A molecular clock for writing systems reveals the quantitative impact of imperial power on cultural evolution
Hiroki Fukui, M.D, Ph.D · 2026
Read A molecular clock for writing systems reveals the quantitative impact of imperial power on cultural evolution by Hiroki Fukui, M.D, Ph.D (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-10957.txt.
Introduction
Fukui compiles a global database of three hundred writing systems, fifty binary structural traits, and two hundred fifty-nine family-tree edges spanning fifty-four centuries. He studies scripts as cultural copies, using tree-building, Bayesian, and neural methods. Scripts show a detectable molecular clock, but political intervention breaks it. Divergence gaps track intervention intensity. He lists thirty major script replacement events. Colonial contact predicts script extinction. The Spanish Empire extinguished the most contacted scripts, six of twelve, followed by Japan with three of nine.
Catalog record
| CREATOR | Hiroki Fukui, M.D, Ph.D |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66491 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.10957
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A molecular clock for writing systems reveals the quantitative impact of imperial power on cultural evolution Hiroki Fukui, M.D, Ph.D Abstract 1 Abstract Writing systems are cultural replicators whose evolution has never been studied quantitatively at global scale. We compile the Global Script Database (GSD): 300 writing and notation systems, 50 binary structural characters, and 259 phylogenetic edges spanning 5,400 years. Applying four methods — phenetics, cladistics, Bayesian inference, and neural network clustering — we find that scripts exhibit a detectable molecular clock. The best-fitting model (Mk+ \Gamma strict clock, incorporating Gamma-distributed rate variation across characters) yields a substitution rate of q = 0.226 substitutions/character/millennium (95% CI: 0.034–1.22; \Delta\text{BIC} = - 4.1 versus relaxed clock; \Delta\text{BIC} = - 1,364.7 versus Mk without rate variation). The shape parameter \alpha=0.51 reveals extreme rate heterogeneity among characters: physical features (directionality, writing medium, stroke construction) change at rates below 0.05/millennium, while social and structural features (pictographic origin, modular composition, user class) exceed 1.0/millennium — a pattern consistent with functional constraint driving clock-like regularity. Political interventions break this clock: deviation from expected divergence times correlates with intervention intensity (Spearman \rho=0.556, p<10^{-4} ), and this association holds when stimulus diffusion events are excluded ( \rho=0.257 , p=5.9\times 10^{-4} ) and when 9 features conceptually linked to political context are removed ( \rho=0.505 , p<0.0001 ). Per-character rate analysis reveals that intervention does not merely accelerate change but selectively rewrites deep structural features. The rate profiles of naturally transmitted and politically mediated scripts are only weakly correlated (Spearman \rho=0.320 ), with the largest divergences concentrated in features defining script architecture — pictographic origin, modular composition, and logographic component — while physical substrate features resist transformation even under political coercion. We identify 30 major script replacement events and rank their destructive impact using a multi-dimensional score validated by sensitivity analysis (Spearman \rho=0.989 across 1,000 weight configurations). A ceiling effect suppresses independent invention wherever writing already exists (Fisher’s exact OR = 0.054, p<10^{-6} ), and colonial contact is a powerful independent predictor of script extinction (Cox HR = 5.25, p = 0.0006). The Spanish Empire extinguished the most scripts (6 of 12 contacted, 50% extinction rate), followed by the Empire of Japan (3 of 9, 33.3%); the Spanish destruction of Maya is qualitatively unique as it eliminated one of only four independent inventions of writing. Feature coding was performed with LLM assistance (Claude, Anthropic) and validated by inter-rater reliability testing on a stratified sample of 40 scripts with two independent human coders blind to script dates and political context (Cohen’s \kappa=0.877 ; human-LLM \kappa=0.929 ; Fleiss’ \kappa=0.911 across all three raters). 2 1. Introduction Writing is the most consequential information technology in human history. It enabled the accumulation of knowledge across generations, the administration of complex states, and the emergence of science, law, and literature as cumulative enterprises. Yet compared to spoken language — whose evolution has been studied quantitatively for decades (Lieberman et al. 2007; Gray and Atkinson 2003; Greenhill et al. 2017; Pagel et al. 2007) — the evolution of writing systems has received remarkably little computational attention. We know far more about how the word went regularized to past tense than about how the letter A rotated from an ox head to its present form. This asymmetry is not accidental. Writing systems present difficulties that spoken languages do not. They are consciously designed, subject to deliberate reform, and can be imposed or abolished by political decree. A language evolves by the accumulated, uncoordinated choices of millions of speakers; a script can be transformed overnight by the signature of a head of state. Atatürk replaced the Arabic script with Latin for Turkish in 1928. The Soviet Union Cyrillicized the writing systems of Central Asia within a single decade. The Spanish Empire destroyed Maya script — one of only four independent inventions of writing in human history — in the course of colonial conquest. These are not gradual evolutionary processes. They are interventions, and any model of script evolution must account for them. The study of writing systems has traditionally been the domain of grammatology and epigraphy, disciplines that produce careful descriptive accounts of individual scripts but rarely attempt cross-system quantitative comparison. The major reference work, Daniels and Bright’s The World’s Writing Systems (1996), is organized as a compendium of individual script descriptions rather than as a comparative framework. Computational approaches have been rare. Hosszú (2024) applied phenetic analysis to four Rovash (Old Hungarian-related) scripts and coined the term "Scriptinformatics," but the dataset of four scripts is too small for phylogenetic inference or statistical testing. We adopt the term "Computational Grammatology" to describe the approach taken here, distinguishing it from Hosszú’s Scriptinformatics in three respects: the scale of the dataset (300 versus 4 systems), the application of phylogenetic and Bayesian methods (versus phenetics alone), and the explicit modeling of political intervention as a variable in script evolution. Phylogenetic methods have been applied to diverse cultural domains — kinship systems, material culture, subsistence strategies — but writing systems have been conspicuously absent from this program (Mace and Holden 2005), largely because no comparative dataset existed. Bentz and Dutkiewicz (2026) used entropy analysis to demonstrate that 40,000-year-old Paleolithic sign sequences have information density comparable to proto-cuneiform — a striking finding that extends the timeline of symbolic notation but does not address the structural evolution of writing systems per se. In biological systematics, the molecular clock hypothesis — the idea that DNA sequences accumulate substitutions at a roughly constant rate — has been one of the most productive tools for dating evolutionary divergences (Zuckerkandl and Pauling 1965). The clock has been applied to cultural data: Lieberman et al. (2007) showed that English irregular verbs regularize at a rate proportional to the inverse square root of their usage frequency, and Gray and Atkinson (2003) used Bayesian phylogenetics to date the divergence of Indo-European languages. But no one has tested whether writing systems exhibit clock-like evolution — whether the structural features of scripts accumulate changes at a predictable rate over time. This paper asks three questions: 1. Do writing systems evolve at a clock-like rate — and if so, what structural properties drive the regularity? 2. Does political intervention break the clock, and does it alter the character of script change as well as its rate? 3. Can we quantify the destructive impact of political power on writing systems? The answers, in brief: writing systems do exhibit a detectable molecular clock (q = 0.226 substitutions/character/millennium under the best-fitting Mk+ \Gamma model; strict clock favored, \Delta\text{BIC} = - 4.1). The clock’s regularity arises from a hierarchy of feature-level change rates — physical and technical features change slowly, constrained by motor habits and material substrates, while social and structural features change rapidly — and the aggregate behavior is well described by a Gamma distribution with shape parameter \alpha=0.51. Political intervention breaks the clock (Spearman \rho=0.556, p<10^{-4} ), not merely by accelerating change, but by selectively altering which features change: the rate profiles of naturally transmitted and politically mediated scripts are only weakly correlated ( \rho=0.320 ), with the largest divergences concentrated in features defining a script’s fundamental architecture. These three questions are addressed in turn in sections 3.2–3.3 (clock and its anatomy), 3.4 (political intervention),