Out of Africa (OOA) has been the main theory of the origin of modern humans since the mid-20th century. Strictly speaking this is only a theory of anthropology. Since language is a human phenomenon, [1] it has however also sprouted a “linguistic Out of Africa” theory alongside.
According to what could be called “the evolutionary theory of language”, we observe that new languages only come about by the spreading and splintering of earlier languages. (Or, perhaps a better biological analogue still is the third tenet of cell theory: that cells only come about from other cells.) This alone already suffices to imply that there exists a family tree of languages, tracing back to the ancient era of glottogenesis. Connected with a relatively late (“recent”) expansion of modern humans out of Africa, we can then in particular infer the highly likely existence of a language that could be called “Proto-Exo-African” (PEA) — the language of the humans who first set on this exodus, which must also have been a common ancestor of all languages spoken in Eurasia, Oceania and the Americas. This is an idea that is in principle sound, even if, in my impression, underappreciated among historical linguists. The smaller number of so-called evolutionary linguists out there do understand it well, at least.
This argument though says nothing about if this common descent would be in any way identifiable from the linguistic data itself. Language does not have a strict analogue of DNA (or any other similarly transferred major biochemical machinery), and is not strictly speaking “transmitted” as much as “constructed” over and over again every generation. No child is born knowing a language, only with the ability to acquire a language. This could add up to the result that every linguistic feature of PEA has been by now either lost or diluted to undetectability. And it happens to be the case that all language families with general approval so far are still at least an order of magnitude younger than the assumed recent OOA spill-out starting some 70,000 years ago. Even the more ambitious proposals like Amerind or Nostratic (that actually have some legitimate comparative evidence backing them, unlike attempts to scrape together things like Proto-World) are only proposed to reach at most some 20-30 millennia of age, i.e. barely a third of the way back.
If PEA might be undetectable by direct means, how error-proof is the indirect demonstration of its existence then? As long as we do not question the underlying OOA theory, there are really only four possibilities under which the assumption of PEA might fail:
- humans leaving Africa did not yet have language, and it has come about only later, possibly several times independently;
- at some point in prehistory, new languages were created from scratch to replace earlier natural languages, and some or all modern languages rather descend from these “new” languages;
- the African exodus population spoke more than one language;
- at some point in prehistory, other (possibly entirely unrelated) language families have secondarily spread out of Africa, to replace some or all descendants of PEA.
Of these, #1 is difficult to directly refute. Spoken language does not fossilize, and hence the study of the biological evolution of language is to a large extent an issue of speculation. The most common opinion around however is that language would have existed at least by the transition to anatomically modern humans (AMHs), as distinct from Neanderthals and the newly found Denisovans, so at least a couple hundreds of millennia ago. In this post series I will continue to follow this assumption as well.
All the others, however, would not make major dents in the hypothesis of monogenesis of non-African languages.
#2 is a priori improbable, and hence not actually a major objection. If we take seriously the rarity of languages being freshly invented (i.e. stick to the principle of uniformitarianism), then even recent glottogenesis events will actually only leave a slightly weakened OOA theory of language, one where we can allow for e.g. Basque or Turkic to have been created from scratch, but all other non-African languages can be still assumed to be descendants of PEA. Similarly #3 would only split the family tree into a small copse of unrelated-at-OOA-time trees, probably at most no more than 3-4. These would have good chances of being still related to one another at some pre-OOA date, so that there is a PEA in the last common ancestor sense, even if confusingly enough it was not itself exo-African (and could be therefore also the ancestor of various African languages).
#4 actually has good chances of being true: maybe the best contender for non-PEA languages spoken in Eurasia today are the Semitic languages in the Levant and Mesopotamia, grouped in the larger Afrasian family, whose homeland is often (though not always) placed somewhere in northeastern Africa. But we can also see that this too would only very slightly push back the boundary of African vs. exo-African languages. Languages spread only step by step. Perhaps some other lineages in the vicinity of Africa, say Sumerian or Dravidian, could be also of yet more recent African origin, but once modern humans had first colonized places like Siberia and Southeast Asia, new intrusions all the way from Africa are unlikely to happen.
Altogether even a buffer zone of maybes doesn’t seem to shake the conclusion that the 100+ exo-African language families known today to linguistics (including isolates) must be only a few top boughs of at most a handful of much larger underlying language families, dating back to the time of the OOA expansion.
But has there really been a recent Out-of-Africa expansion?
There definitely has been at least one OOA event, since also the Neanderthalians, Denisovans and Homo erectus are thought to descend from African hominins. For recent modern human OOA in particular though, the main line of hard evidence has originally rather come from the fossil record. 200k years old Homo sapiens remains from Omo Kibish in Ethiopia, the oldest known throughout the late 20th century, have been some of the best supporting evidence. This is however but a single archeological datapoint — and one that has been overturned even. Since last year, the oldest known remains of AMHs now come instead from Jebel Irhoud in Morocco, dated around 300k years old. Note that this is quite a gap, both chronologically and geographically! The data doesn’t get especially dense going forward either. Only a handful of any modern human remains under 100k years of age are still known. More importantly this slightly extended selection already includes locations also outside Africa, in modern Israel and Oman (discovered not too long ago as well). These have been suggested to represent “failed migrations that died out”, but this strikes me as special pleading. I dout that anyone looking at this scattered early record without the weight of research history (and with understanding of the Signor–Lipps effect) would place the origin of anatomically modern humans within Africa with great confidence. At minimum a Near Eastern origin seems to be entirely within the question as well. Paleoecology could be able to suggest other likely locations still.
Several posts by anthropology blogger Dienekes have moreover drawn my attention to a few interesting additional arguments to consider recent OOA on very shaky ground by now. For one, recall that modern humans’ closest known relatives are the Neanderthalians and the Denisovans — two Eurasian species, with Denisovans branching off first, which would suggest that the common ancestor of the three, and even the Neanderthal-AMH last common ancestor, lived in Eurasia as well. (This does not need to coincide with the LCA of crown group AMHs, however.)
For two, genetics has for long pointed out that the modern human populations of sub-Saharan Africa [2] show altogether greater genetic diversity than those of Eurasia (+ with even further rarification in Oceania and America). However with the rapid development of archaeogenetics in the last few years, we have by now first lines of evidence that this could be due to admixture with archaic Homo sapiens groups.
Maybe the Neanderthal and Denisovan OOA event was then also the main modern human OOA event after all?
This would also imply at least two inverse “into Africa” expansions (one leading to archaic African substratal groups, the other for crown group AMHs). This does not seem to be a very costly assumption though, since the distribution of several archaic haplogroups already demands multiple major population movements across the continent. E.g. the archaic mtDNA haplogroup L0 of South Africa is both first-to-branch-off and present in only fractional proportion in the populations that do carry it, clearly requiring multiple admixture events along the way (instead of, say, a Great San Migration that starts 100k years ago followed by them hanging out in South Africa mostly intact after that). There are also haplogroups with primarily Eurasian distribution but some inroads even into sub-Saharan Africa, requiring their own more recent but still prehistoric into Africa or out of Africa movements or gene flows (e.g. Y-haplogroup T, mtDNA haplogroup U).
An option to be also kept in mind is haplogroup extirpation: various today-African haplogroups may have once existed in Eurasia too, but eventually died out, e.g. under later population movements. The same could be the case anywhere of course, but to me it seems that Eurasia is a priori the most likely location for this. For one already due to size, for two due to the historically well-documented extensive population movements, particularly across and around the major “crossroads” that is the Near East. [3] A third and maybe the most powerful candidate for wiping out genetic diversity from Eurasia would be the latest glaciation period. Still, the human Y-chromosomal and mtDNA haplogroup trees both start off with a remarkably large series of exclusively African early groups. Any theory of AMH origins outside of Africa would have to explain most of these through archaic admixture, with haplogroup extirpation probably only granting some wiggle room around those points in the two family trees where the branches start to turn Eurasian-centric instead.
I’m aware I’m outside my zone of expertise here, so in case this sounds like I am suddendly a few moomins short of a valley, I do want to note that overturning recent OOA is still not looking cut and dry exactly (and if any of the arguments above have big glaring holes in them, I would appreciate readers pointing them out). If you bear with me for now, though, I will develop in the next post some potentially very interesting corollaries this possibility would have.
[1] It may be at times useful to think of linguistics, likewise most other humanities subjects, as a subdiscipline of anthropology (in a somewhat similar way as how biology could be considered a subdiscipline of chemistry).
[2] Important as a human genetic and cultural area, and to some extent also as a more general biogeographical region. North Africa by contrast on many marks aligns with Eurasia instead. I have wondered if sub-Saharan Africa would deserve its own underived term, along the lines of “Maghreb” for North Africa. There are a few historical candidates, but nothing that really stands out as immediately usable: Ethiopia and Sudan have been already claimed by states (much as also Libya), and Zanj also has been kind of claimed by Tanzania. The analogy of Australia could suggest e.g. “Meridionalia” or “Equatoria”, but outright coinages have quite a few orders of freedom to them.
[3] Later also the “highway” that are the steppes, but my impression is that pseudo-periodic nomad invasions of Eastern Europe / Persia / China have only really been a thing after the domestication of the horse, and related inventions such as chariots, saddles, stirrups etc., all fairly recent in the big scheme of things.
Freelance reconstruction 
That’s just mitochondrial DNA, and just in that one paper that passed me by. Nuclear DNA, according to the same paper as well as any other I’ve seen, shows no trace of this, instead finding Neandertalers and Denisovans as sister-groups, even though their common branch is not terribly long. H. heidelbergensis seems to be part of the Neandertal side, though DNA is not preserved. All of these are considered the result of a separate emigration from Africa some 600,000 years ago.
The discrepancy between mtDNA and nuDNA is most easily explained the way the paper does it: as introgression involving a few, perhaps just one, H. s. sapiens who became the ancestor of all later Neandertalers in, specifically, the direct female line. This looks like (sexual?) selection was involved.
Gene flow back into Africa has been known for a while: all Africans today have a thin veneer of Neandertal DNA, much thinner than the non-Africans. Denisovan DNA does not seem to have made it west of the Tibetan plateau; east of that line, three introgression events into ancestors of modern populations can be found.
I see; it doesn’t really eliminate the key thrust of this argument, but does weaken it (it would pare things down to just one clearly-Eurasian lineage, versus one African-and-at-least-marginally-Eurasian).
Yeah. “Africa” in this biogeographic sense didn’t end at the Suez Canal, but included the Fertile Crescent.
Which for sure sounds odd; if early humans managed to locate the upper Nile in the Ethiopian mountains and then follow it north to reach the Fertile Crescent, how come they didn’t manage to similarly navigate their way into Europe, Southern or Central Asia?
Do any other species have this type of a distribution either? Other Africa/Eurasia-stradding species I can think of OTTOMH like the wolf and lion extend much further out too, while plenty of species are exclusively African despite wide distribution there.
In today’s climate, this kind of thing does happen. There used to be hippos and elephants in Syria in historical times.
Wikipedia doesn’t say what the southern limit of the cave lion (Panthera spelaea) was, except for showing a figure from this paper which looks like it plasters over a lot of gaps in the geographic record. But then, we don’t have Pleistocene humans from Anatolia either that I know of.
I agree on the nomad invasions; but the Denisova cave itself has yielded Neandertal remains and a bone splinter whose father was a Denisovan and whose mother was a Neandertaler, so the absence of a geographic barrier did translate into some amount of gene flow.
“some 600,000 years ago” should be “between 765,000 and 550,000”.
Giving birth to a bone splinter must have been painful, even for a Neandertaler.
maybe 5. language nucleus from Neanderthals or Denisovans populations or a hybrid of those and a PEA langoid?
Hah, yeah, that would do the trick. Though it’s still weak to the possibility that there were just some Eurasian language families that were taken over from Neanderthals, versus some that came from Africa. (It would also suggest that if our relative species also had language, then probably our common ancestors did as well…)
I think the most likely is 2. “at some point in prehistory, new languages were created from scratch to replace earlier natural languages, and some or all modern languages rather descend from these “new” languages;”
70 000 years is a long time, plenty of time for that to happen a couple of times.
But who creates a language from scratch and gets a whole community to speak it to their children?
Indeed. What the example of the modern conlanging scene shows is not just that humans are able to start thinking up languages from scratch (let us ignore for now the literacy and linguistic education evidently required before this can happen); it also shows is that constructed languages will start off with no social cachet and about no chances of catching on as a real live spoken variety! Proportionize this result further with population sizes in the Stone Age, and I think the odds of any new languages arising from scratch, if a fully functional language is already available also, are quite microscopic.
There are two rather more plausible processes though by which coinages can end up proliferating wildly: taboo replacements and ethnolinguistic nativism (the purging of loanwords or even cognates, to emphasize one culture’s difference from another — documented not just in the context of modern nationalism, but also from native cultures in multilingual areas of PNG). These will leave grammar mostly as is, but over time the result could be ultimately indistinguishable from sprachbundlich convergence, maybe at most with some minor difference in the amount of cognate affixes or semantically neutral-ish verbs.
And if there is no community?
What about two people who meet each other from different communities and start from scratch?
Or what about a family that is outcast/flees from a community and wants nothing to do with the old one?
What about children that survive as orphans?
Or, what if a community moves into a new environment that is so different that even the 50 core words from the previous language are abandoned and reinvented?
What if a community experiences a great trauma and interprets it as the necessity to reinvent it’s understanding of the world and itself?
70 000 years is a long time for this to happen several times, these are not languages that have several thousand words, but only hundreds and in some cases probably only tens of words, and relied more on expressions and gestures.
And if there is no community?
The core problem is that there is no cultural continuity without community. Pre-literate hermits or individual outcasts eventually die off without leaving anything behind at all.
For your slightly milder scenarios:
1) Two people who meet without a common language: this results in the first place in a pidgin, not in a new language entirely. If contact remains and both people have a community to go back to, eventually people will also simply learn each other’s language. (This also kind of starts getting at the problem of creolization, though it’s a problem more for various details than the big picture of language families: most creoles can be fairly well considered to be genetically descended from their “main lexifier”.)
2) Outcast family: could maybe happen (not sure if this is actually attested anywhere though), but they’re again very unlikely to re-grow into a full community.
3) Orphan children (or children of deaf parents, etc.): definitely happens, but has the same problem as the previous. Also, if these orphans ever come back into a community out there somewhere, they’ll sooner adopt that language rather than them whatever rudimentary family language they had developed.
— But put enough language-less “outcasts” in one place, and yes, that will result in a new language from scratch. This has been documented unfolding in real time with Nicaraguan Sign Language after all. So actually something similar is pretty likely to have been the case for most sign language families out there. It’s hard to imagine how this would happen with spoken language however; who would be the people they’re unable to learn language from?
4) Environment is too different: doesn’t happen. Core vocabulary is mostly separate from environment after all. People will continue to have eyes and skin and mothers and sisters, to eat and sleep and die, to distinguish me from you from that and one from two from all, no matter where they go. This is only marginally more weakly the case for environmental staples like trees, rocks, water, fishes and louse, found all around the globe. (I notice you recently blogged about “the troubles with Swadesh word lists”, but this seems to miss the point that stability is after all only statistical: Romanian might have a notably messy situation with “belly”, but the other languages of the world on average do not.)
5) Community reinvents itself: doesn’t happen either. War or drought or pestilence or other forms of hardship don’t provoke this response anywhere around the world, if anything they do the opposite: cause a community to circle its wagons tighter around the cultural traditions they still have, such as language. (Cf. Hebrew.)
these are not languages that have several thousand words, but only hundreds and in some cases probably only tens of words, and relied more on expressions and gestures.
No, this is a common misconception. By the uniformitarian principle, every language in the time period where the concept of “language families” makes sense at all was about as multifaceted and rich in vocabulary as all languages are today. We just aren’t able to reconstruct more than a tiny sliver of these details thousands of years later.
The “And if there is no community?” was a question, and bellow it was the attempt to construct answers when that could have happen, however unlikely that might be. All of these scenarios are unlikely but there is a small chance that they happen.
2) If from a hundred outcast families 99 die off or retain there language then there is still 1 left. If that 1 becomes a pioneer family in an empty environment? All hypothetical circumstances but 70 000 years is a long time.
3) We are talking about a world with 0.015 persons per km² instead of 56 persons per km² on average. With such an empty world the probability that an outcast / pioneer / orphan stays isolated is higher.
3a) “who would be the people they’re unable?”
– monkeys, dogs, wolves, ostriches, goats, sheep
See: Feral Children
https://en.wikipedia.org/wiki/Feral_child
4) Swadesh understood the limits of such a large word list. The question in my blog post was which word list makes most sense to use to for a “dumb” comparison of a large number of languages. Belly should be scratched from that list, but it might be a useful to make a “detailed” comparison of a few languages.
5) could a community trauma or new environment trigger taboo replacements. If 9 out of 10 communities double down on their cultural traditions and 1 out of 10 reinvent themselves then that would mean over the course of thousands of years this would happen several times. And maybe not with a clean break in one generation but maybe over three generations?
I think it is likely that there is no last common ancestor. Happy to be proven wrong but that might not happen in my lifetime.
Sure, all of these have a small change. But how small exactly? One in ten vs. one in ten thousand vs. one in ten million are all very different cases.
And while 70,000 years is a long time, what matters is not the number of years per se… Ten villages’ worth of people lead to the same amount of events per decade, no matter if this is ten different villages concurrently, or the same village across one century. Because of this, prehistoric population sizes will work mostly against your case. Weighed by population, most of world history has actually happened within historical times. Let’s be generous and say that world population pre-agriculture was one million on average (judging by the total amount of human genetic variation, it may actually have been less than 10,000 at times); then 70,000 years of this will be expected to generate “the same amount of history” as ten million people within 7,000 years, or hundred million people within 700 years, or one billion people within 70 years; or indeed, seven billion people within a mere 10 years! We have hence gone thru about four “units” of this size already since 1800. This is already equal to on the order of 300,000 years of hunter-gatherer life.
(Actually even more so, since human innovativeness is strongly conditional on connections with other people. All Paleolithic would-be linguists would have been intellectually isolated and unable to build on anyone else’s work and knowledge.)
Hence, if a type of event were possible in human prehistory, we should be seeing it actually happening somewhere today as well, perhaps multiple times. And yet we know from the historical record that e.g. thousands of cases of community upheaval quite simply do not lead to any major linguistic transformation at all. A fortiori, the odds of this particular scenario are not one-in-10, they’re vanishingly small, too small to possibly matter.
We do know of one language that was thought up outside of modern conlanging: Demiin. (That was the secret language of second-degree initiates, but of course we could imagine some huge religious upheaval that turned such a thing into the first language of a community.) It has words out of nowhere, sounds out of nowhere, and a few highly visible grammatical features out of nowhere (e.g. just two personal pronouns, “me” and “not me”, instead of the regionally expected nineteen) – but “the Damin registers of the Lardil and Yangkaal use all the grammatical morphology of those languages, and so therefore are broadly similar, though it does not employ the phonologically conditioned alternations of that morphology.”
And if such a thing happened in an isolating language, the syntax would still be inherited.
I just learned that “linguists have hypothesized that Khoe-Kwadi [aka Central Khoisan] actually originated in Eastern Africa and was carried to the Kalahari Basin by pastoralists around 2,000 BC, before Bantu speakers reached southern Africa. Interestingly, recent studies have confirmed the presence of genetic markers from Eastern Africa in Khoe-Kwadi speakers and neighboring populations…” – so perhaps the linguistic history of even southern Africa is like that of northern Eurasia, with wave after wave of expansions steamrollering the linguistic diversity and leaving no clues to the end-Pleistocene situation.
The similarities with Sandawe already should suggest something of this sort for KK really.
I’ve seen even a stronger claim, though alas I do not recall the source: that paleo-DNA would show all populations of southern Africa to have arrived there less than 10,000 years ago (where from, I’m not up to speed on, but on general migration-typological grounds maybe again more likely through East than Central Africa).