Immediately before European contact, Amazonia was a strongly heterogeneous human landscape: substantial sedentary populations, engineered subsistence, and settlement systems along favored river corridors, floodplains, savannas, and bluffs, embedded in a much more lightly occupied forest matrix.

Both halves are independently supported — dense nodes (AMZ-E001, AMZ-E017, AMZ-E014) and thin matrix (AMZ-E010, AMZ-E011); the open question is proportions, not structure.

The cumulative evidence strongly supports a mosaic model of pre-Columbian Amazonia, where massive engineering and dense populations clustered heavily along optimal river corridors while vast tracts of terra-firme forest remained lightly occupied. This structural synthesis elegantly accommodates both the spectacular LiDAR discoveries and the stark negative paleoecological results from the interfluves. The strongest opposing consideration is that our current "mosaic" heavily reflects modern sampling bias—with LiDAR targeted at known hotspots and soil cores undersampling everywhere—meaning future surveys could radically shift the presumed ratio between dense nodes and sparse matrix.

This interpretive synthesis is directly underpinned by strong evidence for dense, engineered settlement corridors (Aquiry: AMZ-C004; southern rim: AMZ-C006; Casarabe, Belterra, Marajó, Xingu: AMZ-C007, AMZ-C011) and equally strong evidence that interfluvial and many western terra-firme regions lack signals of intensive agriculture or dense occupation (AMZ-C013, AMZ-C014). Together these support a heterogeneous mosaic with substantial populations along favored corridors within a lighter matrix. The main objection is that both positive and negative datasets are biased samples, so the precise proportions remain uncertain, but the qualitative mosaic structure is robustly supported.

Dense nodes are independently documented by earthworks and settlement systems (AMZ-C004, AMZ-C006, AMZ-C007, AMZ-C011); the thin matrix by the river-distance gradient and terra-firme palaeoecology (AMZ-C013, AMZ-C014); and the eyewitness accounts describe exactly this alternation (AMZ-C001, AMZ-C003). Neither pristine wilderness nor wall-to-wall civilization survives that conjunction; the mosaic structure does. Strongest opposing consideration: both the spectacular nodes and the sparse cores are biased samples, so the mosaic's proportions — how much corridor, how much matrix — could still move substantially as promise-blind coverage improves.

This is the synthesis the two opposing evidence families jointly force: earthworks, fortified villages, drained-field agriculture, and contact-horizon towns (AMZ-C004, C006, C007, C011) establish the dense nodes; the river-distance gradient and the fourteen-region palaeoecological negative space (AMZ-C013, C014) establish the thin matrix; and the eyewitness alternation of crowded and empty reaches (AMZ-C001, C003) matches the structure independently. Crucially, the mosaic survives any plausible shift in proportions — even the most skeptical reading of this record concedes the nodes, and even the most maximalist concedes the gradient. The strongest opposing consideration is that both sampling frames are biased in opposite directions — spectacular sites oversampled, tiny cores over-extrapolated — so the matrix density, the model's open parameter, could still move either way. That uncertainty attaches to proportions, which the claim deliberately does not specify. Well supported, high confidence.

The mosaic is the one model that every line of evidence in this ledger fits simultaneously: dense engineered nodes (AMZ-E001, E004, E014, E017, E018), a demonstrably thinner matrix by two methodologically independent routes — site distribution modeling (AMZ-E010) and soil palaeoecology (AMZ-E011) — and an eyewitness geography of alternating crowded and empty reaches (AMZ-E006, AMZ-E007). Because the corridor and matrix evidence come from different instruments with opposite biases, the qualitative structure is robust. The strongest opposing consideration is that both halves rest on biased samples pulling in opposite directions, so the proportions — the only thing still in dispute — could move substantially; "mosaic" is nearly unfalsifiable at the structural level and only becomes a test once matrix density is quantified.

Both halves (dense nodes, thin matrix) independently supported; the open parameter is proportions, not structure.

Both halves independently supported; the new Mojos fisheries evidence slightly complicates the 'lightly occupied matrix' half by showing engineered interfluvial subsistence in at least one region, but the structural claim stands.

The synthesis both sides' evidence now converges on: the earthwork and settlement systems (AMZ-C004, C006, C007, C011) establish the dense nodes; the river-gradient and palaeoecological negative space (AMZ-C013, C014) establish the thin matrix; and the eyewitness accounts describe exactly this alternation (AMZ-C001, C003).

The overarching synthesis of archaeological, paleoecological, and spatial modeling records strongly supports a mosaic model of pre-Columbian Amazonia: dense, sedentary populations, intensive food production, and monumental earthworks were clustered along major river corridors, floodplains, bluffs, and savannas, while the vast interfluvial forest matrix remained lightly settled and largely uncultivated. This framework reconciles the monumental discoveries in Acre, Mojos, Santarém, and the Xingu with the persistent negative paleoecological evidence from interfluvial cores. The strongest opposing consideration is the possibility that widespread agroforestry and mobile foraging in the interfluves left ephemeral, low-visibility archaeological signatures that evade conventional survey techniques.

The heterogeneity synthesis follows jointly from the earthwork, Casarabe, contact-era-occupation, and sparse-interfluve records, which pull in opposite directions and are reconciled only by the mosaic reading. Strongest opposing consideration: the synthesis's central open parameter — actual population density of the forest matrix between corridors — has never been measured by a promise-blind survey, so the hotspot-to-matrix ratio remains an assumption rather than a finding.

Late fortified settlements in AMZ-E004 and contact-era systems in AMZ-E017 through AMZ-E019 support substantial regional populations, while AMZ-E010 and AMZ-E011 support lower occupation in much of the interfluvial matrix. Agricultural and hydraulic evidence further shows that at least some hotspots were intensively engineered. The strongest opposing consideration is chronological unevenness: prominent Aquiry and Casarabe evidence substantially predates contact, and neither the hotspot nor matrix densities are measured basin-wide without sampling bias.

The claim is the structural synthesis of earthwork/settlement positives (C004, C006, C007, C011) with sparse-matrix constraints (C013, C014). Evidence records support both legs without forcing either pristine emptiness or basin-filling civilization. Residual uncertainty is the unmeasured true interfluve density under promise-blind sampling.

This synthesis is the one proposition in the case that both evidence legs support rather than one. Dense engineered occupation is measured at Acre (AMZ-E001), the southern rim (AMZ-E004), Mojos (AMZ-E014, AMZ-E022), Santarém (AMZ-E017), and Marajó (AMZ-E018); sparse matrix is independently indicated by river-distance dominance (AMZ-E010), the terra-firme proxy nulls (AMZ-E011), limited modelled earthwork and ADE extents (AMZ-C020), and roughly 80% of the basin at 0–1% earthwork probability (AMZ-C029). The strongest opposing consideration is that the matrix term is quantitatively unconstrained: every negative record is a biased or erasure-prone sample (AMZ-C021, AMZ-C022) and every positive record is a discovery-biased one, so "much more lightly occupied" is an ordering claim with no defensible density attached. The ordering survives; the magnitude does not.

The synthesis rests on both legs being carried: hotspots by AMZ-E001, AMZ-E004, AMZ-E014, AMZ-E017, AMZ-E018, AMZ-E022; matrix by AMZ-E010 and AMZ-E011. The heterogeneous reading is what survives both the positive and negative records. Its open parameter — the matrix density — is inferred from biased samples on both sides, which is why confidence is moderate rather than high.

This synthesis claim has no anchor and no evidence record of its own; it is graded on the admitted records for its named dependencies — E001, E004, E014, E017, E018, E019 for dense engineered corridors and hotspots; E010 and E011 for a lighter matrix. Both halves are directly evidenced; the magnitude in ‘much more lightly occupied’ is inferred, not measured, and the qualifiers that would sharpen it (C020, C021, C022, C029) are anchor-only.

An attribution claim with no anchor or evidence record of its own; the basis named here is the admitted evidence on its dependency claims — E001, E004, E014, E017, E018, E019 for the dense corridors and E010, E011 for the lightly occupied matrix — which carries the support given. What it cannot fix is the matrix density.

A synthesis with no anchor and no record of its own; the grade is carried entirely by the disclosed basis of admitted records on its dependency claims: E001 (C004), E004 (C006), E014 (C007), E017–E019 (C011) for the dense corridors and hotspots, and E010 (C013), E011 (C014) for the lightly occupied matrix. Confidence is moderate because the matrix density is inferred from survey-biased sites and sparse cores, and the qualifiers that would bound it (C020, C021, C022, C029) are anchor-only.

This claim synthesizes the entire ledger: late pre-Columbian Amazonia was neither a pristine wilderness nor a wall-to-wall civilization, but a pronounced human mosaic of dense, engineered agricultural and urban systems along major rivers, savannas, and bluffs, embedded within a sparsely occupied interfluvial forest matrix. Strongest opposing consideration: Advocates of high-density models contend that low-impact interfluvial agroforestry was ubiquitous and that post-abandonment ecological recovery has masked extensive inland human modification.

All six named dependencies (C004, C006, C007, C011, C013, C014) assess as well supported, and the heterogeneity synthesis is the direct reading of dense engineered corridors alongside sparse interfluves. The strongest opposing consideration: the "lightly occupied matrix" quantifier is the least-measured term in the claim — the positive record (discovered sites) and negative record (sparse cores) are both biased samples, and no promise-blind interfluve survey (AMZ-R003) has been run.

Late and contact-era settlement evidence in AMZ-E004 and AMZ-E017–E019, intensive subsistence evidence in AMZ-E014 and AMZ-E022, and the riverine pattern in AMZ-E010 jointly support substantial favored-location populations. AMZ-E011 supports a much lighter terra-firme matrix, producing the claimed heterogeneous mosaic rather than a basin-wide dense landscape. The strongest opposing consideration is that several most spectacular systems predate contact, while survey and palaeoecological sampling biases leave the exact density of the matrix uncertain.

The claim is the structured roll-up of dense corridor and hotspot evidence (AMZ-C004, C006, C007, C011 and supporting landscape work) with sparse-matrix evidence (AMZ-C013, C014). No single counter-record restores either pristine emptiness or uniform basin-wide intensive occupation; remaining uncertainty is quantitative matrix density and local route representativeness, not the heterogeneous pattern itself.

This synthesis is supported from both directions at once, which is the strongest structure available: positive records establish dense engineered settlement along rivers, floodplains and savannas (AMZ-E001, AMZ-E004, AMZ-E014, AMZ-E017, AMZ-E018, AMZ-E022), while independent negative records establish river distance as the dominant occupation predictor and little intensive-use signal across most sampled interfluvial terra firme (AMZ-E010, AMZ-E011). Neither the pristine-wilderness nor the uniform-civilization reading survives both sets. The chief opposing consideration is that the matrix density is still estimated from biased samples on both sides — discovery bias inflates the riverine signal, thin coring and time-graded erasure (AMZ-C021, AMZ-C022) deflate the interfluve signal — so the contrast's magnitude, though not its direction, remains open.

The convergence of macro-scale spatial predictive modeling (McMichael 2025; Peripato 2023), paleoecological core surveys (Piperno 2024), and targeted regional archaeology (Pärssinen 2026; Lombardo 2025; Stenborg 2018) strongly supports a heterogeneous mosaic model: densely populated, highly modified riverine and peripheral savannah corridors embedded in lightly utilized, largely intact interfluvial forests. The strongest opposing consideration is that significant portions of the terra-firme interfluves remain completely unsurveyed by LiDAR, meaning undiscovered non-mound settlement forms could alter the apparent density contrast.

Every dependency in the synthesis is individually supported, and the heterogeneity picture is the only reading consistent with both the engineered-hotspot record (C004, C006, C007, C011) and the sparse-matrix record (C013, C014). The strongest opposing consideration is that the matrix-density parameter — the quantity that separates 'heterogeneous' from 'mostly full' — is currently estimated from biased samples on both sides, with palaeoecological negatives further discounted by the Upano erasure finding; AMZ-R003 is the direct measurement and has not been run.

Late and contact-era settlement evidence in AMZ-E004 and AMZ-E017 through AMZ-E019 supports substantial sedentary hotspots, while AMZ-E014 through AMZ-E016 and AMZ-E022 demonstrate engineered subsistence in favored settings. AMZ-E010 and AMZ-E011 independently support lower occupation away from rivers and weak intensive-use signals across many sampled terra-firme locations. The strongest opposing consideration is that both positive-site discovery and negative-core sampling are spatially biased, so the density of the surrounding matrix remains uncertain even though the heterogeneous mosaic is strongly supported.

The claim is the explicit synthesis of the supported measurements in its dependency list (dense earthwork and settlement systems C004/C006/C007/C011 versus river-dominated and terra-firme-sparse patterns C013/C014). Strongest opposing consideration is residual uncertainty over absolute matrix density pending promise-blind surveys, yet the qualitative heterogeneity is already required by the existing positive and negative records.

Both halves are supported, and by different research networks. Dense, organized occupation at or near contact is documented on the southern rim (E004), at Belterra, Cacoal, and the Upper Xingu (E017–E019), and on the middle Solimões by sixteenth-century witnesses (E007). A lighter matrix is indicated by the river-distance gradient (E010), terra-firme negatives (E011), and basin models whose authors concede earth-builders 'may not have occupied all of Amazonia' (AMZ-C029). The strongest opposing considerations are timing and sampling. Several showcase hotspots predate contact (Aquiry AD 100–300, E003; Casarabe ending around 1400, E014), and both positives and negatives are biased samples, leaving the matrix density unmeasured (AMZ-R003). Qualitatively, it is the only pattern consistent with every line in the ledger; quantitatively, it remains open.

A synthesis with no records of its own, graded on its dependencies: earthworks (AMZ-E001 on C004), southern-rim villages (AMZ-E004 on C006), Casarabe (AMZ-E014 on C007), and contact-era systems (AMZ-E017, AMZ-E018, AMZ-E019 on C011). Against them stand the river gradient (AMZ-E010 on C013) and the negative cores (AMZ-E011 on C014). Evidence of both signs converges; the matrix density is inferred, not measured.