Assessment result
The recovered evidence supports local path and trace-order analysis, but it does not yet support a preservation-adjusted change between TH2 and TH3. The cross-paper surface-label ambiguity remains unresolved; it blocks a dated cross-horizon comparison, while local geometry can still be studied. This is a completed feasibility assessment, not a statistical detection of changed movement.
Review date: October 28, 2026. Scope: White Sands Locality 2, principally the trench-adjacent Area B. Source labels are retained exactly by publication; the 2021 archive, 2025 behavioral paper and 2026 compilation are not treated as three independent site samples.
Current label: LF-5 — horizon assignment unresolved.1 The focused field/sample check is complete for the accessible public record. The issue is parked pending new primary evidence, not scientifically resolved. Local geometry remains usable; no 2026 horizon or model age is assigned to LF-5. The next active research target is the shared model T038S.
1. The same photographic base has different horizon labels
Step 1 completed: Read the photographic registration audit. Numerical alignment and selected landmark checks strongly support a shared photographic base. The figures carry the following labels:
| Location in the photographed exposure | Bennett et al. 2025, Fig. 13A, p. 14 | Springer et al. 2026, supplement Fig. S12, p. 28 |
|---|---|---|
| Central patch containing the prominent groove | TH3, with groove LF-5 | TH2; groove described in caption without an LF identifier |
| Patch immediately above it in the image | TH4 | TH3 |
| Strip along the right edge | TH5 | TH5 |
A similarity transform fitted outside the central patch aligns all 1,281 withheld image-feature matches within two target pixels (median 0.297 pixels). Eight selected landmark patches peak at zero additional displacement; a separate fit restricted to the unannotated lower-right photograph gives a consistent result. These are correlated image details, not counts of independent footprints. The result establishes photographic correspondence; it does not explain whether the differing labels reflect a revised assignment or a figure error. The two publications remain separately attributable, and no author-confirmed correction has been obtained. [S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025)]
This matters directly to the proposed test. We cannot simply take LF-5 from the 2025 “TH3” figure, give it the 2026 TH3 age, and compare its direction with the 2026 “TH2” figure: those figures use the same photographic base and show the same central patch. Such a comparison could turn a labeling difference into a false behavioral change. Neither the raw counts nor the existing ages have been changed. Review issue WS-SURFACE-001 records the unresolved stratigraphic assignment, with the photographic match now documented as WS-PHOTO-001.
Step 2 completed: Read the stratigraphic audit. Fig. 13A’s section marker connects the matched LF-5 patch to the detailed Fig. 13B profile, supporting a source-specific TH3 interpretation above TH2. However, the 2025 and 2026 overview plans appear to assign the corresponding north-side polygon TH2, exposing an apparent internal plan/detail conflict already in 2025. The 2026 sampled profiles do not provide a registered LF-5 connection. TH3 is retained as the 2025 source assignment; our current cross-publication label is “horizon assignment unresolved.” [S1 · Bennett et al. (2021); S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025)]
Step 3 completed: Read the scope and consequence audit. The review includes TH4 because the neighboring photographed patch also changes labels. Published counts and ages remain source records, while unverified patch-to-inventory and patch-to-age joins are blocked. An unqualified Fig. S12 geometry-to-TH2 link in our horizon comparison has been corrected; numerical source values are unchanged.
2. What the original 3D archive contributes
The public BORDaR archive was downloaded and its original track-key workbook inspected. It supplies 13 individually keyed TH2/TH3 prints. The two TH2 entries point to one shared model, also used for TH1 records. Eight of the eleven TH3 entries link to seven distinct models; three TH3 entries have no model reference. The archive is a selected close-up collection, not a complete excavation map. [S320 · Bennett, M. (2021)]
| Source track ID | Source trackway ID | Heel–first-toe length (mm) | Model-key reference |
|---|---|---|---|
| TH2-1 | Unreported | 187.84 | T066 |
| TH2-2 | Unreported | 218.33 | T066 |
| TH3-1 | Unreported | 258.31 | T065S |
| TH3-2 | Unreported | 231.73 | Unreported |
| TH3-3 | TW3-1B | 204.06 | T027S |
| TH3-4 | TW3-1A | 263.39 | Unreported |
| TH3-5 | TW3-1A | 264.62 | T089S |
| TH3-6 | TW3-1A | 263.70 | T038S |
| TH3-7 | TW3-1B | 187.46 | T038S |
| TH3-8 | Unreported | 167.90 | Unreported |
| TH3-9 | Unreported | 286.20 | T070S |
| TH3-10 | TW3-1A | 268.50 | T032S |
| TH3-11 | TW3-1B | 194.08 | T034S |
A useful positive finding: the source assigns four prints to TW3-1A and three to TW3-1B. Their recorded lengths occupy different ranges: 263.39–268.50 mm and 187.46–204.06 mm, respectively. Model T038S is shared by TH3-6 and TH3-7, one from each named trackway. That gives a concrete place to investigate local path relationships. The assignments and measurements are the source authors’ records; we have not independently identified those trackmakers, inferred their ages, or established that they walked together. [S320 · Bennett, M. (2021)]
The seven ASC files were parsed and the PLY header inspected. An internal height rendering of T038S confirmed that the data can be examined as a local surface. However, the inspected files/key supply no verified transform to a common site map or geographic north. The 2025 methods also explicitly describe scaling and rotating individual models to a principal plane. Native model X/Y coordinates therefore cannot be pooled as geographic directions without registration. [S6 · Bennett et al. (2025); S320 · Bennett, M. (2021)]
Some ASC grids contain nonfinite Z values. These may describe crop/background gaps rather than damaged track-bearing sediment, so their fraction is not a preservation score. The inventory preserves model sharing, hashes, headers and finite-row counts; it does not label missing model references as missing ancient activity. A reproducible, read-only parser is included in scripts/audit-white-sands-models.py. The original models and workbook are not redistributed.
3. What can be said about movement now
These observations use the 2025 paper’s own feature and horizon labels until the crosswalk is resolved.
| Feature / relationship | Published evidence | Supported inference | Limit |
|---|---|---|---|
| Area B LF-1–LF-4, labeled TH2 | A group of 1–2.5 m features, preserved in dolomite positive relief and truncated by erosion or covering deposits | Local linear trajectories survive | Their surviving lengths are not complete trip distances; relief and excavation affect visibility |
| LF-3, labeled TH2 | Human prints and a groove have locally complex overprinting/cross-cutting relations | More than a simple undisturbed association needs to be considered | A single unambiguous event sequence has not been reconstructed here |
| LF-19, discussed with the TH2 Area B features | The authors report the groove cutting at least four human prints longitudinally | Those identified prints precede that cutting trace | It does not establish one person, one trip, load, elapsed time, or population |
| LF-5 — horizon assignment unresolved1 | A bifurcating groove, underlying sediment deformation, and footprints with similar orientations on both sides | Local alignment of walking and a dragging-like trace is documented | It is a selected example, and its transfer to a 2026 horizon is blocked by WS-SURFACE-001 |
| TH3-keyed TW3-1A and TW3-1B in the 2021 archive | Named trackway assignments and a shared close-up model for one print from each | A tractable candidate for comparing two local paths | Shared space does not establish simultaneous travel, following, kinship, or event order |
The 2025 authors favor human-drawn transport for at least some linear features. This assessment records their interpretation separately from observed alignment, morphology and cross-cutting. A groove bend need not measure a person’s turn; foot orientation need not equal walking direction. Figures 12C/E also caution against assuming that all illustrated prints belong to one individual. [S6 · Bennett et al. (2025)]
Turning and stopping: no matched, per-trackway turning distributions were recovered for the two dated horizons. Human–animal overlap: taxa and isolated overprints are reported, but no complete paired spatial inventory, taxon-specific detection estimate, or comparable observed polygon was recovered. Neither question currently supports a cross-horizon trend. [S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025)]
4. The density difference is modest and remains unadjusted
Using only the 2026 trench-adjacent counting windows:
| 2026 unit | Human prints | Listed area (m²) | Calculated prints/m² |
|---|---|---|---|
| TH2 | 45 | 3.82 | 11.78 |
| TH3 | 19 | 1.35 | 14.07 |
TH3 has fewer recorded prints, but its calculated density is 19.5% higher. This is arithmetic from S5 Table S1, not an estimate of changed traffic. In a deliberately simplified equal-production scenario, a factor of 1.195 in the combined duration of track recording, preservation and detection would reproduce the same density contrast. Those factors have not been measured; the calculation identifies the scale of a possible confounder without claiming that such a bias has been demonstrated.
The 2025 study reports different cementation/excavation conditions on its TH2 and TH3 surfaces, so a correction cannot be assumed negligible. The nominal 2026 horizon ages, 22.66 ± 0.10 and 22.61 ± 0.09 cal ka BP (2σ), are model outputs. Their 50-year midpoint difference is not an observed duration between walking events, and uncertainties from the shared age model should not be treated as independent. [S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025)]
Count definitions also differ across releases: the 2021 key lists 2 TH2 and 11 TH3 prints; the 2025 Fig. 9B TH3 size sample has 61; the 2026 trench-adjacent counts are 45 and 19. No matching print-level inventory or identical-area definition was recovered. These numbers must not be summed or read as a rise/fall in ancient population. [S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025); S320 · Bennett, M. (2021)]
5. Research directions advanced by this assessment
First, test local route reuse within a documented surface. Start with source-keyed prints TH3-6 and TH3-7 in their shared model T038S and verify their local identifications and any physical overlap. This first inspection does not require a calendar age or geographic north. Register the additional models and source IDs to the source orthomosaic before testing a multi-print route or geographic direction. Distinguish following an existing route from walking together. Nearby independent paths, mud conditions and the excavation window remain alternatives. This question can make progress without pretending to resolve decade-scale climate effects.
Second, measure how documentation changes the observed record. The 2025 paper includes repeated documentation of Area A before and after removal of infill. A comparison of exactly the same polygon could estimate which traces become identifiable after cleaning. This would directly address detection bias. Different photographed bounds, further excavation and intervening erosion must be tracked, so the result would not automatically be a pure preservation experiment. Area A’s relationship to the trench must remain source-qualified. [S6 · Bennett et al. (2025)]
Third, use sediment deformation to distinguish trace formation from later preservation. Compare excavatable grooves with dolomite casts after feature identities are reconciled. Cross-sections may help separate the original action from later cementation or erosion. Different dragging devices or loads remain alternatives; depth alone cannot distinguish them. [S6 · Bennett et al. (2025)]
For a later dated cross-horizon comparison, the missing evidence remains specific: (1) a figure/feature/horizon crosswalk resolving the 2025–2026 labels; (2) original surface maps and model-registration transforms with scale and north; (3) trace IDs and uncertainty in connections/overprints; and (4) matched masks for excavated, unexcavated, eroded and ambiguous areas. These would enable trackway-level tests and sensitivity to observation windows. No author or institution was contacted during this review.
Final field/sample check and disposition
Result: inconclusive; further label investigation is deferred. The reference seed group G05 comprises USGS-1808.1, USGS-1808, USGS-1830 and USGS-1810, from field samples 20WHSA1-18.2a and 20WHSA1-18.3a in Table S4. Those names make the missing connection precise, but neither the inspected section nor the sample table connects LF-5’s original formation surface to that level through verified field contacts. The different deposition/Sequence model treatments remain recorded in the scope audit. Deformation extending down beneath a groove does not locate its formation surface. [S5 · Springer et al. (October 14, 2026); S6 · Bennett et al. (2025)]
| Check completed October 28, 2026 | Finding and access limit |
|---|---|
| Published LF-5 section, sampled columns and seed/model records | No field-supported LF-5-to-G05 connection recovered; the original source labels remain conflicting. |
| 2026 data release | The article describes downloadable Tables S1–S15; relevant sample, count and age tables were examined in its supplement. USGS/DataCite metadata were read, but the separate ScienceBase files returned access errors and were not inspected. Their contents cannot be ruled out as a source of further evidence. |
| 2025 data release | USGS/DataCite descriptions identify modern travois experiments. The release files were not inspected; the description provides no ancient feature/sample crosswalk. |
| Correction search and Crossref records for both article DOIs | No relevant correction located. Empty article relation fields and an absent update-to field do not prove no correction exists. No author-confirmed clarification has been obtained. |
Scope remains undetermined: we cannot establish that this affects only figure annotations, nor that inventory or sample assignments actually changed. Published measurements, source IDs, counts and age outputs remain intact; feature-to-inventory and feature-to-age joins stay blocked. The neighboring TH4 label remains flagged too.
Reopen only for new primary evidence: a surveyed section or field/sample crosswalk, or an explicit author/publisher correction, identifying both photographed patches and stating which annotations, inventory memberships or sample relationships change. Repeating the existing figure comparison is not the next research priority. No author or institution was contacted.
Sources and verification
- S5: Springer et al. 2026, especially supplement Table S1, Fig. S1, and Fig. S12 (PDF p. 28). The original supplement was retrieved through Europe PMC.
- S6: Bennett et al. 2025, full published PDF from the Bournemouth repository, methods, section 3.2 and Figs. 9, 11–13. Figs. 11–13 were visually checked. The associated USGS release describes modern travois experiments; it has not been established as the missing ancient spatial inventory.
- S320: Bennett 2021, 3D Track Data, original archive and workbook key, sheet “TH1, TH2, TH3, TH5 + TH6”, rows 10–22. All eight referenced model files were inspected for structure; the seven ASC files were parsed. Source trackway assignments were not independently reclassified.
- S1: Bennett et al. 2021, author manuscript accessed. The original Science supplementary PDF remained inaccessible; speed estimates found in secondary copies were not imported.
Canonical records are th2-th3-assessment.json, photographic-registration.json, stratigraphic-audit.json, label-scope-audit.json and track-model-inventory.json in research/data/white-sands. Source labels, absent fields, and review issues remain explicit. This assessment strengthens the local evidence trail and defines the missing comparison inputs; it does not establish a population trend or climate-caused behavioral change.
Footnotes
-
Bennett et al. (2025), section 3.2 and Fig. 13, assign LF-5 to TH3; Springer et al. (2026), Fig. S12, labels the same photographed patch TH2. The detailed section supports the 2025 description, but the inspected record does not connect LF-5’s formation surface to the dated seed level through verified field contacts. No clarification was located establishing whether the discrepancy is confined to annotations or also affects inventory/sample assignments. The canonical horizon and feature age remain unassigned. Review checked October 28, 2026; see WS-SURFACE-001 and the final field/sample check. ↩ ↩2