Glottochronology: Vocabulary Divergence Tree

Vocabulary Drift

Word replacement rate r 8% / gen

Simulation speed 4 gen/s

Live Readouts

Generation

Leaf languages

Cognate retention — tracked pair

—

Est. gens since split

—

Actual gens since split

—

Click Split language to divide the proto-language into two daughter branches. The most recently split pair is tracked automatically.

How it works

Every branch keeps a 24-slot core-vocabulary list (a stand-in for a Swadesh list). Each generation, every word independently has probability r of being replaced by a brand-new, unrelated form — this is the "constant rate of lexical replacement" that Morris Swadesh proposed in the 1950s. A word survives unchanged with probability (1 − r) per generation.

When you split a branch, both daughters inherit an identical copy of its current vocabulary, then mutate independently. Because a replaced word can never coincidentally match its sibling's replacement, the fraction of matching ("cognate") slots decays smoothly:

C(t) = (1 − r)^(2t)

C(t) = fraction of vocabulary still cognate after t generations
(1 − r) = retention rate per generation per branch
the exponent is 2t because BOTH branches are independently losing words

Solving for t gives the classic glottochronology dating formula, used here to recover elapsed generations purely from the measured retention percentage — no knowledge of the actual split time required:

t_est = ln( C_measured ) / ( 2 · ln(1 − r) )

Word replacement rate — raise it to watch languages diverge (and the tree's colour patches scramble) faster.
Split language — branches every current leaf into two; the newest split becomes the tracked pair for the retention/estimate readouts.
Est. vs. actual gens since split — with a known, constant r the formula recovers the true elapsed time almost exactly; real glottochronology is far noisier because r is never truly constant across languages or word classes — this is the idealised textbook case.

Real-world relevance: this constant-rate assumption is exactly what underlies (and what critics of) dating language splits like Proto-Indo-European's breakup from measured cognate percentages between attested daughter languages.

Tracked pair (retention readout) Other leaf branches

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.