The Indo-European Wars
March 10, 2015 7:18 AM Subscribe
Over the past few years, some researchers have been arguing using mathematical tree-building and dating techniques, that the Indo-Europeans originated in Anatolia. In an article [.pdf] in the latest issue of Language, a group of historical and computational linguists using similar techniques say otherwise .
The Indo-European language family includes a wide range of languages spoken over a span of territory that reaches from Iceland to India and beyond. An ongoing debate centers around the birthplace of Indo-European, with the two main positions being the Steppe (or Kurgan) Hypothesis and the Anatolian Hypothesis (with other ideas, like the "Out of India" hypothesis placing the homeland in India being thoroughly discredited by most mainstream researchers).
Under the Steppe Hypothesis, broadly accepted by most linguists and outlined in books like The Horse, the Wheel, and Language , the Indo-Europeans lived in the steppe north of the Black and Caspian seas, and gradually spread out, taking technology like horse domestication and wheeled vehicles (and the words for them) with them, placing the date of Proto-Indo-European (PIE) about 6000 years before present.
However, others have argued for Indo-European starting in Anatolia, with the Indo-Europeans spreading out long before the invention of the wheel and the horse, about 8000 years before present. Under this hypothesis, what looks like a PIE word for wheel is actually an artificial construct, back-created from loan translations (in the same way that we can "reconstruct" a non-existent Proto-Algonquian word for fire water ). Nicholas Wade is a popularizer of this theory, and includes an outline in his book, Before the Dawn .
A group of researchers led by Quentin Atkinson and Russell Grey have dated the age of the Indo-European family tree to around 8000 years before present making the Anatolian Hypothesis more likely [Science]. However, historical linguists have expressed doubt (see comments) for those findings, citing, among other issues, inaccuracies with the model's geography and the the model's language tree .
This new paper [NYT] provides evidence for the Steppe Hypothesis, with the findings mostly hinging on whether or not older written languages like Latin can be used as an acceptable substitute for their contemporary spoken languages-- if they are, the proposed date given by the model for Indo-European becomes younger, and the Steppe Hypothesis becomes more likely. And, a new paper in Nature using DNA evidence [Nature, but should be accessible for free via first link in next link] also suggests that the Steppe Hypothesis may be right .
The Indo-European language family includes a wide range of languages spoken over a span of territory that reaches from Iceland to India and beyond. An ongoing debate centers around the birthplace of Indo-European, with the two main positions being the Steppe (or Kurgan) Hypothesis and the Anatolian Hypothesis (with other ideas, like the "Out of India" hypothesis placing the homeland in India being thoroughly discredited by most mainstream researchers).
Under the Steppe Hypothesis, broadly accepted by most linguists and outlined in books like The Horse, the Wheel, and Language , the Indo-Europeans lived in the steppe north of the Black and Caspian seas, and gradually spread out, taking technology like horse domestication and wheeled vehicles (and the words for them) with them, placing the date of Proto-Indo-European (PIE) about 6000 years before present.
However, others have argued for Indo-European starting in Anatolia, with the Indo-Europeans spreading out long before the invention of the wheel and the horse, about 8000 years before present. Under this hypothesis, what looks like a PIE word for wheel is actually an artificial construct, back-created from loan translations (in the same way that we can "reconstruct" a non-existent Proto-Algonquian word for fire water ). Nicholas Wade is a popularizer of this theory, and includes an outline in his book, Before the Dawn .
A group of researchers led by Quentin Atkinson and Russell Grey have dated the age of the Indo-European family tree to around 8000 years before present making the Anatolian Hypothesis more likely [Science]. However, historical linguists have expressed doubt (see comments) for those findings, citing, among other issues, inaccuracies with the model's geography and the the model's language tree .
This new paper [NYT] provides evidence for the Steppe Hypothesis, with the findings mostly hinging on whether or not older written languages like Latin can be used as an acceptable substitute for their contemporary spoken languages-- if they are, the proposed date given by the model for Indo-European becomes younger, and the Steppe Hypothesis becomes more likely. And, a new paper in Nature using DNA evidence [Nature, but should be accessible for free via first link in next link] also suggests that the Steppe Hypothesis may be right .
The Horse, the Wheel, and Language offers up a whole bunch of archaeological evidence for the Steppe Hypothesis. You can actually follow the break-up of PIE into various language families as the culture propagates out of the steppe region. On the other hand, there is basically zero physical evidence for the Anatolian Hypothesis.
posted by 1970s Antihero at 7:55 AM on March 10, 2015 [3 favorites]
posted by 1970s Antihero at 7:55 AM on March 10, 2015 [3 favorites]
From the DNA evidence article in the last link:
As expected, the researchers found traces of ancient hunter-gatherers and the first Neolithic farmers. But there was an unexpected wrinkle: a massive population movement coming from the plains and grasslands of Eurasia, where Russia and Ukraine are today, beginning around 4,500 years ago.
This is pretty interesting, considering that I have been trained to be skeptical of "mass migrations" when they are supposed to explain changes in language and culture (since language can be learned from your new neighbors without them actually "replacing" you). Here it looks like we may have evidence for Proto-Indo-Europeans as a people, migrating into and populating a Europe already occupied by earlier peoples.
This is such exciting stuff. A story that turns everyone from the English, to the Greeks, to the Danes, including the ancient Romans, Gauls, and Hittites, into one big language/culture family.
posted by General Tonic at 8:42 AM on March 10, 2015
As expected, the researchers found traces of ancient hunter-gatherers and the first Neolithic farmers. But there was an unexpected wrinkle: a massive population movement coming from the plains and grasslands of Eurasia, where Russia and Ukraine are today, beginning around 4,500 years ago.
This is pretty interesting, considering that I have been trained to be skeptical of "mass migrations" when they are supposed to explain changes in language and culture (since language can be learned from your new neighbors without them actually "replacing" you). Here it looks like we may have evidence for Proto-Indo-Europeans as a people, migrating into and populating a Europe already occupied by earlier peoples.
This is such exciting stuff. A story that turns everyone from the English, to the Greeks, to the Danes, including the ancient Romans, Gauls, and Hittites, into one big language/culture family.
posted by General Tonic at 8:42 AM on March 10, 2015
> This new paper [NYT]
The "NYT" link is by Nicholas Wade, who has never had a sensible thing to say about language in his entire career. The fact that the Times still allows him to cover linguistics shows some combination of ignorance and contempt I can't be bothered to sort out, but if you have to read his pieces, take them with huge helpings of salt.
Everybody was all excited about the original work by Gray and Atkinson, but they're biologists, not linguists; fortunately, actual linguists have been following up on this stuff, and of the names mentioned in the Times piece I particularly draw your attention to that of Don Ringe, an excellent historical linguist—anything he says you can trust as the current state of the art in the field.
I had a LH post about this a few weeks ago, which produced this helpful comment by the very knowledgeable John Cowan:
The "NYT" link is by Nicholas Wade, who has never had a sensible thing to say about language in his entire career. The fact that the Times still allows him to cover linguistics shows some combination of ignorance and contempt I can't be bothered to sort out, but if you have to read his pieces, take them with huge helpings of salt.
Everybody was all excited about the original work by Gray and Atkinson, but they're biologists, not linguists; fortunately, actual linguists have been following up on this stuff, and of the names mentioned in the Times piece I particularly draw your attention to that of Don Ringe, an excellent historical linguist—anything he says you can trust as the current state of the art in the field.
I had a LH post about this a few weeks ago, which produced this helpful comment by the very knowledgeable John Cowan:
I’ve read the article twice now (barring the appendixes) and I think I follow most of it now. My comments:posted by languagehat at 8:55 AM on March 10, 2015 [26 favorites]
1) The “steppe” and “Anatolian” theories are pretty much reduced to “late breakup of IE” and “early breakup of PIE” respectively. There’s no archaeological evidence discussed here, and the only geographic point is that if PIE broke up in Anatolia, it’s weird that the Anatolian branch is the most out of touch from the others: if the non-Anatolian languages dispersed to the four winds, how did they stay in touch and share sound-changes that didn’t affect Anatolia?
2) The input to the phylogenetic software is something unusual, but in this case justified. The goal here is to find out how long ago PIE broke up, not to classify it. Indeed, a small amount of pre-existing classification is presumed, of the form Latin > French, Latin > Italian > Spanish, Old Irish > Irish, Old Irish > Scottish Gaelic, etc., and this is used to constrain solutions. Elapsed time is estimated using lexical replacement (see below), and it does not matter whether the replacement is a new word for a meaning, or a new meaning for a word. Consequently, the unit of interest is the pair (root, meaning), where “root” is a PIE root. For classification you want to eliminate loanwords, as they are noise, but here, loanwords are just as interesting as internal semantic change.
3) The authors face head-on the well-established fact that lexical replacement does not happen at a constant rate. Instead, they assume it is log-normal, and allow the values of the parameters to arise out of the data rather than a priori. The choice of a log-normal distribution may be incorrect, but it’s not obviously wrong nor perverse.
The "NYT" link is by Nicholas Wade, who has never had a sensible thing to say about language in his entire career. The fact that the Times still allows him to cover linguistics shows some combination of ignorance and contempt I can't be bothered to sort out, but if you have to read his pieces, take them with huge helpings of salt.
Yup. Wade had a big role in publicizing the original Boukaert et al (Pro-Anatolian) study which most linguists thought was...not great, to say the least.
You'll also note that he allowed Atkinson to get the final word in the article, in re: whether or not Latin is an acceptable analogue for what gave us the Romance languages. Which, for me as a linguist, is really the most interesting part of this whole debate. I have told my students in undergraduate classes "Latin is not Proto-Romance" and "Vedic Sanskrit is not Proto-Indo-Aryan"-- because it's (mostly) true!-- but that doesn't mean that it's necessarily a good idea to put that hypothetical sister language in your model (which Grey and Atkinson did), or that it's not reasonable to say "Hey, the basic vocabulary-- which we're building our model with---would have been essentially the same, so let's put Latin is as a proxy", especially when you have the data from graffiti, etc., to prove it.
posted by damayanti at 9:14 AM on March 10, 2015 [3 favorites]
Yup. Wade had a big role in publicizing the original Boukaert et al (Pro-Anatolian) study which most linguists thought was...not great, to say the least.
You'll also note that he allowed Atkinson to get the final word in the article, in re: whether or not Latin is an acceptable analogue for what gave us the Romance languages. Which, for me as a linguist, is really the most interesting part of this whole debate. I have told my students in undergraduate classes "Latin is not Proto-Romance" and "Vedic Sanskrit is not Proto-Indo-Aryan"-- because it's (mostly) true!-- but that doesn't mean that it's necessarily a good idea to put that hypothetical sister language in your model (which Grey and Atkinson did), or that it's not reasonable to say "Hey, the basic vocabulary-- which we're building our model with---would have been essentially the same, so let's put Latin is as a proxy", especially when you have the data from graffiti, etc., to prove it.
posted by damayanti at 9:14 AM on March 10, 2015 [3 favorites]
I'm going to celebrate with some excellent Turkish food!
posted by Renoroc at 9:17 AM on March 10, 2015
posted by Renoroc at 9:17 AM on March 10, 2015
From the ScienceMag link:
Indeed, if "even small differences in the true ancestral language could throw off the timing estimates" by 2000 years, maybe the margin of error's a little bigger than we'd like to admit?
posted by edheil at 9:19 AM on March 10, 2015 [4 favorites]
But many supporters of the Anatolian hypothesis remain staunchly unconvinced. Paul Heggarty, a linguist at the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, questions Garrett’s methods, arguing that, for example, linguists cannot be sure if the Latin attested to in written documents really was the direct ancestor of later Romance languages, rather than some dialect of Latin for which no record remains. Even small differences in the true ancestral language, Heggarty insists, could throw off the timing estimates.If you have to summon into existence a completely unattested form of Latin (I assume he's not talking about Vulgar Latin, since that is attested) in order to support your hypothesis, maybe your hypothesis has some issues.
Indeed, if "even small differences in the true ancestral language could throw off the timing estimates" by 2000 years, maybe the margin of error's a little bigger than we'd like to admit?
posted by edheil at 9:19 AM on March 10, 2015 [4 favorites]
damayanti and languagehat:
At risk of a minor nerdy derailment in a quiet thread, could someone expand upon "Latin is not Proto-Romance" for the layman? I understand why many (like myself) would think that Latin is "Proto-Romance"? And I'm savvy enough to know that I should be skeptical of such a simple idea. But why, really?
posted by General Tonic at 9:25 AM on March 10, 2015
At risk of a minor nerdy derailment in a quiet thread, could someone expand upon "Latin is not Proto-Romance" for the layman? I understand why many (like myself) would think that Latin is "Proto-Romance"? And I'm savvy enough to know that I should be skeptical of such a simple idea. But why, really?
posted by General Tonic at 9:25 AM on March 10, 2015
The way it was taught to me in high school Latin is that the formal Latin we were taught was not the language of the common people and the provinces, though it was closely related. So the new languages that developed came from that language.
posted by tavella at 9:32 AM on March 10, 2015
posted by tavella at 9:32 AM on March 10, 2015
At risk of a minor nerdy derailment in a quiet thread, could someone expand upon "Latin is not Proto-Romance" for the layman? I understand why many (like myself) would think that Latin is "Proto-Romance"? And I'm savvy enough to know that I should be skeptical of such a simple idea. But why, really?
Essentially, it boils down to the fact that spoken languages are different from written languages-- so, what was being spoken and actually gave rise to the Romance languages was likely, to some degree, not quite Latin. Just like how you talk is likely not quite like what you read in, say, the New York Times (but is, likely, fairly close).
We also have things like Latin grammarians complaining about people using "improper" constructions, and, like I alluded to above, we can compare more and less formal types of writing (like graffiti) to get an idea for the extent of the differences between how people actually talked and how they wrote.
posted by damayanti at 9:44 AM on March 10, 2015 [3 favorites]
Essentially, it boils down to the fact that spoken languages are different from written languages-- so, what was being spoken and actually gave rise to the Romance languages was likely, to some degree, not quite Latin. Just like how you talk is likely not quite like what you read in, say, the New York Times (but is, likely, fairly close).
We also have things like Latin grammarians complaining about people using "improper" constructions, and, like I alluded to above, we can compare more and less formal types of writing (like graffiti) to get an idea for the extent of the differences between how people actually talked and how they wrote.
posted by damayanti at 9:44 AM on March 10, 2015 [3 favorites]
Ha, small world: the spoken PIE parable in the Sciencemag link also was my Intro to Indo-European prof. He's a good instructor, and used Snoop Dogg lyrics to explain the middle voice in English.
posted by eclectist at 9:58 AM on March 10, 2015 [5 favorites]
posted by eclectist at 9:58 AM on March 10, 2015 [5 favorites]
especially when you have the data from graffiti, etc., to prove it.
You mean, like, "Romani ite domum"?
posted by A dead Quaker at 10:21 AM on March 10, 2015 [5 favorites]
You mean, like, "Romani ite domum"?
posted by A dead Quaker at 10:21 AM on March 10, 2015 [5 favorites]
> ... used Snoop Dogg lyrics to explain the middle voice in English.
Link, please. Not doubting, just very curious.
posted by benito.strauss at 11:35 AM on March 10, 2015
Link, please. Not doubting, just very curious.
posted by benito.strauss at 11:35 AM on March 10, 2015
> Here it looks like we may have evidence for Proto-Indo-Europeans as a people, migrating into
> and populating a Europe already occupied by earlier peoples.
Colin Renfrew's interesting (to me, anyway) notion was that the first IE-speakers in Europe were also the first agriculturalists, with the spread of IE piggybacking on the spread of people who understood agriculture. Since farming supports a much larger population than hunting/gathering does, this would have been less like a typical wildebeest migration and more like a slow generation-by-generation extension of range (like what coyotes are doing where I live right now.)
posted by jfuller at 2:42 PM on March 10, 2015 [1 favorite]
> and populating a Europe already occupied by earlier peoples.
Colin Renfrew's interesting (to me, anyway) notion was that the first IE-speakers in Europe were also the first agriculturalists, with the spread of IE piggybacking on the spread of people who understood agriculture. Since farming supports a much larger population than hunting/gathering does, this would have been less like a typical wildebeest migration and more like a slow generation-by-generation extension of range (like what coyotes are doing where I live right now.)
posted by jfuller at 2:42 PM on March 10, 2015 [1 favorite]
The steppe-dwellers were not hunter/gatherers. They were pastoral nomadic cattle herders, who also had knowledge of agriculture.
posted by 1970s Antihero at 3:50 PM on March 10, 2015
posted by 1970s Antihero at 3:50 PM on March 10, 2015
Link, please. Not doubting, just very curious.
Sorry, it was over five years ago and I don't remember the particulars, just the anecdote. It was essentially "I'm gonna (or) I'mma verb me some object". I much better remember when I encountered it again, and thought, hey, middle voice: PUSA's Peaches.
I'm movin' to the country, gonna eat me alot of peaches
posted by eclectist at 3:51 PM on March 10, 2015 [3 favorites]
Sorry, it was over five years ago and I don't remember the particulars, just the anecdote. It was essentially "I'm gonna (or) I'mma verb me some object". I much better remember when I encountered it again, and thought, hey, middle voice: PUSA's Peaches.
I'm movin' to the country, gonna eat me alot of peaches
posted by eclectist at 3:51 PM on March 10, 2015 [3 favorites]
« Older "the uncanniness of recorded music" | Karzai's life is split between Baltimore and his... Newer »
This thread has been archived and is closed to new comments
Well, there can be only one correct model, right? Out of India getting eliminated conveys an impetus to narrow it down to a single dominant theory. A quickening, if you will, towards a champion.
But seriously, a nice and thorough post!
posted by XMLicious at 7:49 AM on March 10, 2015 [10 favorites]