An interesting Facebook post from Claire Bowern:
I promise I will stop posting about the Dixon book shortly and go back to #chookbook updates, fieldwork book edits and complaints about email, but I was thinking this morning about what Australianists do and don’t seem to agree on, particularly the linguists. (“we disagree” here means “different people think different things, not “I think one thing and other people think something else”, just in case that’s not clear).
I’m pretty sure almost all of us agree that Pama-Nyungan is a language family, in the same way that Austronesian or Indo-European are language families. We don’t all agree on the composition of the family or its internal structure. We have radically different estimates of how old the family is (4-15kya!). We pretty much all agree that language change works the same way in Australia that it does elsewhere, but I’m pretty sure we don’t agree on how language change works and what processes are most important. Pretty much all of us are puzzled by the relative lack of sound change in Australia, but we don’t agree on what that implies and how to deal with it. We’ve all done fieldwork and understand the complexities of multilingual and multilectal communities and what that means for change, but we disagree about how that might scale up to the Holocene. We agree that all sorts of different data are important for reconstructing history, but we use different material in practice and place different weights on it.
(After the Routledge 2nd edition I said I would never edit another book ever ever again, but now I’m wondering if something that explores these questions from all different angles by people who disagree but can actually talk to each other might be worth doing.)
I’m not sure what “the Dixon book” is, but here’s her previous post about it:
One of the complaints about me in the new Dixon book is about challenging the number of “250 languages” (which appears to have struck a nerve because he reiterates this figure many times). All previous 20th century classifications have about 250 languages. That’s right. But they don’t have the same 250! They each miss different languages. So when you total them up, you get something like 380. Then you take into account more recent work in the Top End and another group of languages turn up. (And include palawa languages.) Then there’s a few known from name and report of intelligibility only, which I included but many don’t. Then on top of that you have the points where people differ about whether two varieties are “same” or “different”, which gives somewhere between about 420 and 480 languages. Summary: 250 is wrong. Definitely higher. How much higher depends on all the things linguists usually argue about. (All this is in Chapter 7 of OGAL but if you just kvetch at the classification at the front and admire the 80-odd footnotes about sources then maybe you might miss it.)
This extremely laudatory review contends that what one might have thought was simply a 2025 revised edition of an influential work from much earlier in Dixon’s career “is not merely an update but a monumental event in linguistic scholarship.”
https://linguistlist.org/issues/36/3476/
Whether it’s a language-specific grammar written in a broad enough style to include negative comments about the contentions of other scholars (who should no doubt feel honored to be deprecated rather than simply ignored) is not clear to me from the review, but maybe?
I am myself not up to speed on what advances in the analysis of Dyirbal grammar may have been made in the 40-odd years since I first became acquainted with Dixon’s then-views, so maybe I should get myself a copy.
I’m not sure what “the Dixon book” is
It’s Australian Comparative Linguistics: An Evaluation (2026). I have not seen it yet. Publisher’s page here, for LH readers who are interested.
May I recommend Claire Bowern, ed., The Oxford Guide to Australian Languages (here) to LH readers who are unfamiliar with Australian languages and wish to learn more about them? (This is the OGAL in the Facebook post that was quoted.)
Claire Bowern:
I found the Gambay first languages map with 780 indigenous Australian languages, apparently following OGAL.
It’s Australian Comparative Linguistics: An Evaluation (2026).
Thanks!
Dixon’s position, more or less, is that eons of diffusional changes between the Australian languages have made it pretty much impossible to establish higher-level language groups rigorously in Australia (he doesn’t deny that lower-level groups can be identified.)
This is a somewhat isolated position; (many) other Australianists have written entire books taking issue with it.
He does make quite a few perfectly valid points, though I think he goes somewhat overboard with them. (This is the Dixon who doesn’t believe in Afroasiatic, after all.)
I haven’t seen this 2026 book yet. Dunno if he’s moderated his views any. Sounds like he hasn’t …
Dixon makes me look like a Nostraticist, but he’s not yer run-of-the-mill ubersplitter; his reasoning is different in many ways, and a lot of his reasons are Australia-specific. (Which is one of the very things his critics object to: he seems to be implying that the usual rules don’t apply in Australia, which looks methodologically suspect.)
As a descriptivist, his reputation is solid. And if I had sixpence for every good descriptive work that references his Basic Linguistic Theory … well, I’d have a lot of sixpences.
I actually have his Dyirbal Mark 2 book; it’s OK, but hardly a new dawn for linguistics.
Well, Prof. Bowern is saying that Dixon’s number for total languages-on-continent (250) is way too low, which would not be the mark of a run of the mill ubersplitter … Although I guess you could simultaneously be a “splitter” who denied the coherence of “Afroasiatic” while being a “lumper” who said that there were fewer languages in the freestanding “Chadic” family than someone else contended there were in the Chadic subfamily of Afroasiatic.
Yeah, there’s no contradiction; the Way of the Splitter lies not in how many distinct languages you recognise, but the extent to which you think it is possible to show that they are genetically related. As I say, Dixon’s take on this is rather idiosyncratic: he doesn’t exactly deny that languages are related, but rather asserts that there has been so much diffusion between neighbouring Australian languages that it’s often impossible to say one way or the other whether they were genetically related to start with.
He certainly has persuasive evidence about Australian languages being very subject to diffusion; whether that is enough to justify his general view about Australian languages and relatedness is … controversial. To an Athanasius contra mundum level, I get the impression.
I suspect Australianists wouldn’t spend a lot of time on his comparative views if he weren’t (justifiably) eminent as a descriptivist. But he does seem to have spurred them to try harder, which is probably all to the good.
His basic view on the diffusion thing goes right back to his original Dyirbal grammar, which has a fairly elaborate statistical argument about this, IIRC (I’m away from my books at present, and may have misremembered.)
He’s right, at any rate, that there are a lot of awkward lacunae and difficult problems even with the generally-accepted Pama-Nyungan family, as Prof Bowern herself is actually implying above.
Now my eye is caught by “in the same way that Austronesian or Indo-European are language families.” Are there no Splitterist doubters of Austronesian as currently proposed?
No. And there’s a reason for the difference.
I have no opinion of any value at all on Pama-Nyungan (I’m just reporting what I’ve read there, and the consensus certainly does seem to be that it’s entirely pukka), but I think there is an unfortunate tendency on the part of those who are primarily familiar with properly-established language families like Indo-European or Uralic or Austronesian to suppose that works which look like they have applied the same standards of rigour to other putative families really have done so. Someone unfamiliar with previous work on the supposed family and unfamiliar with the nature of the primary data is easily misled about this. Ahem …
It occurs to me that comparative-linguistics cranks normally go for unsubstantiated long-range hypotheses, but there is no reason a priori to suppose that there can’t be splitterist cranks, who deny genetic relationships accepted by all competent mainstream scholars …
(Those who do it out of nationalistic chauvinism don’t count. They are unworthy of the proud name of crank.)
Claire Bowern is trying to have her cake and eat it too. The following two statements-
1-“We pretty much all agree that language change works the same way in Australia that it does elsewhere”, and
2- “Pretty much all of us are puzzled by the relative lack of sound change in Australia”
are utterly incompatible with one another. As a non-Australianist who has read some of the scholarly literature and followed some of the debates, I do agree that there is something bizarre/suspicious about the uniformity of the reflexes of the proto-forms in most of the daughter languages of “Proto-Pama-Nyungan”. Almost as bizarre/suspicious as the fact that Proto-Pama-Nyungan is reconstructed as a language whose phoneme inventory and phonotactics are well-nigh indistinguishable from those of a typical Pama-Nyungan language today.
My own uncharitable impression is that almost all of the linguists who work on the historical linguistics of Australian languages have very little knowledge of the diachrony of any other language (including English) or language family, and as a result have no yardstick which could allow them to evaluate how (im)plausible various reconstructions of Pama-Nyungan are.
For example, if Australianists agree that Proto-Pama-Nyungan (assuming it indeed existed) cannot be younger than 4000 years, then it would have to be conceded that it is unlikely in the extreme that it was phonologically and phonotactically well-nigh indistinguishable from a typical Pama-Nyungan language today. Why? Because in the case of much younger (1500-2000 years old) Proto-languages (Proto-Slavic, Proto-Germanic, Latin…) we know that several of their phonological features failed to survive into any of their daughter languages (No Slavic variety remains a language of open syllables only, no Romance variety has preserved /h/ or Latin vowel length…).
It is such a pity historical linguistics is in such a state of (terminal?) decline: my suspicion -it is no more than that!-is that if just a few scholars trained in the comparative/historical linguistics -including dialectology- of some well-studied non-Australian language families were hired in Australia and given (a) research grant(s) to go over the scholarship on the historical linguistics of Australian languages, they would find most if not all of it in dire need of thorough revision if not total replacement.
Hmm, a thought: If Claire Bowern decides to edit a book on the topic, she might consider bringing in some outside contributors thoroughly versed in historical linguistics and the history of scholarship + classification of some well-studied languages families (If the above examples seem too Eurocentric, I hasten to add that specialists in Algonquian, Semitic, Dravidian, Turkic, Indo-Aryan or Bantu -for example-would be just as suitable). It would definitely make for a more interesting book, methinks.
(Why, yes, I might be interested in being such an outside contributor, how DID you guess?)
No one here seems yet to have considered the obvious explanation: borrowing by the proto-language from daughter languages by means of time travel! The speakers of the daughter languages spend some of their time outside the present in dreamtime. Why shouldn’t the speakers of the proto-language have done so? One can imagine them marveling at their more sophisticated descendants and plundering their word stock, just as Uralic borrowed words for woman, water, and, I don’t know, 2nd cousin on the mother’s side from Proto-Indo-European.
Again, I’m away from my books at present, and may have misremembered, but I’ve seen adduced as a proof of the reality of Pama-Nyungan as a genuine subgroup within (presumably) Australian, that all initial alveolars have become retroflex. But the absence of initial alveolars is a synchronic constraint in the modern Pama-Nyungan languages …
It is such a pity historical linguistics is in such a state of (terminal?) decline
There’s actually a good bit of proper bottom-up comparative work going on with several groups of African languages. (It tends to be of value in proportion to how low-level the group in question is – unsurprisingly.) The damage done by Greenberg’s classificatio praecox is gradually healing.
Unfortunately, it’s mixed in with … other stuff … trying to set the clock back.
It occurs to me that comparative-linguistics cranks normally go for unsubstantiated long-range hypotheses, but there is no reason a priori to suppose that there can’t be splitterist cranks, who deny genetic relationships accepted by all competent mainstream scholars …
If you want some fun, google Angela Marcantonio; she doubts that even IE and Uralic are properly established language families.
My own uncharitable impression is that almost all of the linguists who work on the historical linguistics of Australian languages have very little knowledge of the diachrony of any other language (including English) or language family, and as a result have no yardstick which could allow them to evaluate how (im)plausible various reconstructions of Pama-Nyungan are.
I don’t know about Australianist historical linguists in general, but you know, even ignoring the rest of her lengthy CV, Claire Bowern majored in Classics; it would be absurd to suggest that she’s unaware that phonologies elsewhere tend to change over the millennia. In my limited experience, all the Australianist linguists I’ve ever met have seemed to be well-versed in at least one non-Australian language family.
It seems more pertinent to note that Claire Bowern hasn’t really tried to reconstruct proto-Pama-Nyungan in the first place – only the much smaller group proto-Nyulnyulan. I’m no Australianist, but I get the impression that Pama-Nyungan is a bit like Niger-Congo, in that efforts at high-level reconstruction are basically stuck until the unglamorous low-level reconstruction they need gets done (if it ever does.)
Angela Marcantonio
I remember she came to speak to us once at SOAS, arguing – among other things – that Indo-Europeanists’ claim that kw and gw could change to p and b was just absurd; we never observe such things in real life!
There was a Romanian in the audience. (Sadly no Sardinians, though.)
“he seems to be implying that the usual rules don’t apply in Australia, which looks methodologically suspect”
I also got that impression from Dixon, and had the same reaction.
There are quite a few language families that have very transparent cognate sets with relatively few and straightforward sound changes. They tend to be “young” families, that haven’t been diversifying that long. I have no idea what the estimates of PPN’s dates are based on, but wouldn’t one possibility be that even the 4000 BP date is rather too old?* Or is that a firm boundary, for some reason? I can imagine there being a perceived conflict between the amounts of sound and other kinds of change (lexical replacement, morphological restructurings, etc.), but even if that’s the case, should it be a given that this is resolved in favour of privileging, say, lexico-statistics? (My own view is that we should never privilege lexico-statistics, or even take them seriously for much of anything, but it could still be legitimate to note that there are mismatches in expectations for how quickly different linguistic domains usually seem to change.) I have no detailed knowledge of Australian linguistics, though, so there could be an obvious angle I’m missing here.
*15k at the other end strikes me as truly absurd, and not worth taking seriously unless there’s some astonishingly spectacular evidence to support it. Which presumably there isn’t, or there wouldn’t be so much disagreement.
I do like Basic Linguistic Theory quite a lot. A bit tendentious in places, but I found it very helpful as grad student bombarded by various linguistic frameworks. Haven’t gone back to it in some years, but I probably should.
Lameen-
Your statement-
“Claire Bowern majored in Classics; it would be absurd to suggest that she’s unaware that phonologies elsewhere tend to change over the millennia.”
-is one that is directly contradicted by my own experience: L1 French-speaking Classical scholars (I am talking here about ones of the older generation, whose knowledge of Latin and Ancient Greek was such that could almost sight-read either or both languages) of my acquaintance were ALWAYS surprised when I explained to them that the similarity between (say) Latin “fragilis” and French “fragile” was not due to the latter being a later, changed form of the former (which it is not: it was borrowed from written Latin into French), and always surprised if not floored when I told them that the French adjective “frêle” is the true, later, changed form of “fragilis” (Okay, “fragilem”, technically).
And I never denied that Australianists in general are aware of sound changes: my point is that they are seemingly unaware of how deeply and radically phonologies can and indeed do change (For instance, Colloquial French, Lisbon Portuguese, Sursilvan Romansh, and Tuscan differ FAR MORE radically from one another -and from Classical Latin, might I add- in phonotactics and phoneme inventory than two typical Pama-Nyungan languages, despite the latter family -assuming it is a real one-being at least twice the age of Romance).
I can imagine there being a perceived conflict between the amounts of sound and other kinds of change (lexical replacement, morphological restructurings, etc.), but even if that’s the case, should it be a given that this is resolved in favour of privileging, say, lexico-statistics
If I remember correctly, one of my university professors (Norbert Boretzky) wrote an article back in the 80s about what Dixon’s findings in Australian linguistics meant for Indo-European linguistics (the idea was that IEan society may have been closer to the family bands of Australia than to later agricultural societies, but I digress).
What I remember from this article about the features of Australian language history would very much lead to faster replacement of vocabulary than elsewhere – widespread loaning, widespread taboo replacement of words, group decisions on replacing certain words by other words, etc. That seem to be the peculiarities of Australian language development that DE alluded to, and they could conspire to make language families look much older when you look at word replacement rates than when you look at sound changes.
But that’s all from memory, and I know neither whether those really were Dixon’s findings, nor how good they hold up.
Hans: If you want some fun, google Angela Marcantonio
I was thinking of her, too, but it must be 20 years since the Uralic splash. Is she still at it?
FAR MORE radically
Personal favourite Bantu implausible-but-true sound (and meaning) development: Swahili mzee “old man”, cognate with Kusaal biig “child.”
Proto-Bantu can hardly go back as far as some of these estimates for proto-Pama-Nyungan, though there is an awful lot of guesswork going on there too.
I was thinking of her, too, but it must be 20 years since the Uralic splash. Is she still at it?
I see that her critique of IE was published in 2009… time flies. No idea if she’s still at it or what she does nowadays.
But I remember her approach to both Uralic and IE vividly, because it’s pure crankery – detecting some small flaws in the opposing theory is proof that it’s all wrong, while her own explanations for the observed phenomena that led to the theory she opposes are basically handwaving (e.g., according to her the similarities between Greek and Persian are due to contact after Alexander’s conquests. No, I’m not making that up.)
I remember diving into her Uralic article because some nationalist crackpot pestering sci.lang found it and cited it as proof that historical linguistics was All Wrong. I never looked at her IE stuff.
This is cool. I always enjoy hearing what she has to say, and how she says it.
One thing which is not mentioned is what people think of a (mostly) all-encompassing Proto-Australian. Proto-Australian was the title of a book published in 2024 by Mark Harvey and Robert Mailhammer, which focuses on reconstructing the morphology of the protolanguage. I haven’t found any reviews of it. My guess is that it’s at least controversial, but not as much as Dixon’s ideas on Australian being an unresolvable soup.
My understanding is that Australian languages tend to replace vocabulary at a fast rate, while the morphology is more stable. Something like a mirror image of Uralic (if I understand correctly). All odd language families are odd in their own way, or something like that.
A lot of Pama-Nyungan does look the same, but some does not. Ken Hale’s 1964 paper on Northern Paman is a pretty famous exposition of a subfamily where a number of phonological processes managed to obscure the etymologies to casual inspection, including initial consonant deletion. Decades later, Harvey and Mailhammer (in “Reconstructing remote relationships: Proto-Australian noun class prefixation”, Diachronica 34(4):470, 2017, pp. 477–8) wrote:
Etienne, I think you are doing injustice to Australianist historical linguists in general, and to Bowern in particular. There is a long tradition of very good historical linguistics in Australia, with meticulous and innovative work not only on Australian languages but also Austronesian, Papuan, and SE Asian ones. Ross, Pawley, and Enfield are a few names that come to mind.
The phonological conservatism of Pama-Nyungan is unexplained, but I don’t think that the problem is with Australianist methodology. I think the problem is that in general, while historical linguistics and sociolinguistics have discovered many mechanisms for language change, there really isn’t any general theory for linguistic conservatism. If lenition or final devoicing are natural, why do some languages show neither after thousands of years? Why do Maori and Rapanui preserve almost all of the Proto East Polynesian consonant inventory, while Hawaiian merges a third of it away?
I think Bowern’s statement that “language change works the same way in Australia that it does elsewhere” is reasonable. I read it as, language change is not the same everywhere, but Australian is within the global range of oddness. I read that statement as a counter to Dixon’s Australian exceptionalism.
It’s been advanced as an argument against projecting the agglutinative prefixing structure of (non-NW) Bantu verbs back to “proto-Niger-Congo” that if all this was flexion from the beginning, it’s been preserved implausibly well over many millennia.
Myself, I think it’s a factitious problem anyway: the only part of “Niger-Congo” that really “preserves” this system, taking proper account of both form and meaning, is non-NW Bantu itself, and the time-depth there is probably quite shallow (closer to Romance than to Indo-European.)
Moreover, it seems quite plausible that a relatively transparent agglutinative structure like that could be (as it were) self-renewing, with phonological changes being continually undone by analogy and paradigm pressure.
I dare say prefixes may well be less liable to phonological attrition than suffixes in any case (though those Volta-Congo languages that prefix their noun-class affixes have actually often done a pretty thorough attritition job on them anyway.)
But in any case, the fact that English and Lithuanian are both (one presumes, unless one is Marcantonio) outcomes of PIE should make one leery of getting too uniformitarian.
@de
You don’t have to give this question any time if this is just stupid….
Kusaal
kʋdʋgɔ old
kʋrʋg/k (B-F Kusaal) g= become old, k = old
yaap(B-F) grand father
bii (B-F) child, niece, nephew
bii (B-F) fruit
biili (B-F) grain
Swahili
mbegu (Sw) seed
mbege (Sw) finger millet
kua (Sw) grow
mzee grandfather
zeeka get old
Why is the Kusaal child/fruit word not related to the Swahili “seed” word instead of the Swahili “old” word?
I agree this may be crap, but compare
Kus (B-F)
dii surprise
giil nerve, vein, tendon
lii fall
siit honeycomb
Sw
dege convulsions/ stomach pains or dropping of food while eating believed to be caused by envious look.
gegedu cartilage, disc
gegereka crustacean
leg(e)a be loose
sega honeycomb
A thing that’s often struck me doing Oti-Volta comparative stuff is that it seems to be easier to find cognate series with nouns than with verbs.
I don’t know if this is just some blind spot of my own*, or a true fact about verb vocabulary turning over faster than noun vocabulary in Oti-Volta; if it is a real phenomenon, I don’t know if this is also a feature of other language families.
* I find that there seems to be more polysemy with verbs than nouns; but even that might be an impression created by the fact that much of the lexical material is in French, and French verbs often don’t neatly align with English verbs in meaning either. (A surprising number of the English glosses in Niggli’s generally-good Mooré dictionary are actually mistranslated from the French glosses.)
@PP:
I stacked the deck with Swahili mzee: the clever Bantuists long ago traced this to an agent noun derived from proto-Bantu *bíad- “give birth”, (Swahili zaa), itself the only reflex in Bantu of the extremely widespread Volta-Congo *bi “child.” The semantic development is “begetter/bearer” to “elder.” The agent noun of the (unrelated) Kusaal verb du’a “bear/beget” has gone a similar route: du’ad “elder relative.”
The Swahili *b > z change is actually regular before protoBantu *i *u.
(Wiktionary just seems to have got this wrong.)
With regard to the other words:
Agolle Kusaal kʋdʋg(ɔ) “old” and Toende kʋrʋg “become old” do indeed share a root*: the d is original (Toende has lost the d/r contrast, preserved in Agolle Kusaal and Mooré.) The primary sense of the verb is “dry out, wither” (discouragingly for those of us of a certain vintage.)
The core Kusaal word for “elder” (as in “born earlier”) is kpɛɛnm [k͡pɛ̃:m]. There are actually potential Bantu cognates for this, but you run up against a problem that I’ve continually had with cognate-hunting in proto-Bantu: there are just too many candidates, making it fatally tempting to cherry-pick.
In Agolle Kusaal, biig “child” is not used for “fruit”, but the Mooré cognate bíiga is used for both. However, comparative evidence definitely shows that “child” is the primary sense. “Niece, nephew” are inaccurate glosses; biig means “younger male or female relative one generation down, reckoning in the male line”, or just “child” (also “messenger.”) “Man’s sister’s child” is ansiŋ.
In biig, the stem is bi-, and -g is a singular noun-class suffix, not part of the root (cf biis “children.”) In Swahili mbegu “seed”, the stem itself is -begu. Apart from that mismatch, the vowels of the first/only syllables also fail to match.
To make life more complicated, Agolle Kusaal also has a verb bi’ “ripen, mature.” However, not only does this have a glottal vowel, as opposed to the modal vowel of biig, but the actual vowel phoneme differs (/ɪ/ versus /i/; the standard Agolle orthography fails to mark this phonemic distinction.) So the roots can’t be equated synchronically; whether they have some deeper connection, I don’t know. It’s tempting to match it with Bantu *-begu, but I’ve no other evidence to support the idea that Bantu *VgV corresponds to a Western Oti-Volta glottal vowel, so that’s really just pure speculation.
Kua doesn’t mean “grow” in the intransitive sense; it’s specifically “cultivate, farm, break ground with a hoe.” A variant form of the root turns up in Kusaal pʋkpaad “farmer”, an old compound literally meaning “field-cultivator.”
* But they don’t share a suffix: the -g in kʋdʋg “old” is a singular noun-class suffix (cf pl kʋda “old”), whereas the -g in Toende kʋrʋg, Agolle kʋdig “wither, dry out, grow old” is a verb-deriving inchoative suffix.
reckoning in the male line
Actually, the Kusaasi kinship system is not so patriarchal: I should have said, “reckoning along descent lines that don’t cross over from one sex to another.” A woman’s children are her biis and so are her sister’s children.
Kusaasi kinship terms are actually surprisingly unisex; for many key terms, the basic word is actually sex-neutral, like yaab “grandmother/grandfather.” It’s very much in the spirit of the system that men and women use different terms for their “parent-in-law”, but neither basic word for “parent-in-law” actually specifies the sex of the said parent-in-law. (You can specify it, of course, but the language doesn’t compel you to do so.)
Kusaal has three core words for brother/sister, none of which is (in itself) sex-specific.
The technical anthropological term for a system like this is “neat.”
The primary sense of the verb is “dry out, wither”
Actually, the truth may be less depressing: it’s possible that in kʋrʋg, two originally distinct verbs “grow old” and “dry up” have fallen together as a result of historical sound changes.
(I’m in the middle of rethinking my analysis of proto-Oti-Volta root-final stops, which is what all this turns on.)
L1 French-speaking Classical scholars (I am talking here about ones of the older generation, whose knowledge of Latin and Ancient Greek was such that could almost sight-read either or both languages) of my acquaintance were ALWAYS surprised when I explained to them that the similarity between (say) Latin “fragilis” and French “fragile” was not due to the latter being a later, changed form of the former (which it is not: it was borrowed from written Latin into French), and always surprised if not floored when I told them that the French adjective “frêle” is the true, later, changed form of “fragilis” (Okay, “fragilem”, technically).
None of this has anything to do with Bowern, who is very well educated as a linguist — they don’t let the kind of ignorance you’re describing run riot at Yale. Your impression is not only uncharitable but seems to be prejudice based on a few examples you happen to have run into. You can disagree with Bowern without imputing to her every negative characteristic in your mental storehouse.
sega honeycomb
I’ll see your Swedish honeycomb, and raise you Yom sīɣà “bee.”
(Toende Kusaal sĩit is “honey” rather than “honeycomb.” The relationship between the words for “bee” and “honey” in Oti-Volta is actually very interesting … must resist … temptation to explain …)
However, Scandi-Congo has long been solidly established on the basis of irrefutable Chomskyan evidence, so lexical similarities like this are unsurprising.
Oh. Swahili. Not Swedish …
The bee/honey words are problematic once you get outside Oti-Volta. There’s a peculiar alternation between palatal and alveolar initials; that doesn’t absolutely rule out cognacy, because there are similar alternations with other words too (like “horse” and “shea tree”), but the alternations don’t seem to lend themselves to a nice principled set of historical sound changes, and I have an uneasy feeling that we may in reality be looking at (very old) Wanderwörter.
So comparison with Bantu is jumping the gun rather. And again, the g in sega is part of the actual stem, whereas in e.g. Yom sīɣà the second syllable is a class suffix and the actual stem is just sī-. (And this stem is not glottalised in WOV.)
Oh, Yale has had plenty of Classics faculty (and faculty in English and pretty much every other department touching on human language other than hopefully Linguistics and maybe Anthropology) who would tend to confirm Etienne’s sweeping claims …
And even the Linguistics dep’t was known historically to hire people with doctorates from MIT, if you know what I mean.
(That’s the Gur “bee/honey” words that are problematic once you get outside of Oti-Volta, sorry. Ran out of edit time.)
Her dissertation, done at Harvard under the Indo-Europeanist Jay Jasanoff, is here. I shouldn’t worry about her qualifications as a histrical linguist.
Hat: I have nowhere imputed ANY “negative characteristic” to Claire Bowern or indeed to any individual linguist working on Australian languages. I merely first pointed to a contradiction in her Facebook posting (And I am certain I did not state, and do not believe I even implied, that said contradiction implied ignorance or incompetence on her part).
Subsequently I gave the example of my experience with Classicists (as individuals knowing little of substance involving historical linguistics, in my experience at any rate) in answer to Lameen’s remark in his 6:08 posting July 8 (““Claire Bowern majored in Classics; it would be absurd to suggest that she’s unaware that phonologies elsewhere tend to change over the millennia.”). And incidentally, I agree with Lameen: it would indeed be absurd to suggest such a thing. Which is why I never did (I refer Hatters to my clarification at 2:39 PM July 8).
Whether Claire Bowern (or indeed any other individual Australianist) is knowledgeable or not about historical linguistics is something I do not know: I have never had the pleasure of extensively exchanging ideas with her or indeed with any Australianist (be it in person or online). My point was SOLELY that it is possible (not inevitable, nota bene!) to major in Classics and not know much about historical linguistics. I believe this broader point is true (J.W. Brewer’s remark about Classicists at Yale seems to indicate as much), and was (and remains) the only point I wished to make. I apologize to all if this was unclear.
Why do Maori and Rapanui preserve almost all of the Proto East Polynesian consonant inventory, while Hawaiian merges a third of it away?
It’s too darn hot and steamy in Hawaii to put the effort into all that articulation. Especially when you’re reclining in a hammock.
Etienne, you did say, “my point is that [Australianists] are seemingly unaware of how deeply and radically phonologies can and indeed do change”. In some places phonologies can and do change a lot, but in some places (like much of Pama-Nyungan) they don’t. A lot has been written about why Pama-Nyungan and other Australian phonologies don’t change much while lexicons do, and those discussions are quite well informed by linguistic knowledge from the rest of the world.
For instance, Colloquial French, Lisbon Portuguese, Sursilvan Romansh, and Tuscan differ FAR MORE radically from one another -and from Classical Latin, might I add- in phonotactics and phoneme inventory than two typical Pama-Nyungan languages, despite the latter family -assuming it is a real one-being at least twice the age of Romance
Lisbon to Tuscany is, what, 2000 km?
Australia is 4000 km wide.
If the languages grouped together as Pama-Nyungan are lexically diverse enough for the very existence of a common ancestor to be doubtful, then such phonological homogeneity across such a distance remains equally puzzling from a Eurasian perspective whether they are a genuine family or not. In fact, if they aren’t a family it becomes even more puzzling – what kind of language contact situation could homogenize the phonologies of most of a entire continent without any common lingua franca or any large-scale language replacement? And if they are a family, then if their phonology can end up synchronically homogeneous across such distances of space and time, why should we expect its common ancestor to be the odd one out?
@de
Thank you for your patience. If you do not feel it is a further waste of your time ☺️, in your “proto O-V ” PDF, you cite a word equation
[PB]-bʊ.à ‘dog’ [POV] *bò- Kusaal bāa, Moba bg$ , Mbelime būāk$
So here Swahili mbwa corresponds to Kusaal bāa. What is the conditioning factor for Swahili mzee (instead of *mbV) Kusaal bii?
EDIT: you already answered this.
Yeah, the Bantu fricativisation thing is old news (to Bantuists, anyhow); it’s fun for throwing up implausible-looking cognates.
Another one I like is Kusaal kum “death” (kpi “die”) beside Swahili kufa “die”, where ku- is not the root, but a noun-class prefix secondarily incorporated into the finite verb: the bit that actually means “die”, and is cognate to the Kusaal, is -f-.
Although the changes induced by Bantu fricativisation have been well mapped out, there are still mysteries about its distribution: it’s not universal within Bantu, and doesn’t seem to follow subgrouping well; but subgrouping within Bantu is another thing which is very much a work in progress, in any case.
Although I think Swahili mbwa “dog” probably is cognate with Kusaal baa, I don’t understand the precise phonological relationship if so. In Oti-Volta, there is an alternation between rounded and unrounded vowels in this etymon; from an Oti-Volta-internal point of view, the unrounded vowel is likely to be the original (rounding after labial consonants is common), and the Grũsi forms support this.
There’s also an issue with the b-: Grũsi cognates seem to show that two originally distinct consonants have fallen together in Oti-Volta *b, and likewise in proto-Bantu *b. John Stewart proposed a widespread fortis-lenis opposition in “proto-Niger-Congo” stops, but his examples are mostly not in core vocabulary, and the reflexes are inconsistent when you start looking beyond his Potou-Akanic starting-point. Initially, he thought he’d got evidence that this opposition survived into proto-Bantu, but he abandoned the idea eventually (it was based on split reflexes in Northwest Bantu, which most Bantuists nowadays think can be explained by conditioning factors – some of them tonal.)
One difficulty in sorting all this out is that there are too few cognate sets to work with; and critical evidence often depends on “Kwa” languages whose own prehistory has not been rigorously worked out.
Most of the relevant languages have a lot of flexional prefixes, too, which means that root-initial variation might go back to sandhi effects from prefixes.* Unfortunately, this idea is difficult to develop rigorously, and can easily turn into a sort of get-out-of-jail-free card to handwave away inconsistent correspondences.
* Which is a perfectly real thing in itself, of course. John Merrill’s PhD thesis “The Historical Origin of Consonant Mutation in the Atlantic Languages” is a very nice account of how it has operated historically in some Atlantic languages. Merrill has gone on to do proper comparative work in other Atlantic groups.
Speaking of nationalist crackpots, I did google Marcantonio and she believes Sanskrit is indigenous to India.
As I recall, she did argue that every similarity between Uralic languages was superficial and due to contact. Now that you say it, her actual beef to grind in the game may well have been autuchtonous Sanskrit, but that would have been a couple of steps ahead for the Uralic paper.
Lameen: The “Australian paradox” you so clearly define (the lexical/grammatical diversity of Pama-Nyungan languages versus their phonological uniformity) is not a puzzle solely from a Eurasian point of view. A language family of the New World, such as Algonquian, exhibits a great deal of diversity among its member languages, including -crucially-phonological diversity (Micmac, Arapahoan, Cheyenne and Blackfoot are the ones that have changed the most, phonologically, from the reconstructed Proto-Algonquian system -indeed I think Arapohoan has been called “The French of Algonquian” by one linguist).
Thus, an Algonquianist casting a glance at the Pama-Nyungan family would be just as puzzled, and would define an “Australian paradox” in much the same way as an Indo-Europeanist would or a Berber/Arabic specialist such as yourself just did.
About this phonological homogeneity spanning a continent -I think we need to be careful here. A distinction should be made between the break-up of Proto-Pama-Nyungan (In what follows I will assume it indeed existed) into its constituent languages/subfamilies on the one hand and the spread of Pama-Nyungan over most of the Australian land mass on the other.
Allow me to propose the following scenario: The Pama-Nyungan family originally covered a small part of Australia, long after the disintegration of Proto-Pama-Nyungan, giving members of the family enough time to diverge heavily from one another in grammar + lexicon. At a point in time immediately before the spread of various daughter families/languages of Pama-Nyungan across most of the continent, some contact event(s)* within the area where the Pama-Nyungan family was originally spoken caused massive phonological convergence. The subsequent spread of these daughters of Proto-Pama-Nyungan across most of Australia meant that most of Australia was phonologically uniform, but without this phonological uniformity having arisen as a result of any kind of contact event *which extended over the length of the Australian continent*.
I would maintain that the above scenario is the most reasonable hypothesis. An analogy: imagine the year is 3026, and a group of historical linguists are puzzled by the presence of a number of typological features -A definite and indefinite article, a great deal of shared vocabulary, for example- shared between Patagonian Spanish, New Zealand English, New Caledonia French, Afrikaans, Angolan Portuguese and Surinamese Dutch (These are the only surviving Western European languages in 3026). These linguists might wonder what kind of a global contact event could have caused these geographically far-flung languages to converge. And perhaps some linguist might consider the possibility that the convergence event(s) preceded the dispersal of these languages, and thus was not a global one.
*Whether this contact event was in any way related to the later massive spread across most of the continent of said daughters of Proto-Pama-Nyungan is not relevant to the scenario.
The distance argument is deceptive: where the land is less productive, spread is wider and therefore faster. The Inuit languages are spread over a similar distance to that of Pama-Nyungan (about 4000 km great circle distance from the Bering Straits to Davis Strait), with a comparable uniformity in phonological inventory.
Y: But the Inuit languages are very similar to one another, practically a dialect continuum over much of the vast territory they cover. The varieties differ from one another in phonology as much as they do in lexicon and grammar. It is telling, too, that the genetically unrelated neighboring languages Inuit was in contact with in pre-Columbian times (Athabaskan languages, plus the Westernmost members of the Cree dialect continuum) are phonologically quite un-Inuit-like (In the case of Athabaskan the phonological dissimilarity is rather extreme).
One could say the same of the Cree dialect continuum, just South of Athabaskan and North of various other Algonquian languages: there is as much diversity in grammar and vocabulary as there is in phonology, broadly speaking, and the phonologies of the languages in contact with the Cree continuum in pre-Columbian times (Athabaskan, Blackfoot…) more often than not were quite unlike Cree phonology.
If the Canadian indigenous linguistic situation were analogous to the Australian one, you would need to imagine most of the country covered by a dozen or so language families which may be distant genetic relatives of Inuit (or Cree…), most of whose vocabulary and bound morphemes are un-Inuit-like (or un-Cree-like…)…but all of which have a phonology that is nearly identical to that of Inuit (or Cree) languages.
@Y:
This reminds me rather of Johanna Nichols’ spread zone/residual zone thing.
Whatevs, Australia seems to behave (in some ways, at least) like one big Sprachbund. In a way, I suppose that’s what Dixon is saying; still, it’s not impossible to do orthodox comparative linguistics even within a Sprachbund – it just makes things more complicated.
You need to be more wary of leaping to conclusions about similarities being due to common inheritance, but in any case, you ought to be paying attention to the differences at least as much as the similarities.
most of whose vocabulary and bound morphemes are un-Inuit-like…but all of which have a phonology that is nearly identical to that of Inuit languages
Oti-Volta is not remotely as geographically dispersed as Pama-Nyungan, so it’s not a close analogy, but even so I’ve often been struck by how not-particularly-closely-related Oti-Volta languages can share not only generally similar phonological systems but even specific non-trivial morphophonemic processes which don’t seem reconstructable to the protolanguage. (For example, both Western Oti-Volta and Gurma geminate initial stops of noun-class suffixes after monomoraic CV-stems.)
This paper
https://www.researchgate.net/profile/Jason-Rogers-12/publication/392077129_A_Model_of_the_Origins_and_Development_of_Aleut/links/684748add1054b0207fadea7/A-Model-of-the-Origins-and-Development-of-Aleut.pdf
makes an argument for pretty extensive prehistoric Athabaskan influence on Aleut, including adoption of verb vocabulary. (This is rather difficult to picture, on typological grounds, and the authors suggest that it might date to a period when Athabaskan verbal morphology was rather more transparent than in the modern languages.)
Etienne: but back to your earlier argument, Inuit is roughly as old as Romance, so you can’t just say that a language family must have diversified more over that period of time.
DE, Could you say that, say, 90% of the Bantu languages (minus the southern edge) are as alike phonologically as the P-N languages are, and as similar to the (maybe unrelatable) languages of West Africa as P-N is to the other Australian languages?
Could you say that, say, 90% of the Bantu languages (minus the southern edge) are as alike phonologically as the P-N languages are, and as similar to the (maybe unrelatable) languages of West Africa as P-N is to the other Australian languages?
I’m no Bantuist, but to the first question I’d tentatively say yes (with the major caveat that the NW Bantu languages tend to be a good bit more aberrant phonologically overall than the southern languages, which have acquired clicks but otherwise mostly have fairly ordinary Bantu phonotactics.)
Second question: definitely not. Kusaal and Yoruba are very unlike their relatives Swahili and Luganda, for example, both phonologically and typologically (though they do share some things, like a complete lack of grammaticalisation of biological sex, just two grammatical numbers, nominative-accusative alignment etc.) I’d say the phonological differences are (on average) a good bit greater than between Pama-Nyungan and The Rest (though as you pointed out, there are some pretty spectacular deviants within Pama-Nyungan too.)
Well-known discussions of this are in several papers of Tom Güldemann’s:
https://www.cambridge.org/core/books/abs/linguistic-geography-of-africa/macrosudan-belt-towards-identifying-a-linguistic-area-in-northern-subsaharan-africa/E6B107C0FD7AFA6B575EFEE0E65D8786
https://www.researchgate.net/publication/319527953_Sprachraum_and_geography_linguistic_macro-areas_in_Africa
Güldemann’s zones certainly cut across genetic families.
The reason is sheer ignorance coupled with creationist-level personal incredulity that knowledge she doesn’t already have might exist out there somewhere. Last time she was mentioned here (…I think that was Lameen reporting from the same visit to SOAS he mentioned above…), she hadn’t heard of Verner’s law and was triumphantly holding up “exceptions to Grimm’s law” as obvious proof that Grimm’s law, if not IE as a whole, didn’t exist.
Lexical replacement seems to be truly horrible in Australia generally. The most obvious factor is the widespread taboo on dead people’s names, which Bowern herself has explained as “imagine someone named Bill dies and you can’t ask for the bill in a restaurant anymore”.
If I’m not confusing him, Mailhammer has done perfectly mainstream work in IE.
I’d expect “horse” to be a Wanderwort because horses were imported not that long ago; I have no clue how old apiculture is in that part of the world, but it doesn’t seem to be terribly old anywhere, and indeed the “bee” and “honey” words generally reconstructed for Proto-Uralic are considered Pre-Indo-Iranian loans; and given the economic and apparently religious importance of shea trees, their names could have done some wandering as well even though the referent is comfortably native (…though I can’t find in Wikipedia if perhaps the cultured variants are different from the wildtype and might have separate names).
Non-morphologized automatic lengthening of consonants after short stressed vowels also crops up in a few Germanic varieties in the last few hundred years (for at least three different historical phonological reasons). I can imagine that getting morphologized given properly agglutinative morphology.
Y, David Eddyshaw, and other interested hatters: I may have a Romance comparandum for Pama-Nyungan. That is to say, the relationship between two typical Pama-Nyungan languages (Yet again, I will for now assume it is a real language family) may be analogous to the relationship between (INSERT DRAMATIC DRUMROLL HERE…)…Campidanese Sardinian and Western Catalan.
Both are Romance, but Campidanese Sardinian has been heavily influenced by (Western) Catalan. As a result both have nearly identical segmental phoneme inventories: for example, both have seven stressed vowel phonemes (the five cardinal ones, plus a mid-high/mid-low distinction for front and back vowels) and only three (/a/, /i/ and /u/) in unstressed positions. Tellingly, neither Eastern Catalan (the closest genetic relative of Western Catalan) nor Logudorese Sardinian (the closest genetic relative of Campidanese Sardinian) share this system. More importantly, the two systems are diachronically distinct and cannot be reconstructed back to an identical vowel inventory in their common ancestor.
Now, despite this close similarity in segmental phoneme inventory, and the massive influence of the latter upon the former, Campidanese and Western Catalan remain quite sharply unlike one another. Compare Campidanese*-
/narami kun kini andaza e ti nau a kini sezi/
and Catalan
“Dies-me amb qui vas i et diré qui ets.”
(“Tell me whom you go with and I will tell you who you are”), where all three verbs -including the copula- are not cognates. The “é” of “diré” is NOT a cognate of the first person /u/ ending of Campidanese “nau”, either: the former is a synthetic future and the latter a present tense, due to the fact that Sardinian lacks a synthetic future. The “s” of “vas” and the /za/ of of /andaza/ are genuine cognates, however, albeit ones which do not correspond with phonological regularity (the final /a/ of the Campidanese morpheme should correspond to a final /a/ in Catalan).
Campidanese /kini/ does look like Catalan “qui”, but that is because it is either a Catalan loan or a form restructured via Catalan influence. The only clear cognates are e/i “and” and the object personal pronouns. However, while comparison of the entire person-marking paradigm of bound and unbound morphemes in the two languages would indeed uncover many similarities, it would prove difficult if not impossible to reconstruct a coherent proto-system, not least because many of these morphemes in each language can be traced back to distinct Latin etyma.
This seems very similar to the similarities between typical Pama-Nyungan languages: near-identical phoneme inventories, with some similarities in grammatical morphemes and plenty of (sometimes pan-Australian) possible cognates or Wanderwörter.
*I owe this example and most of what (little) I know on Campidanese/Catalan language contact to Eduardo Blasco Ferrer’s STORIA LINGUISTICA DELLA SARDEGNA, which I recommend to any hatter(s) interested in delving more deeply into the matter.
Non-morphologized automatic lengthening of consonants after short stressed vowels also crops up in a few Germanic varieties in the last few hundred years (for at least three different historical phonological reasons). I can imagine that getting morphologized given properly agglutinative morphology.
Interesting!
Having come to a tentative conclusion that my existing reconstruction makes a spurious distinction between non-initial POV *k and *g, I was naturally wondering how the omni-velar-stop found non-initially was realised in the actually-spoken protolanguage.
Most of the subsequent developments are more cross-linguistically plausible if it was [g]; the Eastern Oti-Volta languages have /k/ here, but they also devoice initial *g to /k/, so that’s not really a problem.
The trouble is that Gurma, which isn’t part of the voiced-plosive-hating Atakora Sprachbund that the EOV languages are all in, also has /k/ (subsequently voiced to /g/ in Gulimancema and Moba just to make quite sure of confusing any passing comparativists.)
But Gurma does the gemination thing with class suffixes after short root vowels, and Gurma *k normally only follows short root vowels (with some systematic exceptions which can be plausibly explained away as secondary developments, I think.) So if *g were geminated here too, you then just need to posit *gg > k(k), which, conveniently, is a rule that regularly turns up anyway across Oti-Volta.
It all seems reasonably plausible, but pulling at that one thread actually has other knock-on effects on the reconstruction, and probably entails ditching one of my pet theories about the origin of POV long vowels. (But in retrospect, there are other problems with that theory anyway. Probably past its sell-by date.)
@Etienne:
Nice analogy.
As I recall, she did argue that every similarity between Uralic languages was superficial and due to contact. Now that you say it, her actual beef to grind in the game may well have been autuchtonous Sanskrit, but that would have been a couple of steps ahead for the Uralic paper.
As far as I can tell from the two articles I read by her (one on Uralic and one on IE), she had a beef with Uralic first; her looking into the foundations of Uralic seems to have been triggered by discussions of whether Hungarian is Turkic or Uralic, where people told her that it had been proven long ago that Hungarian is Uralic and she didn’t trust that. Then she came across defenses of Uralic saying that it was established by similar methods as the unequivocally proven Indo-European language family, and she decided to look into that, with the results we discussed. Her support of autochthonous Sanscrit seems to be the kind of mutual attraction of cranks that makes anti-vaxxers believe in pizzagate – if the establishment doesn’t like an idea, it must be right…
i don’t have the historical-linguistics chops to have meaningful opinions on this, but i do wonder whether, or to what extent, the differences here could have to do with differences in the contexts and kinds of language contact and priorities shaping communication between speakers of different lects that stem from the rather different social worlds of western asia (including the european peninsula) and (non-north-central) australia. i’m thinking in particular about the combination of greater baseline spatial mobility and (to my understanding) complete or near-complete absence of the state form in australian societies compared to, say, IE-speakers (who tend towards versions of settled agricultural and mobile pastoral structures, with both producing state-based societies from a pretty early date). and also – though much more hesitantly – of the function(s) of intersecting orally-transmitted geographical/mythological accounts in intercommunal relationships.
Specifically:
– Mainstream Swedish has reinterpreted the outcome of vowel lengthening in stressed open syllables as synchronic consonant lengthening after short stressed vowels. In native vocabulary, you only ever get 1) long stressed vowels followed by short consonants (spelled single, and phonemically short in Old Norse) or 2) short stressed vowels followed by long consonants (spelled double, and phonemically long in Old Norse) or clusters, but the French footballer Benzema, ending in a short stressed vowel, got a genitive in [sː], showing that consonant length is allophonic nowadays.
– Some of the southernmost German dialects lengthen the final consonant instead of the vowel in monosyllabic words that ended in a single short consonant in MHG. Paper here, relevant part starting on p. 239; especially pp. 243 (Apical Alemannic) and 245 (Ultramontane Bavarian).
– Mainstream High German merged /t/ into /tː/ (in the positions where they contrasted), not the other way around, around the end of the Middle High German period. (The result is almost consistently spelled tt.) This happened before the spread (from Low German) of vowel lengthening in stressed open syllables and the spread (from somewhere not too high up in Switzerland) of vowel lengthening in monosyllabic words that ended in a single short consonant, but after the Central Bavarian lenition that (among other things) merged postvocalic /t/ into /d/. Short /k/ and /p/ aren’t native in the positions in question, and the loans that do contain them are mostly or entirely younger.
Seconded.
Pizzaghazi. -gate for scandals, -ghazi for manufactroversies.
There are different periodisations of the history of German; according to one (going back to Grimm) the end of the MHG period is about 1500, according to another (going back to Scherer), MHG ends about 1350 and the period between 1350 and 1650 is Frühneuhochdeutsch. The latter periodisation (“Viergliederung”) is the one currently used by the grammars published in the Sammlung kurzer Grammatiken germanischer Dialekte (they are no longer short, btw).
Hans: Her support of autochthonous Sanscrit seems to be the kind of mutual attraction of cranks that makes anti-vaxxers believe in pizzagate – if the establishment doesn’t like an idea, it must be right…
Cranks will always support eachother’s theories, even (or especially) when those theories are mutually exclusive.
The project of an ethno-linguistic crank is to demonstrate that their preferred ethnos is genetically incapable of innovation.
~There are different periodisations of the history of German
Welsh, too. Simon Evans’ grammar of Middle Welsh counts Dafydd ap Gwilym (about 1320 – 1370) as Early Modern Welsh (and Morgan’s 1588 Bible as the beginning of Late Modern Welsh), but even proper academic sources nevertheless tend to call ap Gwilym’s language Middle Welsh.
I prefer Evans’ periodisation, on the grounds of sheer wilful perversity.
[In fact, from a present-day perspective, the big gulf is between any kind of traditional Literary Welsh and the modern spoken language, rather than between Middle and “Modern” Literary Welsh. Stephen Williams’ Elfennau Gramadeg Cymraeg, a grammar of ostensibly “modern” Literary Welsh (which, as is characteristic of such works, it just calls “Welsh”), discusses the neutral SVO word order characteristic of Middle Welsh prose, because that still appears in the 1588 Bible.]
Oh, I didn’t even know that one. I did know that Frühneuhochdeutsch was comparatively newfangled, but had thought that was just the result of cutting Grimm-era Neuhochdeutsch in two, and that the story behind all this was that on both sides of MHG there were periods when writing was mostly done in Latin, so there were neat convenient gaps between the original three periods, while there wasn’t any between ENHG and NHG.
Anyway, I didn’t intend any precision in absolute dates. I’m sure t > tt has been researched in detail, I just don’t know that research.
Anyway anyway: Luwian. As part of an areal reinterpretation of the obstruent voice contrast as a length contrast, Luwian (or a larger part of the Luwic branch; but definitely not Hittite) underwent Čop’s law, which lengthened all intervocalic consonants after short stressed… at least *e, but probably not only.
Not that new — Scherer’s proposal is from 1878, about a quarter of a century after Grimm’s introduction of Mittelhochdeutsch and Althochdeutsch (for earlier Altdeutsch). But it seems that in the past few decades philologists have decided that it is a useful alternative to Grimm’s periodisation.
t in the past few decades
This periodisation seems to have been already standard when I started to be interested in language history and was used in the books about German language history I read back then, which had been published in the 70s and early 80s.
@ David Marjanović:
Your July 9, 2026 (4:19 pm) statement that you would expect the word for “horse” to be a Wanderwort (because horses had been introduced recently) may not be true: in Western North America the spread of the horse was not accompanied by the spread of a word but of a concept: Plains Cree MISTATIM “horse” (literally “big dog”: MIST- is the augmentative prefix, ATIM the lexeme for “dog”. Because horses were originally introduced as draft animals, in societies where hitherto dogs had been the sole draft animals, the name was unsurprising). Some neighboring languages borrowed the Plains Cree word (this is true of some Ojibwe varieties), but others, instead, calqued the Plains Cree word: in Chipewyan (Athabaskan), the word for horse is “tłįcho”, where “tłį” is the word for “dog” and “-cho” the augmentative.
Surprisingly, from a Wörter und Sachen kind of standpoint*, “horse” actually looks reconstructable to proto-Oti-Volta:
Yom sāmɣā; Gulimancema tāāmō; Moba tāānm̀;
Byali sāngə̄; Ditammari tāsā̰ntà; Nateni sāǹtā; Mbelime sā̰nkɛ̀.
All that works for a proto-Oti-Volta *càm-gá (“child” gender) or *càm-wá (“animal” gender.) Miyobe, probably the closest relative of Oti-Volta, has ì-sáŋ̀.
Western Oti-Volta has an unrelated word, but then, it has an unrelated word for “water”, too, and one imagines that the speakers of POV had water.
The t- in the Gurma languages is irregular, but Gurma does that kind of thing, including in etyma which surely aren’t borrowed: cf Moba tūōnǹ “work” (noun), sūn̄ “work” (verb.) The Gurma long vowel is a thing that often happens in Oti-Volta nouns for animals, and probably reflects compensatory lengthening after the loss of an “animal names” derivational suffix.
* Or maybe not. I’ve no real idea how long horses have been around in the West African savanna zone. In the Mossi-Dagomba states they are strongly associated with chieftaincy, and thus with the foundation of those states less than a thousand years ago, but of course that doesn’t prove that horses were previously unknown there. The Ghana empire apparently had cavalry.
Buli wùsùm “horse” looks a bit like some sort of lightly mangled compound of Western Oti-Volta *wɪ̀d- “horse” (Kusaal wief, plural widi) and the *càm- etymon. No doubt a Mass Comparison aficionado could match it with both …
@David Eddyshaw: “comparative-linguistics cranks normally go for unsubstantiated long-range hypotheses”. You are right. I came across one such person when I was least expecting it. I was on a chairlift in a ski resort and started chatting with the woman beside me, who told me she was Turkish and explained that her language was connected to practically all the non-Indo-European languages of Eurasia, including not only Finnish and Hungarian but even Japanese. I knew that the Ural-Altaic group was no longer regarded as a ‘real’ language family but possibly just a Sprachbund, and mentioned that to her, and she got quite angry. I see now that I had encountered a pan-turanist.
I think this is essentially accepted (lay) wisdom in Turkey. I encountered it myself in a Turkish colleague who was far from a crank in general: it was just what he’d always been told, and not being at all interested in linguistics, he’d never actually come across any differing opinion. I think the supposed link with Hungarian and Finnish is important to the self-image of the more Westernising Turks who continue the Atatürk outlook regarding the position of Turkey in the world.
But there are Hatters who will Actually Know.
I don’t, but it wouldn’t surprise me at all.
Altaic still has its defenders. Downloadable from but not viewable on Kassian’s Academia page, all brackets in the original:
This is innovative in that the classic version of 2003 (Etymological Dictionary of the Altaic Languages) still had Turkic and Mongolic as closest relatives despite acknowledging thick layers of loans between them. I guess Turkish Westernizers will be less unhappy with this version…?
https://en.wikipedia.org/wiki/Martine_Robbeets
springs to mind as a high-profile believer, though she doesn’t actually call it “Altaic.”
There’s a splendid takedown of Altaic out there by Alexis Manaster Ramer (a former believer, who has repented and seen the light.) Can’t find it just now …
Peak Pan-Turan[ian]ism was found among certain post-WWI Hungarian royalists/nationalists who, having been deprived of their own monarchs by the victorious Allies, developed a fascination with the Japanese imperial family as their long-lost kin. Don’t know what they thought about Ataturk …
And farther afield, “horse” in Navajo is just the repurposed “dog” word, and “dog” has been replaced – an independent application of about the same metaphor.
Italian radoppiamento s[ː]intattico may actually be the simplest comparison.
Italian radoppiamento s[ː]intattico may actually be the simplest comparison.
Yes. Good thought.
Interesting that the Italian gemination applies after unstressed non-root vowels, too.
The reflex of the consonant which I am assuming was *g in proto-Oti-Volta was probably *k in proto-Gurma not only as a root-final consonant, but in derivational suffixes (“probably”, because I have only scanty lexical materials for Gurma languages other than Gulimancema and Moba, both of which revoice non-initial *k to /g/ anyway.) Gemination doesn’t seem phonologically natural in that context.
Something a bit similar has happened in Tiberian Hebrew; while the original short /a/ and /e/ vowels become long in an open syllable before the stress, the original short /o/ vowel remains short, but the following consonant is geminated instead. (Holem in a pre-stress open syllable is always “long by nature”, never “long by position.”)
The Tiberian rules for word-initial dagesh after a closely-connected word ending in an unstressed/destressed vowel (which I’ve descanted on before) probably also reflect gemination after an original short vowel. I hadn’t thought of an analogy with the Italian phenomenon.
To Davidayim-
(Yes, the two Davids, with a Hebrew nominal dual suffix).
Actually, the Navajo example strengthens my point: if one compared Navajo “łį́į́”and the first syllable of Chipewyan “tłįcho”, proving that they are cognates on the basis of known sound changes would not be difficult.
And the natural hypothesis (in an alternate universe with no surviving knowledge about the post-Columbian linguistic history of the continent) would be that its original meaning was “horse” (Since horses went extinct about 11 000 years ago in the Americas, it would therefore -with an elegant inevitability- also be concluded that Proto-Athabaskan is older than 11 000 years. QED).
About “Raddopiamento sintattico”: it may interest you both to know that, if I recall correctly*, (some?) Greek varieties of Southern Italy have, through contact, acquired very “raddopiamento sintattico”-looking gemination word-initially, which seems to derive historically (as does “raddopiamento sintattico” in Italian) from a word-final /s/ lost in the preceding word. And which may offer a further typological parallel to the Oti-Volta contact phenomena mentioned upthread.
All: At least Pan-Turanism does take genuine linguistic similarities as a starting point, so the possibility exists -by showing that the similarities are best explained as a product of contact rather than inheritance from a common ancestor- of getting pan-Turanian believers to see the light. Some beliefs about linguistic (non)-relationship are pure products of ethnic/religious/ideological beliefs unrelated to any feature(s) of the languages themselves, and thus simply cannot be countered/put in doubt/proven wrong by linguists (or indeed anyone). As a satirist so aptly put it, “Reasoning will never make a Man correct an ill Opinion, which by Reasoning he never acquired.”
*My source must have been Gerhard Rohlfs’ “Historische Grammatik der unteritalienischen Gräzität”.
and thus simply cannot be countered/put in doubt/proven wrong by linguists (or indeed anyone).
Except, of course, by the Führer or the party.
And of course, Comrade Stalin conclusively disproved the views of Nikolai Yakovlevich Marr, not by appeal to the bourgeois so-called “science” of comparative linguistics, but by the truly scientific method of demonstrating that Marr’s theses were incompatible with a correct understanding of historical materialism.
Exactly, comrade! That method never fails.
it would therefore -with an elegant inevitability- also be concluded that Proto-Athabaskan is older than 11 000 years
Several Oti-Volta languages (and their neighbours) use their inherited word for “horse” to mean “bicycle.” Older materials tend to specify “iron horse”, but the “iron” part often gets dropped nowadays; not that surprisingly, I suppose, given that bicycles are a good bit commoner than horses thereabouts, in these prosaic days.
But again, a purely mechanical reconstruction of proto-Oti-Volta, devoid of any cultural historical information, could easily conclude that POV speakers had bicycles.
This conclusion, would, however, be unsurprising in view of other linguistic evidence.
The etymon instantiated in e.g. Kusaal lͻr, Mooré lórè, Kabiyè lɔɔɖɩyɛ “four-wheeled motor vehicle” is readily reconstructable to proto-Central Gur.
Kusaal alͻpir “flying machine” is clearly cognate with Moba lúópīl̀ “flying machine”; although this etymon is sparsely attested in Oti-Volta, as Kusaal and Moba belong to quite different branches of the family, “flying machine” must surely be reconstructed to proto-Oti-Volta, at least.
Wakanda, nothing.
Kasem alapɩlɩ “flying machine” confirms that proto-Gur speakers had this technology; I was too hidebound by linguistic Luddism in timidly only assigning this to proto-Oti-Volta.
The relationship to Mande (e.g. Dyula aviyɔn) is not certain, but it is surely significant that the words belong with the same phoneme. The Mande form appears to have undergone some metathesis and lenition of postvocalic consonants: tentatively, the original form can be reconstructed as *aʎubõ (for the second vowel, cf Farefare alupele “flying machine.”)
The relationship to Mande (e.g. Dyula aviyɔn) is not certain,
But must involve Afro-Asiatic; cf. Hebrew ‘aviron.
Excellent point!
These are deep waters …
Mande, Hebrew and Aymara surely constitute a subgroup within the family, as the forms for “airplane” (Mande /aviyɔn/, Hebrew /‘aviron/ and Aymara /awjuna/) are far closer to one another than to any of the Oti-Volta reflexes of PAAA (=Proto-American-African-Asian).
Taking David Eddyshaw’s proposed form */aʎubõ / as a point of departure, we can see that Proto-Subgroup-Hebrew-Aymara (PSHA for short, named after the two geographically most distant members of the family) innovated in shifting intervocalic /ʎ/ to /v/ or /w/, deleting intervocalic /b/, and fronting /u/. One could tentatively reconstruct a PSHA form */av(w)iõ/.
The PSHA form yields the Hebrew form via addition of /’/ to the initial vowel, shift of the sequence /i/ + V to /ir/ + V, and shift of /õ/ to /on/. It yields the Aymara form via shift of /õ/ to /un/, then /una/, with the preceding /i/ becoming /j/. And it yields the Mande form via shift of /õ/ to /ɔn/ and of prevocalic /i/ to /ij/.
Prima facie it might seem unlikely, on geographical grounds, that Mande could be part of a subgroup including Aymara and Hebrew to the exclusion of Oti-Volta, but if we assume that PAAA speakers had aircraft, as seems indeed to have been the case (see above), then the difficulty vanishes.
Of the non-Mande African forms quoted by David Eddyshaw (alͻpir, lúópīl̀, alapɩlɩ, alupele), all appear to differ from PAAA */aʎubõ/ via strengthening (Verschärfung) of intervocalic /b/ and depalatalization of */ʎ/. To this can be added the addition of an additional morpheme with a rhotic/lateral. ALL THREE changes are alien to PSHA, tellingly (In support of the Scandi-Congo hypothesis one could point to the first and the last changes yielding a final syllable which, via typologically unremarkable sound changes, are readily comparable to Swedish/Norwegian/Danish “fly(g)”, “airplane”).
Now, the fact that PSHA is spoken both in and out of Africa (Leaving aside speculation on Scandi-Congo for the moment), whereas non-PSHA languages are found in Africa only, points to the African continent (i.e. the area with the greatest genetic diversity) having been the homeland of PAAA. It follows, as naturally and indeed inexorably as night follows day, that if the PAAA homeland was in Africa and if a word for “airplane” existed in PAAA, then the invention of the airplane is likely to have taken place in Africa.
Science is truly wonderful.
P.S. I am uncertain about one thing: How should I react if sometime in the future a student of mine is so pleased with the above hypothesis that (s)he presents it in a term paper as their own? Thoughts, anyone?
Etienne, if they do, chances are they got it through some diligent AI. Since AI is never wrong, the student should get a perfect grade.
…Oh. That’s completely different from the Germanic examples, then.
It’s been going on in some Spanishes, though, so I’m not too surprised…
Some of the examples in Treccani involve dropped final consonants other than s:
(d) followed by m => mm
Perché mai? ▶ Perché mmai?
A merenda▶ A mmerenda
(d) followed by f => ff
Che fai? ▶ Che ffai?
(d) followed by p => pp
da + prima ▶ dapprima
(c) followed by p/d => pp/dd
così + detto ▶ cosiddetto
né + pure ▶ neppure
Some examples seem to imply an emphatic radoppiamento after final vowel
sopra + tutto ▶ soprattutto
o + dio ▶ oddìo.
Sarò franco ▶ Sarò ffranco
https://www.treccani.it/enciclopedia/raddoppiamento-sintattico_(La-grammatica-italiana)/
I am not arguing with the statement that the dropped consonant is mostly s and would note that the other mutations are similar to ones seen within compound words in Latin.
I suppose the idea is that the consonant stretching was generalized to “after a final vowel” after the -s was forgotten.
What do we see within compound words in Latin? What do you mean?
raddoppiamento
…Oh. It’s self-demonstrating.
@dm
I was thinking about stuff like
ad+firmō > affirmō
I can’t think of any with dropped final c, however, although I suppose maybe hodie could have been pronounced hoddie (I don’t know if Welsh heddyw is just a spelling thing or reflects interference by a BL form like *(h)oddi(e)).
No: dd is just the digraph for /ð/. It’s the soft mutation of /d/: historically, *d after a vowel.
The he- part of heddiw is from an old demonstrative *se, and diw is an old oblique case form of dydd “day”; it still turns up in constructions like dyw Calan (Ionawr) “(on) New Year’s Day.”
Welsh intervocalic written p t c actually are geminates (and derive historically from consonant clusters, most often with the second consonant *h, from earlier *s); old versions of the 1588 Bible still write e.g. atteb for ateb “answer.” Some Welsh people carry this gemination over into English.
Although Latin grammars mark the vowel long in e.g. hoc “this” (neuter), IIRC the form was really *hŏcc, (from *hod-ce); the geminate /kk/ pronunciation was preserved before a following vowel in verse, making the syllable heavy, despite the vowel being short.
de Vaan writes this under hodiē:
The -c in Cl. L. hic etc. is the particle –ce, not part of the inflexional endings. It’s irrelevant for an old compound like this.
According to Sihler, *hoiē (without -d-) is the source for Romance forms like It. oggi.
For a long time, I thought that the gemination of suffix consonants after Western Oti-Volta short root vowels was due to assimilation of a lost root-final consonant, but eventually realised that it couldn’t be, even though, in most cases, an original root-final consonant really has been lost.*
It makes no difference to the outcome which final consonant has been lost, and there is also a small group of CV roots where the evidence seems pretty solid that there never was any root-final consonant to begin with.
The trick is, that proto-WOV had a lot more CV roots than POV: precisely because of the loss of certain root-final consonants.
* What was confusing me, mainly, was that I hadn’t really registered that Mooré, which has diphthongs where POV *r has been lost after a root vowel, has short diphthongs in such cases: zoeta “run” (imperfective) is CVCCV: zŏe-tta. Mooré, like Kusaal, has length distinctions in diphthongs: this is apparently rare enough cross-linguistically that I’ve encountered papers denying that it ever happens. I expect it all comes from not learning classical Greek at school. (Or Old English. Or Tohono Oʼodham.)
@ulr
Thanks, as I said I really can’t think of an example where elision of c causes doubling of following consonant in Latin. Although ct > tt all over Italian, the only Latin possibility I can suggest now is the cognomen “Cotta”.
David Eddyshaw, Plasticpaddy-
One thing about “Italian”* gemination is that, word-internally, it is not entirely “Lautgesetzlich”: the phenomenon seems to be more common in post-stressed vowel position than in other positions, but it cannot be predicted which Latin etyma will yield geminates in Italian and which will not**. Whatever the exact phonological conditioning originally was, it is possible that it applied across word boundaries as well. In which case there may well be instances of “raddoppiamento sintattico” which never involved a lost final consonant in the first place.
*Which should be called “Late Central Latin” rather than Italian: the Latin etyma of several French words must have undergone gemination corresponding to Italian gemination: cf. TOUTE, “all” (feminine), which, because of its final /t/, must go back to a Latin form with a geminate (A simple /t/ would have been deleted intervocalically in French), just like Italian TUTTA and unlike Spanish TODA (Whose /d/ can only go back to a simple stop, not a geminate).
**Comparative scholars who insist that demonstrating a genetic relationship must entail a perfect understanding of the diachronic sound changes connecting a language to its relatives should be consistent in their belief and deny that Italian can be shown to be genetically related to Latin and the other Romance languages. Considering the amount of written “work” seeking to “prove” the non-existence of numerous established language families these days, they would have plenty of company, too.
If hodie is old enough, it should (p. 2) actually have become hoie (with /jː/), and hodie would be restored from dies like odium from odi.
Mooré, like Kusaal, has length distinctions in diphthongs: this is apparently rare enough cross-linguistically that I’ve encountered papers denying that it ever happens.
Mabaan is a fun one: not just length distinctions in diphthongs, but in rising diphthongs. With the length showing up on the first element to boot.
Kusaal has that too, e.g. pia “dig up” (short) versus sia “waist” (long); tua “pound in a mortar” (short), versus sabua “girlfriend” (long.)
Sia and sabua are straightforwardly [sia], [sabua]; you could try to analyse pia, tua as [pja], [twa], but there are no word-initial consonant clusters apart from cases like these, and word-initially, these short rising diphthongs actually contrast with glide + vowel: ia “seek”, ya [ja] “houses”; uak “flood”, wak “run dry” (though the latter pair differ in tone as well as segmentally.)*
(Though there have been historical confusions: Agolle Kusaal has [saa] iank, apparently “[the sky] jumped”, for the [saa] yank “[the sky] flashed with lightning” which would be the “correct” cognate of other Western Oti-Volta verbs.)
* You could try to get out of the minimal-pair contrast by positing e.g. /ʔjā/ “seek” versus /jā/ “houses”, /ʔwāk/ “flood” versus /wàk/ “run dry.” Vowel-initial full words in Kusaal are in fact (optionally) realised with a phonetic [ʔ] onset, and historically they actually have lost former initial consonants. But you still have the problem that this analysis introduces word-initial consonant clusters, which otherwise are not a feature of the language at all.
In the opposite direction from Lithuanian, where ai etc. are phonetic diphthongs, but Vaiana was apparently a bridge too far and is being marketed as Vajana.