The trouble with “lexomics” is, as some of the commenters on the Nature article pointed out,is that the process is Lamarckian, not Darwinian; it’s driven, not followed. If I was a lithping king, I could make all my thubjects lithp without too much trouble.

Sure, but this is entirely irrelevant. The methods we’re using (phylogenetics) are top-down views of evolution and just reconstruct the history. They’re relying on a system that evolves through “descent with modification” and not necessarily “survival of the fittest”. We discuss this in more detail in a paper called The Pleasures and Perils of Darwinizing Culture.

The same goes for fossils – phylogenetic methods absolutely don’t need fossils. If we have them, then we can incorporate them. However, modern languages carry around remnants of their ancestors in them. That is, two languages that are closely related should share a lot of features with their ancestor. One of the benefits of phylogenetics is that we can infer ancestral states using probabilistic models to estimate the most likely ancestral variant. In fact the Bayesian likelihood methods we’re using deliberately sum across all possible ancestral state assignments at every node on every tree. Yes, it is just an estimate, but it’s the closest we can get without a time machine. I’m working on a paper explaining this in more detail, but it’s quite a way off yet, but Quentin and Russell’s paper here might help explain a bit.

Hope this helps!
–Simon

The other major problem is that there aren’t any fossils*. All the ancestors are hypothetical proto-languages. If you take all the most common characteristics of an existing clade (or as many as you can find) and distill them down to the lowest common denominators, you’ll end up with a ‘proto-language’.

But you can’t be at all sure that major characteristics of the original ancestral language have not been entirely lost, or preserved in only a minority of the existing remnant languages. (Which you ignored, just because they were a minority).

Then, to trace the ‘descendents’ from this hypothetical language is absurd.

Even then, though, I hope some of the newer generation of linguists (you, Simon? – please)can use the mechanical/statistical techniques used by geneticists to resolve some major ‘language family tree’ problems, like the star-like pattern of supposed descendents from proto-Austronesian and proto-Oceanic.

regards

Richard
*Except where we have surviving scripts. But it was pointed out a long time ago that if the Comparative Method was used, retrospectively, on the Romance languages, the resulting proto-language would NOT be Latin.