not ancient behavior, and the model cannot fully escape that.

The McMichael et al. 2025 result is verified verbatim: across 7,363 AmazonArch sites, distance from river is the most important predictor at 21.5% permutation importance, with interfluves modeled as more sparsely populated. As the basin-scale distributional fact, it is the single most discriminating record in the case, simultaneously contradicting pristine-wilderness and uniform-density readings and predicting exactly the corridor-heavy world the eyewitnesses describe. The strongest opposing consideration is discovery bias: rivers are where archaeologists can reach, so part of the gradient is survey effort rather than ancient behavior, and the model updates on biased input without fully escaping it. Well supported, high confidence as a model result — held short of established as a real-world gradient by the sampling confound.

AMZ-E010 reports the result from 7,363 database locations with a quantified effect (21.5% permutation importance for distance from river) and an explicit statement that interfluvial areas were more sparsely populated. The pattern is also mechanistically expected — transport, protein, floodplain soils — and it matches the independent bluff-model geography (AMZ-E006), so priors reinforce the measurement here rather than substituting for it. The strongest opposing consideration is that site databases inherit survey effort, and rivers are how archaeologists reach the forest; some unknown fraction of the gradient is discovery bias, which modeling can mitigate but not eliminate. Only promise-blind transects (AMZ-R003) can decompose it.

Strong result over the largest site database, with survey bias acknowledged as partially irreducible.

Strong result over the largest site database (AMZ-E010) with acknowledged survey bias; the new interfluvial fisheries evidence (AMZ-E022) qualifies the gradient's interpretation without contradicting the statistical pattern.

Verified verbatim from McMichael et al. 2025 (PNAS): "Distance from river… was the most important predictor variable" (21.5% permutation importance); interfluves "more sparsely populated." Note: the paper contains no explicit "not empty" disclaimer — sparser presupposes occupied, and this case quotes only what is there.

A machine-learning spatial modeling analysis of 7,363 archaeological sites across Amazonia by McMichael et al. (2025) identified distance to rivers as the single most critical predictor variable, accounting for 21.5% of model permutation importance and showing marked demographic thinning in interfluvial zones. This quantitative finding validates the classic ecological model of human settlement clustered along river corridors. The strongest opposing consideration is that the distribution of known archaeological sites is inherently confounded by modern discovery and survey biases, as accessible waterways receive vastly more fieldwork than remote terra-firme interfluves.

The 7,363-site model with river distance as the dominant predictor (21.5% permutation importance) and sparser interfluvial occupation is verified verbatim. Strongest opposing consideration, acknowledged in the record itself: site databases inherit survey-effort bias toward accessible riverine areas, so the model partly measures where archaeologists have looked — which is precisely the gap AMZ-R003's promise-blind surveys are designed to close.

AMZ-E010 reports a database of 7,363 locations, identifies river distance as the most important predictor with 21.5% permutation importance, and describes lower modeled occupation probabilities in interfluvial areas. The strongest opposing consideration is survey bias: known-site databases are likely enriched near accessible rivers. Thus the model result is well supported as stated, while its translation into true historical density remains less secure.

McMichael et al. 2025 (AMZ-E010) over 7,363 sites finds distance from river the top predictor (21.5% permutation importance) and interfluves more sparsely occupied than near-river areas. Survey-effort bias toward accessible rivers is acknowledged but does not reverse the ranked predictor result.

AMZ-E010 verifies both parts verbatim: distance from river as the top predictor at 21.5% permutation importance across 7,363 AmazonArch locations, and interfluvial areas modelled as more sparsely populated than riverine ones. The result is consistent with independent lines — Denevan's bluff model (AMZ-E006), Walker's limited earthwork and ADE extents (AMZ-C020), and the 0–1% probability mass over most of the basin (AMZ-C029). The strongest opposing consideration is circularity risk, which the record itself names: presence-only site databases inherit survey-effort bias toward accessible riverine areas, so a river-distance predictor is partly a map of where archaeologists have walked. AMZ-R003's promise-blind sampling frame is the only clean way to break that.

AMZ-E010 quotes the 7,363-site model, river distance at 21.5 percent of permutation importance as the top predictor, and the finding that interfluvial areas were more sparsely populated than those near rivers. The site database inherits survey-effort bias toward accessible riverine areas, which the source acknowledges as a limitation.

E010 (strong) quotes the 7,363-site model: distance from river is the top predictor (21.5 percent permutation importance) and interfluves were more sparsely populated. Site databases inherit survey-effort bias.

E010 (strong) states river distance as the top predictor across 7,363 sites and sparser interfluvial occupation; the database inherits survey-effort bias.

AMZ-E010 (strong, supports) quotes the 7,363-site model in which distance from river is the most important predictor (21.5% of permutation importance) and interfluvial areas are differentiated as more sparsely populated. Stated limit: site databases inherit survey-effort bias toward accessible riverine areas.

A predictive model analyzing 7,363 archaeological sites across Amazonia found distance from a river to be the single most important predictor of pre-Columbian human occupation (accounting for 21.5% of model permutation importance), demonstrating that interfluves were markedly less populated than river corridors (McMichael et al. 2025). Strongest opposing consideration: The site database inherits unavoidable geographic biases from historical survey efforts focused disproportionately along accessible navigable waterways.

AMZ-E010 verifies the 7,363-site database and the 21.5% permutation importance for river distance verbatim. The strongest opposing consideration, acknowledged in the record itself: site databases inherit survey-effort bias toward accessible riverine areas, so river-distance dominance partly reflects where archaeologists have looked as well as where people lived.

AMZ-E010 directly reports a model using 7,363 sites in which river distance has the greatest permutation importance and interfluvial occupation probabilities are lower than those near rivers. This strongly supports the statistical claim as stated. The strongest opposing consideration is that archaeological site databases are themselves biased toward accessible riverine survey areas, so the inferred population contrast may be overstated.

McMichael et al. 2025, using 7,363 AmazonArch sites, find distance from river the most important predictor (21.5% permutation importance) and interfluves more sparsely occupied than near-river areas (AMZ-E010). Survey-effort bias toward accessible rivers is acknowledged and cuts against over-reading emptiness, but does not reverse the dominant predictor result.

AMZ-E010 quotes a 7,363-site analysis in which distance from river is the top predictor at 21.5% permutation importance, with interfluvial areas modelled as more sparsely populated. The sample size and the explicit variable-importance statistic make this a measurement rather than an impression. The countervailing consideration is stated in the record and is serious: site databases inherit survey-effort bias toward accessible riverine areas, so some unknown portion of the river signal is a discovery artefact — which is precisely what AMZ-R003's promise-blind survey would resolve.

McMichael et al. (2025) analyze 7,363 archaeological sites and find that distance from river channels accounts for 21.5% of permutation importance in predictive modeling, with interfluves consistently exhibiting lower occupation density. The strongest opposing consideration is that archaeological databases are heavily biased by survey effort concentrated along easily navigable waterways and deforested transport corridors.

The 7,363-site model and the 21.5% permutation importance for river distance are verbatim-verified. The strongest opposing consideration is the record's own stated limitation: site databases inherit survey-effort bias toward accessible riverine areas, which could partly manufacture the very river-distance signal the model reports; the paper also contains no sentence disclaiming that distant forests were empty.

AMZ-E010 directly reports a model using 7,363 site locations in which distance from rivers was the most important predictor and interfluvial occupation probabilities were lower. The strongest opposing consideration is accessibility-driven survey bias, because known-site databases can overrepresent river corridors. The statistical result is strong, but the inferred magnitude of the true settlement contrast remains model-dependent.

McMichael et al. 2025 (E010) state distance from river is the most important predictor (21.5 % permutation importance) and that interfluves were more sparsely populated, based on 7,363 sites. Strongest opposing consideration is survey-effort bias toward accessible rivers, which the models attempt to address but cannot fully eliminate.

E010 verifies the model over 7,363 AmazonArch locations. Distance from river is the most important predictor (21.5% of permutation importance), and the models separate interfluvial areas 'that were more sparsely populated than those close to rivers'. The strongest opposing consideration is the record's own limitation: site databases inherit survey effort biased toward accessible rivers, which inflates exactly this predictor. Moreover, 21.5% is the largest share, not a majority. The ledger also documents substantial upland and interfluvial occupation (Belterra, E017; Loma Salvatierra, E022), so the gradient is a tendency, not a rule, and it is consistent with a riverine account like Carvajal's.

AMZ-E010 reports that across 7,363 sites, river distance is the strongest predictor (21.5% permutation importance), with interfluves more sparsely occupied. Site databases inherit survey bias toward accessible rivers, which the record states.