Schedule
The summer school runs 19–24 July 2026 on Naxos, Greece. This course meets daily from Monday 20 to Friday 24 July, 14:10–14:55 (see the full summer school programme).
Programme
| Date | Session | Topic | Materials |
|---|---|---|---|
| Mon 20 July | Day 1 | Introduction: the comparative method and the phylogenetic turn | slides |
| Tue 21 July | Day 2 | Data: Swadesh lists, Lexibank, cognate detection, PMI alignment | slides · Lab 1 |
| Wed 22 July | Day 3 | Probabilistic models of language change: CTMCs, substitution models, rate variation | slides |
| Thu 23 July | Day 4 | Bayesian phylogenetic inference: MCMC, MrBayes, priors, convergence, divergence dating | slides · Lab 2 |
| Fri 24 July | Day 5 | Reading two methods sections: Heggarty et al. (2023) on the Indo-European homeland, and Blasi et al. (2019) on labiodentals | slides |
Slides go up on the morning of each session. The labs are available now and stay available.
Hands-on labs
All three labs run inside your web browser. No Python, no R, no MrBayes, no account, and nothing you write leaves your machine. Open a lab, edit a cell, press Run Code.
Everything runs client-side, so once a lab’s setup cell has printed ready it keeps working even if the WiFi drops.
The labs tell one story, in order:
- Lab 1 — Cognate detection (LingPy): where the character matrix comes from, and why automatic cognate detection misses a third of the real cognates.
- Lab 2 — Bayesian phylogenetics (MrBayes): what you build on top of that matrix, and how to decide whether to believe it. Includes a demonstration that MrBayes’s default branch-length prior fails on this data. They are yours to keep. They will still run, with no setup, on any machine with a browser.
A third lab is also online, Lab 3 — Beyond cognacy: building a tree without any cognate judgements at all, the method behind the 3,397-language world tree. It covers material that did not make it into the course in the end, so it is not part of the programme and nothing else depends on it. Have a look if you are curious.
Software
None required. But the code in the labs is not a browser-only dialect — it is ordinary Python (lingpy) and ordinary R (ape, treeio, ggtree), so you can copy any cell straight into a script or RStudio at home and it will run unchanged. MrBayes itself is a compiled program and is the one thing the browser cannot run; its output is precomputed for you, and if you want to run your own chains afterwards it lives here.
Please do not spend the session installing anything.