Skip to content
Induscript

RC-002: Do particular signs show statistically significant terminal-position bias?

Research compilation

Research compilation from a published ICIT-style digitization — not a complete or licensed photographic corpus

Structural finding

Terminal bias is measured as P(end)/P(any) with a minimum-occurrence threshold. Labels stay “ender-biased”, never “suffix” as fact.

Background

  • This case is a demonstration of the platform's research-case format.
  • It uses the seeded demonstration dataset, not a licensed photographic corpus.

Dataset

  • Demonstration dataset version demo-2026.09.
  • Inclusion: complete and partial inscriptions as generated; duplicate groups are flagged.
  • Exclusion: no claimed museum objects; no unlicensed images.

Method

  • Define the pattern in analytical order.
  • Count raw, independent-object and duplicate-reduced frequencies.
  • Compare with a length-preserving shuffle null model.
  • Report limitations and interpretation boundaries in public view.

Results

  • See the live analysis modules for counts derived from the current demo seed.
  • Directional retention and PMI are descriptive until a verified import is locked.

Negative controls

  • A motif-only explanation is not supported as a translation.
  • The demo generator itself planted some structure; that is disclosed.

Limitations

  • Demo data cannot support historical conclusions.
  • Sample sizes for smaller sites are too small for percentage claims.

Interpretation boundary

Do not read this case as a decipherment or as a statement about a real excavated pair until verified plates replace the demo.

Induscript, RC-002, 18 September 2026. Demonstration case.