Spec 027 — Movement, Culture, and Language in the Greater Indus World: Frozen Design and Pre-Analysis Plan (owner-adjudicated freeze, 2026-10-11)

Abstract

Frozen study design, owner-adjudicated and frozen 2026-10-11 (repository: specs/027-movement-culture-language/, freeze commit 92711e9d; adjudication record adjudication.md in the same directory). An OSF Registration of this frozen design is filed from this component as its pre-analysis plan. Question: how do population movement, mobility, material culture, and language relate in the greater Indus world before, during, and after the Harappan period — and can sign-repertoire variation be tested against that record without smuggling the answer into the design? Design (adjudication answers, verbatim effect): Stage order 0A — Stage 0 builds a source-asserted settlement gazetteer AND a chronology spine together (descriptive infrastructure only; every row carries its source, licence state, and coordinate-precision/restriction state; the contested Mutin et al. 2025 Mehrgarh redating stays labelled CONTESTED). Stage 1 is the Q2 gate test: does site explain recorded sign-repertoire variation after rarefaction to common sample size and object-type stratification, under a pre-declared permutation null, with discriminating controls and leave-one-site-out jackknife? M77 (Mahadevan) numbering governs the primary test; Parpola/CISI appears only as a separately labelled crosswalk sensitivity. Corpus rules as adjudicated: provenanced records only; pottery graffiti NOT pooled with seals/tablets; damaged/multi-line objects only under the frozen minimum legible-token rule; deduplication under the program's dedup artifact; unmapped signs retained as UNMAPPED under one frozen pooling rule. DESIGNED NEGATIVE CLOSURE: if a valid Stage 1 test does not reject the null, the bounded claim is CONTRADICTED and all downstream repertoire-geography work under Spec 027 stops — that is a successful outcome. Stage 2 (Q7 material networks, Q6 standardisation timing) is conditional and separately frozen. Registered-but-blocked questions, with blockers named in the spec: Q1 mobility arm (isotope/aDNA site coverage ≤3 sites; ecological-fallacy fence) — NOT-YET-TESTABLE as designed; Q3 Cemetery H/PGW paired data — not in hand; Q4 Late Harappan transect — linked dataset not assembled; Q5 Mehrgarh chronology replication — no primary dataset; Q8 substrate stratigraphy — requires at least two human etymologists coding blind (AI coding is one origin group and is not a substitute); Q9 Brahui discriminator — may be not-yet-testable, declared in advance. Hard limits: no gene-to-language identification; no ethnic labels projected backward; no object-to-mass-migration inference; no later text as a Harappan transcript; mobility/context evidence never enters decipherment (PRED) scoring; no package deals; no teleology joins.