Glottochronology (Lexical Dating)

The canonical Swadesh formula is t = log(C) / (2 log r). Here C is the proportion of basic-vocabulary cognates two related languages still share (between 0 and 1), r is the assumed retention rate — the fraction of core vocabulary kept per millennium, conventionally about 0.86 for the 100-word list — and t is the estimated time since the two languages separated, expressed in millennia. The 2 appears because both daughters have been independently losing vocabulary since the split.

Its central premise is that basic vocabulary is replaced at a single constant rate everywhere and always. Empirically this is not true: replacement rates differ across languages, individual words, and historical situations, and language contact injects loanwords that distort cognate counts. The clock was also calibrated on only a handful of language pairs with documented histories, and the reported dates carry large but usually unstated uncertainty. For these reasons the method is regarded as unreliable for absolute dating.

Yes. Modern computational phylogenetic methods, especially Bayesian inference as in the Indo-European studies by Gray and Atkinson and by Bouckaert and colleagues, infer divergence dates while estimating rates from the data and allowing them to vary across the tree, rather than assuming one fixed Swadesh clock. These methods also quantify uncertainty explicitly. They are not free of controversy, but they address the core methodological flaws that discredited classical glottochronology.