d'Huy's phylogenetic (phylomemetic) reconstruction of myth, and its critics
A body of work applying the algorithms of biological phylogenetics — maximum parsimony, Bayesian tree inference, neighbour-joining — to folktales and myths, treating shared narrative traits as "characters" the way a biologist treats homologous genes. Julien d'Huy is the most prolific proponent; the program overlaps with berezkin-motif-database's areal statistics but goes further, claiming it can reconstruct ancestral states (proto-versions of a tale) and date deep descent. It is the concrete answer the vault needs for Q1 — how one would empirically tell descent from convergence — and its limits are exactly why Q1 stays open. Class 3 by nature: it produces trees and ancestral reconstructions, not dated witnesses.
The method codes each version of a tale as a string of present/absent traits, then builds the tree that requires the fewest independent gains/losses of traits (parsimony) or the most probable tree under an explicit model (Bayesian). For "Polyphemus" and the "Cosmic Hunt", d'Huy reports trees whose geography tracks known human migration routes, and infers Paleolithic antiquity for the root.
The decisive logical move — the same one that licenses the descent/contact/convergence trichotomy this vault runs on: shared structure across lineages that no contact route connects is the signature of common descent or convergence, not borrowing; congruence between a myth tree and an independent tree (genetic, linguistic, geographic) is the evidence that the myth tree is tracking real history rather than noise.
Tehrani's quantitative work on "Little Red Riding Hood" (PLoS ONE 2013) is the cleaner demonstration of the same toolkit: it statistically separated the European and East Asian (Tiger Grandmother) branches, showing the method can distinguish a single descent lineage from independent traditions that look superficially similar.
The standing objections are severe and unresolved, which is why the vault treats every output as class 3. (1) Horizontal transmission breaks the tree model: genes mostly descend vertically, but myths are borrowed sideways constantly, and a borrowed trait masquerades as inherited — exactly the descent-vs-contact confusion the method is supposed to resolve. (2) Character coding is the analyst's choice: what counts as "the same motif" (lumping/splitting) determines the tree, and there is no agreed protocol — the same critique that hits Berezkin. (3) Rooting and dating are model-dependent and uncalibrated: there is no mythological equivalent of a fossil or a sound law to anchor the molecular clock, so "Paleolithic" roots are extrapolations, not measurements. Ross, Greenhill & Atkinson (2013) showed that for European folktales, geography and language both structure the data — i.e. areal diffusion and descent are entangled in the real signal. The method is genuinely useful for generating falsifiable hypotheses and for the negative result (showing two traditions are not a single lineage); it does not deliver dated descent on its own.
distinguishing-descent-contact-convergence-methods, great-flood, new-world-flood-convergence-test