15 September 2009

Murray Gell-Mann On Physics and Linguistics

Science News has an interesting interview with quantum physics pioneer Murray Gell-Mann, who discovered the quark, on the occasion of his 80th birthday, mostly about physics, linguistics and scientific revolutions. He spent most of his life at Cal Tech, but is now at the Sante Fe Institute, a think tank designed to develop unconventional scientific ideas.

I'll summarize the physics, briefly.

While there are reasons that preon theory could be attractive and it is a possible theory, the evidence is there. In his view, "at the moment nothing seems to point to composite quarks or composite leptons."

His is inclined favorably towards superstring theory, and in particular, the theoretical graviton, and he feels that this theory could explain the cosmological constant (a.k.a. dark energy) and the mass ratios of generations of quarks and leptons, he notes that colliders are approaching the point where supersymmetric particles should appear, yet they have not. His intuition is that everyone is failing to identify some key broken symmetry from which formulas would flow naturally.

He also expresses interest in a "bootstrap theory" as a particle physics solution:

[B]ack in the days when the superstring theory was thought to be connected with hadrons rather than all the particles and all the forces, back in that day the underlying theory for hadrons was thought to be capable of being formulated as a bootstrap theory, where all the hadrons were made up of one another in a self-consistent bootstrap scheme. And that’s where superstring theory originated, in that bootstrap situation. Well, why not investigate that further? Why not look further into the notion of the bootstrap and see if there is some sort of modern symmetry principle that would underlie the superstring-based theory of all the forces and all the particles. Some modern equivalent of the bootstrap idea, perhaps related to something that they call modular invariance. Whenever I talk with wonderful brilliant people who work on this stuff, I ask what don’t you look more at the bootstrap and why don’t you look more at the underlying principle[.]

I will quote at length what he has to say about linguistics (an interest of mine in a similar vein as well):

For some reason in this country and in Western Europe, most tenured professors of historical and comparative linguistics hate the idea of distant relationships among human languages, or at least the idea that those can be demonstrated. And they set up extraordinary barriers to doing so, by making anyone who thinks that languages are related, what they call ‘genetically related’ in linguistics — it has nothing to do with biological genetics — in other words, one language descended from another and so on, so you get a sort of tree. They put a tremendous burden of proof on anyone who wants to say that languages are related in this way, by this common descent. . . .

But the evidence is actually pretty convincing for quite distant relationships. And in this collaboration between our group at the Santa Fe Institute and the group in Moscow, both consisting mostly of Russian linguists — because in Russia these ideas are not considered crazy — in that collaboration we seem to be finding more and more evidence that might point ultimately to the following rather interesting proposition: that a very large fraction of the world’s languages, although probably not all, are descended from one spoken quite recently…, something like 15 to 20,000 years ago. Now it’s very hard to believe that the first human language, modern human language, goes back only that far. For example, 45,000 years ago we can already see cave paintings in Western Europe and engravings and statues and dance steps in the clay of the caves in Western Europe, and we can see some quite interesting advanced developments in southern Africa that are even earlier, like 70 or 75,000 years ago or something like that. It’s very difficult for some of us to believe that human language doesn’t go back at least that far. So I don’t think that we’re talking about the origin of human language — I would guess that that’s much older. But what we’re talking about is what you can call a bottleneck effect, very familiar in linguistics….

And we see that with Indo-European, which spread widely over such a tremendous area; we see it with Bantu, which spread over a big fraction of the southern two-thirds of Africa. We see it with the spread of Australian. The Australian languages seem to be related to one another. Although there were people in Australia 45,000 years ago, it doesn’t look as if the ancestral language of the Australian languages could go back more than 10 or 12,000 years, maybe even considerably less. So there was a huge spread also. I personally suspect that it was some group from a certain area in New Guinea that we can point to that gave the language ancestral to the Australian languages. Communication between Australia and New Guinea was easier then.

Anyway, since these things are so common in history, in the history of language, why not believe that a big fraction of the world’s languages all go back to a single one, relatively recently? Anyway, we’re accumulating more and more and more evidence. Eventually I think everybody will be convinced that these relationships really exist. . . .

The more the work continues, the more things keep pointing, and they keep pointing toward superfamilies composed of the known families, the acknowledged families, and then super superfamilies, and then finally one single super–super-superfamily that covers a very big fraction of the world’s languages. Provisionally we call that Borean [from the north wind, Boreas, an admittedly likely inaccurate word].

He also touches on the sociology of acceptance of new scientific paradigms, citing plate tectonics (finally accepted in 1962-1963), and the demise of the idea of Fred Hoyle "that there was no early universe."

The linguistics problem shares a character with the particle physics problem.

Modern physics has been able to reduce quantum physics to a score of particles, but has had trouble making the final unification. Modern research has been able to fit the vast majority of modern languages into a fairly modest number of established language families (fewer than twenty), with a sprinkling of language isolates spoken in the aggregate by significantly fewer than 155 million people in a world with 6,707 million people (perhaps five dozen language isolates are still in use, all but fifteen of which are dying; Korean and Basque are the most widely spoken; dozens more that are now extinct; the cluster of these languages among indigeneous peoples of North America, South America and greater New Guinea, moreover, suggests that lack of study as much as true language isolate status, keeps them standing alone), but is shying away from a final bold unifying step that might link those last few pieces.

In both cases, the evidence needs to make a definitive statement is elusive, for a reason that is inherent in the nature of the beast. The linguistic connections Gell-Mann is interested in happened well into pre-history. The phenomena needed to go beyond the standard model of particle physics are largely extremely high energy phenomena (he notes that the Standard Model is a low energy approximation of the physical world).

Footnote: This month's Popular Science actually had an interesting science article (usually, notwithstanding the name, it is about technology, not science per se). It discussed an idea new to me called "dark flow," a new companion to cosmological conundrums dark matter and dark energy. It suggests that galaxy clusters in some of the universe are heading towards a point outside what should be the boundaries universe, rather than in a universal outward direction as Big Bang Theory would tend to lead one to expect. A blog post on the topic explains:

Kashlinsky et al. have examined data from the cosmic microwave background—the relic radiation left behind by the big bang—measured by the Wilkinson Microwave Anisotropy Probe (WMAP). They noticed that a specific microwave signal from galaxy clusters (known the kinetic Sunyaev-Zeldovich effect) observed in WMAP microwave maps is consistent with the idea that everything out to a billion light years from us is moving coherently with the velocity of ~ 1000 km/s in some mysterious direction (see image, right). This speed is around 10 times larger than the standard cosmological model with dark energy would predict. Since they cannot explain the motion with anything within the observable uinverse, they go on to speculate that matter from beyond the edge of the observable universe—some pre-inflationary super structure—could possibly be tugging on the galaxies and causing this super-Dark flow. . . .

Mike Hudson at the University of Waterloo and his collaborators, have recently submitted a preprint that obtains a similar dark flow on slightly smaller scales, but from a completely independent method using distance measurements to optical galaxies.

The easiest way to explain this result would be that the Big Bang universe isn't all that is out there and that something outside the Big Bang sphere is exerting a gravitational pull on big objects within it (something that could also explain dark energy, although not dark matter). Of course, that is a huge intellectual leap to take.

There are concerns, however, about the methodology and assumptions that go into the result.

Footnote Two, September 16, 2009: Borean, the proposed language superfamily discussed by Gell-Mann and proposed by Russian linguist Sergei Starostin (1953–2005) would include almost all of the languages on the map below except the Bantu related languages of Sub-Saharan Africa (a.k.a Niger-Congo) shown in orange and the languages of Australia and New Guinea shown in purple (Papuan and Pama-Nyugan)

In the Borean scheme, languages would be grouped into three main language superfamilies: Nosratic, Dené-Caucasian and Austric. The Nostratic superfamily of languages would in turn group together a Eurasiatic superfamily, and outside the Eurasiatic superfamily, Afroasiatic and Dravidian language families.

The Eurasiatic language family within the Norastric superfamily would include Altaic, Chukchi-Kamchatkan, Eskimo-Aleut, Indo-European, Kartvelian, Nivkh, Uralic, and Yukaghir. "Joseph Greenberg considered his Amerind superfamily to be the closest relative of Eurasiatic, which would place Amerind in Nostratic and all languages of the Americas in Borean."

The controversial view of the Altaic language family in this view include in addition to the less controversial Turkic, Mongolic, and Tungusic, also Korean, Japanese, and the Ryukyuan languages.

In addition the Nostraic language superfamily would include Afroasiatic and Dravidian language superfamilies.

A second Dené-Caucasian language superfamily would inclue Basque, Burushaski, Na-Dené, North Caucasian, Sino-Tibetan, and Yeniseian.

A third Austric language superfamily would include the Austroasiatic, Austronesian, Hmong-Mien, and Tai-Kadai language families.

The Borean proposal has not included the other African macrofamilies (Nilo-Saharan, Niger-Congo, and Khoisan), Australian languages, or the Indo-Pacific languages. It has been suggested that Indo-Pacific might include two languages, Kusunda and Kalto, on the Eurasian continent, which would then be the only languages in Eurasia not genetically descended from Proto-Borean.

As Wikipedia explains in its article about Bornean:

In total the languages in these families are spoken by over 90% of the . . . world population

Nikita Krugly took credit for suggesting the name in 2000[1] but Georgiy Starostin cautioned that the name should not encourage a premature assumption that the African families are more distantly related.

Research on this hypothesis is in its early stages and even its component families have not been established to the satisfaction of most historical linguists, who remain highly skeptical or simply withhold judgment.


Michael Malak said...

The archeology conspiracy website viewzone.com calls the common human language "First Tongue". Wikipedia quotes nineteenth century sources saying pretty much the same thing.


Of course, the Bible story of the Tower of Babel also refers to a common human language.

I think it's foolish to think that Christopher Columbus or even the Nordic were the first fools to try to cross the oceans. I further think it's foolish to think that we had to wait until the Internet to have memes. There are certain concepts that can be expressed best in foreign languages -- those ideas get exported and imported along with the symbolic representations for those concepts.

That's why English is so expressive. It has words and phrases like "je ne sais qua", algebra, kindergarten, chutzpah, and blitz.

Andrew Oh-Willeke said...

The real kicker of the idea Gell-Mann is talking about is not that that all people once spoke one language and that all languages since are descended from it.

The novel part is the idea that there was widespread linguistic diversity perhaps 12,000 years ago, and that then some single upstart language swept rapidly across much of the world (presumably as a companion to the Neolithic revolution) and wiped almost all of the old tongues out while starting to fragment itself.

Indeed, this erasure of old languages and with them presumably old cultures, could help explain why the writers of the Bible embraced a counterfactual YEC worldview. Not only did they lack a written history of their true legacy, they also lacked even an oral tradition as some pre-Semitic language wiped out that old way of thinking.

Anonymous said...

I think it's beautiful that a physicist has teemed up with the dying Soviet linguists to present their ideas. Dixon argues it best ("The Rise and Fall of Languages" Cambridge UP, 1998), but the problem with Nostratic or even positing the existence of an 'established' language family such as Sino-Tibetan (now often referred to as Sinitic and Tibeto-Burman), is that it ignores the other common process of language change: convergence. Sometimes completely unrelated languages mix: this model better explains loan words and the apparent syntactic similarities between Mongolic, Turkic, and Tungusic far better than positing the existence of an Ur-language, Proto-Altaic. Therefore, if unrelated languages can mix, how could we posit ur-languages? Only if we see all language evolution like tree branches which radiate but never re-emerge or cross, not to mention establish sound change as an absolute, could we do this.

English, too, although Germanic, is noticeably the odd man of Germanic with its flirtations with Celtic and its romance with Romance. Simply using the exploding divergence model is too, too simplistic to explain how languages interact.

What is really jaw-dropping silly of Gell-Mann is when he makes broad, generalizing statements like "for some reason... [linguists]...hate the idea of distant relationships." Well, clearly they are hated for the same reason they fail romantically: they just don't work. If we could actually find written evidence for far-flung connections from 80,000 years ago, I promise you the champagne would be flowing in many departments. The problem is that the NeoGrammarian Hypothesis, although generally accepted for practical reasons, has been shown to be uneven in the last 100 years we actually have recordings. We can document that some dialects change faster than other dialects, which appear to change like glaciers, even when they occupy the same space; ergo, how can we clock theoretical language changes with zero evidence backwards by millenia? Answer: we can't. Sound change in human language isn't like carbon testing. It isn't a fixed constant that can be measured. So... until we find some better graffiti, we're screwed to know what human language was like 12,000 years ago.

Finally, it would never be considered foolish to think we had to wait until the Internet to have memes--the word itself was coined in 1976. To say 'there are certain concepts that can be expressed best in foreign languages' buggers the mind. Why is 'computer' superior to 电脑 dian4nao3 'electric-brain'? The first, an English word (Latin origins) was borrowed whole-sale into Russian and dozens, if not hundreds, of languages. The second was a calque into mainland Mandarin 计算机 jisuanji 'calculating machine'. However, the Mandarin speakers on Taiwan thought a native compound would be better, so they made diannao, and that word spread like wildfire over the rest of the Mandarin-speaking (and, by extension, Chinese-speaking) world. Foreign words often get carried with technology not because they have some superior correspondence of sound to ideas but because that's what the damned thing is called when it arrives. "Vat eez dees?" "Dees eez a kompyooter." "Vow, dats sooo kool!" Hell, even the Old Peking word for 'dog' is from Indo-European (PIE: kwon [ Sinitic qwon > quan 犬), domestication of such being an imported technology and, not to mention, a delightful snack.

Anyway, Gell-Mann is another scientist who enjoys rocking the boat of other fields with his physics creds. He's teamed up with an embarrassing crew of guys who made their name in the Soviet days and get avoided by the younger batch in Russia. Thanks, Murray, for the Quark, and the Joyce references; now go find a retirement resort and leave the linguists alone!

Anonymous said...

I want to debunk the singularity.
Are you in or out?