If only my results were this pretty

This is a midpoint-rooted phoneme tree produced using a Bayesian approach, from a paper whose sole purpose is to debunk a recent-ish high profile paper that claimed a language's phoneme inventory is correlated with how far away it is from Africa, thus providing support for the out-of-Africa story of human origins.

What this picture shows: something to do with phonemic relatedness and geographic clustering and how they don't correlate particularly well. I'm not really sure, to be honest, I've never used that technique or read a paper that used it before, so my being a linguist isn't particularly helpful here. Although I will assume that the different colours correspond to different language families, since that's pretty standard.



Source: Language Log, which also goes into lots more detail about the paper and has lots of other colourful graphs, although none as cool as this one.