« Genitives in the Future | Main | The Unification of Genitives! »



Glen Gordon

This is a big topic, definitely. I read your article but have been fluttering about with busy offline stuff. I'll have to quickly message you right now and I'll have to try to brief... I'll try!! But this made me recall Duane's post on Abnormally Interests on the topic of Building a Semitic Tree where he says that he was "thinking of attempting to construct a phylogenetic tree for a set of Semitic languages using Markov chain Monte Carlo methods." Eeeek!

I have a programming background but my real passion is linguistics. So knowing both subjects reasonably well, I can see the problems that develop when both the CompSci and Ling departments clash and threaten to produce evil mutant babies together based on mutual misunderstandings of either field. A linguist will typically deify computer programming as the saviour of modern academia whereas the computer programmer will typically misunderstand human language to be something that can be easily reduced to ones and zeros like Cobol. (If you remember Cobol, you're too damn old and you also need to get out more, hehehe.)

So anyways, this mad phylogenetic game is tricky because of course language doesn't nicely branch off into nice splits. Unlike animals, languages have no problem merging given the right conditions. There is no linguistic version of an "interspecies barrier" so while tigers and baboons obviously can't produce offspring, pidgins can arise between any group of languages.

Well, that's all I can think up for now.

Ivan Štambuk

Starostin's StarLing program can generate phylogenetic trees for selective languages with Swadesh list input. It already comes preloaded with dozens of lists of most major families, and the end result is not too bad either. It can also generate various simple phoneme correspondences (like Grimm's law) and date divergences between languages/families. Unfortunately, it doesn't come preloaded with Semitic database.

Like Glen says, the line between biological species is discrete (i.e. you cannot cross it once the mismatch between DNA surpasses certain threshold), while natural languages (which are, even today, just like formal languages, defined in terms of their lexis)  have no trouble acquiring vocabulary/morphology en masse from neighbours, due to intensive cultural contact. The only modern correspondence would be analogical to implanting a fish gene into a tomato (gee, that already happened ^_^).

NB, the biological species barrier operating on DNA mismatch level has been discovered and broken recently - they managed to crossbreed Escherichia coli and some Salmonella which had common ancestor 150 million years ago, by suppressing a certain enzyme or something.


Starostin's method doesn't appeal to me. Exactly because of the major problems with equating language trees to phylogenetic trees as used in biology it seems near futile to use this. Besides that it seems pointless, because one can only use Swadesh lists for such methods with words that are known cognates, rather than eyeballed 'cognates'. And once you've already established known cognates, chances are you're not going to have to make phylogenetic trees since it's pretty clear already.

Glen Gordon

PhoeniX: "And once you've already established known cognates, chances are you're not going to have to make phylogenetic trees since it's pretty clear already."

Yes, exactly. It's too bad that these linguist-programmer teams can't just write a good program and run it with reasonable results. Instead they have to "tweak the data" to make it get whatever they want (aka "Pay no attention to that man behind the curtain."). By the way, now I'm interested in what a fish-tomato tastes like.

The important thing I guess is that their names get easy publicity even if they don't deserve it. Hooray for free market capitalism in academia! Note in that link: "One could argue that the increased emphasis on careerism by academics, using the currency of degrees and publications, reflects the influence of the commodity form generally, as well as the increased role of direct government funding of individual academics. Rather than adopting an intellectual commitment to the legendary community of scholars, most academics think and act in terms of an individual career." So in other words, it's not how logical your claims are but rather how marketed your name is. Get out there, scholars, and sell, sell, sell! :)

The comments to this entry are closed.