Tracing the roots of syntax with Bayesian phylogenetics

In 2014, Tom Griffiths and I published a paper entitled "Tracing the roots of syntax with Bayesian phylogenetics" in PNAS. The paper uses Bayesian phylogenetic modelling to investigate the relative stability of different basic word orders and the preferred directions of change between them. This page serves as a space for errata and general discourse surrounding this paper. If I receive or anticipate frequently asked questions about the paper, or see frequent misconceptions about the paper "in the wild", I will attempt to address them fairly and honestly here. On this page, I speak for myself; not Tom, not the University of California, where I wrote the paper, and not the University of Auckland, where I now do similar work.

Computational phylogenetics and historical linguistics have had a short, rocky relationship thus far, to say the least. A lot of previous work in this field has been heavily criticised. From what I have seen, there has been some valid concern about potential shortcomings of early work, as well as gross misunderstandings of the same work, and an apparent aversion to "non-linguistic" methodologies in general.

I believe that the future of historical linguistics is computational. I believe that "traditional" linguists and the computational newcomers (many of whom, myself included, were trained in other disciplines) both have valuable insights which will help to realise this future, and that each should strive to learn what they can from the other. I believe that the current level of vitriol surrounding this new methodology is regrettable, and I hope to see more productive engagements between the two camps in the future. I have set up this page in this spirit.

Frequently asked or anticipated questions

Q: Some of the text in your figures is tiny and I can't read it clearly, can I see full size figures?

A: Yes! Some of the figures really are quite bad. Here are full size versions as PNGs: