Published October 28, 2026 | Version v2
Preprint Open

Distinct Morphological Topologies in Ritual Masks: Evidence for Geographically Bounded Encoding of Non- Human Craniofacial Traits

Authors/Creators

Description

Updated Paper 2026-10-27

This study presents a computational morphological analysis of N=239 ritual masks and carvings from six cultural groups: Seneca False Face (North America), Māori wheku (New Zealand), Pacific Northwest Coast (Tlingit/Haida), Mesoamerican, European, and Japanese (Noh/Kyogen).

Using a Vision-Language Model (Qwen-VL) to extract standardized morphological descriptors, followed by rigorous manual verification to correct automated false positives (e.g., distinguishing traditional chin markings from protruding tongues), we performed UMAP dimensionality reduction and per-corpus trait prevalence analysis.

The results reveal two distinct morphological topologies rather than a single, universal "monster" archetype:

  1. The Trans-Beringian/Coastal Complex: The Pacific Northwest Coast and Mesoamerican corpora form a shared, distinct cluster, characterized by elevated rates of nasofrontal depression and anomalous eye morphology.

  2. The Eurasian/Isolation Gradient: The Seneca and Māori corpora form a separate morphological bridge, characterized by elevated rates of extreme oral anomalies (macroglossia and the "bow-tie" lip shape). The Seneca masks act as a bridge between standard European morphology and the distinct Māori cluster.

These findings indicate that the encoding of non-human or archaic craniofacial morphology in ritual objects is not a universal human cognitive default, but follows distinct, geographically bounded pathways. This provides preliminary material-culture evidence consistent with the Marginal Interactor Model, suggesting that different populations encoded memories of physically distinct hominin competitors based on their specific ecological and historical contexts.

Files included in this deposit:

  • The full manuscript (PDF/Markdown).

  • The corrected, per-corpus trait prevalence data (trait_prevalence_MANUAL.csv).

  • The high-resolution UMAP clustering visualization.

  • The Python scripts used for trait extraction and dimensionality reduction.

  • The Qwen-VL extraction prompt

 
Original Description- 
 
This paper presents an exploratory computational analysis of N=150 ritual masks and carvings to test the "Myth as Compressed Protocol" hypothesis. As part of the broader Hominin Mythology Project, this study investigates whether the material culture of populations with plausible archaic hominin overlap (Seneca False Face and Māori whēku) preserves congruent, non-normative craniofacial signatures compared to comparison groups (European and Japanese mask traditions).
 
Using the Qwen-VL vision-language model, nine standardized morphological descriptors were extracted from the dataset. TF-IDF vectorization and UMAP dimensionality reduction were then applied to visualize semantic clustering and calculate automated trait prevalence.
 
Key Findings:
  • The Archaic Signature: Experimental groups encode specific archaic traits at descriptively higher rates than comparison groups, most notably macroglossia (16× higher prevalence), prognathism (3.5×), and nasofrontal depression (3×).
  • Biomechanical Anomalies: Manual verification confirmed the presence of the "bow-tie" lip morphology (extreme lateral eversion with midline constriction). This is interpreted not as a resting face, but as a frozen articulatory posture linked to a divergent hominin vocal tract and socially salient "signature sounds."
  • Independent Encoding: UMAP topology reveals a "bridge" effect, suggesting that Seneca and Māori carvers were independently encoding a shared category of deviation from the Homo sapiens norm, rather than copying a single diffused template.
 
These findings provide preliminary, independent material-culture evidence that triangulates with the project's textual "Survival Manual" motifs, suggesting that global folklore and ritual objects may preserve fragmented memories of physically distinct hominin competitors.

Files

EXPANDED_STUDY_MANUAL_CORRECTION_UMAP_stretched.png.png

Files (1.3 MB)

Name Size
md5:aad15953f93c467f48d0f80d90c7df82
3.6 kB Download
md5:ef962c214c5c8d791f4d3708d9470a49
324.6 kB Preview Download
md5:a59c8f2e0c0e82f0bcba00f819a4eb8d
655.7 kB Preview Download
md5:1ef2da6dceedac6aa71141171204b2bb
271.7 kB Preview Download
md5:7725105e5242e70c7bb9187e1b37b82c
5.6 kB Download
md5:94726238df72f249a0bf56828c688353
2.8 kB Preview Download
md5:5ab806468d92ff5ec4ee5711c31becc5
2.0 kB Preview Download