28 Jan 2010

How to make a mockery of Proto-Japanese

"Nikolayev and Starostin 1994 offer many putative cognate sets and reconstructions for Nakh-Daghestanian and its branches, as part of a reconstruction of a putative North Caucasian comprising Northwest Caucasian and Nakh-Daghestanian. These two families have not in fact been shown to be related. Nikolayev and Starostin proceed by assuming relatedness and then assembling cognate sets so as to maximize recurrent correspondences." (Berkeley Linguistics Society. Proceedings of the Annual Meeting of the Berkeley Linguistics Society (1975), p.12)
Every once in a while I still get a few zealots from the long-ranger camp coming to the defense of Sergei Starostin's work with misguided and condescending comments to the effect of "How dare you criticize Starostin!" and "But look at all these correspondences". As to the first line of reasoning: a theory above rational criticism is called a cult. I don't do cults nor do I respect their followers. I also find it frustrating to talk to someone that persists on confusing academic criticism with personal attack and by so doing, creating arguments out of nothing. As to the second line of reasoning, it's the very implausibility of Starostin's correspondences that I object to, so showing me more of his paltry 'sound correspondences' is only being irritating and grievously wastes all of our time.

Recently, this letter was about the origin of Japanese numerals and it was evident that the commenter failed to absorb any of the simple facts I explained in The hidden binary behind the Japanese numeral system. So with guiltless glee I dropped his comment in the wastebasket with all the other troll trash. However, it's a chance to gloss over more pedantic silliness written by doctors of linguistics who should know better. Be brave; be self-sufficient; be curious; open your eyes wide and keep your brain closed to fantasy.

On page 341 of Choi, Japanese/Korean Linguistics, Vol. 3 (1993) (see link), in the article Notes on Some Japanese-Korean Phonetic Correspondences by Alexander Vovin, we see a list of reconstructed numerals comparing Proto-Japanese, Proto-Manchu-Tungus and Proto-Altaic. Notice, reader, how Proto-Japanese *mi 'three' is claimed to come from *ñi and that the attachment of *[ñ] to the Proto-Manchu-Tungus etymon is unexplained and ad hoc, together with the fact that a change of [ɲ] to [m] neighbouring a front high vowel is absurd and completely unmotivated from the perspective of rational notions of phonology. Compare also Starostin's *ŋ[i̯u] with Vovin's *ñïl₁ï and how fellow Altaicists aren't agreeing even on numerals or which cognates to use. We're led to believe that *mu 'six' is to be from *ñu as well and one might remark how curious it is that so many marked phonemes are being reconstructed for simple number terms. God knows why both his *-l₁ï in Altaic 'three' and *-ŋu in 'six' so conveniently disappear in Japanese, why *ñ- in '3' and *i- in '5' need to be added in Proto-Manchu-Tungus to smooth things over, et cetera ad absurdum. Indeed, maybe only God could explain such absurdity; one absurdity for another. Meanwhile, the *d- in 'four' and the same *d- in 'eight' are derived from different sources which, by doing so, ignores the unignorable vowel harmony inherent in the attested Old Japanese set (ie. mi- '3'/mu- '6'; yö- '4'/ya- '8')! No reasonable person can claim credibly that the consistent 'binary' pattern in Japanese numbers simply happened by chance from originally dissimilar roots as Vovin, Starostin et alia are effectively claiming by the shape of these reconstructions.

In other words, theory trumps facts just like we find in all religions. It's amazing what gets passed peer review (assuming such a thing was ever effective in weeding out nonsense).


  1. Ah! The wonderful world of Altaic linguistics. Before I started my study in Comparative Linguistics I was convinced that such a language family existed and that it included Korean and Japanese, since, how else could languages be so similar?

    Needless to say, I was unaware of the enormous effects regional influence can have on languages. And really say, Mongolian and Japanese aren't all that similar to begin with.

    Korean and Japanese still puzzles me though. I have seen some reconstructions that I have not found completely unconvincing, but a genetic relation is far from obvious, yet these languages, grammatically and systematically are so absurdly similar. You can often make long complex sentences, interchange the morphemes with the other language's counterparts and get intelligible sentences every time.

    It boggles the mind (and yes I'm sure there are plenty examples where it doesn't work all that nice, but these languages are a lot more similar grammatically than say, Dutch and Lithuanian).

    Is that really all regional influence? At least to some extent it must be, but if it's that I'm not sure if it's sensible to even consider comparing these two as being genetically related, there's no real obvious cognates.

    -End of Rant-

    -Start of second Rant-

    I really wonder how a guy like Starostin has ever been taken seriously by anyone. This guy got payed for publishing the trash.

    Probably his biggest problem was, that he simply stretched himself out too far. There is no way he was able to specialise in all the language families he wrote about. This is a problem that is already typical for anyone who would want to attempt to reconstruct Indo-Uralic, Starostin went so much further than merely Indo-Uralic and not just for one end of the world, but several.

  2. Concerning Proto-Altaic itself, I'm convinced it exists and is unavoidable. It's necessary to explain the similarities that do exist between these languages, particularly visible in the shared pronominal systems.

    As for Sergei Starostin, he no doubt delighted in the linguistic diversity of his own very large country. Perfectly understandable - it is fascinating. So his involvement in Altaic, Abkhaz-Adyghe, Nakh-Daghestanian, Burushaski and Uralic, while far-reaching, makes sense from his own geographical point-of-view. It's just too bad he didn't like the KISS principle very much and just winged it by feel. And then there's the name of his project, The Tower of Babel, that helps to erode confidence in his willingness to separate personal/religious convictions from strict research and further draws fervent attention from puritanical sorts who are by nature afraid of facts that conflict with their fantasies.

    Back to the craziness, in his Altaic *tṓj- '4' and *ǯa- '8', we see that Starostin expected us to believe that the two just magically became Old Japanese yö- '4' and ya- '8' showing the same vowel harmony as in the fitö- '1'/futa- '2' pair. We're also supposed to believe that **-pkun is a magical suffix in Proto-Tungus-Manchu that just happened to show up to confound the non-Starostinian linguist. What's this suffix based on? What evidence? What is it supposed to mean? Apparently it came out of thin air from what I can see. When we look at the evolution of this Altaic numeral system *as a whole*, it just looks contrived as contrived can be with suffixes of all sorts just showing up without decent explanation.

  3. I am not fully convinced of an Altaic grouping. I do think the languages are related, but not in an exclusive "Altaic" grouping without Uralic, Indo-European, and your Aegean family.

    As someone interested in the evolutionary history of life I learned the principles, called Cladistics, evolutionary biologists and paleontologists use to get a grip on the timing, relatedness, and distribution of traits in evolutionary lineages. on important principle is the distinction between shared derived traits and shared retained archaic traits (our having 5 fingers on each hand, for example, is an archaic trait inherited from the common ancestor of all land vertebrates). I believe a similar principle can and should be used for languages.

    I would not be surprised if many shared features of "Altaic" languages are of this "archaic" nature.

  4. Cladistics only help us understand language in a generalized sense. There are many cases where a wave model is more appropriate. Unlike animal species, languages don't have any genetic barriers that prevent new radical blends. Consider areal influence, creoles and code-switching which clearly defy cladistics.

    A wave model would be much more appropriate in understanding Proto-Altaic too.

  5. Starostin deserves a tip of the hat for compiling the data and making it freely available. But he also deserves a wag of the finger for his amateurish reconstructions, and making his database searchable by reconstructed meaning instead of actual meaning.