Phonemic Diversity paper

[Pellonpekko]

I'm linking this to suggest this be discussed, and do not intend to advocate the ideas. I am not very well qualified to evaluate them anyway.

"Phonemic Diversity Supports a Serial Founder Effect Model of Language Expansion from Africa"

http://www.sciencemag.org/content/332/6027/346.full

What are your thoughts about this?

EDIT: Since it turned out not everyone can access the above URL, here are some links on public websites.

Abstract: http://www.sciencemag.org/content/332/6027/346.abstract
NY Times popular article: http://www.nytimes.com/2011/04/15/scien ... .html?_r=1
Criticism: http://www.cslu.ogi.edu/~sproatr/newindex/atkinson.html
Language Log: http://languagelog.ldc.upenn.edu/nll/?p=3090

[Drydic]

"The content you requested requires a AAAS member subscription to this site or Science Pay per Article purchase. To find out what content you currently have access to - view your access rights. If you would like to recommend that your institution subscribe to this content, please visit our Recommend a Subscription page."

Fascinating.

That said, I did heard of this on the news. I really would like to read more, since the article on it on the BBC site has several interesting tidbits (and, conversely, several alarm bells). Given their history of accuracy, I'd rather read the original before commenting.

[Soap]

It was mentioned on the Linguistic Quackery thread, where Risla suggested that maybe the version of the story we're hearing is not a good representation of what they actually came up with (mass media is notorious for that). Personally I think it's just a coincidence. Everyone always seems to assume that everything that we see in Africa has "always been that way", and only the rest of the world has changed. But really we dont know what kind of phonologies might have existed in Africa for the past 50000 years before going extinct.

[Pellonpekko]

Uhm, I'm sorry about this. I apparently used my university-granted access powers for the paper without being aware of them in any way.

I can't do anything about this right now. I'll look for a summary or similar when I have the chance.

EDIT: Language Log is going to discuss the paper soon. It's mentioned at the bottom of this entry on another paper.

http://languagelog.ldc.upenn.edu/nll/?p=3088

[Matt]

I read this last night, and have mixed feelings about it.

I think linguists have reached (or will soon reach) the limits of what comparative reconstruction can determine, because linguistic relationships are so quickly obscured, and I think this sort of biology-inspired analysis will become more common. That said, I have real doubts about some of the assumptions in this paper. Are all phonemes equally susceptible to loss, for instance? It seems to me that clicks might behave differently than pulmonic consonants in terms of phonological change, but I don't know enough to state that with any confidence. And Atkinson's findings seem to depend on Proto-World (or at least the common ancestor of the 504 languages he analyzed) having an awful lot of phonemes, if phonemic diversity in fact decreases as populations break off from a speech community and go their own way. I find that hard to believe; it seems more likely to me that Proto-World would have an unremarkable number of phonemes.

I had an interesting facebook exchange with a friend of mine about this, and she made the comment "I tend to pin language to a fundamental human capacity, so in my historical imagination if humanity came from Africa then language came from Africa, it was just a matter of time until the science caught up."


I've read a lot of articles dealing with the diversity of language in the Americas, and I think the most convincing research has been done by Johanna Nichols.

The following was published in Science magazine, vol. 279, no. 5355, 27 Feb 1998, pp. 1306-7, in an article called "Mother Tongue Traces Steps of Earliest Americans":

From 12 to 17 February, some 5400 people descended on Philadelphia for the annual meeting of the American Association for the Advancement of Science (AAAS, which publishes Science), celebrating its 150th anniversary this year. President Bill Clinton addressed a packed hall, unveiling Neil Lane as his next science adviser and Rita Colwell as the next NSF director (Science, 20 February, p. 1122). But there were more reasons to celebrate: symposia on everything from the earliest Americans to martian life-forms, two of the topics featured in this special news section.

When several prominent archaeologists reached a consensus last year that humans lived in South America at least 12,500 years ago, their announcement struck a lethal blow to what had been a neat picture of the peopling of the Americas--that the first settlers were big-game hunters who had swept over the Bering land bridge connecting Asia and North America about 11,000 years ago. But this revised view of prehistory, based on 2 decades of study of the South American site called Monte Verde in Chile, has spawned a new mystery: When did the ancestors of Monte Verde's inhabitants first set foot in North America? Archaeologists trying to address that question have come up empty-handed, as there are few reliably dated digs in America older than the Chilean site.

At the AAAS meeting, however, a possible answer emerged from another field--linguistics. Using known rates of the spread of languages and people, Johanna Nichols, a linguist at the University of California, Berkeley, estimates that it would have taken about 7000 years for a population to travel from Alaska to Chile.
Because that would put the first Americans' arrival squarely in the middle of the last major glacial advance, Nichols proposes that "the first settlers began to enter the
New World well before the height of glaciation"--earlier than 22,000 years ago.

That date is early but is in accord with recent genetic studies suggesting that the diversity of DNA across American Indian populations must have taken at least
30,000 years to develop (Science, 4 October 1996, p. 31). In addition, Nichols's extensive analysis of Northern Hemisphere languages also suggests that several
groups of Asians entered the New World, where they adapted rapidly to a range of habitats and adopted diverse ways of hunting and gathering.

This picture is winning favor with linguists. "I believe that her general analysis of the linguistic situation in the Americas is essentially right," says linguist Victor Golla of
Humboldt State University in Arcata, California. "We need a much longer period of diversification among American linguistic stocks than the 11,500 years" allowed
by the old view, he says. And although not totally embracing the linguistic findings, archaeologists acknowledge that, combined with other recent findings, Nichols's
results indicate that the old, simple view of the peopling of the Americas is dead. "The bottom line," says University of Kentucky, Lexington, archaeologist Tom
Dillehay, who excavated Monte Verde, is that "the picture is a lot more complex than it was."

To try to get a better fix on how long it would have taken people entering the New World to get to Monte Verde, Nichols surveyed 24 language families that had spread over vast distances, such as Eskimoan languages that traveled from Alaska to Greenland and Turkic tongues that migrated from Siberia to central Europe. She found that the fast-moving languages that spread on foot--the only way the first American settlers could travel--moved 200 kilometers per century on average.
With this yardstick, Nichols calculated that even if early Americans made a beeline, taking the shortest routes over the 16,000 kilometers of varied terrain from Alaska to southern Chile, the trek would have taken at least 7000 years. This would have put the Monte Verdeans' ancestors in Alaska when glaciers made it "probably impossible" to enter the continent, she says. Instead, Nichols argues, the evidence "strongly suggests" a migration before a major glacial advance began 22,000 years ago.

Nichols checked her result against those obtained by other methods. For example, the New World has 140 language families--almost half of the world's total--and she estimated how long it would have taken this rich diversity of tongues to develop. Nichols began by surveying nearly all the language families of the Northern Hemisphere, from Basque to Indo-European, to see how often new language families have split off from an ancestral stock. She found that, on average, 1.5 new language families arose in each ancestral stock over the last 6000 years. Plugging that rate into computer models--which included an allowance for new migrations that carried in new languages after the glaciers retreated--yielded 40,000 years as the minimum time required to produce so many language families.

Nichols also found that languages along the coasts of the Pacific Rim, from Papua New Guinea north to Alaska and then down the west coast of the Americas, share
a remarkable set of grammatical and phonological features, such as the sound "m" in the second-person pronoun (the singular "you" in English), verb order, and -numerical classifiers--words used in some languages when a number modifies a noun. These features set apart the coastal language families from those farther inland, indicating that coastal tongues were probably imported by later settlers.

These kinds of features prompted Nichols to propose the following scenario: The first immigrants from Asia crossed the Bering land bridge "well before" 22,000 years ago and made it to South America. After the glaciers retreated, some people spread north, where they gave rise to the Southwest's Clovis culture, perhaps, and to other peoples. Meanwhile, human beings were again on the move along the Pacific Coast in Asia, with some language families heading south to Papua New Guinea and others north over the land bridge into Alaska--where they could have crossed once the ice sheets melted 12,000 years ago. Yet another group arrived at least 5000 years ago, she argues, giving rise to the Eskimo-Aleut family of languages.

These early dates from linguistics and genetics are prompting archaeologists to reexamine and take more seriously their earliest sites of human occupation, including possible signs of a human presence at Monte Verde as early as 33,000 years ago, says Dillehay. "These findings of great antiquity from linguistics and genetics help us out, but in the end, we have to get the actual time dimension from the archaeological record." To linguists, however, a thousand words are worth a fossil.

[TaylorS]

I think the authors of this paper are chasing an illusion. The combo of clicks and huge phoneme inventories is an areal feature of pre-Bantu southern Africa. The tiny phoneme inventories of the Oceanic languages is a fairly recent development; Indonesian, Tagalog, and the Formosan languages are not phonetically impoverished and certainly neither was Proto-Austronesian itself. Hmong-Mien languages, which originated from not too far from the place the Austronesian languages originated from, have huge phoneme inventories. On the other hand many Niger-Congo languages have fairly small inventories.

The linguistic tsunami that was the Bantu Expansion, IMO, destroyed any way of determining if what these authors seem to be claiming is meaningful or not in any case.

[Tropylium⁺]

You see any global tendencies to this map? Me neither.

[zompist]

My take on the subject. In brief, I'm highly skeptical.

[Yiuel Raumbesrairc]

You see any global tendencies to this map? Me neither.

WALS, strangely... does :

2. Geographical distribution

Languages with average size consonant inventories are found in most areas of the world, suggesting that this size truly is a representative of something typical for spoken human languages. The languages with larger or smaller inventories on the other hand display quite marked regional disparities in their distribution.

Those with smaller than average consonant inventories predominate in the Pacific region (including New Guinea), in South America and in the eastern part of North America, with particular concentrations of “small” inventories in New Guinea and the Amazon basin. The degree of typological similarity with respect to consonant inventory size between the languages of New Guinea and Australia is intriguing. The received idea is that the population ancestral to speakers of today’s Australian languages reached the continent when New Guinea and Australia were connected by dry land in the now partly-submerged landmass known to geologists as the Sahul shelf. Since the landbridge linking New Guinea and Australia was severed around 7000 years ago, contact between Australian and New Guinea peoples is believed to have been strictly limited except in the immediate region of the Torres Straits. Could this similarity represent the conservation of a trait common to languages spoken long ago when the lands were joined?

Those with larger than average consonant inventories are particularly strongly represented in Africa, especially south of the equator, as well as in an area in the heart of the Eurasian landmass, but are most spectacularly concentrated in the northwest of North America. The languages in this latter area belong to a number of different language families with no demonstrable genealogical relationship, including Eskimo-Aleut, Na-Dene, Salishan, Tsimshianic and Wakashan, among others. There is no evidence that the predominance of large consonant inventories in this area is a consequence of direct borrowing of words between these languages although cultural contacts between the peoples concerned are in many cases intense and deep-rooted. The situation is clearly different in one part of the African zone where large consonant inventories occur. Several Bantu languages (part of the larger Niger-Congo family) in the southern part of the continent, such as Zulu and Yeyi, are known to have enlarged their consonant inventory by borrowing clicks and other sounds which they did not previously use from languages of the Khoisan group, which already had many consonants (see, for example Louw 1975).

[Tropylium⁺]

You see any global tendencies to this map? Me neither.

WALS, strangely... does :

No, all those hotspots of large or small inventories are pretty regional. By "global tendency" I would understand something like "small inventories in all of the Americas" or "large inventories in all equatorial areas".

[Pellonpekko]

I updated the first post with some more links, which I hope are accessible for all.

I believe we could challenge the supposition that more phonemes means more phonetic complexity in a language. Finnish would be a good example, and Estonian even a better one. Both languages have a relatively small inventory of basic sounds, but have a lot of diphthongs and make heavy use of length contrasts. Surely that should figure out somehow into the supposed phonetic complexity of these languages?

Well, I look forward to what LL has to say about this and I will now read Zomp's thoughts.

[Matt]

I believe we could challenge the supposition that more phonemes means more phonetic complexity in a language. Finnish would be a good example, and Estonian even a better one. Both languages have a relatively small inventory of basic sounds, but have a lot of diphthongs and make heavy use of length contrasts. Surely that should figure out somehow into the supposed phonetic complexity of these languages?.

Cayuga might be an even better example of a language with few phonemes but lots of phonetic complexity. There are nine consonants and 7 vowel qualities (five oral and two nasal, and depending on whom you ask length is phonemic as well). Phonetically there's all sorts of interesting stuff going on. Vowels in unstressed syllables are realized with voicelessness next to /h/ and creaky voice next to /ʔ/. Stops in unstressed syllables preceded by /ʔ/ can be realized as ejectives. I don't really have more than a basic understanding of the phonetic rules involved, but there's a lot more going on than is obvious from just looking at the phonemes.

[zompist]

More thoughts, as I've now read the original paper.

[AnTeallach]

Now a full discussion on Language Log
http://languagelog.ldc.upenn.edu/nll/?p=3090
and also mentioned on John Wells's blog:
http://phonetic-blog.blogspot.com/2011/ ... nemes.html