The Great Proto-Indo-European Thread

[Frislander]

Wait, so if we assume a traditional reconstruction of stop series with breathy stops, could we explain the missing voiceless aspirates by saying they spirantized into laryngeals? This isn't necessarily an explanation for all laryngeals in all positions, but it might perhaps at least act as a source for those found at syllable boundaries.

[WeepingElf]

There's this idea of mine I don't know of whether it is of any value, which I have nicknamed the "fricative theory": the breathy-voiced stops originally were voiced fricatives, so we get, with the idea I posted above, nice quadruplets of voiceless and voiced stops and fricatives at each of the places of articulation. Things would have started to slide when the three voiceless velar fricatives were weakened to voiced pharyngeals with the palatalization and labialization features intact, which were the sounds that coloured the vowels.

I know that this is not without problems; it especially does not explain the stop co-occurrence constraints which the glottalic theory makes such a good job of (though such constraints do not hold in all languages with ejectives; for instance, they do not hold in Kartvelian, where we have words like Geo. k'op'e 'ladle' - two ejectives in a single root, one of them labial). Also, such a system would probably be highly stable, and one wonders how it got to "slide" in the way I laid out above in the first place!

Hence, I am not at all convinced of the "fricative theory" myself! Though I'll perhaps pursue it in a Para-IE conlang some day. I'd rather consider some sort of the glottalic theory likely, though not for PIE at the time of breakup (definitely not PIE3, as I call the common ancestor of the non-Anatolian IE languages) but perhaps PIE1 (the pre-ablaut stage, parent of PIE2 and the hypothetical Aquan languages which I explore in my Hesperic conlang family) or at most PIE2 (the ancestor of PIE3 and Anatolian). However, I prefer the agnostic term "emphatic stops" over "glottalized stops" or even "ejectives" - we are apparently dealing with a highly marked manner of articulation here, but that may have been something else than ejective.

If we could establish regular sound correspondences between PIE and some external comparandum (the best candidate is IMHO Uralic), we'd perhaps know more - and face new riddles

[Howl]

I don't have a preferred phonology of the laryngeals. I find that the more I learn about them, the more uncertain I become about them. But if I had to make a guess, it would be something like this:

H (uncolored, e.g. adjacent to apophonic o) was 'x' but turned into 'h' before the syncope.
h1e was 'çe' but turned into 'he' some time after the syncope.
h2e was 'Xa'.
h3e was 'xʷo'.

Wait, so if we assume a traditional reconstruction of stop series with breathy stops, could we explain the missing voiceless aspirates by saying they spirantized into laryngeals? This isn't necessarily an explanation for all laryngeals in all positions, but it might perhaps at least act as a source for those found at syllable boundaries.

If the laryngeals were velar fricatives, PIE had four series of obstruents. So, something like that is very much possible. Also note that PIE must have had at least one syncope of vowels, and there would have been more syllable boundaries in pre-syncope PIE.

[KathTheDragon]

There are no "missing" voiceless aspirates. The notion that the system *t *d *dʰ is missing anything is a misconception based on the misleading notation used for the breathy-voiced stops, which is a Sanskritism.

[Howl]

There are no "missing" voiceless aspirates. The notion that the system *t *d *dʰ is missing anything is a misconception based on the misleading notation used for the breathy-voiced stops, which is a Sanskritism.

I also don't think that deriving the laryngeals from Kʰ is the right way. It would leave too much unexplained.
But I thought you supported the Cao Bang theory (the one with the implosives), and in that theory the system is the standard *t *d *dʰ at the time PIE splits up (but *t *ɗ *d in pre-PIE).

If we could establish regular sound correspondences between PIE and some external comparandum (the best candidate is IMHO Uralic), we'd perhaps know more - and face new riddles

I could tell you what I see in my IE-Uralic comparison, but (1) it is a long story and (2) I don't like being called a junk-etymologist. But for a starting point, I refer you to Adam Hyllested's work:

https://www.academia.edu/377087/Interna ... Laryngeals

[KathTheDragon]

But I thought you supported the Cao Bang theory (the one with the implosives), and in that theory the system is the standard *t *d *dʰ at the time PIE splits up (but *t *ɗ *d in pre-PIE).

That's right.

Edit: I just found and read the powerpoint outlining the theory, and there's a lot there I didn't previously know! Good to have an actual direct parallel.

[hwhatting]

I don't think that the laryngeals need to parallel any other series. The three different outcomes of the syllabic larnygeals (and of the combination of syllabic resonants + laryngeal) in Greek and the colouring effects on adjecent vowels show that there were three, but I don't think we need to assume more than three just to get a parallelism or to satisfy token frequencies.

[Šọ̈́gala]

Wait, so if we assume a traditional reconstruction of stop series with breathy stops, could we explain the missing voiceless aspirates by saying they spirantized into laryngeals? This isn't necessarily an explanation for all laryngeals in all positions, but it might perhaps at least act as a source for those found at syllable boundaries.

I happen to have posted just such an idea here a few months ago: http://www.incatena.org/viewtopic.php?f ... 4#p1139844

[Šọ̈́gala]

The model I've been tending toward lately is to think of h₁, h₂, h₃ as places of articulation. This explains why h₂ and h₃ color vowels while no other consonants (with the possible exception of *k) do: the coloring is conditioned by place of articulation. h₁ could easily be more than one POA: any of the places known from other consonants plus glottal perhaps. Within any specific POA, there could have been multiple phonemes. Maybe that gets us something toward explaining the frequency of h₂: maybe it contains /X/, /ʁ/, /q/, /ɢ/, and /ɢʱ/ (although including the stops would require us to surrender the idea that traditional *velars were really uvulars).

[Howl]

I think the unvoiced stop series was aspirated in earlier PIE. This would give a much saner stop system: Tʰ : T~D : Dʰ.

Such a stop system can easily explain all the peculiarities of the late PIE stop system. And it does not need any exotic glottalic or implosive stops to do so. Early PIE would have allowed the following kinds of roots: TʰeTʰ (late PIE TeT), TʰeT (late PIE TeD) TeTʰ (late PIE DeT), DʰeDʰ and the rarer DeDʰ and DʰeD.

- So the constraint against late-PIE DeD roots means that in roots with two stops, one stop must have been aspirated.
- And the constraint against late-PIE TeDʰ and DʰeT roots means that the stops in the root must have agreed in voicing.
- The constraint against late PIE 'b' could be explained by assuming that the labial stops were always aspirated in early PIE.

The first step to the mainstream PIE stop system is a merger of T~D to D. Germanic and Armenian (and possibly also Tocharian and Anatolian) derive their stop systems from this stage: Tʰ : D : Dʰ.

The next step is the deaspiration of Tʰ. The rest then derives from the late PIE stop system from that stage: T : D : Dʰ.

[KathTheDragon]

exotic glottalic or implosive stops

Ejectives, implosives, or glottalised stops are not "exotic", whatever that's supposed to mean.

Moreover, I don't see how your system is at all "sane", or even necessary.

[Soap]

I think the unvoiced stop series was aspirated in earlier PIE. This would give a much saner stop system: Tʰ : T~D : Dʰ.


- The constraint against late PIE 'b' could be explained by assuming that the labial stops were always aspirated in early PIE.

according to _ evidence of a consonant shift in 7th century Japanese _, stop systems with b/p/ph...are likely to aspirate the b. It's ambiguous as to whether he means it for labials only, or is using those 3 as shorthand for all stops. The letters were not capitalized.so I think it means labials only. I'm not sure ..
....t I will fill it in when I can get to a bi.g screen pc. I'm also not sure what the author used to come to that conclusion... if pie itself is part of the basis , it means little or nothing to us. There's also old Chinese,which has in some setups an unpaired bh, but I think that the reconstructions of
Old Chinese may themselves be influenced by pie.

[Howl]

exotic glottalic or implosive stops

Ejectives, implosives, or glottalised stops are not "exotic", whatever that's supposed to mean.

See the map at http://wals.info/chapter/7

In the Indo-European languages, ejectives are only attested in some Eastern Armenian dialects. And implosives are only attested in Sindhi and Saraiki. These sounds are rare in Europe and North-Asia. Ejectives do occur a lot in the Caucasian language families, but the Caucusus is kind of an island in a wide sea of languages that do not have anything like that. Implosives do occur a lot in the Austroasiatic languages, but the area where these languages are spoken is a long way from IE territority.

I am biased against reconstructions that include such phonemes. And I think that trying to explain the peculiarities of the PIE stop system with the properties of such phonemes is a mistake. Explaining them in terms of the sounds that these phonemes may have had, answers the 'why' question, but not the 'how' question. There must have been sound changes in pre-PIE that caused these peculiarities. And the reality is that many sound changes do not have a particular reason.

[WeepingElf]

Well, similar arguments could be levelled at the traditional reconstruction as well. Breathy-voiced stops are rare. They occur in many languages of India - which all have either inherited or borrowed them from a single language, Sanskrit. They occur in Wu, a Chinese dialect spoken in and around Shanghai. They occur in a small sub-sub-branch of Bantu in southern Africa. Otherwise, they hardly occur at all. On the other hand, this rarity makes it very difficult to arrive at typological generalizations about them.

But evidence of PIE breathy-voiced stops does not come from Sanskrit alone. The developments in Armenian and Greek can be accounted for by these as well. The Old Armenian voiced stops IMHO probably were breathy-voiced, as they still are in some Eastern Armenian dialects, and if they were, the Western Armenian change where they merged with the aspirated stops becomes a simple loss of voicing. The Greek aspirated stops would likewise just have lost their voicing. So far, so Greco-Aryan. But the Germanic development could also be explained that way. First the same change as in Old Armenian, then a spirantization that affected all stops with [+breath]. A similar thing would have happened in Italic after the breathy-voiced stops lost voicing in initial (but not in medial) position.

Yet, we indeed cannot be sure that the PIE breathy-voiced were something else, such as the voiced fricatives (see my post about the "fricative theory" a few days ago, which I consider not particularly apt though).

What regards ejectives, the Caucasus is not really far from the Pontic steppe where PIE most likely originated. Is it out of the question that both the Caucasus and the Pontic steppe once were part of a larger Sprachbund for which ejectives were characteristic? Yet, the language family Gamkrelidze and Ivanov used as a model for their glottalic theory, Kartvelian, actually doesn't show the constraints the theory is based on - a word like Georgian k'op'e 'ladle' could not exist in "Glottalic PIE": two ejectives in a root, one of them labial!

Yet, the low frequency of the *D set, the apparent lack of a labial member and the constraint against two of them in a root point at a manner of articulation that was 1) not intermediate between the *T set and the *Dh set (because these were subject to a harmony rule) and 2) highly marked. I prefer the agnostic term "emphatic" for these. Similar constraints were in place with the emphatic stops in Akkadian (which probably were ejectives but could have been something else). In Akkadian, the lack of a labial emphatic is a Semitic family trait, while the "no two emphatics in a root" constraint is an innovation of Akkadian - other Semitic languages have roots with two emphatics. Also, root structure constraints in a Semitic language can of course hardly be compared to those in a language like PIE, which unlike Semitic was not a triconsonantal root language (though not utterly unlike them, either).

However, while some kind of glottalic theory is plausible in an early stage of PIE, at the time of breakup, this would have shifted to something close to the traditional reconstruction, if you ask me.

[Richard W]

Breathy-voiced stops are rare. They occur in many languages of India - which all have either inherited or borrowed them from a single language, Sanskrit.

Well, strictly speaking, I doubt the breathy voiced consonants in Pali derive from Sanskrit. (You're on safer ground if you claim Old Indic as a dialect continuum.) Is there no evidence of breathy consonants in Iranian? Some feature seems to have spread in Bartholomae's Law, but possibly breathy consonants are just a notational convenience for Proto-Iranian.

They occur in Wu, a Chinese dialect spoken in and around Shanghai.

That looks like a old areal feature (Mon-Khmer, Tai, Cham, Chinese), associated with tone-splitting in the tone languages. The simple story is that initial voiced obstruents became breathy, the breathiness moved to the vowel, and from there affected register (e.g. oldish and conservative Khmer) and sometimes aspirated the stop as it lost voicing - aspiration is a characteristic of the Siamese-Lao area, where the explanation given is that breathiness of the vowel merged with aspiration, but with a tonal contrast.

[Vijay]

I'm not entirely sure there aren't Indian languages that have borrowed words with breathy stops from other languages that inherited them, either. And Wu isn't one dialect; it's a dialect group.

[WeepingElf]

Breathy-voiced stops are rare. They occur in many languages of India - which all have either inherited or borrowed them from a single language, Sanskrit.

Well, strictly speaking, I doubt the breathy voiced consonants in Pali derive from Sanskrit. (You're on safer ground if you claim Old Indic as a dialect continuum.)

I should have written "Old Indic" rather than "Sanskrit", of course. That's what I meant. Sanskrit, in the strict sense of the word, is the somewhat artificial language the ancient Indian grammarians abstracted from the not quite homogenous language of the Vedas, and of course, no vernacular language of India descends from it!

Is there no evidence of breathy consonants in Iranian? Some feature seems to have spread in Bartholomae's Law, but possibly breathy consonants are just a notational convenience for Proto-Iranian.

I don't know much about Iranian, but you may be right.

They occur in Wu, a Chinese dialect spoken in and around Shanghai.

That looks like a old areal feature (Mon-Khmer, Tai, Cham, Chinese), associated with tone-splitting in the tone languages. The simple story is that initial voiced obstruents became breathy, the breathiness moved to the vowel, and from there affected register (e.g. oldish and conservative Khmer) and sometimes aspirated the stop as it lost voicing - aspiration is a characteristic of the Siamese-Lao area, where the explanation given is that breathiness of the vowel merged with aspiration, but with a tonal contrast.

This may be the case. Again, I don't know much about the diachronics of Chinese and other East Asian languages.

[Vijay]

Sanskrit can mean both ("Old Indic" is also called "Vedic Sanskrit," after all). It just usually refers to Classical Sanskrit.

[Tropylium]

Is there no evidence of breathy consonants in Iranian?

Martin Kümmel suggests that some cases of *Tʰ in Indo-Iranian can be explained by "aspiration throwback" in Iranian: *TanDʰ > *TʰanD, followed by spirantization (*pʰ *tʰ *kʰ >*f *θ *x). (This suggests loss of aspiration after nasals would probably be earlier than general *Dʰ > *D.)

Speaking of aspirates, here is a weird PIE fact: LIV has a number of roots that violate the supposed contraint against *DʰeT- roots, if we count roots with a medial resonant as well… but they are all *bʰ plus a velar:
*bʰelk- 'stützen'
*bʰlewk- 'blaß werden'
*bʰreḱ- 'schärfen' (II only)
*bʰrekʷ- 'zusammendrängen'
*bʰrenḱ- 'abfallen' (II only)
*bʰrenk- 'anschwellen'
*bʰrewḱ- 'abschaben'
*bʰrewk- 'streichen, streifen'
*bʰreyHk- 'sich sträuben, sich aufstellen'
The only idea I would have about what's going on here is that maybe there is a *bʰ- prefix (something of the sort is well attested in Germanic), but why does it then combine with roots ending in velars only?? For comparison, 8/14 of the *bʰ(R)e(R)Dʰ- roots have *-dʰ-; and 8/24 of the *bʰ(R)e(R)D- roots have *-d-.

(It's also weird how many of these begin with *bʰr-.)

[Soap]

Because there was no /b/. Whether you believe in b>bh, allophonic aspiration, or neither , I think this is best explained by assuming /bh/ is taking the place of */b/.

Sorry for bump, I will quote the post when I get a chsnce.

Is there no evidence of breathy consonants in Iranian?

Martin Kümmel suggests that some cases of *Tʰ in Indo-Iranian can be explained by "aspiration throwback" in Iranian: *TanDʰ > *TʰanD, followed by spirantization (*pʰ *tʰ *kʰ >*f *θ *x). (This suggests loss of aspiration after nasals would probably be earlier than general *Dʰ > *D.)

Speaking of aspirates, here is a weird PIE fact: LIV has a number of roots that violate the supposed contraint against *DʰeT- roots, if we count roots with a medial resonant as well… but they are all *bʰ plus a velar:
*bʰelk- 'stützen'
*bʰlewk- 'blaß werden'
*bʰreḱ- 'schärfen' (II only)
*bʰrekʷ- 'zusammendrängen'
*bʰrenḱ- 'abfallen' (II only)
*bʰrenk- 'anschwellen'
*bʰrewḱ- 'abschaben'
*bʰrewk- 'streichen, streifen'
*bʰreyHk- 'sich sträuben, sich aufstellen'
The only idea I would have about what's going on here is that maybe there is a *bʰ- prefix (something of the sort is well attested in Germanic), but why does it then combine with roots ending in velars only?? For comparison, 8/14 of the *bʰ(R)e(R)Dʰ- roots have *-dʰ-; and 8/24 of the *bʰ(R)e(R)D- roots have *-d-.

(It's also weird how many of these begin with *bʰr-.)

[KathTheDragon]

That's hardly an explanation. How does that account for the lack of *bʰ...t roots? And why the overrepresentation of *bʰr-, when it only accounts for about a quarter of all roots beginning with *bʰ in total?

[Soap]

That's hardly an explanation. How does that account for the lack of *bʰ...t roots? And why the overrepresentation of *bʰr-, when it only accounts for about a quarter of all roots beginning with *bʰ in total?

Well, you could ask the same questions about other root shapes. E.g., why are roots ending in /t/ rare in general? It may be due to the past tense marker , or it may have some entirely unrelated explanation that left no imprints on the daughter languages. Likewise /r/ in the 2nd position seems to occur mostly or entirely after stops.

÷÷÷÷÷÷÷
The numbers we're dealing with are so small that I thunk chance is the best explanation. It's similar to how English has words like _flip_ , _flutter_, _ flounder _ , etc, but no
Productive affix _ fl- _.
.

[KathTheDragon]

I'm sorry, this dismissal is completely laughable.

Roots ending in *t are not "rare", there are only slightly fewer of them than there are ending in *ḱ. And then there is no "past tense marker". It's wholly unsurprising that *r should mostly appear after stops, given that a) there's a known avoidance of *sr in a few daughters and b) stops comprise 60% of non-sonorant root initials.

It's much more likely that this is just the LIV being over-zealous about making connections - indeed, when checking the etymological dictionaries I have for the words listed under *bʰelk-, de Vaan calls the connection with the given Greek word problematic, and Beekes doesn't even mention it! Both call the etymology of their respective words unclear. It's a similar story at *bʰrekʷ-, *bʰrenk- (Kroonen suggests a compound, Derksen doesn't even list the Slavic word), while *bʰreḱ- is restricted to Vedic, *bʰrenḱ- is restricted to Indo-Iranian, *bʰrewḱ- is restricted to Slavic, *bʰrewk- is restricted to Balto-Slavic, and *bʰreyHk- is only certainly in Greek (Matasović doesn't list the Celtic words), so none of these can even be posited for PIE anyway.

[Howl]

Also, all but one (*bʰrekʷ) of these bʰRk roots have a question mark before them in LIV. The question mark means 'that the material from the individual languages is not enough to accept this root with certainty'.

Because there was no /b/. Whether you believe in b>bh, allophonic aspiration, or neither , I think this is best explained by assuming /bh/ is taking the place of */b/.

It is possible to figure out what happened to /b/ because there are words in the individual daughter languages that reflect /b/. And often there is also a form without /b/ in another group. This can be interpreted as traces of a previous sound change.

The examples I know of are:
* the previously mentioned kewp/kewb 'to boil, to desire': kewb in Germanic versus kewp in Balto-Slavic and Italic
* *h₂ébōl 'apple' in Germanic, Balto-Slavic, Celtic and probably also Italic versus Greek ampélos 'vine'
* *gʰeh₁bʰ- 'to grab, to take' versus Italic *gʰeh₁b as attested in Oscan and Umbrian.

So, /b/ became /p/ or /bʰ/ depending on the shape of the root.

[Tropylium]

Also, all but one (*bʰrekʷ) of these bʰRk roots have a question mark before them in LIV. The question mark means 'that the material from the individual languages is not enough to accept this root with certainty'.

Probably yes, but that makes the issue all the weirder. The individual daughter languages generally didn't continue to maintain the PIE root constraints, and if they have innovated some new roots mixing the former *Dʰ and *T series, there definitely should be also some similar cases looking like **bʰert-, **dʰenk-, **gʰret- etc.

I thought of one possibility that might explain the abundance of *bʰr-, though. Many branches turn *mr- into *br-, right? Including Germanic, but only after *bʰ > *b (brain < *mregʰmno-). So e.g. *bʰrenk- (Germanic + Slavic) and *bʰrewk- (Balto-Slavic only) could be actually rather *mrenk-, *mrewk-.

This may even explain why there's only two **bʰReT- roots: if these were to come from *mReT-, then the zero grade would be *mR̥T-, and *mr- > *br- did not affect *mr̥ in at least some branches (Germanic *murguz 'short' < *mr̥ǵʰus). The zero grade of something like *mrenk- would be however *mrn̥k- with consonantal *r, and denasalization would be able to apply in general.

It is possible to figure out what happened to /b/ because there are words in the individual daughter languages that reflect /b/. And often there is also a form without /b/ in another group. This can be interpreted as traces of a previous sound change

In principle maybe, but the picture looks too inconsistent to attribute this to sound changes. Instead it looks as if stops from non-native sources were substituted in various ways so that some branches end up pointing to PIE *bʰ, some to *b, some to *p. E.g. if a pre-Germanic *kʰewp- were to be loaned into Balto-Slavic, it would naturally come out as *kewp-, and therefore end up looking like a case of "*b ~ *p alternation".

Compare how e.g. English hour versus German Uhr is not a case of Proto-West Germanic *ū ~ *ō alternation, it's a case of the word being loaned into English before *ū > /au/ but into German after it.