Zompist Bboard Again

Austro-Tai is fait accompli by now, even if many details are still to be worked out (I suppose including even: whether Austronesian and Kra-Dai are two separate branches at all, or just two geographic areals within the family). Here are two brand-new works from Blench:
– Tai-Kadai and Austronesian are related at multiple levels and their archeological interpretation
– Mapping Austronesian and Tai-Kadai language relationships against Taiwan/mainland archeology

Whoa wut? Nowai!

Blust, writing in 2009, says that a "historical connection" between Austronesian and Tai-Kadai "now appears virtually certain", but that whether it's genetic or just extensive borrowing remains "a point of contention". Though given that the borrowing would involve at least the numbers 5-8 and probably the numbers 1-10, I think it's fair to suspect that it's actually genetic.

He savages Sagart's Sino-Austronesian, though. He accepts that it's 'rigorous', but also basically says that Sagart's imagining things due to how incredibly broad-minded he is about accepting matches (Sagart apparently compares obscure Old Chinese words of unclear meaning with semantically vaguely connected forms in a reconstructed language he's apparently invented for the east coast of china, ancestral to PAN, which he doesn't detail, many of which items are hard to find in PAN itself).

He points to Sagart's 2004 paper on PAN cognates in Buyang, a conservative language that hasn't entirely lost the initial syllables yet. He also mentions Ostapirat's work on the hypothesis.


However, he also notes the suggested connexion with Austroasiatic. This it seems is a sort of Indo-Uralic kind of thing: there's virtually no lexical similarity, other than a few words that are probably loans, but there is very striking agreement in the morphology, which has only become more convincing as further research into conservative AA languages has revealed more and allowed the proto-forms to be more precisely determined. These agreements are even more striking because most of them are infixes, and systematic infixation is extremely rare outside these two families. See Reid. He even mentions Diffloth's argument that the lack of cognate vocabulary is almost proof of a family connexion: in that the total lack of apparent shared vocabulary strongly suggests no substantial borrowing between the two families, which makes the similarities in morphology less likely to be borrowing, so more likely to be inherited. Although I don't think the argument convinces him.

So it's possible there's a broader Austric that include TK. Or, it's possible that, for example, PAN (or its ancestor) was an Austroasiatic language that almost completely replaced its vocabulary by borrowing from the ancestor of TK, but kept its morphological elements.

Or the AA-AN correspondences are just coincidence. A small handful of morphological items only one or two phonemes in length hardly seems a robust basis on which to build a family...

Salmoneus wrote: ↑Wed Oct 17, 2018 7:03 am Blust, writing in 2009, says that a "historical connection" between Austronesian and Tai-Kadai "now appears virtually certain", but that whether it's genetic or just extensive borrowing remains "a point of contention". Though given that the borrowing would involve at least the numbers 5-8 and probably the numbers 1-10, I think it's fair to suspect that it's actually genetic.

Blench's data in the first paper, assembled across earlier works ups the ante much further than just numerals:

• pronouns: 1PS, 2PS, 1PP, 'this', 'that'
• body parts: 'head', 'eye', 'nose', 'tooth', 'hand', 'shoulder', 'thigh', 'ribcage', 'blood', 'fart', 'navel'; 'skin ~ fish scale'; maybe 'mouth', 'tongue', 'liver', 'wing'
• relatives: '(grand)mother', 'ancestor', 'child'
• animals: 'bear', 'otter', 'bird', 'crow', 'leech', 'louse'; maybe 'snake', 'shrimp', 'cicada', 'crab', 'fish'
• plants: 'taro', maybe 'flower', 'chaff', 'leaf'
• life: 'to die', 'to eat', 'to vomit', 'to weep', 'alive', 'blind', maybe 'sick', 'meat'
• 'water', 'fire', 'moon'; maybe 'rain', 'mist', 'star', 'cloud'
• 'name'
• 'boat', 'road'; maybe 'house'
• 'black', 'sharp', 'sour'
• 'to cut', 'to pinch'; maybe 'to bury', 'to cover', 'to plant', 'to slap', 'to weave'
• 'to arrive'; maybe 'to borrow', 'to sell'

"Maybe" items are either restricted to Formosan or one Kra-Dai branch (usually Hlai), and some could very well be loans. Even these aside though this still leaves not much chances of the two not being related.

-

mae wrote: ↑Wed Oct 17, 2018 7:40 pmHe also says that the classical reconstruction of 'child' for PAn is *aNak, but instead the evidence prefers a root *-lak. But anyone with even a trivial knowledge of Austronesian historical linguistics knows that the symbol *N describes a lateral;

The regular correspondence however seems to be Austronesian *N ~ Kra-Dai *n ('freshwater', 'moon', 'rain', 'bury'). Combined with *N > *n in Malayo-Polynesian, it looks doubtful if the analysis as an original lateral is really correct after all.

-

Tropylium wrote: ↑Wed Oct 17, 2018 6:16 pm Blench's data in the first paper, assembled across earlier works ups the ante much further than just numerals

I'm aware - when I said "would involve at least", I did not secretly mean "would involve only".
Even these aside though this still leaves not much chances of the two not being related.
[/quote]

Hardly. Leaving aside that a bunch of these may be coincidence, there are many languages that have undergone massive lexical borrowing in their histories, and I don't see how this can be ruled out here - indeed, Blench's "multigenetic model" (aka "I can compare any branch of Tai-Kadai with any branch of Formosan, because that's way easier") seems more compatible with a borrowing scenario than a genetic one.

mae wrote: ↑Wed Oct 17, 2018 10:56 pmNasal reflexes of a lateral are not "doubtful" at all. If we look at the different branches of Austronesian, they overwhelmingly represent lateral reflexes for that particular proto-phoneme.

Cognates no, the reconstruction, yes. You might be aware that Blench also proposes that the primary division into lots of small branches on Taiwan versus everything else in a huge Malayo-Polynesian branch is not warranted on archeological grounds. The linguistic evidence has turned out to be not solid either: there's for example recent work from Norquest which shows that also some "MP" languages could be used to set up "crypto-phonemes" that are lost widely, including demonstrating that *C is retained as distinct from *t in some of these. With this, the possible evidence for a MP subgroup falls to nothing but *N *S > *n *h. At the same time, by Norquest's data a similar large subgroup could be also set up where instead *f *g > *p *k, but which would then include also the Formosan languages. So clearly we have no reason to privilege the former hypothesis.

One possibility is that *N was something close to *n to begin with (e.g. *[ɳ]), so that *N > *n, if not outright an archaism, would be too trivial to use for subgrouping, while it would be rather the *N > lateral languages, all clustered on Taiwan, that form a single subgroup.

(There probably still will be some sort of a narrower "Bulk Austronesian" group left after realignments of this type, though.)

mae wrote: ↑Wed Oct 17, 2018 10:56 pmAlso, 'rain' is not an example of PAn *N ~ Kra-Dai *n (…)

Whoops, clearly going too fast right now. Key point being regardless that we have two different correspondences: *N ~ *l in words that are restricted to a single group, but *N ~ *n whenever there are cognates more widely across Kra-Dai. The latter surely is more likely to be native and the former more likely to be loans.

Salmoneus wrote: ↑Thu Oct 18, 2018 6:33 amLeaving aside that a bunch of these may be coincidence, there are many languages that have undergone massive lexical borrowing in their histories, and I don't see how this can be ruled out here

Apparently it needs to be pointed out once again: massive borrowing of core vocabulary is rare. Can you name even a single language that has borrowed some two dozen Swadesh-100 list items, all of them from the same source? Or indeed, more than a third of the "Yakhontov list" of the most genealogically stable 35 meanings. Even discounting single-branch examples, AN and KD have 14/35 matches.

Salmoneus wrote: ↑Thu Oct 18, 2018 6:33 amBlench's "multigenetic model" (aka "I can compare any branch of Tai-Kadai with any branch of Formosan, because that's way easier") seems more compatible with a borrowing scenario than a genetic one.

They're not mutually exclusive. It seems to me that Blench is heading towards interpreting the locally distributed words mostly as loans vs. the widely distributed ones as common inheritance (but yes, "multigenetic" seems like bad terminology for this). As noted above, we can already see hints that these layers have separate by sound correspondences, too.

One additional thing to note is that Blench is usually not the most trustworthy source for detailed phonological reconstruction. For starters, he mixes here different systems of reconstruction in the Austronesian and Hlai proto-forms. In AN *e, *C, *S and *ə, *ʈ, *ʂ, in Hlai *b, *əy and *ɓ, *əəy are the same thing. If he weren't citing stuff from previous studies, I'd want to check correspondences first before trusting the data to not involve superficial look-alikes.

I've done a lookover regardless though, since most earlier Austro-Tai sources don't seem to be available online. I've also only used the sub-branch reconstructions of Tai, Kam-Sui and Hlai that Blench gives, not Ostapirat's proto-KD (for which I'd like to see what branch-level data it's really based on).

1. Final consonants are well preserved in both families.
– Trivial correspondences for *p *t *k *m *n *ŋ *w *y.
– Clear non-trivial correspondences are AN *ʀ ~ Hlai ɦ (Tai and Kam-Sui ∅) and AN *N, *l ~ KD #n.
– AN *q corresponds to both KD #k, and Hlai *ʔ (the latter with no Tai or Kam-Sui cognates, but maybe only accidentally). Hlai *ʔ also corresponds to AN ∅.
– AN *i, *u have both #close-type and #ə + glide -type correspondences in KD. Apparently in AN, earlier word-final *əy, *əw > *i, *u. (However, this could also be a post-AN areal change. I'm pretty sure I've seen *-i, *-u > *-əy, *-əw posited for some AN descendants.)

2. Basic vowel correspondences between final syllables are roughly AN *a *i *u *ə ~ KD #aa #ii #uu #ə.
– Makes me think AN may have gone through a Berber-style vowel collapse: short vowels > *ə, long vowels > short.
– One non-trivial correspondence is short *a rather than long *aa in Tai when AN has the root structure *CaCa ('eye', 'thigh').
– Correspondences for E-type and O-type vowels in KD are too few to establish much of anything. Feasibly they could be conditioned by penult vowels: e.g. AN *qudaŋ 'shrimp' ~ Tai *kuŋ (if from *kr-), Hlai *raaŋ; AN *bulak ~ Tai *ɓlook 'flower'; AN *duʂa ~ Tai *sooŋ, Hlai *hluuʔ '2'. Further fragmentation within KD is evident though, and actually even just the correspondences of Hlai *uu in Tai are pretty heterogeneous according to this data.
– Tai has also *ɯə ~ AN *a, Hlai *aa, always with a close penult in AN ('moon', 'vomit', 'taro'). Maybe we should reconstruct *ɯ…aa in Austro-Tai.

3. Medial/initial consonants are slightly more munged, but often still recoverable.
– *m *n are stable. AN *l, *r, *ʀ all ~ KD #r, except AN *bVl ~ KD #ɓl.
– *p, *t, *k seem to be roughly preserved (with some latter sound changes in Hlai). *ʈ merges into *t in Tai and Kam-Sui, but gives *tʃʰ in Hlai ('eye', 'louse'). '3' might have developed as AT *tələw > KD #trəw > #ʈəw > Hlai *tʃʰʷəw.
– AN *qVt ~ KD #ɗ in 'liver', 'black', presumably through *ʔt. AN *qətut ~ Tai *k.tɤt 'fart' seems to be instead built on the above-mentioned *q ~ *k correspondence. Compare also AN *qudip ~ Tai *T.dip 'alive'.
- AN *ʂ and *s both ~ KD #s.
– 'Snake' shows AN *ʂVl ~ Hlai *ly. Perhaps this was rather *sl (the actual reflexes are l ~ ɬ ~ s ~ ts).
– Similarly, '2' shows AN *dVʂ ~ Hlai *hl > most Hlai /ɬ/, some /ɗ/ or /tθ/. Perhaps rather KD #d(V)s > Hlai *tɬ.
– AN *b mostly ~ KD #ɓ, but #b in 'shoulder' (or is this just an intervocalic allophone?), #pʰ in 'mist ~ rain' (?!)
– AN *d corresponds in Hlai with both *ɾ ('alive') and *r ('shrimp'). The latter only gives /r/ in one Hlai language while other languages have velars, which could suggest reconstructing instead KD #d, #kəd > Hlai *r, *kr. However this "*kr" also comes from AT liquids > KD #r, so probably not.
– AN *z, *dž, *ď, *w only occur once in the data (the first three correspond to various voiced coronals in KD, the last to Tai *k ~ Hlai *ɣ) and don't allow for any conclusions yet.

4. Penult vowel qualities seem to be mostly lost in KD, but not entirely.
– Hlai penult *u (not preserved anywhere, reconstructed for cases of /Cʷ-/ in some but not all varieties) corresponds to AN penult *u ('plant', 'alive', 'head').
– Proto-Hlai also has a similarly reconstructed penult *i, but this doesn't come up as such in Blench's data.
– Penult *i seems to trigger *p > Kam-Sui *py-, Hlai *fy- in 'tooth'.

This looks pretty damn convincing, particularly when we don’t have any super archaic forms of Tai-Kadai or Austronesian attested as we do with IE. Imagine having to reconstruct PIE with no Ancient Greek, Avestan, or Vedic Sanskrit evidence, let alone Anatolian evidence.

Forget Blench; it's Ostapirat that interests me.

Also, Sagart, Blust, et al.
I just want Ostapirat's 2018 papers/presentations without an account or paywall.

I have been thinking again about Indo-Uralic and genetics. It seems as if the Yamnaya culture, who probably were the PIE speakers, was dominated by two Y-DNA haplogroups, R1a in the north and R1b in the south, at least according to the Eupedia maps of which I don't know how reliable they are (they probably are imprecise but not utterly wrong). Now, R1a and R1b are related to each other, but at a much deeper level than the IE languages, and both originate south of the Caucasus if I am not mistaken. Uralic, in contrast, seems to be connected to a completely different Y-DNA haplogroup, N1c, originating in Siberia.

So what happened? A possible explanation would be this: R1a and R1b came from the south and originally spoke languages unrelated to IE. R1b would become the Dniepr-Donets culture and R1a, Samara. The N1c north of Samara spoke Proto-Indo-Uralic, a Mitian language. Then the R1a shifted from their original language to Indo-Uralic, altering the language in the process to become PIE0 (the stage of Pre-PIE before the Great Vowel Collapse). Samara developed into Khvalynsk, which expanded south into the Dniepr-Donets area, imposing their language which underwent further changes, such as the Great Vowel Collapse, and became PIE1. Meanwhile, the N1c people in the north continued to speak their (more conservative) Indo-Uralic language which evolved into Proto-Uralic.

I have to confess, as I did before, that my knowledge of human population genetics is limited. This is a difficult and fast-moving science, where laypeople like me have to rely on popular summaries which tend to be out of date before they reach the public. Accordingly, the above may be utter hogwash

Let me add some more. While Proto-Uralic is a fairly typical Mitian language, quite similar to Eskimo-Aleut and to a lesser degree the Altaic languages, PIE is quite different. The four main phonological changes can be summed up as follows:

1. Development of a third phonation type of stops (maybe ejective or some other highly marked phonation).
2. Loss of the Proto-Mitian palatal consonant series except */j/, probably by merger with the dentals.
3. Development of palatalized and labialized velar stops.
4. The Great Vowel Collapse: all non-high vowels (including former high vowels that had been lowered before sonorants) merge into */a/ (which later splits up into *e, *o and *a, and their long counterparts, according to the rules of ablaut and the vowel-colouring effects of laryngeals).

What is the relative chronology of all these? It is IMHO quite likely that #2 predated #3, as otherwise, one would expect some kind of interference between the old palatal series and the new palatalized velars. On the other hand, #3 is probably closely connected with #4 but logically precedes it; it is IMHO best to assume that the three velar series first developed as allophones next to the different vowels, and became phonemic when the GVC took place. #1 could have happened at any point.

I'd therefore guess that #2 happened at the transition from Proto-Indo-Uralic to PIE0, and #3 and #4, at that from PIE0 to PIE1. I don't know yet where to put #1, but like #3 and #4, it looks like the influence of a substratum typologically (not necessarily genealogically) associated with Caucasian (especially NWC) languages, which would be the Dniepr-Donets language, so I'd ascribe it to the transition from PIE0 to PIE1 as well. It is likely that PIE0 had already lost the palatal series, because the lack of such sounds is not a typically Caucasian feature, to the contrary! This probably happened farther away from the Caucasus, i.e. in the PIU > PIE0 transition.

It should be pointed out that, even in the extremely unlikely event of there being a close Uralic-IE relationship:

- 1. may never have happened (there's no reason to think Uralic didn't instead lose its third phonation)
- 2. may never have happened (there's no evidence of IE ever having had a fourth dorsal series and you're postulating one purely to connect it to Uralic; yet even if the connexion is true, it could be Uralic that is innovative)
- 3. may never have happened, although to be fair I suspect it did (it seems an unusual arrangement and one that quickly decays in IE languages)
- 4. may never have happened (there is no reason to think there ever were any other vowels, and you're just postulating them to connect it to Uralic; nor, for that matter, is it clear that there was ever a 'collapse' to one vowel, given that PIE shows some *o that can't easily be explained by ablaut, even by analogy, and given that PIE does of course also have *i and *u, even if you'd prefer to call them something else).


As a general point, I think it needs to be said that if you invent enough huge swathes of lost phonology for every language, you can make everything look like everything else. I mean, to make your relationship look appealing, you're having to:

- increase the stop inventory by 1/3rd
- decrease the stop inventory by 1/4ter
- increase the stop inventory again by 2/3rds
- decrease the vowel inventory by, what, 2/3rds, or 4/5ths?
- increase the vowel inventory again by a further 200%

... all for two languages that are supposedly close relatives living right next to one another.

Tropylium wrote: ↑Thu Oct 18, 2018 11:49 amCan you name even a single language that has borrowed some two dozen Swadesh-100 list items, all of them from the same source? Or indeed, more than a third of the "Yakhontov list" of the most genealogically stable 35 meanings. Even discounting single-branch examples, AN and KD have 14/35 matches.

Isn't this assuming that those words already have the same source? If we have no regular sound correspondences, then coincidence becomes very easy to rationalize as meaningful data. You talk about the Swadesh list as if we had any such thing for reconstructed forms of either language family, or as if each language had a tidy set of semantically exclusive terms. We have wildly varying attested modern forms, and reconstructed forms that sometimes look similar, if you cast a wide enough net semantically, but which do not correspond in any systematic way. Blench seems to change his mind constantly about whether a monosyllabic Thai word needs to match the first or second syllable of a reconstructed PAN word. And we can only take his word for it that none of these languages have any other semantically relevant words which he discarded for not fitting the mold.

The length of your correspondence list is not what's impressive; it's the statistical significance of it.

Salmoneus, your objections may be valid, but you are really behaving like a plonkster. If you don't know what a "plonkster" is, that is someone who joins a discussion only to tell the participants that their discussion is meaningless. He may be right (depending on what the discussion is about), but his behaviour is impolite and pointless. For instance, I don't believe in astrology, but I don't join astrology fora just in order to tell the people there that I don't believe in astrology. I leave them to themselves and refrain from joining their discussions - because they won't want to hear it and nothing would be won that way.

We all know that you are extremely sceptical against all sorts of macrofamily hypotheses or attempts at recovering information about languages that are utterly lost in history. Many linguists share your opinion, and I at least am aware that I am just speculating and that my hypotheses may be wrong. (Hence, I don't boldly trumpet them on a web site, but submit them to discussion in the appropriate threads here.) Nobody here wants to hear your ignoramus et ignorabimus ad aeternum routine, I think; we all know how you tick, you have shown often enough, and you need not carve that out over and over again each time someone posts an idea here.

So you'd only like to "submit your ideas for discussion" with people who promise to agree with them? Because rather than defend your ideas, you'd like to harp on yet again about your personal dislike of me and my terrible failings? How original.


There's speculation, and then there's crackpottery. It's possible to speculate on the basis of evidence and reason, and it's also possible to clearly distinguish between these, and random stuff you invent for conworlding purposes. My problem isn't that you say things that aren't provable - as you may have noticed, I've frequently engaged in speculation on exactly the same topics - it's that you mix together facts, theories, wild hypotheses and fiction and you don't distinguish between them, and that's irresponsible. I'm not skeptical of macrofamily theorising, I'm skeptical of how you do it. As, for example, the way you constantly treat your 'great vowel collapse' with seemingly exactly the same weight as consensus reconstructions. To use your analogy: when somebody comes to an astronomy group and happily mixes together astronomy and astrology in the same sentence, then yeah, you can bet somebody's going to want them to try to demarcate those topics more clearly.

If you can't recognise that an approach that begins by assuming a series of massive phonological collapses and massive phonological splits is kind of methodologically perilous, that's not my problem. If you don't think that that series of massive processes is at least slightly incongruous beside the equally precious assumption that these two languages were particularly closely related, then I think that's your problem.

Because, in the end, if you don't acknowledge the validity of criticism, how can you hope to make "progress" in your "research"? If it's not OK to point out the implausibility of one idea, how will you ever improve your ideas? If you genuinely want to link two languages together, then you shouldn't leap to the easiest, least unfalsifiable approach, one that reshuffles the pack of one language so dramatically in order to fit another - because if that's your first line of attack, what evidence could ever disprove it? What evidence could ever cause you to reevaluate that theory, when you can retool that methodology to turn anything into anything?

If you actually want to make progress, begin with an open mind. Identify problems, don't just assume whatever solution Old Albic prefers. If, for example, Indo-Uralic is valid, then the fact that one language has a series that the other doesn't have is a problem, with several possible solutions - but instead, you state as fact, but without any stated reason or evidence, that "the phonological change", the solution is of a particular sort. Why? That's not reasoning about macrofamily relationships, that's macrofamily fan-fiction!

And I mean, OK, maybe you're right and maybe you have great reasons for thinking that. But if you gave us a problem, suggested a solution, and shared your brilliant reasons for that solution, you might actually persuade people! By instead just stating your pet theory as fact, you're liable to drive away some, and mislead others. And that's not in anybody's interests. You seem to keep getting carried away and forgeting what's fact and what's fiction. Each time, you pretend to believe my objections are "valid", but since you then ignore them, clearly you don't! You act as though I'm trying to cancel Christmas and outlaw speculation, but all I've asked for is for you make yourself more clear in your wording - what you think we can agree on, what you're proposing, why you're proposing it, instead of blanket assertions in which your opinion is treated as gospel. It astonishes me that you're so unwilling to do this.


And don't fucking flame me. I didn't shout insults at you, don't shout insults at me. We're neither of us eight, nor President, I'd hope we could do without that.

Sal, I have pretty similar views as you on the topic but the discussion between me and the other posters here has never turned ugly

mèþru wrote: ↑Wed Jan 16, 2019 6:12 pm Sal, I have pretty similar views as you on the topic but the discussion between me and the other posters here has never turned ugly

Well congratulations.

And I'm sorry if that post was overly honest, but I'm getting fed up of his my-way-and-no-criticisms approach - even before the personal flaming started coming out.