The Great Macrofamily thread: Indo-Uralic, Altaic, Eurasiatic, Nostratic etc.

[WeepingElf]

This paper by Tropylium (thanks for posting the link on the Great PIE Thread!) makes me think again about possible sound correspondences between IE and Uralic (which of course can only be established by the drudgework of the comparative method, but speculations like this may tell us where to look). My idea is that both the *T set and the *Dh set of PIE stops correspond to Uralic stops, with the difference between the two due to a lexically specified (no influence from ablaut!) prosodic feature within Pre-PIE (maybe a continuation of a pre-GVC distinction in vowels?), while the *D set corresponds to the Uralic spirants (if that's what they were, I mean the *ð *ð' *x set). Which happens to lack the labial member (perhaps merged with *w in Pre-PIU, which may explain why some PIE words have *w where one would expect a stop) which may or may not have been missing from the PIE *D set as well.

Alas, as I said above, this needs to be explored by means of the comparative method.

[Nortaneous]

I wouldn't expect breathy voice to become creaky voice or vice versa like that -- unless you had interchange between a glottal stop and a glottal fricative in pre-PIE.

[WeepingElf]

As Howl has pointed out in the Great PIE Thread, one problem with my idea is that Uralic doesn't allow *ð and *x in initial position, so one would have to assume that these consonants were lost, either by changing into something else or by dropping (vowel-initial words seem to be very common in PU, so the missing initial spirants may hide there). As I said, this needs to be examined by means of the comparative method.

@Nortaneous: I don't understand why you bring "creaky voice" to the discussion of this (I didn't even understand that your post was a reply on mine - perhaps it isn't after all?). Neither the IE *D set nor the Uralic *ð/ð'/x set is creaky voiced in the standard reconstructions, so I see no reason why this should stand in the way of my idea.

But after all, this is just a spontaneous idea, and I don't know how good it is yet. It may turn out to be utter bullfrogs.

[Tropylium]

As I think I mentioned last thread already, Uralic *d just has very few IE correspondences at all, and some semantically good-looking cases don't offer any regularity: #śedäm ~ *ḱerd- 'heart', *edə- ~ *h₂ant- 'front' (well, both have an *RT cluster in IE, but *nt ~ *nt would seem more expected).

one problem with my idea is that Uralic doesn't allow *ð and *x in initial position, so one would have to assume that these consonants were lost, either by changing into something else or by dropping (vowel-initial words seem to be very common in PU, so the missing initial spirants may hide there).

Sounds possible, but yes, I would like to have examples around. Initially I can only think of comparisons like *deh₃- ~ *amta- 'to give' where dealing with the medial consonants would requre a lot of assumptions.

[WeepingElf]

I admit that my idea is adventurous and probably misguided; it was just an idea I had which I wanted to share with you. Actually, I'd rather guess that the PIE *D set is, somewhat like the Itelmen ejectives, the result of cluster developments - or all words with such stops are borrowed from a (Caucasian?) language with ejectives, and there is no native phonological pathway to these sounds at all (somewhat like the clicks in southern Bantu languages). I have been hypothesizing for quite some time (and I am not the first; C. C. Uhlenbeck did so in the 1930s already) that PIE emerged from a language related to Uralic on a Caucasian-like substratum. The Uralic-like language could have been that of the Khvalynsk culture and the Caucasian-like one that of the Dniepr-Donets culture. When the Khvalynsk took over Dniepr-Donets, resulting in the Sredny Stog culture, their language turned into (Early) PIE.

[Znex]

A bit more of a crackpot idea here perhaps, but this sparked in mind after seeing that many Turkic languages have the same change *b > m / _...N as hypothesised for Basque, and then noticing how similar the pronouns for the Mitian languages and the "Altaic" languages are under this light and so tried comparing them and a few other words. If anything, this might at least be a indicator just to show the extent of an older Sprachbund between them all, predating the newer common features of stuff like vowel harmony and agglutinativity:

Code: Select all

{1SG}
PIE *éǵ, (e)me-					PB *ene- << **ebene				PU *(e)mi-n-						PT *(e)pen					PM *pi							PTN *pi
> Hit. ʔuk, ammu-, amm-				> en-, ene- {1SG}				> Mansi am, ānəm-; Hung. én 				> Chuv. eb(ě), man-				> Mong. pi, na-, min-; Sant. pi				> Manc. pi, min-; Sib. pi, min-
> Sans. ahám, ma-				> mB enne- {1SG.OBL}				> En. mod'i; Ngan. mənə					> Turk. ben, ban-, -im; Crim. men, mań- 	> Mogh. bi, min-; Dag. bī, na-, min-			> Evn. bi, min-; Udeg. bi, men-
> Anc.Gr. egṓ, (e)mo-, (e)me			> B en- {VRB 1SG.OBL}, ni {1SG.ABS}		> Fin. minu- > Coll.Fin. mä~mie, mu-			> Uygh. men, mën-, -im; Yak. min
> Lat. egō, me-											> Sami mon, mu-

{2SG}
PIE *tī́, t(w)(e)- < **tī, tw-			PB *tʰi, tʰo					PU *ti-n-, ?tu-n-					PT *(e)sen					PM *tʃʰi						PTN *si < **tʰi
> Hit. tsīk, tu-, twē-				> B hi {2SGFAM}, to {2SGFAM.M.VOC},		> Hung. te						> Chuv. es(ě), san-				> Mong. tʃʰi, tʃʰin- {2SGFAM}; Sant. tʂʰi, tʂʰin-	> Manc. si, sin- 
> Sans. tvám, tú-, t(a)v-			ho-re {2SGFAM-GEN}, txo(txo) {boy.VOC}		> En. toďi, Ngan. tənə					> Turk. sen, san-, -in; Crim. sen, sań-		> Mogh. tʃi, tʃin-; Dag. ʃī, ʃa-, ʃin-			> Evn. si, sin-
> Anc.Gr. sý, so-, se										> Fin. sinu- > Coll.Fin. sä~sie, su-			> Uygh. sen, sën-, -iŋ; Yak. en
> Lat. tū, tu-, te-										> Sami don, du-

{1PL(EXCL)}
PIE *wey, n̥s- (< **m̥-s-) 			PB N/A						PU *me							PT *pī-ŕ/z					PM *pi-t						PTN *pi-t
> Hit. wēs, ants-										> Hung. mi						> Chuv. ebir, bir-				> Mong. pit; Sant. pit					> Manc. pe, men- 
> Sans. váyam, asma-										> Ngan. mɨŋ						> Turk. biz, -imiz; Crim. biz			> Mogh. bid; Dag. bʲed					> Evn. mit
> Lat. nōs‚ nes-/nos-										> Fin. me(i)-						> Uygh. biz, -imiz; Yak. bih-igi
> Germ. wīz, uns-										> Sami mi(i)-

{1PL(INCL)}
PIE ?		 				PB *gu??					PU ?							PT ?						PM *pa							PTN *pu

{2PL}
PIE *yuH, us- (< **tw-s-?)			PB *s-tʰu?					PU *te							PT *sī-ŕ/z					PM *tʰa							PTN *su
> Sans. yūyám, yuʂm-				> AB zu {2PL}					> Hung. te						> Chuv. esir, sir-				> Mong. tʰa, tʰan- {2SGPOL;2PL}; Sant. tʰa, tʰan-	> Manc. suwe, suwen-
> Lat. wōs, wes-/wos- 				> B zu {2SGPOL}, zu-ek {2PL}			> Ngan. tɨŋ						> Turk. siz, -iniz; Crim. siz			> Mogh. to(d), ton-; Dag. tā, tān-			> Evn. su, sun-
> Germ. yūz, us-										> Fin. te(i)						> Uygh. siz, -ingiz; Yak. ih-igi
> Balt.-Slav. yūs, yu-										> Sami di(i)-

{2}
PIE *dwóh₁, dwi-					PB *bi						PU *ka~k-ta?						PT *tvi >> ki					PM *tʃy-r						PTN *tʃö-r
> Toch.A. wu, we				> B bi,						> Mansi kit; Hung. ket					> O.Bulg. tvi-rem {second}			> Khit. tʃyren						> Manc. tʃuwe; Sib. tʃu
> Sans. dvá					bigarren {second}, hamabi {12},			> Nen. sidja, Komi kɨk					> Chuv. i-kě												> Evn. dʲūr; Udeg. dʒū
> Anc.Gr. dýo					biki {twin}, biri {lung}, bihotz {heart}	> Fin. kaksi						> Turk. i-ki; Crim. e-ki 
> Lat. duo											> Sami guokte						> Uygh. i-kki; Yak. i-kki
> Arm. er-ku

{DEMDIST}
PIE *so, to-					PB ? *tʰa-r?, *tʰo?				PU *tä, to						PT *-ti-					PM *tʰe-r						PTN *tʰe-r
> Sans. sa, ta-					> CB *har {DEMDIST} > B -a {DEF}, etc.,		> Hung. túl {too}, tegnap {yesterday}			> *em-ti {now} > Chuv. ěntě,			> Mong. tʰer, tʰȳ; Sant. tʰere				> Manc. tʰere, tʰe-
> Anc.Gr. ho, to-				hartara {so}, -tara {MODDEM}			> Nen. tā, tʲei {yesterday}				Turk. ʃ-imdi, Crim. indi,										> Evn. tar, ta-
> Lat. sī {if}, tam {so}, iste {DEMH}, etc. 	> to {2SGFAM.M.VOC}				> Fin. tuo {DEMDIST}, tämä {DEMPRX} > Coll.Fin. tää	Uzb. endi 
> Germ. sa, θa-											> Sami duot {DEMDIST}, dāt {DEMPRX}

{WH}
PIE *kʷi-, kʷo-					PB *s-kʰe-r??					PU *ku, ke						PT *kʰe-m					PM *kʰe-n						PTN *kʷʰe-n
> Hitt. kʷis, kʷi-				> B zer {what?,how?}				> Hung. ki, hol {where?}, hoɟ {how?}			> Chuv. kam					> Mong. xen; Sant. kʰiən				> Manc. we
> Sans. ka, kim {what?,how?}									> Nen. ʃā, ku-						> Turk. kim; Bash. kem				> Mog. ken; Dag. xən					> Evn. e-ku-n
> Anc.Gr. tis, ti- 										> Fin. kuka, ke(n)-, ku-				> Uzb. kim 
> Lat. quis, qui-, quo-, cu-									> Sami kie
> Germ. hwis/hwos, hwi-/hwo-

{WHorNEG}
PIE *ne, meh₁					PB *na, no-r					PU *ne, mi						PT *ne, -ma					PM ?							PTN ?
> Toch. mā {NEG} 				> B non {where?}, nor {who?},			> Hung. nem {NEG}, mi {what?;how,so}			> Chuv. měn {what?}
> Sans. na {NEG}, mā {PROH}			noiz {when?}					> Udm. ma {what?}					> Turk. ne {what?}, -ma {NEG}; Crim. ni {what?}
> Anc.Gr. mē {PROH} 				> inon {no/somewhere}, inor {no/someone}	> Fin. mi-kä, mi- {what?}				> Uygh. nëme {what?} 
> Lat. ne {NEG}

{out,beyond > NEG}
PIE *eǵʰ-s 					PB *e(s)gi < **egi-s				PU *e-							PT ?						PM ?							PTN ?
> Anc.Gr. eks, ek {out of,from,by} 		> B -egi {EXCS}, ez- {extra-},			> Fin. ei {NEGVRB}, Sami ii {NEGVRB}
> Lat. eks {out of;throughout;PRIV}		ez {NEG}
> O.Ir. a {out of,from}

[Tropylium]

People here might be interested to hear that Alexis Manaster Ramen has recently been uploading several backcatalog (mostly 90s) review/method papers on Nostratic and Altaic on academia.edu. I'd recommend as a starting point the following two:
On Illich-Svitych's Nostratic Theory
Glass Houses: Greenberg, Ringe, and the Mathematics of Comparative Linguistics

A selection of others:

Nostratic from a Typological Point of View
The use of reconstructed forms in Nostratic studies
On the Nostratic inclusive/exclusive
On Indo-European Triune Velars and Nostratic Front Rounded Vowels
Clusters or affricates in Kartvelian or Nostratic
On "Some Thoughts on Indo-European–Kartvelian Relations"
Body Part Terms as Evidence in Favor of the Altaic Theory
The Altaic Debate and the Question of Cognate Numerals

[WeepingElf]

Thank you, these look interesting!

[Ketsuban]

On the other hand, if you like me prefer an incrementalist approach to Altaics you might be interested in Alexander Francis-Ratte's 2016 dissertation which presents reasons to reject Vovin's radical skepticism about the question of Korean-Japonic common origin and builds atop prior work by the likes of Samuel Martin and John Whitman to reconstruct a common ancestor. The author informs me that he intends to publish a "more edited version" of the dissertation in the future.

[WeepingElf]

Yes, this is IMHO a healthier business than the Nostratic (or large macrofamilies in general) thing where a large family tree is set up on flimsy evidence, and then languages from maximally distant branches are compared to get down to the root. I call that the "top-down" approach, and as I have said here several earlier times, I prefer the "bottom-up" approach which starts at closer relationships such as Indo-Uralic or Korean-Japanese in order to get more and more parts of the tree reconstructed. Of course, with the bottom-up approach, there is the problem of finding out just which languages form such nodes, but I don't think this is unsurmountable.

[Richard W]

Of course, with the bottom-up approach, there is the problem of finding out just which languages form such nodes, but I don't think this is unsurmountable.

Nor strictly necessary, PIE was reconstructed well enough without resolving the branching order. The large-scale branching order of Afroasiatic seems to be irrecoverable,

[Tropylium]

Irrecoverable from casual inspection, perhaps less so once we have some reliable reconstructions.

I would think the merits of top-down versus bottom-up depends a lot on what we're doing. Reconstructing a family top-down is generally a good idea since, for language families that don't have an established reconstruction yet, "subgroups" are quite often casually eyeballed typological-geographical areas that cannot be relied on to be always real genetic units. (Re Afrasian, Cushitic and Chadic still have this problem.)

The establishment of new relationships by contrast kind of needs to be bottom-up: even if you compared Wolof directly to Zulu with extensive effort (or, perhaps, a bit more widely, Atlantic with Bantu) and manage to find them related, this proves nothing whatsoever about the five hundred intervening languages in Nigeria & Cameroon if you haven't even examined them in any detail. And sometimes a closer look will turn up cases like hey, Chuvash is actually an outlier of Turkic that just happens to look typologically a bit Finno-Ugricized.

[WeepingElf]

Irrecoverable from casual inspection, perhaps less so once we have some reliable reconstructions.

I would think the merits of top-down versus bottom-up depends a lot on what we're doing. Reconstructing a family top-down is generally a good idea since, for language families that don't have an established reconstruction yet, "subgroups" are quite often casually eyeballed typological-geographical areas that cannot be relied on to be always real genetic units. (Re Afrasian, Cushitic and Chadic still have this problem.)

It tells a lot that exactly in the best-studied language family of all, namely Indo-European, there is no consensus on subgrouping above such entities as Italic, Germanic, Balto-Slavic or Indo-Iranian. How then can people be sure about subgroups in families about which they know less? This has been puzzling me for quite a while. It seems to me as if many subgrouping schemes are based on lexical statistics, which is rather easy to do, but has turned out to be not too reliable; the Indo-Europeanists, in contrast, try to rely on shared phonological and morphological innovations for subgrouping - and run into a wall because the isoglosses intersect in so complex ways that no uncontroversial family tree can be drawn. With most other families, we don't know enough about the phonological and morphological developments in the individual languages to even try this, so everyone still works with lexical statistics, with all its shortcomings. One family where the shared innovation approach has become fairly recently viable is Uralic - and the traditional classification based on lexical statistics has been challenged.

The establishment of new relationships by contrast kind of needs to be bottom-up: even if you compared Wolof directly to Zulu with extensive effort (or, perhaps, a bit more widely, Atlantic with Bantu) and manage to find them related, this proves nothing whatsoever about the five hundred intervening languages in Nigeria & Cameroon if you haven't even examined them in any detail. And sometimes a closer look will turn up cases like hey, Chuvash is actually an outlier of Turkic that just happens to look typologically a bit Finno-Ugricized.

Yep.

[Richard W]

It tells a lot that exactly in the best-studied language family of all, namely Indo-European, there is no consensus on subgrouping above such entities as Italic, Germanic, Balto-Slavic or Indo-Iranian. <snip>... the Indo-Europeanists, in contrast, try to rely on shared phonological and morphological innovations for subgrouping - and run into a wall because the isoglosses intersect in so complex ways that no uncontroversial family tree can be drawn.

Isn't that exactly what one should expect from a dialect continuum? Wasn't there a similar mess with 'Western' Romance (sensu lato, i.e. including Sicilian) and modern continental West Germanic a hundred years ago?

[WeepingElf]

It tells a lot that exactly in the best-studied language family of all, namely Indo-European, there is no consensus on subgrouping above such entities as Italic, Germanic, Balto-Slavic or Indo-Iranian. <snip>... the Indo-Europeanists, in contrast, try to rely on shared phonological and morphological innovations for subgrouping - and run into a wall because the isoglosses intersect in so complex ways that no uncontroversial family tree can be drawn.

Isn't that exactly what one should expect from a dialect continuum? Wasn't there a similar mess with 'Western' Romance (sensu lato, i.e. including Sicilian) and modern continental West Germanic a hundred years ago?

Exactly that. Core IE evidently was a dialect continuum for quite some time, perhaps as late as 2000 BC. So at that time, you could travel from the Rhine to the Indus without ever crossing a language boundary separating mutually incomprehensible languages, though an Indo-European from the Rhine probably couldn't understand one from the Indus.

[jal]

Exactly that. Core IE evidently was a dialect continuum for quite some time, perhaps as late as 2000 BC. So at that time, you could travel from the Rhine to the Indus without ever crossing a language boundary separating mutually incomprehensible languages, though an Indo-European from the Rhine probably couldn't understand one from the Indus.

Which could also explain why innovations could spread in patterns that can't be explained if the language development followed a strict tree model.


JAL

[Salmoneus]

There's a couple of ways in which the IE analogy is special pleading, though...


- it's easy to say that we can't know any groupings above the level of the primary branches. But that's a circular definition. In particular, we talk about Indo-Iranian as a 'primary branch', but that's only because we can reconstruct it - it's a grouping of three different branches, and it's dated to thousands of years before some of the other 'primary branches'. Being able to reconstruct the Indo-Iranian branch - which, let's not forget, includes 2 out of 3 surviving Indo-European languages and about 1 in every 2 Indo-European speakers - is a massive accomplishment, too easily fobbed off.

- most Indo-European branches no longer exist, and many were never attested, largely due to the secondary expansions of Celtic, Slavic and Iranian (as well as the intrusion of Uralic and Turkic languages).

- we have pretty much worked out the phylogeny of IE. It's obviously a strawman to expect very early dialects of a massively-and-rapidy-expanding language to follow a pure tree structure, but that doesn't prevent us having a very good idea of the relationships:

- Anatolian split off first
- then Tocharian, which went east while everything else was going west
- next probably Greek and Armenian split off, going south, or remaining south, occupying the western steppe (or maybe even early pannonia?) while everyone else formed the corded ware culture up north
- the corded ware culture more or less split into two dialect groups (or possibly full languages), which we can call western and central-eastern
- western split into a northern branch and a southern branch
- the southern branch split into Italic and Celtic
- the northern branch, which remained more closely connected to its baltic neighbours to its east, eventually yielded germanic
- the central-eastern branch re-invaded the steppe (the Sintashta culture), coming back in contact with the southeastern languages (Greek, Armenian, their extinct relatives), sharing some grammatical and phonological innovations with them
- the Sintashta languages broke into Indo-Aryans, moving south rapidly, and Iranians, staying north for longer, and the Nuristani languages (it's not clear what they did)
- meanwhile, the remaining central group continued to expand to the south (or had done so already), yielding eventually Albanian. Unfortunately, this area has all been secondarily bulldozed, so where exactly Albanian, Illyrian, Dacian, Thraco-Cimmerian and Phrygian fit in is unclear, only one language surviving.
- the remaing core languages are Balto-Slavic, Slavic just happening to be a language that exploded many centuries later.

[mèþru]

I think that much of what you wrote on branching is disputable and people are still far from a consensus there.

I think there is pretty good evidence that Thracian, Dacian and Slavic are subbranches of Baltic. I also think that Italo-Celtic is not a proper clade but a sprachbund that probably included several para-Italic and para-Celtic languages like Ligurian and Lusitanian.

[Richard W]

Where do "the remaining core languages" come from? The "remaining central group" also springs from nowhere.

I don't believe there is consensus on the existence of 'Italo-Celtic' as a reasonable node.

[Moose-tache]

I think there is pretty good evidence that Thracian, Dacian and Slavic are subbranches of Baltic. I also think that Italo-Celtic is not a proper clade but a sprachbund that probably included several para-Italic and para-Celtic languages like Ligurian and Lusitanian.

All of this is also subject to the same skepticism that we just dumped on Salmoneus.

Thracian shares some sound changes with Baltic (d and dh collapse, satem, partial merger of a and o), but these are all common changes found in multiple "branches." And let's not forget: Dacian doesn't exist. Anyone who presents you with a list of Dacian words is a charlatan.

As for Italic, the Italic languages do share a few sound changes, but not many. And as with Baltic and Thracian, they could be a coincidence ("Oh, you don't like pronouncing voiced aspirates either, huh? Small world!"). But "sprachbund?" How would we measure that? If this sprachbund also explains the small number of sound changes and grammatical features that the Italic languages share, then it quickly becomes a circular argument.