John Emerson sent me a link to Bob Yirka’s Phys.org piece “First languages of North America traced back to two very different language groups from Siberia,” saying “Looks dodgy to me”; I responded “Well, Johanna Nichols is well respected, so I wouldn’t dismiss anything she says out of hand, but this is surprising. I’ll toss it over the fence at the Hattery and see if it gets torn to shreds.” Yirka’s executive summary:
Nichols’ techniques involve the use of linguistic typology, a field that involves comparing languages and organizing them based on shared criteria. To learn more about early North American languages, she compiled lists of language characteristics and applied them to all known languages. She then scored each of the languages based on the revealed qualities. This allowed her to compare the languages as a way to find resemblances among them and spot patterns.
Nichols found that she could trace the languages spoken in early North America back to just two lineages, both of which originated in Siberia. They came, she notes, with the people who made their way across land bridges during Ice Age glaciation events.
Those two main groups she found evolved into different languages as people moved to different regions—she focused most specifically on 60 of them. She found that many of those languages were also impacted by multiple waves of Siberians arriving in North America. She concludes that some of the characteristics of the original languages have been retained through the years and are now in the current linguistic population.
Nichols’ paper (open source) is here; back in 2012 marie-lucie wrote:
As for the age of this alleged group [Amerind], Johanna Nichols pointed out that its diversity was such that it would have taken perhaps 50,000 years to achieve it, a far greater length of time than the current estimate of the earliest entry into the continent.
I think that the reason there is such a large number of acknowledged families in the Americas (amounting to much greater linguistic diversity than would be expected from the recentness of its population) is that most of the accepted relationships between the languages of these families are very shallow (the languages are very similar to each other), and that some genetic relations between families have not been recognized because they were not obvious.
(We’ve discussed Nichols many times over the years.) Have at it!
I believe the entry estimates have since doubled.
https://phys.org/news/2023-12-north-america-people-sea-ice.html
That’s far less than doubling.
I saw the news piece and read the original BTW, but I was sheepish about offering it for LH, because I’m afraid of overstretching my welcome with all the mathematical, physical, and genetics papers offered for a discussion. So sometimes I take a break LOL
I skimmed through the article, and there are several red flags.
First of all, why on earth was this not published in a LINGUISTICS journal? Nichols has published a great deal of respectable articles in linguistics journals, why this sudden change?
Second of all, I am struck by the ahistorical basis of many of the criteria whereby she divides different languages in the Americas into different groups. For example, she focusses on whether languages have an N/M “first/second person” pronominal paradigm, and she is clear: the language must have both an /n/ for first versus /m/ for second person opposition (and no other consonants!) somewhere in the person-marking system (free pronouns, bound pronouns, verb affixes…).
She seems to assume that once a language has such a system it cannot change: she stresses that pronominal paradigm borrowing is vanishingly rare. I agree with her there.
But the notion that the surface distribution of N/M versus non-N/M pronoun paradigms in *present-day* American Indian languages could reveal migration patterns dating back to the Ice Age is, not to put too fine a point on it, preposterously laughable.
Are we to understand that over the millennia *none* of the N/M languages underwent denasalization of consonants, shifting /n/ and /m/ to /d/ and /b/, respectively? That none turned /n/ to /l/ or /r/, /m/ to /v/ or /w/? That none neutralized the /n/ -/m/ distinction word-finally? ANY such change would have removed a language from one group (Languages with /n/ + /m/ first + second person pronouns) to the other (Languages WITHOUT /n/ + /m/ first + second person pronouns).
Do please note that such a change need only have applied to either /n/ or /m/ in a given phonological context: analogical spread could easily have subsequently eliminated the original /n/ or /m/.
If you were to apply this criterion to the languages of Northern Eurasia with the frequent /m/ -/t/ “first singular”-“second singular” pronoun pair, one would have to conclude (assuming Grimm’s law was recognized and thus that -for instance-German “du” is recognized as a cognate of the Indo-European /t/-initial second person pronoun) that Maltese, Finnish, Turkish, Basque, English, Dutch, Greek and varieties of Spanish and Portuguese without “tu” must be the product of a settlement wave distinct from the one which was linguistically ancestral to Western Frisian, Hungarian, (the other) Romance languages, Slavic, German, Saami, Baltic and the Scandinavian languages.
Frankly, I do not believe this article is worth anybody’s time. I would very much like to be proved wrong.
Nichols is a serious scholar worthy of respect, and some of her ideas are of great interest (she’s also a proper descriptive linguist, which gains her many Brownie points in my book.)
But this is yet another paper based on the fundamentally false idea that typology is a reliable guide to historical linguistic relationships.
Nichols has produced numerous works of this kind over the years. When I’ve been in a position to check, I’ve often found her actual data to be unreliable (the most fundamental Greenberg problem), but the main trouble is the unsound overarching methodological assumptions.
Asia-America long-range relationship studies seem particularly prone to this. Michael Fortescue, some years ago, had a very interesting but essentially empty treatment of such issues along very similar lines in his Language Relations across the Bering Strait.
It is essential to real science to be able to accept that the data may just not be there to answer fascinating questions that we would love to have firm answers to. Watering down our standards of proof is not the right response.
On the question of personal pronouns: Western Oti-Volta languages are at least as closely related to one another as the Romance languages. In the second person singular some of the languages have a form going back to proto-Oti-Volta *fʊ, some to *ɲɪ (both of which have cognates even outside Oti-Volta.) Their distribtion within Western Oti-Volta very clearly fails to correlate with any other criteria for subgrouping.
I am (always) surprised by so-called ‘scientists’ ability to to misunderestimate human will to survive and ingenuity in avoiding mortal danger. (This applies equally to the discussion about Polynesians or South AmerInds setting off into the wild blue yonder with no idea what they’ll find.)
If it’s sea ice for as far as the eye can see, and it’s exhausting work carrying your whole possessions across it, why would you even set off? Is there going to be anything to eat along the way? Is there going to be somewhere and something to make a shelter when get to the end of the sea ice? That is, if there’s an end to the sea ice.
Presumably the current is from the melting of the ice, and being fresh water would ‘float’ on the ocean, so would be pushing the boats off shore. Then how would they get back to land? Supposing there is any land beyond the as-far-as-you-can-see sea ice.
But this is yet another paper based on the fundamentally false idea that typology is a reliable guide to historical linguistic relationships.
That’s what I was afraid of!
@AntC – not entirely applicable to the current article, but your comment reminded me of a passage I just read in Felipe Fernandez-Armesto’s “1492”, about the (paradoxical?) timidity of most sail-powered human sea exploration:
“Most would-be explorers have preferred to sail into the wind, presumably because, whether or not they made any discoveries, they wanted to get home. Phoenicians and Greeks, for instance—dwellers at the eastern end of the Mediterranean—explored the length of that sea, working against the prevailing wind. In the Pacific, Polynesians colonized the archipelagoes of the South Seas, from Fiji to Easter Island, by the same method.
Generally, however, fixed wind systems inhibit exploration. Where winds are constant, there is no incentive to try to exploit them as causeways to new worlds. Either they blow into one’s face, in which case seafarers will never get far under sail, or they sing at one’s back—in which case they will prevent venturers from ever returning home. Monsoon systems, by contrast, where prevailing winds are seasonal, encourage long-range seafaring and speculative voyages, because navigators know that the wind, wherever it bears them, will eventually turn and take them home.”
Thanks AG. I think we’ve thrashed all those options to a point beyond death and shipwreck on the ‘Muskogean …’ thread. (Which is what I was thinking of.)
is the explanation for the boats plying the South American West coastal trade never going far off shore. Certainly nothing as foolhardy as Kon-Tiki. (IOW if a boat drifted like that, it would be an accident due to the sudden cessation of the onshore breeze.)
It wouldn’t need to be as extreme as monsoons. The Southern Ocean ‘Trade winds’ are seasonal (north of the Roaring Forties); the North Atlantic has quite varied episodes within the prevailing Westerlies: ideal for exploring little-by-little to Faroes, Iceland, Greenland, Labrador, …
“1492”
Remember Columbus had woefully miscalculated the size of the globe. And every authority told him so. He had no inkling there was a whole vast continent and another vaster ocean to cross to get to the actual Indies. By all precautionary seafaring rules you’ve outlined, he should have sailed into oblivion never to be heard of again.
“time to develop” for Linguists to develop the language families’ typology? Or time for the languages themselves to develop diversity? If the latter, do we really know what timescale it takes for languages to diversify? If the language groups are widely-scattered and quickly lose contact with their linguistic forbears. Which is Etienne’s point in the 2012 thread.
If timescale for Linguists to develop typologies, perhaps we can let Linguists work at it for longer (and avoid wasting time on thinly-evidenced speculation), to reveal closer connections so amalgamate into fewer distinct families? (Something like having to rejig IE to plug Tocharian into the tree.)
And we’ve discussed on the S.Am. indigenous languages threads how many languages there must have been of which we know nothing. Then how can we say so many lineages are distinct? Perhaps the extinct languages would have filled in the gaps and revealed much tighter family connectedness?
able to accept that the data may just not be there to answer fascinating questions
Yeah. What @DE said.
Before recent extinctions the Americas hosted close to half of the known linguistic lineages
Certainly American languages are very diverse, and it is generally accepted that the usually posited time depth for human settlement is not long enough for it to be likely that this all developed in situ. So there is nothing controversial about the idea of multiple movements into America of peoples speaking very different languages already.
But I think this particular claim is a bit of an artefact. Americanists have a praiseworthy splitter tradition: it didn’t take them long to demolish Greenberg’s Amerind. But further work may well yet demonstrate some more long-range relationships rigorously.
In the case of Africa and New Guinea the position is quite different. Greenberg’s African lumpery is still to a great extent taken to be basically correct by most. In particular, his “Nilo-Saharan” is taken to be real, but it actually consists of a great number of families which have not been shown to be related by anything even approaching proper comparative methods. (Don’t get me started on “Niger-Kordofanian.”) In New Guinea, “Trans-New-Guinea” is a similar Frankenstein monster.
@Dmitry Pruss
While I wasn’t being particularly precise, there are some things I left implicit. Most important in context, the 13,000 years in the article I linked already constitute a modestly higher estimate. For example, here’s an article from 2009 stating that “Another American record has been shattered. […] The bone has a serrated edge and, judging by radiocarbon dating and the deposit in which it was found, is 12,300 years old.”
I looked at the paper. It does some things right, but doesn’t convince me of its conclusions.
First, there is the purely linguistic evidence. I’m less pessimistic than DE about using typological characters. They are not completely arbitrary. First, the paper picks typological characters which have some claim to be more stable (i.e. correlated with genetic relationship) than others, based on the survey of Wichmann and Holman (preprint; also this.) Second, the features do weakly corrlate, as the NeighborNet diagram shows. Broadly, this confirm the old impressionistic observation that the Northwest and the Southeast of North America are typologically distinct. However, I don’t see that this more quantitative approach extracts any more details.
Beyond that, the paper wanders into thin air. It takes its kinda-convincing clusters and attaches them to presumed episodes of entry into the Americas, based, as far as I can tell, on their present-day geographical distribution. It falls very short of convincing.
Worse yet, if you are going to go back and reconstruct the full history of the peopling of the Americas, North America won’t do by itself. South American languages are distinct from those of the North, and themselves fall broadly into two typological clusters (see Aikhenvald’s summary here, p. 193). Those imply either more entries, or that the entries were of people speaking typologically diverse languages, which goes against the underlying assumption of the paper.
I haven’t had time to read the paper yet; I suspect I’ll agree with Y. Basically, Nichols is looking for a shortcut that simply may not work, instead of putting in the work that is known to bring results, e.g. Zhivlov’s work on Hokan or m-l’s ongoing reconstruction of Proto-Penutian that could eventually be compared to other large families. However:
Why “none”? Why not “few enough that after 15,000 years we’ll still catch most of the big picture”?
That’s a very rare sound change. I know two examples: Southern-or-so Min and Nitinaht > Ditidaht.
That’s not terribly common either, although the complete lack of nasal consonants in Iroquoian could obviously have happened this way (…the obvious alternative is that it’s why Iroquoian has nasal vowels).
That is common, but would only affect languages with person-marking suffixes; as a typological matter that’s rare in North America as far as I know.
Yes, but again it’s not likely to grossly distort the picture. We may not get the exact “migration patterns dating back to the Ice Age”, but may well end up much closer to that than to the “preposterously laughable” end of the spectrum.
All of that gets much easier if living on the ice edge is the goal – if you’ve got the hang of living off seals, fish and seabirds and of sheltering in snow. Once you’ve managed that (…don’t eat polar-bear liver!), you just spread, and at some point you discover land accidentally. The question is how old this lifestyle really is.
That would surprise me.
The latter. The former probably wouldn’t even be called “develop”.
That’s another work I have yet to read.
I very much agree. For starters, N/M languages continue right into South America; AFAIK they make up about half of the languages there, not much unlike in North America.
underwent denasalization of consonants, shifting /n/ and /m/ to /d/ and /b/
That’s a very rare sound change. I know two examples
Also a change in progress in present-day Korean varieties. See for example the studies here, here, and here. This is one of the many reasons I have found dabbling in Korean challenging but also fun.
Not just Ditidaht. Several languages around the Puget Sound underwent the same sound change, including Makah (Southern Wakashan, like Ditidaht), Lushootseed (Salishan), Quileute (Chimakuan).
Pawnee has no nasals. Its reflex of Proto-Caddoan *n is [ɾ].
(Pawnee and Makah are not autonyms, obvs.)
…don’t eat polar-bear liver!
I recall reading somewhere that Inuit have in the past done this on purpose, because the Vitamin A poisoning gives you quite a high. Brain-expanding, I suppose, if not actually mind-expanding …
Peels the skin right off your scrotum.
>On the question of personal pronouns: Western Oti-Volta languages are at least as closely related to one another as the Romance languages. In the second person singular some of the languages have a form going back to proto-Oti-Volta *fʊ, some to *ɲɪ (both of which have cognates even outside Oti-Volta.) Their distribution within Western Oti-Volta very clearly fails to correlate with any other criteria for subgrouping.
We just discussed the range of Spanish second person singular pronouns in another thread. Is second person singular mire subject to respect taboos* and other factors that result in mutability or is that just a coincidence between Romance and Western Oti-Volta?
* not quite the right word maybe?
South American languages are distinct from those of the North, and themselves fall broadly into two typological clusters
So did the S.Ams. initially settle in the North, then get pushed South by later migrations? Or did they arrive later and spread along the Western seaboard by boat, bypassing N.Ams? (wikip seems to be hedging its bets)
I’m finding it hard to think there would be no mixing of languages over that sort of distance/timescale. Are there no elements of S.Am. languages to be found in N.Am. languages or v.v.? Is there any sort of pidgin/creole?
Or is it all so remote no detectable traces remain/can be securely identified? In which case doesn’t that also apply to Nichols’ claims?
The Clovis-era Anzick child, found in Montana, is genetically akin to South American Indians, not Northern ones.
Some South Americans may have come by sea, others by land; but when you get to the isthmus in Panama, there isn’t that much difference, is there?
As far as I know, no one ever suggested that any North American language has anything to do with a South American language, except maybe Yurumanguí (Colombian, poorly documented)-Hokan, which no one believes; plus some languages suggested as recent back-migrations (Timucua, not a popular idea either).
Languages get wiped out by later migrations, most commonly without leaving a trace. I imagine that is the case for the ancestors of South American languages on their way down.
Is second person singular mire subject to respect taboos* and other factors that result in mutability or is that just a coincidence between Romance and Western Oti-Volta?
There is a possibility that the you-singular *ɲɪ in WOV is indeed a repurposed plural. The you-singular form *fʊ is probably the original; there is a difficulty with explaining the *f initial, but it may be the regular development of *k͡p in clitics, which would open the way to comparison with the Bantu you-singular object prefix *kʊ and the like.
It’s not just second person forms though. For example, take these first-person plurals:
Gyeli (Bantu) bí
Swahili (Bantu) tu- (verb subject prefix)
Kulango (“Gur”) bɩ̀
Toende Kusaal (“Gur”) tʋn
These languages are undoubtedly all related (even I, smug in my splitterism, agree), and both the *tʊ and *bɪ forms surface in yet other Volta-Congo branches.
The problem is that Gyeli and Swahili are closely related to one another, and Kulango and Toende Kusaal, though not close, are both part of Greenberg’s “Gur.”
[In the case of Kulango, it presumably got called “Gur” because it has noun-class suffixes rather than prefixes, though I myself think that this is unlikely to reflect a real primary genetic division within Volta-Congo: it’s been given an unwarranted significance because of the way attempted Volta-Congo reconstruction has too often assumed that the protolanguage must have been very like proto-Bantu. However, Kulango is spoken far the the west, on the other side of Oti-Volta (real “Gur”) from Bantu. ]
The first-plural subject pronoun in Actual Spoken French, of course, is /ɔ̃/, unrelated to the forms found in any other Indo-European language, but ultimately derived from a word meaning “earth”, because why not?
The first-plural subject pronoun in Actual Spoken French, of course, is /ɔ̃/
Which, of course, takes singular verbs. La logique, quoi!
Which, of course, takes singular verbs
Who’s to say it does? In Actual French, the great majority of verbs only inflect* for number and person in the second person plural. (Some dialects, not even that, I gather.)
* Inflect by suffixing, at any rate. You can, of course, make an excellent argument that, in fact, Actual French verbs inflect for subject and object using agreement prefixes. Bantu influence, I would think. Must be one of those substrates we keep hearing about.
Sorry, by whom? IME the full proposal gets these days approximately the same highly skeptical response as Altaic (though including, sure, being taken as real e.g. by a variety of those anthropologists whose knowledge of linguistic classification drags decades behind state of the art).
But I guess there might be some terminological vaguity involved where you’re counting anything other than “they’re ~30 separate lineages”, while I would argue that, for starters, a family proposal that does not include at least one of Saharan and Songhay cannot be called “Nilo-Saharan“, and should be pared back to something like “Macro-Sudanic” or maybe indeed Greenberg’s “Chari-Nile”. (Which I find gets strangely little attention considering that he explicitly claims it to be a better-supported unit than the full NS.)
“Niger-Kordofanian” seems basically obsolete, I’ve never run into any support for it out in the wild. Even megatrees from Blench that keep including “Kordofanian” as an early branch do not use the term. More common might be versions that move e.g. Heiban–Talodi under the Atlantic-Congo perimeter instead (as one should, if following its typological definition consistently).
More similar, I would think, to maximalist versions of Niger-Congo or Sino-Tibetan: fairly clearly has a large family to it but probably has had various nearby isolates / smaller families unwarrantedly lumped into it too. Core TNG also already has reconstruction work for it going and all.
Who’s to say it does?
Anyone who says “on va” but “ils vont.”
Sorry, by whom?
Dimmendaal (no flake, and no uberlumper in other domains, though wrong in this case.)
“Niger-Kordofanian” seems basically obsolete
I don’t think it’s gone away as a concept: it’s rather that “Niger-Congo” is used instead. Bits of Kordofanian have been excluded as time has gone by, but the general idea of including (some of) Kordofanian in “Niger-Congo” has not been abandoned. A veritable poster child for thinking that typological resemblance alone is enough to demonstrate genetic affinity, in fact.
Core TNG also already has reconstruction work for it going and all.
Sounds like you know a lot more about it than I do. It sounds very interesting.
Anyone who says “on va” but “ils vont”
Yes, there are still some common irregular verbs that distinguish three persons in the plural (syncretism between various singular and plural forms in the verb is, of course, extensive in ASF.) A largish (though unproductive) group are verbs like
ʒfini
tfini
ifini
ɔ̃fini
vfinise
ifinis
But really: why should /va/ be considered any less (or more) plural than /alɔ̃/?
The fact that the one of the singular forms is homophonous is surely not enough to prove this in French (of all languages!)
Dimmendaal excludes Songhay but makes a reasonably promising case for Saharan being related to Eastern Sudanic, and in particular Nilotic (that being the branch he’s most familiar with). If that pans out, Nilo-Saharan might remain a defensible label.
Another way of looking at it:
Supposing that the French, with their near-rosbifesque gift for creative spelling, had decided to write on aimh, on parlh, on finih, on vah in these cases (the h, is, of course, silent, as always; here it replaces e but is added after other vowel symbols.) This is (of course) entirely in line with their existing habits like writing il aime but ils aiment, and thus in accord with the highest principles of French orthography.
Who would then have the audacity to claim that vah was singular? An obvious error!
David Eddyshaw: Actually, colloquial French verbs are substantially more complicated than that. It turns out that the number of verbs with distinctive third singular and third plural forms of the present indicative…is growing. Yes indeed.
Verbs of the /ifini/ -/ifinis/ class are not alone: a majority of verbs whose infinitives do not end in /e/ likewise oppose -in the standard!- a singular and plural form via an added consonant on the latter form (/idi/ versus /idiz/, /idɔʁ/ versus /idɔʁm/): as a result verbs with a third person singular ending in a vowel, no matter what their verb class is, are liable to have a plural form created through addition of a consonant: Standard French has “il joue, ils jouent”, both /iʒu/, but in colloquial Montreal French an opposition exists between singular /iʒu/ and plural /iʒuz/.
The latter stem /ʒuz/-is also used in the present subjunctive (all persons), highlighting the indicative-subjunctive distinction. A similar singular-plural distinction is being created in the subjunctive of some verbs too: Standard French “qu’il soit/qu’ils soient”, both /ki(l)swa/, is in colloquial Montreal French singular /kiswa/, plural /kiswaj/.
The last of these forms is attested in colloquial French in Europe: the tendency is the same, but interestingly the choice of plural-marking final consonant sometimes differs on each side of the Atlantic. For example, l’Académie française actually called out/condemned a plural form /krwav/, plural of /krwa/, and in fairness clearly explained why French speakers make this mistake, but reminded users that singular and plural alike in the standard should be /krwa/-
https://www.academie-francaise.fr/croivent-pour-croient
-However, on this side of the Atlantic, the only plural form of /krwa/ I have ever heard is /krwaj/
(Ah, I can see it already: the year is 4024, and some ambitious young linguist is exploring the history of neo-French languages and has discovered that some morphological divergences can be dated back to the late twentieth/early twenty-first centuries).
(The similarity to the morphology of feminine adjectives is no accident, I suspect: indeed, in one attested variety of non-standard French a third plural present subjunctive /vɔn/ is attested, which stands to the third plural present indicative form /vɔ̃/ in much the same way that the feminine form /bɔn/ stands to the masculine form /bɔ̃/).
In short: the old verbal suffixes are not quite dead yet: they are holding their own and spreading even as the new prefixed morphology is evolving.
I used to find the morphology complexity of Basque or Hungarian frightfully exotic, but am beginning to realize that from the vantage point of a non-Romance speaker, Colloquial French verb morphology is quite comparable.
AANNDD…returning to the original topic of this thread, I very much doubt that a linguist comparing (say) Colloquial French, Colloquial English and Colloquial Welsh verb morphology would be able (on the basis of this comparison) to deduce much about the settlement history of Indo-European in Western Europe (assuming they even managed to see that English, French and Welsh make up a family which does not include Hungarian, Basque or Turkish. Which I doubt).
Wow, that stuff on Colloquial French verb morphology is mind-boggling!
The Higher Purpose of French is to make Latin look ridiculously simple.
It is good that this great project is continuing unabated.
Is there a good descriptive grammar of contemporary Spoken Colloquial French, free of any allegiance to the standard language? Bearing in mind that some of it is expected to quickly become obsolete.
It’s in no way a grammar, but I was just reading Knud Lambrecht’s Topic, Antitopic and Verb Agreement in Non-Standard French (by “Non-Standard French” he means Actual Spoken French), which is well worth looking at. It’s well-written and logical and also has the great merit of being short. It’s pretty old though (1981.)
I was actually reading it because of an interest in focus and topicalisation, on which Lambrecht is one of the notorious experts. So ignorant am I that I hadn’t in fact twigged that he is a L1 French speaker. I suppose the name “Knud” should have tipped me off …
It would appear that the creation of a third person plural present indicative form by addition of a final consonant, with the new “long” stem also used for the subjunctive, is a living tendency, if l’Académie française is to be believed…
https://www.academie-francaise.fr/la-prononciation-de-voient
Both “faulty” sentences with /vwaj/ would be unremarkable in the streets of Montreal (if realized with the right accent, that is…)
Having just discussed this with some colleagues, it would appear that in Québec the trend affects more verbs than in France, but in France the verbs which do have a non-standard plural form are (seemingly) used more consistently than in Québec. In Québec /z/ and /j/ seem to be the only consonants that may be added to the singular verb, whereas in France the choice appears to be between /v/ and /j/.
And the trend may affect more verbs in Québec than in France because of a peculiarity of Québec French morphophonemics: “ils” and “elles” do NOT trigger /z/ liaison: thus, “il aime” and “ils aiment” are homophonous (/jɛm/, versus standard singular “i(l)ɛm”, plural “i(l)zɛm” or European non-standard singular /jɛm/, plural /zɛm/).
So, a contrastive pair (leaving aside the specific realization of certain vowels, a matter of phonetics more than phonology mostly):
“Ils ne les croient pas et ils n’ont jamais cru en cette histoire” (/ilnəlekrʁwapa e ilnɔ̃ʒamɛkʁy ɑ̃sɛtistwaʁ/, in the standard) would be /ilekrʁav pa e izɔ̃ʒamɛkʁy ɑ̃sɛtistwaʁ/ in non-standard European French and /jekrʁaj pa e jɔ̃ʒamɛkʁy ɑ̃stistwaʁ/ in non-standard Québec French (with the typically Québec fusion of “ils” + “les” as /je/ and the demonstrative /st/- in pre-vocalic poisition).
The notion that this change could be a case of Sapirian “drift” (i.e. that because of the homophony of third singular and plural forms for many verbs, both varieties made use of the same “repair” strategy, i.e. analogically extending a stem -final consonant addition – already found with a large number of verbs) is made all the more likely to my mind by the fact that one of the most persistent non-standard features of Acadian French is…the third person plural marker “-ont”, so that even in the speech of educated Acadians you will have an opposition between /ikʁa/ versus plural /ikʁajɔ̃/. Francophones everywhere seem to feel the need to maintain or create a singular-plural distinction involving verbs marked in the third person, and will use whatever morphological material is available to do so.
In answer to Y: I do not know of any good descriptive grammar of Colloquial French in ANY language, unfortunately.
For the record, I’ve encountered croivent and even perpetrated it myself…
That’s trivial. You need the quadrivium… I mean, there is only one mathematically possible fully resolved unrooted tree with three terminal taxa. (With four you get two.)
The Uralists, faced with an almost entirely unwritten family at the time, figured out pretty early in the 19th century that you have to document and compare dozens of dialects without even worrying about which ones, if any, should be classified as belonging to the same language.
Hungarian and even Turkish I can perhaps grant you, but Basque is a few steps too far. It… just… smacks you upside the head with how different it is.
@ “any good descriptive grammar of Colloquial French in ANY language, unfortunately.”
Neither of these publications is recent but they are informative:
Le Français tel qu’on le parle (E. Billaudeau, 1963)
Le Français sans fard (A. Martinet, 1974)
Francophones everywhere seem to feel the need to maintain or create a singular-plural distinction involving verbs marked in the third person, and will use whatever morphological material is available to do so
That is extremely interesting.
It reminds me a bit of Welsh noun plurals, where there has over the centuries been a grand kidnapping of other nouns’ inherited plural endings by nouns that hadn’t planned ahead for what would happen when the Romans went off with all the Brits’ final syllables when they finally evacuated.
Kusaal does this regularly in nominals where the historically expected singular and plural forms would have ended up identical because of the loss of final short vowels.
In fact, this loss is not unconditional, though it affects words in the great majority of contexts where they appear, including in citation. There are still a few nouns which show the etymologically “correct” ending exclusively in environments where the final vowel loss is regularly inhibited.
Thus bi’em “enemy” (< *bi'emʊ) makes the plural bi’emnam: structurally, this is actually a noun compound with a bound word meaning “plurality”: “a plurality of enemies.” But at the end of negated phrases, where the final vowel loss is suppressed, you get e,g.
Ba ka’ bi’emma.
“They are not enemies.”
with bi’emma from the expected plural *bi’emba: in other contexts this would have become bi’em, identical with the singular form.
>I suppose the name “Knud” should have tipped me off …
Well, he does pronounce it /ɔ̃/
No, that’s the Danish pronunciation.
(Of everything.)
Bolding mine: no not really, the intentional “dozens of dialects” phase only comes in from circa 1890 on. The only unwritten languages documented in multidialectal detail before that were the ones like Sami, Mansi, Khanty or Selkup that are by any sensible metric really subfamilies of several mutually unintelligible varieties, and where it’s hence required already for making any damn descriptive sense of things either.
Right, but the idea that there’s a split between “Kordofanian” as a small outlier versus “Niger-Congo” as its much larger sister does seem to have been. This is what I would see as defining the notion of “Niger-Kordofanian”, and it hence already disappears as soon as “Kordofanian” is no longer held to be a single entity. One could in principle propose also entities like “Niger-Katla” splitting into Katla + “Niger-Talodi”, keeping the topology but moving the root of the family to be within “Kordofanian”, but that does not seem to be happening either.
Dimmendaal’s approach to Nilo-Saharan slightly confuses me — he has good-looking work about morphological correspondences, but then faced with the seemingly heavily phonologically reduced Central Sudanic, instead of looking for that many traces he just declares it outside the core that would’ve innovated most of the morphology. I guess it’s one way to keep prioritizing typology over comparison (i.e. compare only things that remain typologically equal)?
Honestly not that much; Wikipedia has a pretty decent article about the situation.
Ah. The “Hamito-Semitic” misleading bipartite name problem. I get you.
I agree that if Talodi etc are related to Volta-Congo at all, it’s as one of several peripheral (from the V-C standpoint) groups with no close genetic links amongst them pairwise (same with the various parts of “Atlantic.”) Though at that level, you’re really talking about something like “Altaic” rather than e.g. Afroasiatic (much less, Indo-European or Uralic.)
If I’m getting the correct gist of the situation, one might well call Atlantic-Congo kind of an Ural-Altaic of Africa; with the exception that no competing “Indo-Uralic” or “Uralo-Siberian” or “Austro-Japonic” etc. theories have yet emerged to outright contradict it. And I must say a proliferation of prefixal & number-specific noun classes looks at least, to field a Nicholsism, an awful lot more individual-identifying to me than the fairly worldwide-common SOV-CVC-agglutinating Ural-Altaic typology; even if it is most probably still no proof in the absense of sufficient actual material identities in the system.
For that matter, if there really are stragglers in Kordofan (or whereever else), then they must surely go thru basically the same process that e.g. Albanian has, before they can become established members of the family: first reconstruct e.g. Proto-Atlantic-Congo from other evidence, and then show that some core chunk of e.g. Proto-Talodi can be derived from it. (I may not believe in the notion of the comparative system “stopping working” after some specific arbitrary time-depth, but I do think it stops working if the timesteps between reconstruction nodes are allowed to grow arbitrarily long.)
…and getting finally to the OP though, yeah the new paper strikes me as rather make-belief as it does not seem to even establish a particularly unique typological profile for any of her statistically identified areas. But the starting anthropological idea of linguistic diversity surviving already from a 24kya entry event seems worth of further development at least.
J Pystynen: On your last point, I disagree. I am unaware of ANY part of the world where present-day typology matches/confirms/indicates anything known (thanks to written records) to have existed at a much shallower point in the past than what Nichols, in her article, claims to uncover.
What is there in present-day Turkish (standard or dialectal) which would allow a scholar working on them to deduce anything about the structure of Anatolian languages? Or, indeed, their very existence? What is there in present-day Egyptian Arabic (standard or dialectal) which would allow a scholar working on it to deduce anything about the structure of Old Egyptian? Or, indeed, its very existence? What is there in present-day Iraqi + Kuwaiti Arabic (standard or dialectal) which would allow a scholar working on them to deduce anything about the structure of Sumerian? Or, indeed, its very existence? What is there in the Italian dialects of Tuscany today which would allow a scholar working on them to deduce anything about the structure of Etruscan? Or, indeed, its very existence? What is there in present-day Persian (standard or dialectal) which would allow a scholar working on them to deduce anything about the structure of Elamite? Or, indeed, its very existence?
I could go on, but instead I will ask: can any hatter think of even a single clear counter-example to my claim? Is there any instance anywhere on the planet where the typology of a present-day language/dialect cluster matches something about the typology of an earlier (non-ancestral!) language (for which we have written records) spoken over 2000 years ago?
If not, and considering how quickly language change operates (Trying to imagine what linguists would “deduce” about Indo-European if, for instance (colloquial) French, English, Welsh and Albanian* were the only languages compared would make for an amusing exercise…) then I am afraid that my evaluation above was true: Nichols’ article really is a waste of time.
*Considering the huge number of extinct and unrecorded languages in the Americas, any kind of comparative reconstruction cannot be compared to Uralic or indeed to any language family where abundant data, including dialect variation in various languages, is available: instead it must be compared …to comparative Indo-European involving a mere handful of present-day languages, just as in the example above.
And I must say a proliferation of prefixal & number-specific noun classes looks at least, to field a Nicholsism, an awful lot more individual-identifying to me
The Greenbergian Niger-Congo proposal does indeed depend pretty heavily on the notion that the number-specific noun class system is so unusual typologically that common inheritance is the only plausible explanation for the resemblances. (Which didn’t stop him from adding groups like Mande and Dogon with no trace that they ever had such a thing, but there you go. Greenberg was Greenberg.)
“Prefixal” number-specific noun classes, of course, does not apply to “Gur”, a group (meaning by this Oti-Volta, Grusi, Koromfe and Baatonum, at least) where even I think that the relationship to e.g. Bantu is beyond reasonable doubt. Most would-be reconstructors explain this by supposing that Oti-Volta etc have lost the “original” class prefixes, and the suffixes are secondarily developed from concordant demonstratives or the like, but there is zero evidence for these alleged lost prefixes: in fact, where Oti-Volta languages have class prefixes, it is abundantly clear that it is they which have developed from clitic demonstrative elements.
This cycle of replacement of prefixes by suffixes also, to be at all plausible, given that it is so exceptionless throughout “Gur” (and Adamawa, and maybe Kru) would presumably need to be ascribed to a Gur-Adamawa protolanguage: but in fact, even “Gur” alone in the traditional sense is a real baggy-monster group, and it is (believe me) a lot easier to trace real genetic relationships between Oti-Volta and Bantu than between e.g. Oti-Volta and Kulango or Tiefo. I very much doubt that there ever was a “proto-Gur-Adamawa” at the back of all these suffixing languages.
Greenberg himself seems to have subscribed to the view that the class affixes varied in their position relative to their heads from the beginning. This seems vastly more plausible to me.
But of course, it you suppose that they were clitics (pro- or en- by branch) from the beginning, even when you can establish actual sound correspondences, you’re not actually reconstructing the morphology of the protolanguage, but lexical elements, or rather third-person pronominals. (And I’m writing this in a language known to have borrowed all its plural third-person pronouns,)
The “number” thing is also not nearly as persuasive as is sometimes suggested. Atlantic languages typically have substantially more singular than plural classes: there just isn’t anything like the Bantu “genders”, i.e. characteristic fixed sg/pl pairings of affixes. (Whereas Oti-Volta can actually match several of these pairings with those of proto-Bantu.)
The only halfway plausible phonological match between individual class affixes between Volta-Congo and “Atlantic” as far as I can see is the human-plural *be/*ba (there in Dogon too, but not in any of Kordofanian AFAIK, where the lookalikes are typically human singular.)
This doesn’t seem a lot to base wide-ranging theories of relatedness on, especially in the absence of any really compelling lexical evidence (especially as far a Kordofanian is concerned.)
(The usual arguments for Broad Niger-Congo from alleged resemblances in verb derivational morphology are even shakier.)
I don’t think real progress is possible without a proper reconstruction of proto-Volta-Congo.
(There is some exciting stuff going in with the various parts of Atlantic, though.)
David Eddyshaw:
Your statement-
“This cycle of replacement of prefixes by suffixes also, to be at all plausible, given that it is so exceptionless throughout “Gur” (and Adamawa, and maybe Kru) would presumably need to be ascribed to a Gur-Adamawa protolanguage”
-is one I disagree with: one could with equal plausibility claim it to be a feature of a prehistoric “Gur-Adamawa” language area, within which originally typologically different albeit genetically related languages would over time have become typologically more uniform as a group. Which need not imply that the suffixes might not indeed be genuine “Niger-Congo” (?) cognates (whatever their original status -prefixes, clitics? -in the proto-language might have been).
An analogy: the fact that the Modern Greek nominative-accusative neuter singular definite article /to/ is a cognate of German /das/ (same meaning), both going back to Proto-Indo-European */tod/, needn’t mean or imply that their similarity in function (definite article in both instances) is to be traced back to Indo-European. And in an alternate universe where Western Europe is being explored linguistically with no reference to written documentation, the ubiquity of definite articles in Western Europe -including in non-Indo-European languages such as Hungarian-and their near non-existence in Indo-European languages spoken outside of Western Europe, might suggest as much to intelligent scholars.
Yes, fair comment. Although really all I meant to maintain was that this can’t be a matter of reconstructing proto-Volta-Congo morphology (which would be more of a clincher of genetic relationship than just establishing regular lexical correspondences.) If it was all about inherited morphology, then you would need to project the change back to a common protolanguage.
In fact, the relevant morphemes (whether affixes or clitics) clearly are (sometimes) cognate between Oti-Volta and Bantu: the most impressive may be the sg/pl pairing which turns up as *-rɪ̀/-ɣà in proto-Oti-Volta and *dɪ-/ma- in proto-Bantu (the pl *m- is actually very plausibly explicable as a secondary development in pB, and there is a lot of evidence from other Volta-Congo branches to back this up.) This is such an arbitrary pairing that I think its value for suggesting a genetic relationship is quite high, even though I don’t attribute it to pVC morphology.
I actually do suspect that the Gur suffixing of class affixes is an areal feature.
The “Gur” area is very largely one where noun-phrases have head-final order, and the area where the class affixes are prefixes is very largely one where NPs are head-initial (though there are exceptions in both directions), so you could actually attribute the difference to this very general typological feature, which could very well have been areal (though this explanation assumes that the class markers were indeed clitics rather than affixes originally.)
There is some evidence that pre-proto-Oti-Volta may actually have been SOV, but unfortunately for neat hypothesising there is nothing much to support the idea that SOV was formerly usual throughout the whole “Gur-Adamawa” range. Very few of the extant languages are (unless you count the Senoufo languages, which have very clearly fallen under the bad influence of Mande.)
Is there any instance anywhere on the planet where the typology of a present-day language/dialect cluster matches something about the typology of an earlier (non-ancestral!) language (for which we have written records) spoken over 2000 years ago?
Well, yes. Modern Palestinian Arabic is not hugely typologically different from Biblical Hebrew; same basic word order (bar maybe a bit more SVO than VSO), verb-subject agreement through suffixation in the perfective and prefixation in the imperfective, suffixed object pronouns, prefixed definite articles, double object constructions… Old South Arabian is a bit less similar to modern Yemeni Arabic, but still quite typologically similar. And, while we don’t have records worth mentioning for Scythian, comparison to other Indo-Iranian languages of the period suggests that it can’t have been that typologically different from modern Ukrainian or Russian; Slavic is pretty conservative in that regard. Of course, in all such cases the similarity is due to more remote common ancestry, not to substratum effects; in that respect, spread zones could plausibly be more typologically stable on average than accretion zones.
Actually, here’s an example. It’s a single feature, though, and the written records of the earlier language (…family) are very sparse (and not local).
It’s even worse: das and that are cognates, and the latter’s ancestor was the neuter singular definite article in Old English; but the extension from demonstrative pronoun to article is parallel. In early OHG, articles were about as optional as in Homeric Greek, AFAIK; it seems to be generally acknowledged that Proto-West-Germanic lacked articles altogether.
*galaxy brain*
In fairness to Nichols, what examples of written languages native to accretion zones do we even have for 2000 years ago? Urartian, maybe Hurrian, even more doubtfully Elamite; that’s about it really. Not a huge sample to draw any conclusions from, and for all I know there could be Urartian substratum features in Armenian or Zazaki or something.
Ergativity seems to have a higher-than-average frequency in the Armenian highlands and borderlands between contemporary Iran and Iraq (depending on how generous you are): Kurdish languages, NENA languages, Hurro-Urartian, Sumerian, Kartvelian languages?.
This is all very impressionistic, and it’s not immediately obvious how one could turn this into a testable hypothesis…
Can ergativity in fact be spread by language shift or contact?
I think the problem with such observations is that the current distributions are generally strongly areal, so attributing them to common substrates seems superfluous, unless you actually have historical evidence that the languages in question
(a) always shared such features from the earliest times that they were extant in the area in question
and
(b) have unequivocal genetic relatives elsewhere that don’t share the feature.
Otherwise you could just as well be just seeing more-recent Sprachbund effects, or just looking at pure coincidence (a thing that it is usually unwise to discount, and that fancy statistics will not save you from if you’ve made the wrong assumptions before you began.)
In fact, without good evidence for the prehistory of the concerned languages, ideally in written records, but possibly from rigorous internal reconstruction or comparative work, you just have no way of telling.
The Nichols paper would be a lot more persuasive if the features noted did not cluster geographically. As it is, I don’t think they really establish anything. “Neighbouring languages tend to resemble one another more than can be accounted for be chance.” Yup. Yes, they do that.
[This discussion actually has given me some helpful ideas about “Gur”, in the traditional Greenbergish sense of “languages that have class suffixes instead of prefixes, and (by the way) are all more or less neighbours.” There is a “Central Gur” core where there is evidence for an actual genetic branch, basically Oti-Volta/Grusi/Koromfe/Miyobe/Baatonum, but the more westerly parts of “Gur” seem in reality to have no close genetic links to Central Gur. This would make perfect sense if the suffix-last thing was areal, and if the class markers were clitics and not affixes in proto-Volta-Congo, that really is not hard to imagine at all. Whether this is a substrate thing or a more recent convergence phenomenon is probably not now answerable, though. It might even not be a matter of just one or the other.]
Can ergativity in fact be spread by language shift or contact?
I can’t see why it shouldn’t be. (Niger-Congo languages tend to be boringly nominative-accusative, but there are definitely areal patterns of alignment in Eastern Africa, including cross-linguistically unusual things like marked-nominative patterns. Lameen will know more about that, I suspect.)
Well, is there any documentation of Basque or Mayan speakers shifting to ergative Spanish?
@Y”Well, is there any documentation of Basque or Mayan…”
Several contemporary students of the Spanish of speakers of Maya are mentioned here: https://www.bbc.com/mundo/noticias-62572739. Ergative verbs are not mentioned. Maybe the Diccionario del español yucateco, written by two of them, has some information.
This article on the Spanish of speakers of Basque mentions ergativity a few times:
https://www.csub.edu/~tfernandez_ulloa/influencias%20morfosintacticas.pdf
Ergativity definitely can spread through contact, as seen in Neo-Aramaic (for which we do have records going back 2000+ years) under Kurdish and broader Iranian influence.
Y: Ergativity, interestingly, did NOT spread from Basque to any neighboring Romance variety, despite the expansion of Romance at the expense of Basque over the past couple of millennia or so: more broadly, Basque and local Romance varieties remained -and still remain- sharply unlike another, typologically.
Ergativity can also be lost, and I have sometimes wondered whether, in Indo-Iranian (All modern Indo-Iranian languages were once ergative), contact played any role in its loss, as the languages which lost ergativity seem to be in close contact with a non-Indo-Iranian/non-ergative language (compare ergative Hindi -surrounded by other genetically related ergative languages-and Pashto (ditto, leaving aside Persian), on the one hand, and non-ergative Bengali and Persian, which have both been in close contact with genetically unrelated, non-ergative languages.
Didn’t I read here in a thread about “living Sanskrit” that modern practitioners often use the ergative pattern of Hindi, essentially an analytic passive = the conjugated copula + a participle, instead of the cleverly irregular conjugations of the classical language? As if ergativity is spreading backwards in time.
This was already common in Classical Sanskrit, i.e. Sanskrit in the strict sense of the term, which was already a “dead” language in the sense of being nobody’s mother tongue.
The SAE feeling that ergative alignment is basically a sort of weird archaism* is not borne out by the facts.
*This is actually stated outright in Edzard’s (extremely disappointing) Sumerian grammar, coupled with the astonishing notion that this is in some way connected with the fact that the language had a separate register for women’s speech (interpreting Emesal as such), which he imagined to be an archaism too. Women, eh? They’re just so ergative, bless ’em!
Well, in many traditional cultures they are expected to do most of the work, while the men are at cockfights or praying.
This may be the key.
When praying, men use the pergative to maintain purity.
I don’t do no womanly ergativity, Me stay firmly in nominative-accusative alignment.
Spoken by a true man!
underwent denasalization of consonants, shifting /n/ and /m/ to /d/ and /b/
That’s a very rare sound change. I know two examples
Also a change in progress in present-day Korean varieties. See for example the studies…
When I opened YouTube just now tonight, it suggested watching this video on Korean denasalized nasals, intended for your average intermediate student of Korean and K-drama fan. (No doubt Google is using my search data from Chrome for YouTube.) I thought I would share this video here on LH for those who don’t want to read a whole phonetics paper.
Convincing! Interesting, though, that it’s limited to word-initial position (explaining why there isn’t any [ŋ] > [g] mentioned); that must be because word-medial voiced plosives already exist.
Japanese Kan-on (apparently based on the actual Chinese pronunciation of Chang’an,)
Not a denasalising sound change exactly, but:
Several Oti-Volta languages of the Atakora in Benin have lost the initial contrasts *m/*b and *n/*d, with /m n/ appearing before nasal vowels and /b d/ before oral vowels (but initial *g has been devoiced instead, and initial *ŋ has just been dropped.)
So you get e.g. Toende Kusaal dãam “beer” but Nõotre náam.
This is part of an areal conspiracy to get rid of voiced plosives and affricates rather than abolishing nasals, though. (Atakora /d/ is usually actually realised as [l] or [ɹ].)
Eastern Tucanoan languages only have contrasts between voiced plosives and nasals as a manifestation of morpheme-level suprasegmental nasalising prosodies. You don’t actually need /m n/ phonemes. (I think there’s a fair bit of that in Amazonia.)
Go-on, rather; this Go corresponds to Mandarin Wú and starts with [ŋ] in the original.
However, that’s a different phenomenon: it’s [ŋ] being borrowed as the [g ~ ŋg ~ ŋ] phoneme, which doesn’t occur word-initially in native words and is reconstructed, perhaps a bit too mechanically, as Proto-Japonic *nk except where it comes from morphologized intervocalic voicing (-kana ~ -gana).
Pirahã comes to mind, but I didn’t know it was representative of anything…!
No, I mean Kan-on. Plentiful examples of entirely regular m -> b and n -> d.
As I say, it is usually attributed to a (partial) denasalisation already in the Chang’an dialect the Japanese scholars were imitating.
Nasal harmony is in fact common in Amazonia. Even Guarani has it, though not to the extent of having no contrast of nasals with voiced stops, as in Tucano.
Interesting that I missed this about Kan-on. It’s right there in the first table, and the dan/nan example is nán in Mandarin.