zompist bboard

Coincidentally, a new and seemingly pretty definitive paper in Science yesterday, about the amerindian population. A study of 92 samples from western south and central america (Mexico through to Chile), 8600BP to 500BP, based on mitochondria. Results:

- all related
- last contact with non-American populations was 23,000BP
- a period of isolation in a small population of maybe 10,000 people, presumably equating to being trapped in a hostile but survivable Beringian refugium surrounded by ice sheets
- around 16,000BP, a rapid explosion in population size, presumably equating to the peopling of the Americas
- rapid expansion along the coast, so fast it was probably aided by boats, taking only 1,500 years to reach southern Chile
- as new areas were reached, communities settled, and stopped having contact with one another
- the 92 samples represent 84 different linneages - communities were genetically isolated
- none of these 84 linneages correspond to any known post-colombian linneages, suggesting that the vast majority of pre-colombian diversity was wiped out

Things we can draw from this:
- long chronologies can perhaps be ruled out - no evidence of anyone surviving from before the Beringian invasions, and the speed of advance looks like people moving into an uninhabited area. Suggests that if there were any migrations to the new world before 16000BP, it was presumably in very small numbers, and they must have suffered from some major technological or biological handicaps. [The Thule replaced the Dorset quickly with very little interbreeding... but there were big technology gaps (dogs, sleds, boats, bows, etc), the Dorset habitat declined (warm period, Thule whalers took over defrosted arctic, dorset retreated with the ice (thule then redeveloped dorset behaviours to cope with the return of the ice), and there were only very small numbers of people involved anyway (a few huddled ice-hole fishers, vs the whole of the americas)]

- there is an Amerindian language family - descent from a single, homogenous interbreeding, inbred glacial population of only a few thousand people strongly implies a common ancestral language. For central and south america, at least, I suppose it's possible that north americans, particular those in the west, might perhaps have come from a rival beringian tribe. Possibly.

- 16,000BP is not out of the range of reconstruction, so a reconstructed proto-Amerindian may be a possibility. Afro-Asiatic, for instance, is around 10-18,000 years old; Nilo-Saharan seems to at least pre-date agriculture. However, 16k is clearly pushing it - even if a persuasive demonstration of a unified family is eventually produced, it's unlikely to explain everything

- also a problem: long isolation means subfamilies will be hard to demonstrate

- in particular, it suggests that it's a bad idea to expect big subfamilies on the western coast. It may only be later expansions into the west that show large families

- on the other hand, one good sign for reconstruction: isolation probably limits the amount of borrowing going on. Of course, there can be extensive linguistic contact without interbreeding... but non-interbreeding societies are much less likely to have the sort of extensive borowings that make relationships hard to adduce

- California probably wasn't unique in its plethora of families - it was probably like that all the way down the coast (except places like the Atacama), with the diversity obscured first by expansive and/or invading groups and then by European conquests and the population crash

Personally, I think one of two things is true of the American Indian languages:

1. Many new language families developed in the Americas, unrelated to Eurasian languages and each other, with several giant families found over wide regions.
2. The languages are all part of a giant superfamily after all, but so many sound changes and grammatical changes have occurred that the superfamily cannot possibly be constructed.

The first seems more plausible (to me).

mèþru wrote:Personally, I think one of two things is true of the American Indian languages:

1. Many new language families developed in the Americas, unrelated to Eurasian languages and each other, with several giant families found over wide regions.
2. The languages are all part of a giant superfamily after all, but so many sound changes and grammatical changes have occurred that the superfamily cannot possibly be constructed.

The first seems more plausible (to me).

Option 1 doesn't make sense. You're suggesting that new families arose from nothing. One can conceive of multiple languages entering the Americas and giving rise to families now only represented in the Americas (is that what you meant? - call it Option 1a.), or that diversification occurred in the Americas, which is Option 2.

Salmoneus reports an argument that Option 1a didn't happen .

1a is possible, but I don't get why 1 couldn't happen.

Interesting. I'm completely ignorant about genetics, but this seems to fit very well with the archaeological data. Although, I am skeptical about the implications of this research for the existence of an "Amerind" family. First, a population of less than "a few tens of thousands" hunter gatherers could potentially speak several languages, even languages of different families. Second, some or ever all of the languages of the first migration could have been replaced by unrelated languages spoken by later migrants. With that being said, its definitely an intriguing hypothesis, but the linguistic evidence for Amerind seems pretty meager, even it is 16 000 years old.

mèþru wrote:1a is possible, but I don't get why 1 couldn't happen.

So are you suggesting that groups of speakers began speaking new languages spontaneously? Or were you assuming that these were pre-linguistic peoples that first began speaking once in Beringia/the Americas? I don't think you're going to find either of those can be supported.

8Deer wrote:First, a population of less than "a few tens of thousands" hunter gatherers could potentially speak several languages, even languages of different families.

While isolated in glacier-bound Beringia for 5,000 years?

Yeah, while I don't think we'll ever be able to reconstruct languages as old as the original invasion of the Americas to a useful level of detail, I very much doubt that there were any significant populations of non-verbal humans at that late date. So presumably, even if we can never reconstruct the ancestor(s), all the present-day languages of the Americas descend from something that existed at the time of the invasion. The only question is whether there was a single ancestor language at that date or more than one. We might never be able to determine the answer with much confidence. Anyway, mèþru's option #1, as stated, implies that multiple language families developed from non-language during that period, which is almost certainly not true.

As for #1, I was thinking that ritual conlangs like Damin or avoidance languages might have gradually complicated and replaced other speech.

mèþru wrote:As for #1, I was thinking that ritual conlangs like Damin or avoidance languages might have gradually complicated and replaced other speech.

That's an interesting idea. Correct me if I'm wrong, though, but aren't avoidance languages typically solidly derived from their speakers' primary language? IIRC they're generally ciphers of that language with a restricted vocabulary but the same or very similar grammar and phonology, and I think there's evidence that even their lexicon tends to derive in some way from that of the primary language. I don't think I'd really call that the creation of an entirely new language family.

As far as I know, Damin appears to be an exceptional case of a ritual language having a dramatically different phonology from the speakers' primary language, or indeed from any other languages the speakers were traditionally familiar with.

I don't get why mèþru's scenario #1 should be so unlikely. We have real-life examples of spontaneous language generation already from certain sign languages, so why couldn't it have also taken place in the Americas (and thus at least partially explain the proliferation of language families attested in the Americas today)?

So you guys' answer to language families not being demonstrably related-- when their point of common origin would, according to nonlinguistic evidence, be further in the past than is accessible by the comparative method-- is "Maybe they all independently decided to start speaking conlangs".

Vijay wrote:I don't get why mèþru's scenario #1 should be so unlikely. We have real-life examples of spontaneous language generation already from certain sign languages, so why couldn't it have also taken place in the Americas (and thus at least partially explain the proliferation of language families attested in the Americas today)?

The only thing I find seriously implausible is an interpretation where language, as a whole, had not developed yet.* This is apparently not what mèþru meant. I kind of think multiple language families originally developing from ritual/avoidance/other context-specific constructed languages is a less parsimonious explanation than them developing from existing full languages, but it's not inconceivable.

Also, as has been pointed out, given the amount of time separating the present-day languages of the Americas from a hypothetical Beringian common ancestor, we wouldn't expect to be able to reconstruct this ancestor with the methods available today, so nothing about the diversity of the present-day languages rules out common descent as a possibility.

*If you're wondering, I at least find it hard to believe that non-linguistic humans would be able to preserve technological innovations over long periods, and there's quite a bit of technological development that's much older than the first migration to the Americas.

Xephyr wrote:So you guys' answer to language families not being demonstrably related-- when their point of common origin would, according to nonlinguistic evidence, be further in the past than is accessible by the comparative method-- is "Maybe they all independently decided to start speaking conlangs".

If you're including me with that statement, then I think that really depends on how exactly you define "conlang." Is a pidgin in its initial stages of development a "conlang," for example?

Chengjiang wrote:I kind of think multiple language families originally developing from ritual/avoidance/other context-specific constructed languages is a less parsimonious explanation than them developing from existing full languages

Why?

My the idiocy astounds me.

Travis B. wrote:My the idiocy astounds me.

My the judgmentalism astounds me.

Vijay wrote:

Travis B. wrote:My the idiocy astounds me.

My the judgmentalism astounds me.

You guys clearly don't understand linguistics. Languages do not appear de novo (sign languages notwithstanding, considering they arise in quite different conditions from spoken languages), creoles are not languages created de novo, ritual languages are not really created de novo (and never are used as actual languages used by the population), and while taboos on the use of words that had been as names for now-deceased people have in some cases resulted in increased turnover in vocabulary they certainly have not resulted in replacement of the underlying languages themselves. And "open-mindedness" is not an excuse for ignorance and stupidity.

Travis B. wrote:You guys clearly don't understand linguistics.

Funny, given that I have a damn master's degree in it. What do you have to demonstrate that you're such an expert?

Languages do not appear de novo (sign languages notwithstanding, considering they arise in quite different conditions from spoken languages)

Travis, no one knows how every damn language out there arose. There are only so many languages that are attested; there are bound to have been tons of languages over human history that died without ever being mentioned in any written source. We don't know how languages even looked past a certain point in time, let alone how exactly they arose. That's the whole point.

creoles are not languages created de novo,

Never said anything about creoles; I said pidgins. Either way, though, neither pidgins nor creoles are genetically related to any of their source languages, and there are arguably examples of both that were nevertheless created out of their source languages.

Vijay wrote:

Travis B. wrote:You guys clearly don't understand linguistics.

Funny, given that I have a damn master's degree in it. What do you have to demonstrate that you're such an expert?

I am not a linguist, but having a degree does not make one immune to being an idiot. Don't pull the whole "I have a degree and you do not" thing.

Vijay wrote:

Languages do not appear de novo (sign languages notwithstanding, considering they arise in quite different conditions from spoken languages)

Travis, no one knows how every damn language out there arose. There are only so many languages that are attested; there are bound to have been tons of languages over human history that died without ever being mentioned in any written source. We don't know how languages arose hundreds of thousands of years ago. That's the whole point.

Have we ever seen a language created de novo? Do we have any actual evidence for any languages being created de novo? The reasonable assumption is simply that they split up from other documented language groupings at a time depth too early for us to reconstruct, and that their relatives in Eurasia could likely have become extinct before leaving any evidence of their existence. In this case, remember that the spread of humans from Eurasia to the Americas occurred at a time depth well before the greatest time depth within our capabilities to possibly reconstruct any language varieties.

Vijay wrote:

creoles are not languages created de novo,

Never said anything about creoles; I said pidgins. Either way, though, neither pidgins nor creoles are genetically related to any of their source languages, and there are arguably examples of both that were nevertheless created out of their source languages.

I said creoles, but I also meant pidgins. And just because they do not have a single genetic parent does not mean they are created de novo. They almost invariably have one or more substrate languages and a lexifier language (or possibly multiple lexifier languages).

Xephyr wrote:

8Deer wrote:First, a population of less than "a few tens of thousands" hunter gatherers could potentially speak several languages, even languages of different families.

While isolated in glacier-bound Beringia for 5,000 years?

Why not? Languages diversify internally all the time and 5000 years is a hell of a long time for linguistic shift. Isolation isn't a barrier to linguistic change.

Personally, I don't buy the idea of new language families arising from ritual languages/conlangs/Creoles, but I also don't buy the Amerind hypothesis. The idea that multiple languages/language families entered the Americas in the ancient past seems totally plausible to me, especially given what we know about Na-Dene and Eskimo-Aleut.

EDIT: For the record, there is experimental evidence against the "conlang" idea.I remember reading a study where they showed that very young infants couldn't learn languages with "unnatural" grammatical rules in artificial languages, whereas natural rules were easy for them. Since we all know what our early conlangs were like, I have a hard time imagining children learning them as an L1.

8Deer wrote:EDIT: For the record, there is experimental evidence against the "conlang" idea.I remember reading a study where they showed that very young infants couldn't learn languages with "unnatural" grammatical rules in artificial languages, whereas natural rules were easy for them. Since we all know what our early conlangs were like, I have a hard time imagining children learning them as an L1.

That's because we study them. Someone in prehistory who doesn't have a library (or internet) full of knowledge on languages isn't probably going to come up with something much different than their own language(s). Damin is grammatically almost identical to its source language, for example, the difference being entirely restricted to allomorphy (i.e. there isn't any). Even working at it, I highly doubt without exposure to formal study that someone's going to come up with, for example, ergativity if their language is nom-acc, case-marking if their language is head-marking, distinctive aspect if their language is strictly tense-based, and so on, so you're probably not going to have gross violations of "natural" grammatical rules in a conlang created by a pre-literate society.

Also, this is why I pointed out the problem of how you define "conlang." The idea that I'm thinking of, at least, is very different from starting by learning the sort of conlang people may create on a forum such as this one.

vokzhen wrote:

8Deer wrote:EDIT: For the record, there is experimental evidence against the "conlang" idea.I remember reading a study where they showed that very young infants couldn't learn languages with "unnatural" grammatical rules in artificial languages, whereas natural rules were easy for them. Since we all know what our early conlangs were like, I have a hard time imagining children learning them as an L1.

That's because we study them. Someone in prehistory who doesn't have a library (or internet) full of knowledge on languages isn't probably going to come up with something much different than their own language(s). Damin is grammatically almost identical to its source language, for example, the difference being entirely restricted to allomorphy (i.e. there isn't any). Even working at it, I highly doubt without exposure to formal study that someone's going to come up with, for example, ergativity if their language is nom-acc, case-marking if their language is head-marking, distinctive aspect if their language is strictly tense-based, and so on, so you're probably not going to have gross violations of "natural" grammatical rules in a conlang created by a pre-literate society.

The phonology of Damin is not natural by any stretch of the definition. Also, these hypothetical children learning a conlang would also hear their parents language around them all the time. When they grow up, which language do you think they would continue speaking in their community?

This is all really speculative anyways, I feel like we should start a separate "origins of linguistic diversity" thread at this point

8Deer wrote:

Xephyr wrote:

8Deer wrote:First, a population of less than "a few tens of thousands" hunter gatherers could potentially speak several languages, even languages of different families.

While isolated in glacier-bound Beringia for 5,000 years?

Why not? Languages diversify internally all the time and 5000 years is a hell of a long time for linguistic shift.

Unlike the insane American Languages Are Conlangs Invented By Basque Monks Hypothesis, which is a crackpot theory that violates all principles of linguistics and proposes phenomena never observed and inherently implausible, with no clear motivation, this idea is actually possible.

But I don't think it's likely. Remember, the population size calculated (which iirc was 2,000 breeding females, probably equating to 10,000 people in total at any time) isn't for the total number of people in Beringia at that point, it's for the people in the interbreeding population. So there wasn't isolation.

We do see the many-tiny-languages thing in America... but there we also see genetic isolation accompanying it.

In Beringia, for there to be multiple languages, we'd have to be looking at a small community with extensive in-breeding, in a relatively confined space, in a shared environment, probably with a lot of mobility (hunter-gatherers in that sort of marginal environment generally have) somehow managing not to speak to one another. It's possible, but to me it seems much less plausible than a common language or, at most, a handful of related languages (say, inland hunters vs coastal fishers?) that themselves have a common origin.

Vijay wrote:

Chengjiang wrote:I kind of think multiple language families originally developing from ritual/avoidance/other context-specific constructed languages is a less parsimonious explanation than them developing from existing full languages

Why?

Positing a phenomenon that is completely unheard of and implausible, occurring multiple times, in order to explain evidence that would appear more-or-less the same even without that phenomenon, is by-definition unparsimonious and is incredibly ad-hoc. Let me elaborate:

Problem #1: The phenomenon is completely unheard of: No population has ever abandoned their previous language and adopted a new one invented from whole cloth. The comparison to sign languages is not valid: deaf people did not all abandon a previous, perfectly-functioning sign language and start speaking in a new one invented ex nihilo. Nor did the members of the Lardil tribe start speaking Damin in everyday life.
Problem #2: The phenomenon is implausible: Why would a population all stop speaking their old language and start speaking a conlang? How would they be convinced to do so? Damin had a less robust grammar than Lardil, and so even if the Lardils wanted to start speaking Damin all the time it would be highly impractical. Appealing to speech taboos in Australian languages also doesn't solve the problem: the new vocabulary entries were borrowed from neighboring languages or coined from native material; maybe sometimes they straight up made-up a new word, but I doubt this happened often and certainly not often enough to completely stock an entire lexicon with made-up words. Plus, according to Dixon (pace Salm) this made the languages of Australia appear more related, not less, so it's the opposite of what you want for the Amerind Conlangs hypothesis!
Problem #3: The phenomenon would have to occur multiple times. Because if it had just happened once, then the initial ritual conlang would be Proto-Amerind and then you're back where you started. So at least twice, with two completely independent conlangs, is necessary... probably much more than two, given how much alleged non-consanguinity seems to be being assumed for this hypothesis.
Problem #4: The phenomenon is unnecessary to explain the evidence: Languages diverging at 3-4x the time depth of PIE (and which have, for most of the families involved, not had as much comparative reconstructive work done on them) would not appear presently to be related anyway. Put in bayesian terms: it doesn't make what we observe (multiple apparently-unrelated families) any more likely than the null hypothesis-- P(B) and P(B|A) just cancel each other out, and we're left with the P(A), the random, unsupported, out-of-the-blue suggestion that "maybe there were conlangs".

Compare that to the competing hypothesis, which explains the evidence equally well and assumes that nothing happened in the history of the languages involved that hasn't been observed in language families from every other part of the world.