Victoria Falls on the Zambia/Zimbabwe border is near the location where the Last Common Language was spoken.
Science magazine's latest issue has a paper reporting that the Last Common Language was spoken in Africa before Homo sapiens migrated to other parts of the world. All exisiting languages are descended from this one, according to a New Zealand psychologist, Quentin D. Atkinson.
What are some scenarios that this paper appears to rule out?
- Language arose two or more times at different times and places. That is the story of writing, and modern sign languages, but not speech.
- Language was invented by Neanderthals and spread to Homo sapiens once they migrated out of Africa.
- Language was invented by Homo sapiens outside of Africa (say, at Lascaux, France) and then spread around the world and back into Africa. Many technologies do seem to have such a history, including agriculture, but none of them actually conquered the whole species the way language did.
A couple of scenarios that seem to still survive as possibilities:
- Like the human lineage itself, speech is much older than the migration of out Africa and there were a variety of earlier languages or at least protolanguages, but, just as we all share a relatively recent Last Common Ancestor, we also share a relatively recent Last Common Language.
- Linguistic communication began with hand signs which were later replaced by spoken languages, one of which became the Last Common Language.
- Neanderthals had some form of language and there may have even been a little phonemic interchange when they interacted with Homo sapiens (just as there was some genetic input to Homo sapiens from Neanderthals) but there was no wholesale adoption of a Neanderthal tongue. We see something similar in southern Africa where the Xhosa people were already present when Bantu migrants began arriving from the north. Some of the Bantu languages picked up the Xhosa clicks, but retained their basic Bantu structure and phonology.
Atkinson's approach looks for a "founder effect" which was defined by one of the giants of modern evolutionary theory, Ernst Mayr, as, "The establishment of a new population by a few original founders (in an extreme case, by a single fertilized female) which can carry only a small fraction of the total genetic variation of the parental population." Atkinson is making an analogy between genes and phonemes, as basic units passed on from generation to generation. It is a plausible analogy, and I have no quarrel with it, but we should keep in mind that it is an analogical argument and subject to sudden failure when pressed too far.
In linguistics, therefore, the founder effect means that when a new speaking population is created by a small group taking only part of the language with them, its words, phonology, and syntax. For example, America's New England colonies and the Virginia colony were founded by people from distinct parts of England who brought different phonologies with them that persist to this day. \
The out-of-Africa theory posits a series of founders. For example, a founding group begins moving northeast, into what is now the Middle East, from there some move north into central Asia while others follow the coast to India. From India, some move north into the subcontinent while others push still further east into Indochina. We can imagine a series of founders in this account:
- The migrants to the Middle East.
- The migrants to India.
- The migrants to IndoChina.
If we imagine an idealized scenario in which phonemes are stable, the Last Common Language has P phonemes, but the original founder population uses only P – x1 of those phonemes. The migrants to India have P – x1 – x2 phonemes while the founders in Indonesia have P – x1 – x2 – x3 phonemes. If the story really were this simple, the facts would have been noticed years ago: the further you get from Africa, the fewer the number of phonemes in a language. Atkinson's feat was to perform the multiple, complex statistical analyses proving that—after taking the many confounding factors into account—that's exactly what you find. Phonemic diversity shrinks as you move further from Africa.
I'm like the cop in last week's episode of Law & Order: LA who told the prime suspect that he didn't understand tire evidence at all, but the experts said it looked bad for the suspect. I'm really not in a position to discuss the statistical reasoning found in the paper, but it looks bad for the idea that language wasn't already a universal characteristic of Homo sapiens by the time the species began drifting out of Africa.
Atkinson's data comes from 504 languages in the World Atlas of Language Structures. He hypothesized that if language began once, it might have begun anywhere and he chose 2,560 potential sites as possible points of origin. He then generated probable phonemic distributions, taking a variety of factors into account, for each of these possible starting points. Comparing the possible distributions with the actual one, he found the theoretical distributions that best fit the data. And, lo, the sub-Saharan origins are the best fit.
He also tested for a second language point of origin, postulating a first language point in Africa and a second somewhere else. Amusingly, the best fit from this investigation proved to be in South America, although even there the best fit did not appear in every statistical charting tested. The migration into the Americas is believed to have occurred across the Bering strait and, sure enough, the further away from the strait one gets in the Americas the less phonemic diversity is found. This pattern is good evidence that the Bering strait crossers were speakers and no second language invention was necessary.
The South American finding is a methodological warning. If you seek a best fit, you will find one. That doesn't prove the hypothesis. This same warning applies to the overall approach. Akinson sought a best fit and found one. However, the many statistical analyses supported one another, and unlike in the South American case, there was a great deal of robust support for the first fit.
One final matter is the date. Atkinson makes no claims dating various founders, but does say: An origin of modern languages predating the African exodus 50,000 to 70,000 years ago puts complex language alongside the earliest archaeological evidence of symbolic culture 80,000 to 160,000 years ago." [p. 348]
I must confess that while all this work is interesting, none of it seems surprising. If Atkinson had discovered anything else, I would have had to take a long hard look at my forthcoming book and I'm in no mood to quote Emily Litella. ("Never mind.")
eb,
There are some rather large flaws with this paper. I don't have the opportunity to point them all out, but I can point you to Mark Lieberman's excellent discussion of them over at Language Log, http://bit.ly/JUOMv.
Although, I disagree with you, I really like your blog! Keep up the good work; you have challenged me to reevaluate many things.
========================
BLOGGER: Thanks. If you can't learn from the people you disagree with, who can you learn from?
Posted by: Perj44 | April 21, 2011 at 11:47 AM
Two questions:
1. WHY should a "new speaking population ... created by a small group tak[e] only part of the language with them..."?
2. WHY should the original small group create a language more complex (phonemically rich) than any subsequent language over 50,000 years of evolution?
I do not believe that there is any significant evidence for this behavior in languages - other than the gradual elision/deprecation of any language over time, which is independent of migration/exodus.
Atkinson's work is fantastically thorough, although much of the MCMC/Bayesian anlysis is beyond my comprehension. However, while the analysis may have delicately teased a needle from the haystack, it may also have conjured a chain of African rabbits from a Bayesian hat.
Posted by: Geuphrates | May 03, 2011 at 05:45 PM