Code and results for the paper:
"Ecological Override of Lexical Frequency: Sea as a Coordinate-System Origin Stabilises Proto-form Retention in Austronesian"
Submitted to Journal of Language Evolution (2026).
scripts/ Analysis scripts (Python)
results/ Output JSON files from each analysis
figures/ Figures included in the paper
| File | Purpose | Paper Section |
|---|---|---|
scripts/h1c_ab_coding_analysis.py |
H1c two-feature F_AB coding; Fisher exact tests | §3.3, Table 1, Fig. 3 |
scripts/plan_a_continuous_ecological_gradient.py |
AIC model comparison M0–M4 for SEA retention continuous gradient | §3.2, Table 4 |
scripts/plan_b_h1c_power_expansion.py |
Bootstrap power analysis for H1c at OR=4.50 | §3.3, Fig. 5 |
scripts/plan_c_left_right_partial_correlation.py |
Partial Spearman correlations — SEA/LEFT gradient orthogonality | §4.2, Fig. 4 |
scripts/generate_figures_p4_p5.py |
Generate Fig. 4 (gradient orthogonality) and Fig. 5 (power curve) | Fig. 4, Fig. 5 |
| Analysis | Metric | Value |
|---|---|---|
| H1c F_AB (n=25) | OR | 16.5, p=0.005 |
| H1c F_AB (n=33) | OR | 14.08, p=0.002 |
| Plan A: M2 vs M1 | ΔAIC | −30.95 (M2 AIC=149.96) |
| Plan B: power at n=80 | Bootstrap power | 87.7% |
| Plan C: SEA partial r | Spearman r controlling LEFT | +0.459, p<0.001 |
| Plan C: LEFT partial r | Spearman r controlling SEA | +0.131, p=0.003 (−48% drop) |
The analyses use:
- ABVD Lexibank (Greenhill et al. 2008; Blum et al. 2025) — 903 Austronesian languages, 15 spatial concepts
- Glottolog 5.0 (List et al. 2024) — 517-language phylogenetic subset for D statistic
Raw data files are not redistributed here; download from ABVD and Lexibank.
python >= 3.9
scipy
numpy
pandas
matplotlib
statsmodels
MIT License. See LICENSE for details.