The Great Proto-Indo-European Thread

[WeepingElf]

I am once again thinking about the glottalic hypothesis.

One language family which is sometimes drawn into this discussion is Kartvelian. However, the constraints in PIE which led to the glottalic theory do not exist there. I think it suffices to mention one Georgian word: k'op'e 'ladle', which does not only have two ejectives in the root, but one of them is even labial!

Where similar constraints existed is Akkadian. Like all Semitic languages, it lacked a labial emphatic stop, and it disallows (unlike Proto-Semitic) two emphatics in a root (in PS roots with two emphatics, one is dissimilated, see Geers' Law). Alas, we don't know what the Akkadian emphatic consonants really were, but most semiticists assume that they were ejectives.

Whatever the PIE voiced stops started as, they probably were some kind of highly marked type. It would be interesting to sort out what their cognates are in Uralic (if that family is related to IE at all). Proto-Uralic had at least one affricate, a class that did not exist in PIE. Did this become PIE0 *t'? ("PIE0" is my term for a very early stage of PIE before the Great Vowel Collapse and the associated changes; in PIE0 there would have been just two "emphatic" consonants, *t' and *k', as the velar series has not yet split into three. "PIE1" is the stage after the Great Vowel Collapse but before ablaut, with three velar series already but only three vowels; "PIE2" is Pre-Anatolian PIE, and "PIE3" Post-Anatolian PIE, i.e. the PIE of the standard model.) But what about the velar "emphatic", PIE0 *k'?

[Soap]

I dont think that modern Georgian really has much bearing on what Caucasian languages were like 5000 years ago, though. PIE's homeland is usually placed on the other side of the mountains anyway. I think that if the glottalic theory is true, it's reasonable to assume that PIE acquired its ejectives by areal influence from surrounding languages which were not closely related to PIE, and that it may never have acquired the hallmark features of those languages such as (at least today) presence of bilabial ejectives and of ejectives in unstressed syllables.

The Semitic setup is tempting to link with PIE, and Semitic plays a part in many writeups of the linguistic landscape of pre-PIE Europe, but I'm not aware of any convincing evidence of the Semitic family ever being north of the Caucasians or the Mediterranean Sea in prehistoric times; it's all speculation. On the plus side, though, we do have strong evidence that Semitic phonologies were quite similar to their present form even 5000 years ago, and possibly longer, since it has a rich literary history and can be compared to related languages that diverged further back in time.

I believe in Indo-Uralic myself, but I think most people who support both Indo-Uralic and the glottalic theory have the glotticals being lost by Uralic rather than acquired by PIE. This may be motivated by the idea that the shared proto-language of both is usually reconstructed as being somewhat near the Middle East, with the Uralic speakers fanning out across Siberia while the PIE people stayed put. I'm not sure if this idea is well-established or not, though, since I suspect it's motivated largely by the need to reconstruct a language even further back in time that was spoken either in the Middle East or in the Horn of Africa, since Semitic is usually also considered to be related and its homeland is fairly well established.

However, if the Uralic affricate is related to the PIE glottalics, perhaps the glottalics could go back to Proto-Indo-Uralic, and have a (likely conditional) sound change that changes them into affricates in the Uralic branch only. Then it would be likely that the velar affricate would shift to either /k/ or /x/ unconditionally, leaving no trace of its previous existence. The labial affricate, if there ever was one, would likely also shift to /p/ or perhaps /f/ ... though there is no need to reconstruct a labial in the first place if we assume the lack of /p'/ in PIE0 is inherited from Indo-Uralic. That would leave just the coronal series, which may or may not have had multiple members that later merged, and may or may not have had a palatalized form that disappeared in PIE.

[WeepingElf]

The parallel between "Glottalic PIE" and Akkadian is only meant as a typological one; I don't believe in either a "genetic" relationship or a close contact between IE and Semitic/Afrasian. Nor Kartvelian (though the latter may perhaps be an early offshoot of Mitian, but this is very uncertain). (And Geo. k'op'e has cognates in other Kartvelian languages which likewise show those two ejectives - also, it is not the only word of that kind - so we know that the Geers' Law like constraints weren't in place in Proto-Kartvelian either.) I don't know about NWC or NEC, though, of which NWC may have been a neighbour (but an unrelated one!) of PIE.

We two agree that the closest living kin of IE is most likely Uralic (though this cannot be considered proven yet!). Yet, Indo-Uralic probably had further relatives, for which Altaic and Eskimo-Aleut are IMHO the best candidates. Eskimo-Aleut has a similar consonant inventory as Uralic (basically, it adds a uvular series and a lateral fricative), and the - quite similar - inventory Fortescue reconstructs for Uralo-Siberian (Uralic, Yukaghir, Chuktoko-Kamchatkan, Eskimo-Aleut) may give a good idea of what the Proto-Mitian inventory could have been like (better, at any rate, than either Dolgopolsky's or Bomhard's Proto-Nostratic inventories, which IMHO are bloated by the inclusion of Afrasian, Dravidian and Kartvelian and by failure to recognize conditioned changes). This inventory has one class of voiceless stops and one class of voiced fricatives, and no ejectives anywhere. (There are also sibilants, nasals, liquids and semivowels, of course.)

Which means that the Proto-Uralic system is probably more conservative than the PIE one. Now I consider it likely that the PIE "voiced aspirates" really were voiced fricatives (at least in the northwestern dialect), which leaves the voiced (unaspirated) stops (which I refer to by the agnostic term "emphatics") to account for. I noticed that PU (and probably PIU) had at least one affricate, perhaps two (alveolar and postalveolar; which would have merged in the loss of the alveolar/palatal opposition in PIE); where did it go in PIE? To *t' (standard model *d) perhaps?

Alas, all these questions can only be answered by comparing IE and Uralic lexicon in order to find out which sounds actually correspond to which. Saying that "language A and language B both have the phonemes X, Y and Z, so they are probably cognate" may fall into the trap of the phonemes having been permuted in one of them. Consider how Greek and Armenian both have "the same" voiceless/voiced/aspirated triads in their stop inventories, but these do not correspond trivially because Armenian has shifted its stops.

[Šọ̈́gala]

Per Carrasquer Vidal (a Nostraticist), PIE tenuis correspond to Proto-Kartvelian ejective; PIE plain voiced (ejective according to the glottalic theory) to Proto-Kartv. voiced; and PIE voiced aspirate to Proto-Kartv. voiceless aspirate, except for *bʰ which corresponds to Proto-Kartv. *b.

This seems to make sense in terms of the classical model of PIE phonology (probably corresponding to a late phase of PIE development), or even early Indo-Iranian: three stop series: t, d, dʰ etc. Tenuis are treated as unspecified for ejectiveness; voiced aspirates are treated as unspecified for voicedness, except that bʰ loses aspiration to fill the slot of the missing /b/. This makes more sense if, at the time, the "plain voiced" series was less marked than the "voiced aspirate" series. If we assume that the early PIE system was considerably different from late PIE, this seems to imply that shared Kartvelian/Indo-European vocabulary consists of loanwords rather than inherited vocabulary. For one thing, if we posit a phase where PIE had both a tenuis series and an ejective series, then it wouldn't make sense for the former rather than the latter to correspond to Kartvelian ejectives.

[jal]

Yet, Indo-Uralic probably had further relatives, for which Altaic and Eskimo-Aleut are IMHO the best candidates.

Is this supported by genetics? Though we must be careful not to think language change means population change, I would find it pretty remarkable if the current Eskimo-Aleut population would genetically not be related at all to the peoples whose languages share a common origin.


JAL

[Salmoneus]

Well, genetics are best avoided if you want to talk about Indo-Uralic, since genetically it makes very little sense to begin with.

We know that Late PIE was spoken in Eastern Europe by a population who were half Eastern European Hunter Gatherer and half Caucasian Hunter Gatherer in genetics. The EHG population was in turn partially descended from the Mal'ta people of Siberia.

However, speakers of Uralic languages are generally not. Genetically, Uralic speakers are late migrants from the east - they have lots of genetic indicators of east asian ancestry. Most notably, Uralic populations have extremely high densities of Y-haplotype N (almost 100% in Siberian Uralic populations), which is believed to have originated in or near China. Genetically, it seems as though there has been a big migration from the east through Siberia, which has, with the exception of a few isolated groups, largely steamrollered over any earlier siberian cousins of the EHGs.

This in turn is supported by many linguists, who a) see potential family relations with east language groups like Yukaghir and Eskimo-Aleut, and b) see what appear to be areal typological links to other eastern families like Turkic, Tungusic and Mongolic. This all makes perfect sense if we believe that Pre-Uralic developed somewhere to the east, migrated to the region of the Urals, became Proto-Uralic, and then underwent a secondary expansion, perhaps due to technological improvements, perhaps influenced by contact with IE peoples, from whom they clearly borrowed a great many words over a very long period of time.

The only fly in this otherwise very neat ointment is the presence of MT pronouns in both languages, along with a few other single-phoneme similarities (/n/ in a negative, something velar in an interrogative, /m/ in the accusative). But there are three problems with founding Uralic on that basis:

1. these sorts of morphological markers cross-linguistically tend toward 'simplicity' both in number of phonemes and in which phonemes are found. Given that we're drawing from a very reduced pool of single simple phonemes, coincidences will occur.

2. MT pronouns are also present in Mongolic, Tungusic, Turkic, Yukaghir, Eskimo-Aleut, Chukotko-Kamchatkan, Nivkh, and Kartvelian. If they support anything, they support a general Eurasiatic, not a specific Indo-Uralic. Given the known movement of peoples, it may be that "Eurasiatic" is a very ancient family (reinforced by later areal influences), the language of the early repopulators of the tundra (who, being mammoth hunters, would have been highly mobile), and IE and Uralic may be separated by tens of thousands of years (with all the intermediate languages now extinct). The only obvious reason to assume Indo-Uralic is a lack of any other surviving Eurasiatic neighbours for Indo-European. Then again, the fact the same system is found in Kartvelian, which has little if any reason to be related to Nivkh at all, may suggest that there's both familial resemblance AND coincidence: there might be two MT systems, a Eurasiatic one and an Indo-Kartvelian one.

3. Pronoun morphemes aren't necessarily archaic. Iirc in some of the Altaic branches the MT system has seemingly bee calqued later on due to influence from neighbouring groups. While pronouns may not be directly borrowed that often, it's not impossible that existing potential pronominal morphemes may be selected amongst with reference to prevailing areal traits.

Of course, Indo-Uralic is genetically possible. It would presumably suggest that Proto-Uralic at some stage borrowed the language of a neighbouring tribe wholesale, in the same way the Sami have borrowed Uralic. That doesn't sound particularly likely, since the two groups would have lived in different habitats, had different cultures and technologies, and would not have had much genetic contact at the time - making it a much bigger gap for language to jump across than the Finn-Sami gap. But it's possible. The two groups did live vaguely near each other, after all. But then there's another problem: it means that all the Ural-Altaic typological similarity must be total coincidence (since Finns, Estonians, etc, would never have been in contact with Altaic peoples since borrowing their language).

One way around that would be to suggest that Early PIE looked very much like an Altaic language. I don't think that's the commonly-accepted view, however. Although that's an interesting thought-experiment! What would the Pre-PIE Ablaut system look like if we assume vowel harmony!?.

The alternative, which may be more attractive, is to accept the migration of Pre-Uralic from the east, but have it meet "PIE" in the east. We know that there was an almost-certainly "PIE"-speaking population in the east very early on. I put "PIE" in quotes, because it's possible that the Afanashevic languages would have been the most remote branch of the family - they seem to have been around contemporaneous with the Yamnaya, who are most often assumed to be Late PIE. Afanashevic could then have been borrowed by the Uralics in passing. Or the same may have happened with some even more distant relative of PIE somewhere in western siberia - that might fit better with the timing and course of the haplotype N migration.

[WeepingElf]

Per Carrasquer Vidal (a Nostraticist), PIE tenuis correspond to Proto-Kartvelian ejective; PIE plain voiced (ejective according to the glottalic theory) to Proto-Kartv. voiced; and PIE voiced aspirate to Proto-Kartv. voiceless aspirate, except for *bʰ which corresponds to Proto-Kartv. *b.

So a third opinion on the stop correspondences between IE and Kartvelian? Dolgopolsky has:

PIE *t = PK *t'
PIE *d = PK *t
PIE *dh = PK *d

Bomhard has:

PIE *t = PK *t
PIE *d = PK *t'
PIE *dh = PK *d

Now Carrasquer (according to you) has:

PIE *t = PK *t'
PIE *d = PK *d
PIE *dh = PK *t

(All three, of course, using traditional notation for PIE.)

They can't all three be right, which may mean that all three are wrong! Actually, I think that all three are wrong. Kartvelian, if related to IE at all, is IMHO more distantly related than Uralic, and I consider it likely that the Proto-Indo-Uralic consonant inventory was more like the Proto-Uralic one, which makes all the comparisons given above spurious.

Yet, Indo-Uralic probably had further relatives, for which Altaic and Eskimo-Aleut are IMHO the best candidates.

Is this supported by genetics? Though we must be careful not to think language change means population change, I would find it pretty remarkable if the current Eskimo-Aleut population would genetically not be related at all to the peoples whose languages share a common origin.

Drawing genetics into discussions of language relationships is always problematic, and best avoided entirely. What counts are linguistic similarities, and these can be seen between Uralic and Eskimo-Aleut, as well as between IE, Uralic and Altaic (for instance, the famous "Mitian" pronouns). Of course, I am not sure at all that Indo-Uralic constitutes a valid node in the Mitian family tree (if such a tree exists at all, of course!), but IE and Uralic have always been neighbours, and internal reconstruction in PIE leads to something that looks reasonably similar to Proto-Uralic.

Yet, as long as all these relationship hypotheses are unproven, they cannot really be drawn on in order to find out what PIE was like. They can only give us hints on what may have happened, and may be misleading. However, once we have found out which languages IE is most closely related to, we will be able to use this knowledge to look deeper into the prehistory of PIE. Alas, that time has not arrived yet.

[Soap]

Minor reply to a few posts up above:

If we assume that PIE, Uralic, and Eskimo-Aleut form a node, and that the glottalic theory in some form is true,
And that proto-Uralic affricates correspond to pre-PIE ejecdtives at least conditionally,
Then it would be convenient to reconstruct a /q/ for the common proto-language which became /k'/ in pre-PIE, /k/ in proto-Uralic, and stayed /q/ in Eskimo=-Aleut.

[WeepingElf]

Minor reply to a few posts up above:

If we assume that PIE, Uralic, and Eskimo-Aleut form a node, and that the glottalic theory in some form is true,
And that proto-Uralic affricates correspond to pre-PIE ejecdtives at least conditionally,
Then it would be convenient to reconstruct a /q/ for the common proto-language which became /k'/ in pre-PIE, /k/ in proto-Uralic, and stayed /q/ in Eskimo=-Aleut.

Yes, that is an idea I am actually entertaining. Fortescue thinks a Proto-Uralo-Siberian *k split into *k and *q in Eskimo-Aleut (as well as in Chukotko-Kamchatkan and Yukaghir - which would mean that it split in every branch of Uralo-Siberian except Uralic!) depending on the nearest vowel, but he may simply be wrong about that, and *k and *q may have been distinct phonemes in Proto-Mitian (though a vowel-induced split may of course have happened in Pre-Proto-Mitian) which merged in Uralic and remained distinct in Eskimo-Aleut (and in IE where *q became *k'). The shift *q > *k' doesn't look weird to me; as is well-known, the reverse happened in various Semitic languages. (Now there are of course sound changes that usually don't happen in reverse, but I think this one may go both ways.)

[Soap]

Minor reply to a few posts up above:

If we assume that PIE, Uralic, and Eskimo-Aleut form a node, and that the glottalic theory in some form is true,
And that proto-Uralic affricates correspond to pre-PIE ejecdtives at least conditionally,
Then it would be convenient to reconstruct a /q/ for the common proto-language which became /k'/ in pre-PIE, /k/ in proto-Uralic, and stayed /q/ in Eskimo=-Aleut.

Yes, that is an idea I am actually entertaining. Fortescue thinks a Proto-Uralo-Siberian *k split into *k and *q in Eskimo-Aleut (as well as in Chukotko-Kamchatkan and Yukaghir - which would mean that it split in every branch of Uralo-Siberian except Uralic!) depending on the nearest vowel, but he may simply be wrong about that, and *k and *q may have been distinct phonemes in Proto-Mitian (though a vowel-induced split may of course have happened in Pre-Proto-Mitian) which merged in Uralic and remained distinct in Eskimo-Aleut (and in IE where *q became *k'). The shift *q > *k' doesn't look weird to me; as is well-known, the reverse happened in various Semitic languages. (Now there are of course sound changes that usually don't happen in reverse, but I think this one may go both ways.)

Sorry, I actually got confused a bit there. What I meant to say was that the common proto-Mitian language had a phoneme /k'/ which developed into /q/ in Proto-Eskimo-Aleut and into an affricate /kx/ in proto-Uralic, from where it went quickly on to a plain /k/, or possibly in some positions to /x/. Only PIE preserved the initial value, and only in the case that we assume the glottalic theory is correct.

However, I agree that /q/ > /k'/ would also work; the only reason that I propose /k'/ as the original is because I've mostly seen phonologies for Nostratic that also have ejectives.


--------


Also, another idea I've had that I have absolutely no evidence for, but also no evidence against:

Early PIE didnt just have labiovelar stops, it also had rounded bilabials and rounded coronals. These arose through a process of dissociation where previously existing rounded vowels spread their roundedness onto surrounding consonants, as has happened in Micronesian, Aranta, and probably the Caucasian languages which are geographically right next door. It simply happened that the rounded bilabials merged with the plain ones unconditionally early on, and are unnecessary to reconstruct for PIE proper. The rounded coronals probably merged either with labiovelars (which is attested) or with bilabials (which is also attested).

This idea is only useful for the purpose of reconstructing pre-PIE relationships, and may be a good answer of how a vowel-rich language like Proto-Uralic could be closely related to something which is often reconstructed as having just one vowel, /e/, and at most a small vowel inventory of which certain types were grammatically conditioned rather than freely occurring forms.

I dont believe in the three velar series model, but if we propose that it was true, the palatals could have similarly once been distributed across all POA's, with the palatalized coronals almost certainly merging unconditionally with the traditional palatovelars, and the labial ones probably merging with plain labials.

As for why labialization was not contrastive on nasals or fricatives, I'd say that PIE must have lost its /ŋ/ at some point, and likely took /ŋʷ/ and (/nʷ/ > /ŋʷ/) along with it, and as for the fricatives, at least one is commonly reconstructed with a labialization contrast. As for liquids, any preexisting /lʷ/ and /rʷ/ could easily have become a plain /w/. See, for example, https://en.wikipedia.org/wiki/Adyghe_language , which has labialization only on stops and fricatives, and lacks /ŋ/, like PIE.

[Šọ̈́gala]

For anyone who's curious, Carrasquer Vidal describes his version of IE/Kartvelian reflexes at the end of this short article on the glottalic theory: https://www.academia.edu/8605000/The_glottalic_theory

I seem to recall seeing a slightly longer account somewhere else, but I can't find it. In this article, he describes the dʰ : tʰ correspondence as very uncertain and speculative, but the others as more solid. On the surface, at least, mḳerd- : k̂erd- looks like a slam dunk.

[Travis B.]

WTH is that U+E001 supposed to be there?

[Soap]

WTH is that U+E001 supposed to be there?

I think it's /k'/, the voiceless velar ejective. I think it renders appropriately in some fonts but not others.

[Travis B.]

WTH is that U+E001 supposed to be there?

I think it's /k'/, the voiceless velar ejective. I think it renders appropriately in some fonts but not others.

Apparently that is part of a private use area.

[kanejam]

I think this has been posted before but this article argues for a distinction of t: ~ t in Anatolian, and thus suggesting (something like) *t: ~ *t ~ *t' for Early PIE (PIE1), which is even easier to reconcile with the Proto-Uralic stop system.

The Early PIE system would then have developed to *t ~ *d ~ *d' in Core PIE (PIE2) and further to *t ~ *dh ~ *d in Central PIE (Greco-Armeno-Indo-Iranian).

[Šọ̈́gala]

WTH is that U+E001 supposed to be there?

I think I fixed it now. Should be k with dot below, representing the velar ejective. See https://en.wiktionary.org/wiki/Reconstr ... %B8%B3erd- (and note that in this case the modern Georgian happens to be very similar).

[Šọ̈́gala]

So a third opinion on the stop correspondences between IE and Kartvelian? Dolgopolsky has:

Has anyone attempted to explain this as different strata of vocabulary? Certainly no law of god or man says there can't be inherited "Indo-Kartvelian" vocabulary alongside one or more layers of early IE loan words into Proto-Kartvelian.

[WeepingElf]

Sorry, I actually got confused a bit there. What I meant to say was that the common proto-Mitian language had a phoneme /k'/ which developed into /q/ in Proto-Eskimo-Aleut and into an affricate /kx/ in proto-Uralic, from where it went quickly on to a plain /k/, or possibly in some positions to /x/. Only PIE preserved the initial value, and only in the case that we assume the glottalic theory is correct.

Are you positing a Proto-Mitian language with only one ejective, */k'/? Not impossible (my father's idiolect is like that!), but I don't think this is very likely, given that all branches of Mitian show either /q/ or /k/ as the reflex of the phoneme in question, and none (except "Glottalic PIE", of course) /k'/.

However, I agree that /q/ > /k'/ would also work; the only reason that I propose /k'/ as the original is because I've mostly seen phonologies for Nostratic that also have ejectives.

The Nostratic phonologies have ejectives to account for Afrasian and Kartvelian. Now I think that Afrasian is not particularly related to Mitian, and a Mitian membership of Kartvelian is at least highly doubtful. If we disregard these two families, we have no ejectives anywhere in Mitian except in Itelmen (where they are known to be secondary) and "Glottalic PIE" (and Ossetian and some Eastern Armenian dialects, where they are also known to be secondary). Hence, I doubt that Proto-Mitian had them.

Also, another idea I've had that I have absolutely no evidence for, but also no evidence against:

Then it's just a case of Russell's Teapot, and scientifically worthless.

Early PIE didnt just have labiovelar stops, it also had rounded bilabials and rounded coronals. These arose through a process of dissociation where previously existing rounded vowels spread their roundedness onto surrounding consonants, as has happened in Micronesian, Aranta, and probably the Caucasian languages which are geographically right next door. It simply happened that the rounded bilabials merged with the plain ones unconditionally early on, and are unnecessary to reconstruct for PIE proper. The rounded coronals probably merged either with labiovelars (which is attested) or with bilabials (which is also attested).

What led you to this idea - your apparent obsession with such sounds, as evidenced by your conlangs? Gamkrelidze and Ivanov at least proposed labialized coronals, but the idea sank like a stone and was AFAIK never seriously discussed. And labialized labials are rare as hens' teeth.

This idea is only useful for the purpose of reconstructing pre-PIE relationships, and may be a good answer of how a vowel-rich language like Proto-Uralic could be closely related to something which is often reconstructed as having just one vowel, /e/, and at most a small vowel inventory of which certain types were grammatically conditioned rather than freely occurring forms.

I never was much into the idea that PIE had only one vowel at any stage. As I see it now, PIE1 (the pre-ablaut stage) had a system of three vowels */a i u/ in which */a/ was most frequent, and the result of the "Great Vowel Collapse", a merger of at least three vowels */a e o/; so PIE0 would have had at least */i e a o u/, perhaps even more (but probably no front rounded vowels, which would have led to labialized front velars in later PIE, which are not observed). Now, Juho Pystynen (Tropylium on the ZBB) once suggested that Uralic */ü/ may have been due to a chain shift in which old */u/ was fronted and all other back vowels raised one step.

I dont believe in the three velar series model, but if we propose that it was true, the palatals could have similarly once been distributed across all POA's, with the palatalized coronals almost certainly merging unconditionally with the traditional palatovelars, and the labial ones probably merging with plain labials.

I once was skeptical about the three velar series, especially in light of the low frequency of the "plain" velars, but the reconstruction is solid, and my current model explains them well: the three velar series preserve traces of fronting and rounding of PIE0 vowels as they were before the GVC. An example:

PIE3 *k^wel- 'to turn' < PIE1 *k^wal- < PIE1 *kol- or *kul- (*u would have been lowered to *o before the resonant *l, thus falling victim to the Great Vowel Collapse); indeed, *kul- seems to be a Proto-Mitian root with reflexes in Uralic and Altaic.

I see no reason to posit labialization and palatalization across the entire Proto-Mitian obstruent inventory. There simply is no evidence of that anywhere in Mitian; just how many teapots do you want to launch into orbits around the Sun?

As for why labialization was not contrastive on nasals or fricatives, I'd say that PIE must have lost its /ŋ/ at some point, and likely took /ŋʷ/ and (/nʷ/ > /ŋʷ/) along with it, and as for the fricatives, at least one is commonly reconstructed with a labialization contrast. As for liquids, any preexisting /lʷ/ and /rʷ/ could easily have become a plain /w/. See, for example, https://en.wikipedia.org/wiki/Adyghe_language , which has labialization only on stops and fricatives, and lacks /ŋ/, like PIE.

PIE indeed lost the velar nasal in my model (I assume that Proto-Mitian, like Proto-Uralic, Proto-Eskimo-Aleut and Fortescue's Proto-Uralo-Siberian, had nasals at all points of articulation where it had stops). But I don't think there ever were any labialized liquids (a labialized velar nasal may have existed if the loss of the velar nasal(s) happened after the Great Vowel Collaps, but I think the velar nasal was probably lost earlier).

I think this has been posted before but this article argues for a distinction of t: ~ t in Anatolian, and thus suggesting (something like) *t: ~ *t ~ *t' for Early PIE (PIE1), which is even easier to reconcile with the Proto-Uralic stop system.

The Early PIE system would then have developed to *t ~ *d ~ *d' in Core PIE (PIE2) and further to *t ~ *dh ~ *d in Central PIE (Greco-Armeno-Indo-Iranian).

Hmm, I can't say that this was "wrong", but I have my doubts. Why should the most common grade of stops in "classical" PIE have been marked by gemination? It is IMHO quite certain that the voiceless stops of the standard model originally have been the least marked type. Also, my working hypothesis (which may turn out to be wrong) connects the standard model's voiced aspirates to the "Uralo-Siberian" voiced spirants (and I have suggested earlier that they actually may have been just that, at least in the northwestern dialects of PIE).

EDIT: I consider the following model most likely:

Code: Select all

PIE1    PIE2    PIE3
---------------------
*t   >  *th  >  *t
*t'  >  *t   >  *d
*ð   >  *ð   >  *ð/dh
---------------------

Anatolian branched off at PIE2, and merged *ð into *t (former *t'); *th may have split into initial *t and medial *th > *t:.

Another possibility is of course:

Code: Select all

PIE1    PIE2    PIE3
---------------------
*t   >  *t   >  *t
*t'  >  *d'  >  *d
*ð   >  *ð   >  *ð/dh
---------------------

[gach]

Yes, that is an idea I am actually entertaining. Fortescue thinks a Proto-Uralo-Siberian *k split into *k and *q in Eskimo-Aleut (as well as in Chukotko-Kamchatkan and Yukaghir - which would mean that it split in every branch of Uralo-Siberian except Uralic!) depending on the nearest vowel, but he may simply be wrong about that, and *k and *q may have been distinct phonemes in Proto-Mitian (though a vowel-induced split may of course have happened in Pre-Proto-Mitian) which merged in Uralic and remained distinct in Eskimo-Aleut (and in IE where *q became *k'). The shift *q > *k' doesn't look weird to me; as is well-known, the reverse happened in various Semitic languages. (Now there are of course sound changes that usually don't happen in reverse, but I think this one may go both ways.)

If you are counting data points, it's also more attractive to explain the Yukaghir uvulars as a later harmonic split from velars (see Piispanen 2016, pp. 255-256).

[Soap]

Sorry, I think you may have misunderstood me.

Sorry, I actually got confused a bit there. What I meant to say was that the common proto-Mitian language had a phoneme /k'/ which developed into /q/ in Proto-Eskimo-Aleut and into an affricate /kx/ in proto-Uralic, from where it went quickly on to a plain /k/, or possibly in some positions to /x/. Only PIE preserved the initial value, and only in the case that we assume the glottalic theory is correct.

Are you positing a Proto-Mitian language with only one ejective, */k'/? Not impossible (my father's idiolect is like that!), but I don't think this is very likely, given that all branches of Mitian show either /q/ or /k/ as the reflex of the phoneme in question, and none (except "Glottalic PIE", of course) /k'/.

No, Im proposing a full series of ejectives, from the lips to the throat, which survived in PIE only, and only if the glottalic theory is true. You believe in the glottalic theory, am I right? I'm skeptical of it myself, but I do believe that PIE, Uralic, and likely Eskimo-Aleut are related, and I'm proposing one way to account for the mismatch of their phonetic inventories. I also believe that most linguists are too attached to the idea of unconditional sound shifts, but for the moment, I'm willing to speak in generalities myself while acknowledging that the sound changes I'm speaking of were likely polyconditional just like most of the other attested sound changes in each branch.

Also, another idea I've had that I have absolutely no evidence for, but also no evidence against:

Then it's just a case of Russell's Teapot, and scientifically worthless.

Early PIE didnt just have labiovelar stops, it also had rounded bilabials and rounded coronals. These arose through a process of dissociation where previously existing rounded vowels spread their roundedness onto surrounding consonants, as has happened in Micronesian, Aranta, and probably the Caucasian languages which are geographically right next door. It simply happened that the rounded bilabials merged with the plain ones unconditionally early on, and are unnecessary to reconstruct for PIE proper. The rounded coronals probably merged either with labiovelars (which is attested) or with bilabials (which is also attested).

What led you to this idea - your apparent obsession with such sounds, as evidenced by your conlangs? Gamkrelidze and Ivanov at least proposed labialized coronals, but the idea sank like a stone and was AFAIK never seriously discussed. And labialized labials are rare as hens' teeth.

No, I'm just noting that languages with healthy Uralic-like vowel systems that collapse over time into inventories consisting of only one vowel, or even two or three, are more likely to have developed a contrast of labialization at all places of articulation rather than just one, as proposed by the mainstream PIE linguists. If PIE had a richer vowel inventory, then yes, my idea should carry less weight, but the mainstram PIE theories all depend on reconstructing a vowel system dependent heavily on ablaut, which can be analyzed as containing only one vowel, /e/. As for labialized labials, again, look at modern Adyghe or any of the other languages in its family. They are also known from Aranta and various Micronesian languages, all of which have skimpy vowel setups similar to what mainstream theorists propose for PIE.

I dont believe in the three velar series model, but if we propose that it was true, the palatals could have similarly once been distributed across all POA's, with the palatalized coronals almost certainly merging unconditionally with the traditional palatovelars, and the labial ones probably merging with plain labials.

I once was skeptical about the three velar series, especially in light of the low frequency of the "plain" velars, but the reconstruction is solid, and my current model explains them well: the three velar series preserve traces of fronting and rounding of PIE0 vowels as they were before the GVC. An example:

PIE3 *k^wel- 'to turn' < PIE1 *k^wal- < PIE1 *kol- or *kul- (*u would have been lowered to *o before the resonant *l, thus falling victim to the Great Vowel Collapse); indeed, *kul- seems to be a Proto-Mitian root with reflexes in Uralic and Altaic.

I see no reason to posit labialization and palatalization across the entire Proto-Mitian obstruent inventory. There simply is no evidence of that anywhere in Mitian; just how many teapots do you want to launch into orbits around the Sun?

Again I think you misunderstood me; for one thing, the evidence you give there is of a shift from a rounded vowel to a plain vowel with labialization reflected on the adjacent consonant, which is exactly the idea I'm proposing. Wouldnt it be likely that such a shift would happen with other consonants besides just the velar stops?

------

Yes, that is an idea I am actually entertaining. Fortescue thinks a Proto-Uralo-Siberian *k split into *k and *q in Eskimo-Aleut (as well as in Chukotko-Kamchatkan and Yukaghir - which would mean that it split in every branch of Uralo-Siberian except Uralic!) depending on the nearest vowel, but he may simply be wrong about that, and *k and *q may have been distinct phonemes in Proto-Mitian (though a vowel-induced split may of course have happened in Pre-Proto-Mitian) which merged in Uralic and remained distinct in Eskimo-Aleut (and in IE where *q became *k'). The shift *q > *k' doesn't look weird to me; as is well-known, the reverse happened in various Semitic languages. (Now there are of course sound changes that usually don't happen in reverse, but I think this one may go both ways.)

If you are counting data points, it's also more attractive to explain the Yukaghir uvulars as a later harmonic split from velars (see Piispanen 2016, pp. 255-256).

Mostly quoting this because I realize I bumped the thread onto the next page; I dont have a strong opinion here, but what you said makes sense to me, though I'll add that most proposals that include uvulars as part of Proto-Mitian or any similar proto-language probably recognize this and do not rely on Yukaghir for evidence of primordial uvulars.

[WeepingElf]

Sorry, I think you may have misunderstood me.

Oh, I am sorry.

Are you positing a Proto-Mitian language with only one ejective, */k'/? Not impossible (my father's idiolect is like that!), but I don't think this is very likely, given that all branches of Mitian show either /q/ or /k/ as the reflex of the phoneme in question, and none (except "Glottalic PIE", of course) /k'/.

No, Im proposing a full series of ejectives, from the lips to the throat, which survived in PIE only, and only if the glottalic theory is true.

I see. I can't say this can't be right, but I am skeptical. Why does no Mitian language preserve them as such? Probably because they simply weren't there

You believe in the glottalic theory, am I right?

Actually, not! I consider it possible, and think the phonemes in question were marked in some way, but note that I prefer the agnostic term emphatic. They may have been glottalized in some way (ejectivity is just one kind of glottalization!), pharyngealized or even affricates. Or something else I haven't thought of yet.

I'm skeptical of it myself, but I do believe that PIE, Uralic, and likely Eskimo-Aleut are related, and I'm proposing one way to account for the mismatch of their phonetic inventories.

Fine. Then you are much like me. I wouldn't say that I believe in that relationship, but there are so many similarities between those families that descent from a common ancestor seems to be the most likely explanation.

I also believe that most linguists are too attached to the idea of unconditional sound shifts, but for the moment, I'm willing to speak in generalities myself while acknowledging that the sound changes I'm speaking of were likely polyconditional just like most of the other attested sound changes in each branch.

Right - the failure to recognize conditional shifts is part of the reason why inventories of proposed macro-protolanguages are often so bloated (both Dolgopolsky and Bomhard reconstruct fifty consonant phonemes for Proto-Nostratic, for instance!). Take a simple example from Indo-Uralic. There seem to be plenty t:t and s:s correspondences, which allow to reconstruct *t and *s. Yet, there are also correspondences between PIE *s and PU *t, e.g. in the nominative plural and in the verbal plural subject markers. Does this justify a third phoneme, perhaps *θ? I prefer an explanation by a conditional change *t > *s, in conditions that need to be found out. (The dual markers PIE *-h1:PU *-k seem to indicate a corresponding conditional change *k > *x, by the way.)

What led you to this idea - your apparent obsession with such sounds, as evidenced by your conlangs? Gamkrelidze and Ivanov at least proposed labialized coronals, but the idea sank like a stone and was AFAIK never seriously discussed. And labialized labials are rare as hens' teeth.

No, I'm just noting that languages with healthy Uralic-like vowel systems that collapse over time into inventories consisting of only one vowel, or even two or three, are more likely to have developed a contrast of labialization at all places of articulation rather than just one, as proposed by the mainstream PIE linguists.

I see. So the changes accompanying the GVC would have also led to palatalized and labialized labials and coronals - which we do not see. Their absence, however, can be explained otherwise. The split into plain-palatalized-labialized triads affected only the velars because these are most susceptible to influences from neighbouring vowels because they are articulated with the back of the tongue, which is involved with the vowels, too. Coronals are articulated with a different part of the tongue - the tip - and labials with the lips. There are plenty of languages where, for instance, velars are palatalized before front vowels while coronals and labials are unaffected.

The point here is that we have no traces of labialized or palatalized labials and coronals, thus we cannot posit them. Everything else would be a violation of Ockham's Razor.

If PIE had a richer vowel inventory, then yes, my idea should carry less weight, but the mainstram PIE theories all depend on reconstructing a vowel system dependent heavily on ablaut, which can be analyzed as containing only one vowel, /e/.

I know those theories, and I think they are basically right; though I think that in the pre-ablaut stage, the ancestor of the ablauting vowel was */a/, and */i/ and */u/ also were vowels which resulted in the */ei~oi~i/ and */eu~ou~u/ ablaut series where a consonant closes the root after a diphthong (where the diphthong ends the root, I assume */j/ and */w/, respectively, being the old closing consonants). At least, positing */i/ and */u/ seems more natural to me than assuming these were all diphthongs */ai/ and */au/ and no high monophthongs existed. This is another case where external cognates would be helpful.

As for labialized labials, again, look at modern Adyghe or any of the other languages in its family. They are also known from Aranta and various Micronesian languages, all of which have skimpy vowel setups similar to what mainstream theorists propose for PIE.

Point taken. Alas, we have no evidence of them existing in PIE, so you are just launching a teapot into space. It is your turn to show where they have gone!

I once was skeptical about the three velar series, especially in light of the low frequency of the "plain" velars, but the reconstruction is solid, and my current model explains them well: the three velar series preserve traces of fronting and rounding of PIE0 vowels as they were before the GVC. An example:

PIE3 *k^wel- 'to turn' < PIE1 *k^wal- < PIE1 *kol- or *kul- (*u would have been lowered to *o before the resonant *l, thus falling victim to the Great Vowel Collapse); indeed, *kul- seems to be a Proto-Mitian root with reflexes in Uralic and Altaic.

I see no reason to posit labialization and palatalization across the entire Proto-Mitian obstruent inventory. There simply is no evidence of that anywhere in Mitian; just how many teapots do you want to launch into orbits around the Sun?

Again I think you misunderstood me;

Sorry. I now see what you are getting at - palatalization and labialization of other consonants than velar stops on the way from PIE0 to PIE1. See above for my reasons to doubt that.

for one thing, the evidence you give there is of a shift from a rounded vowel to a plain vowel with labialization reflected on the adjacent consonant, which is exactly the idea I'm proposing. Wouldnt it be likely that such a shift would happen with other consonants besides just the velar stops?

See above. Velar consonants are simply more susceptible to influences from neighbouring vowels because they are articulated with the same part of the tongue.

But let's stop bickering about this; I can see what you are getting at, and can't say you are wrong; yet, the burden is on you to provide evidence for your hypothesis.

Yes, that is an idea I am actually entertaining. Fortescue thinks a Proto-Uralo-Siberian *k split into *k and *q in Eskimo-Aleut (as well as in Chukotko-Kamchatkan and Yukaghir - which would mean that it split in every branch of Uralo-Siberian except Uralic!) depending on the nearest vowel, but he may simply be wrong about that, and *k and *q may have been distinct phonemes in Proto-Mitian (though a vowel-induced split may of course have happened in Pre-Proto-Mitian) which merged in Uralic and remained distinct in Eskimo-Aleut (and in IE where *q became *k'). The shift *q > *k' doesn't look weird to me; as is well-known, the reverse happened in various Semitic languages. (Now there are of course sound changes that usually don't happen in reverse, but I think this one may go both ways.)

If you are counting data points, it's also more attractive to explain the Yukaghir uvulars as a later harmonic split from velars (see Piispanen 2016, pp. 255-256).

Mostly quoting this because I realize I bumped the thread onto the next page; I dont have a strong opinion here, but what you said makes sense to me, though I'll add that most proposals that include uvulars as part of Proto-Mitian or any similar proto-language probably recognize this and do not rely on Yukaghir for evidence of primordial uvulars.

Uvulars also exist in Chukotko-Kamchatkan and Eskimo-Aleut; Fortescue assumes that they are developments of Proto-Uralo-Siberian velars next to back vowels (i.e., not much different from Yukaghir). Some Turkic languages also have uvulars, where it is the same story as in Yukaghir.

But we are digressing more and more from Proto-Indo-European matters to Eurasian macrofamilies; I have created a new thread for that.

[KathTheDragon]

Seeing as we're sharing this again, my current model of the pre-ablaut PIE phonology is similar to yours, WE, but differs in a few key points.

Firstly, I believe the three stop series were *t *d *ɗ (for traditional *t *dʰ *d). Like you, I am very skeptical about Kloekhorst's conclusion that PIE must have had *t: *t *t' simply because Anatolian does; I see no reason why we can't have *t *d > *tʰ *t > *t: *t, even if I can't pluck any attestations of it out of the air.

Secondly, I also support three velar series, and reconstruct them as *k *kʷ *q (for traditional *ḱ *kʷ *k). This adequately explains the rarity of traditional *k, as well as motivating both the Centum and Satem treatment of the three series. I don't understand the desire to remove traditional *k from the parent language, on the basis of basically two or three highly specific environments where the series are neutralised.

Lastly, with the vowels, I also posit *a *i *u, but also long *ā *ī *ū as well. *a and *ā are the two ablaut vowels, and generally yield *e and *o respectively, barring the sound changes that redistributed them; unlike you, however, *ey comes from *ai, and original *i instead yields non-ablauting *e, as seen in the nominative plural *-es, for instance. Similarly, *u yields non-ablauting *o, as in *potis (for the fact that it's non-ablauting, see the failure of Brugmann's law in Indo-Iranian). Lastly, *ī and *ū yield *i and *u, as in the instrumental plural *-is and the root "become" *bʰuH-. Note that this is compatible with your "Great Vowel Collapse" scenario, and it even helps to explain why *a is so absurdly common.

[Soap]

Secondly, I also support three velar series, and reconstruct them as *k *kʷ *q (for traditional *ḱ *kʷ *k). This adequately explains the rarity of traditional *k, as well as motivating both the Centum and Satem treatment of the three series. I don't understand the desire to remove traditional *k from the parent language, on the basis of basically two or three highly specific environments where the series are neutralised.

The two series model Im familiar with has a simple contrast of plain vs labialized velars, and no distinction in the primary PoA. I see on WIkipedia that a palatal vs labiovelar model exists, but I would consider that an entirely separate hypothesis rather than a variant of the two-series model Im familiar with, since it would require a very different explanation.

I dont believe there were any uvular consonants in PIE because:
1) there are no uvulars in the early daughter languages, meaning that they would need to have been individually discarded in each of the early IE branches, leaving no trace in any of them even conditionally. We can't simply suppose that /q/ > /k/ happened once in the satem languages and once in the centum languages because some centum languages forked off from early PIE before the satem branches did, meaning we'd need to assume that the same unconditional change happened in those languages independently.

2)It seems odd to propose a plain uvular series without a corresponding labialized uvular series; if one is added, then we'd need to propose an unconditional loss of six consonant phonemes in every branch of PIE, as above.

3)Voiced uvular sounds are rare in languages in general; to propose that PIE had two (or four) contrasting voiced uvular stops would make the language even more odd, phonologically, than it already is. Support for a uvular series seems often to be paired with the glottalic theory, which makes two or in some cases all three of the stops voiceless. But your model seems to propose not only a voiced uvular stop, but also a voiced uvular implosive, which is even rarer than the stop.

All centum languages merge the traditional plain velars with the traditional palatovelars unconditionally. This means that satem languages are the only evidence that the traditional plain velar series ever existed, and according to Wikipedia:

Alternations between plain velars and palatals are common in a number of roots across different satem languages, where the same root appears with a palatal in some languages but a plain velar in others (most commonly Baltic or Slavic, occasionally Armenian but rarely or never the Indo-Iranian languages). That is consistent with the analogical generalisation of one or another consonant in an originally-alternating paradigm but difficult to explain otherwise.

Assuming this is true, I consider this a very strong argument in favor of two dorsal series.

[KathTheDragon]

I dont believe there were any uvular consonants in PIE because:
1) there are no uvulars in the early daughter languages, meaning that they would need to have been individually discarded in each of the early IE branches, leaving no trace in any of them even conditionally. We can't simply suppose that /q/ > /k/ happened once in the satem languages and once in the centum languages because some centum languages forked off from early PIE before the satem branches did, meaning we'd need to assume that the same unconditional change happened in those languages independently.

I do not consider this a very strong argument, because q > k is a perfectly natural change. Besides, in my model the change happened once, in dialectal IE between the departure of Anatolian and the final breakup of the family. In "satem" dialects, the change triggered the original plain velars to front, while in "centum" dialects it did not.

2)It seems odd to propose a plain uvular series without a corresponding labialized uvular series; if one is added, then we'd need to propose an unconditional loss of six consonant phonemes in every branch of PIE, as above.

Also a non-problem.

3)Voiced uvular sounds are rare in languages in general; to propose that PIE had two (or four) contrasting voiced uvular stops would make the language even more odd, phonologically, than it already is. Support for a uvular series seems often to be paired with the glottalic theory, which makes two or in some cases all three of the stops voiceless. But your model seems to propose not only a voiced uvular stop, but also a voiced uvular implosive, which is even rarer than the stop.

Good thing the two stops in question are very rare - a good number of traditional plain velars are based on comparisons not including Indo-Iranian, which leaves open the possibility that they are not really the plain velars, but the palatovelars with their fronting blocked, which is a known phenomenon. So while the real number of voiced plain velars is hard to measure, it's remarkably small - and thus I don't see this as that big of an issue.

All centum languages merge the traditional plain velars with the traditional palatovelars unconditionally. This means that satem languages are the only evidence that the traditional plain velar series ever existed, and according to Wikipedia:

Alternations between plain velars and palatals are common in a number of roots across different satem languages, where the same root appears with a palatal in some languages but a plain velar in others (most commonly Baltic or Slavic, occasionally Armenian but rarely or never the Indo-Iranian languages). That is consistent with the analogical generalisation of one or another consonant in an originally-alternating paradigm but difficult to explain otherwise.

Assuming this is true, I consider this a very strong argument in favor of two dorsal series.

This is but one way to interpret the data - the other, and I believe the more widespread understanding, is that these alternations are due to conditioned blocking of satemisation in the non-Indo-Iranian satem languages. The only environment I know off the top of my head is before a resonant followed by a back vowel in Balto-Slavic, cf. Lithuanian akmuõ, one of the star examples of this change.

Besides, there are other, more pressing concerns with a two-dorsal model, namely the root constraint on having two stops at the same PoA. There are numerous good examples of roots with both a plain and palatovelar, like *ḱenk-, and there is no clear distribution on which stop usually comes where. I have never seen an explanation for this under a two series model, which for me is a much greater weakness than the arguments levied against a three series model.

[WeepingElf]

Seeing as we're sharing this again, my current model of the pre-ablaut PIE phonology is similar to yours, WE, but differs in a few key points.

Firstly, I believe the three stop series were *t *d *ɗ (for traditional *t *dʰ *d). Like you, I am very skeptical about Kloekhorst's conclusion that PIE must have had *t: *t *t' simply because Anatolian does; I see no reason why we can't have *t *d > *tʰ *t > *t: *t, even if I can't pluck any attestations of it out of the air.

*t *d > *tʰ *t isn't weird at all - it happened in Armenian (according to the standard model), Icelandic and Scots Gaelic; I am sure there are many more. The next step - *tʰ *t > *t: *t - doesn't seem too weird to me, either, but I consider it more likely that the Anatolian fortis stops were aspirated, which doesn't rule out that they were phonologically behaving as geminates. The Anatolian languages may have been like Proto-Basque in having fortis-lenis pairs, which manifested as either an aspiration or a voicing opposition in stops, and a difference in duration in other consonants.

There is apparently some faint evidence of an Indo-European substratum language ("Pelasgian") with an Armenian-like consonant shift in Greece; IMHO it is likely that this language was Para-Anatolian, and may hint at the Armenian-like system having been in place in PIE2 after the PIE1 emphatic stops had turned plain voiceless. This system would then have shifted to the standard model in PIE3 in a kind of "reverse Armenian" shift. But this thing is highly controversial; others propose a non-IE substratum, and perhaps both are right: both languages may have existed side by side, or Pelasgian passed on words from the non-IE language to Greek, or both.

Secondly, I also support three velar series, and reconstruct them as *k *kʷ *q (for traditional *ḱ *kʷ *k). This adequately explains the rarity of traditional *k, as well as motivating both the Centum and Satem treatment of the three series. I don't understand the desire to remove traditional *k from the parent language, on the basis of basically two or three highly specific environments where the series are neutralised.

I am by now quite happy with the standard model - palatalized vs. plain vs. labialized - as my model predicts just that. The only thing that still bugs me is that velars with no vowel adjacent seem to come out as palatalized rather than plain. For my part, I prefer the agnostic terms "front velar" and "back velar" for the two unlabialized velar series.

Lastly, with the vowels, I also posit *a *i *u, but also long *ā *ī *ū as well. *a and *ā are the two ablaut vowels, and generally yield *e and *o respectively, barring the sound changes that redistributed them; unlike you, however, *ey comes from *ai, and original *i instead yields non-ablauting *e, as seen in the nominative plural *-es, for instance. Similarly, *u yields non-ablauting *o, as in *potis (for the fact that it's non-ablauting, see the failure of Brugmann's law in Indo-Iranian). Lastly, *ī and *ū yield *i and *u, as in the instrumental plural *-is and the root "become" *bʰuH-. Note that this is compatible with your "Great Vowel Collapse" scenario, and it even helps to explain why *a is so absurdly common.

This is possible; I need to take a closer look at these things. How many instances of non-ablauting *e and *o are there, and how sure are we that they didn't ablaut? I still feel that my model is simpler, but there may be problems with it that I am not yet aware of.