The Great Proto-Indo-European Thread

[hwhatting]

If you go the two-dorsal route, you have problems like the numeral '8', which is of a shape that should inhibit palatalisation.

And you'll have to account for Luwian, which shows reflexes implying three dorsal series, but is otherwise close to centum Hittite. IMO, late (immediately pre-split) PIE had three phonemic doral series; any explanation deriving this from fewer series (two or one) must refer to developments anterior to late PIE.
Concerning RUKI and satemisation, I'd prefer a wave model in geographically adjacent dialects to an explanation based on genetic unity (that would also explain why the more outlying Baltic languages have the most cases of un-palatalised dorsals.)

[vokzhen]

By my understanding (as someone who knows little of PIE beyond what's in this thread), even if the two-dorsal was the original for late PIE, taking it as correct right now we're left shrugging and going "analogy?" to explain many instances of the front-dorsal. The three-dorsal might be an artifact of our ability to reconstruct, but it describes the data without resorting hand-waving nearly as much.

[Pabappa]

I dont have the wordlists in front of me right now, but going off of what it says on Wikipedia (which is similar to what dnghu.org uses),

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). This is consistent with the analogical generalization of one or another consonant in an originally alternating paradigm, but difficult to explain otherwise.

Assuming this is correct it's difficult to explain how it could not be analogy. I would love to see the examples of problematic roots, but if even a single such root exists, it casts the whole three-dorsal theory into doubt since one would have to explain why one language reflects, e.g. /k/ as /kʲ/ and another closely related language has it as /k/.


I dont see why *oḱtō is a problem. The cluster /ḱt/ seems to have been very rare anyway, so Im not sure where a rule that /ḱt/ should always become /kt/ (in a 3-dorsal system) would come from. In fact I dont see any other roots with /ḱt/ so I assume it must mostly occur in verbs or compounds. I would just assume /ḱt/ should just be treated as a normal sequence which would mean that all satem languages would change ḱt to kʲt and then there is no problem.

[hwhatting]

I dont have the wordlists in front of me right now, but going off of what it says on Wikipedia (which is similar to what dnghu.org uses),

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). This is consistent with the analogical generalization of one or another consonant in an originally alternating paradigm, but difficult to explain otherwise.

Assuming this is correct it's difficult to explain how it could not be analogy. I would love to see the examples of problematic roots, but if even a single such root exists, it casts the whole three-dorsal theory into doubt since one would have to explain why one language reflects, e.g. /k/ as /kʲ/ and another closely related language has it as /k/.

One alternative explanation is that when the wave of the /k'/ > sibilant shift passed through the Satem dialect group, the outlying languages did not participate fully (that's what I assume). Another would be dialect mixture with centum languages. Both explanations are compatible with two- and three-dorsal models.

[KathTheDragon]

I dont see why *oḱtō is a problem. The cluster /ḱt/ seems to have been very rare anyway, so Im not sure where a rule that /ḱt/ should always become /kt/ (in a 3-dorsal system) would come from. In fact I dont see any other roots with /ḱt/ so I assume it must mostly occur in verbs or compounds. I would just assume /ḱt/ should just be treated as a normal sequence which would mean that all satem languages would change ḱt to kʲt and then there is no problem.

The problem is more that the velar (under typical two-dorsal theories) is fronted in the first place.

[hwhatting]

The problem is more that the velar (under typical two-dorsal theories) is fronted in the first place.

I wanted to say the same, but then - what if this fronting is a regular sound law? A shift of the point of articulation caused by the following /t/ (which has a more frontal PoA, after all)? As Publipis says, we don't have a lot of instances of /kt/ outside of morphology. Does anyone have other examples for /kt/ or for roots ending in a plain velar, on which we could check how /kt/ behaves when it arises due to morphology?

[KathTheDragon]

Any verbal root ending in /k/ plus one of the action/result suffices -ti- and -tu- would give the required cluster. So would the 3s and 2p of an athematic verb of one of those same verbal roots.

[Salmoneus]

Any verbal root ending in /k/ plus one of the action/result suffices -ti- and -tu- would give the required cluster. So would the 3s and 2p of an athematic verb of one of those same verbal roots.

But those clusters occur across a morpheme boundary.

[Plus deverbal adjectives in -to]

[Sleinad Flar]

Dorsal + velar clusters outside of morpheme boundaries are rare anyway (heck, double plosives within morphemes are rare in general), and ATM I can only think of two cases ("eight" and "night"), so you can't avoid looking at morpheme boundaries. Even the eight-word looks like a morpheme boundary to me. For this reason, looking at derived words with -k't-, -kt- and even -k^wt- seems like a good idea to me, especially since the relation between the derived word and the original root/verb is often obscured.

(There's a sequel to LIV called NIL, which is a lexicon of nominal derivations on verbal roots. I'll a have a look in it later on when I find the time and see what I can find.)

[KathTheDragon]

ATM I can only think of two cases ("eight" and "night")

According to Kloekhorst, the Hittite verb nekuzi "to become evening" indicates that *nókʷts is actually a derivative in -t- from a root *negʷʱ-, and so this too is an example across a morpheme boundary.

[Pole, the]

Also, I've seen sowewhere PIE "eight" analyzed as "two fours".

[KathTheDragon]

Also, I've seen sowewhere PIE "eight" analyzed as "two fours".

I've seen that, too, but I can't seem to find who suggests it.

[Dewrad]

Also, I've seen sowewhere PIE "eight" analyzed as "two fours".

I've seen that, too, but I can't seem to find who suggests it.

Adams and Mallory discuss the suggestion in The Oxford Introduction to Proto-Indo-European and the Proto-Indo-European World, but they don't cite whose idea it was originally. They do point out that the PIE for 'four' obviously wasn't *hₓoḱto-, but rather suggest a derivative from *hₐeǵ- 'to stick out', meaning something like *hₐoḱtoh₁(u) meaning "two sets of four fingers". But they concede that *hₐoḱto- 'set of four fingers' is unattested. They also suggest a possible link with Proto-Kartvelian *otxo- 'four'.

[Zju]

The idea was that eight was a dual of the Hittite word for four. The exact word was mentioned in one of these articles(which are a good read), I'll find it a bit later if someone doesn't do it before me.

[Soap]

It says *meju- is the Anatolian root for the number four. The seven blog posts are a very interesting read, indeed, but I dont see how Anatolian *meju- could be related to the numeral eight.
See http://langevo.blogspot.com/2014/10/two ... party.html

[hwhatting]

Any verbal root ending in /k/ plus one of the action/result suffices -ti- and -tu- would give the required cluster. So would the 3s and 2p of an athematic verb of one of those same verbal roots.

But those clusters occur across a morpheme boundary.

[Plus deverbal adjectives in -to]

I had a peek at the reverse index of roots in LIV, there's quite a lot of roots ending in a simple velar - much more than roots starting with one, is my impression. A glance at formations with -t- of a few of them in NIL show the usual velar reflex, not the palatal reflex. That was to be expected - I imagine if it would be otherwise, it would have been noticed before. There are two logical conclusions:
(1) There was a sound law /kt/ > /k't/, but its effects were removed by analogy already in Late PIE, except in isolated words like *ok'toH -> this means that the three dorsals were phonemic already in late PIE.
(2) There was no sound law /kt/ > /k't/. In that case, KathAveara's objection holds - why, based on a two-dorsal system, is there a palatal in *ok'toH, without any obvious palatalising trigger and in an isolated word that doesn't seem to show any obvious starting point for analogy? If, OTOH, we assume late PIE was three-dorsal and any developments triggering palatalisation happened at some earlier point (perhaps even pre-ablaut), then there is no problem.*)
Personally, I tend towards (2) - late PIE had three phonemic dorsals, and the developments creating that system happened before the earliest splits of the family.

*NB: This objection works only against proposals for two-dorsal systems where palatalisation is seen as secondary. It doesn't wok for two-dorsal systems where the palatalised velars are seen as original and the simple velars or the labiovelars are the derived phonemes. But those would run into similar problems - I haven't seen any cogent account on what would be the trigger factors for the required allophony in the reconstructable Late PIE system.

[KathTheDragon]

It doesn't wok for two-dorsal systems where the palatalised velars are seen as original and the simple velars or the labiovelars are the derived phonemes.

It is fairly easy to demonstrate that the labio-velars were distinct to the other two series. IIRC there are any number of pairs of roots which contrast.

[jmcd]

Fixsme: The creole usage in the article seems suspect: I tried googling for French creole C and French Creole D and got two Swadesh lists: http://ielex.mpi.nl/language/French_Creole_C/, http://ielex.mpi.nl/language/French_Creole_C/ but there's still hint as what distinguishes them and the forms cited in the second list seem unlikely what with the incomplete sound changes ('wespire' when 'r' becomes 'w' unconditionally otherwise)

I admit I can't find the reference to Haiti myself but there is only one creole language in Haiti anyway.

Loanwords aren't that much of a problem with creoles because the creole languages with the most substrate (West-Central African) influence, such as Saramaccan, still only have

[KathTheDragon]

I've been re-reading about the sound changes to Albanian, and it struck me that the palatalisation of the labio-velars before front vowels, but not of the back-velars, combined with the fact that the result of this palatalisation is identical to the outcome of *ḱw, implies that the labio-velars had the same PoA as the front velars. The logic is that rounding is a feature that would inhibit palatalisation, rather than encouraging it, so it is very implausible that a rounded consonant would palatalise, while its unrounded counterpart stays put. This effectively rules out the possibility of the labio-velars sharing a PoA with the back-velars.

Thoughts?

[vokzhen]

Couldn't /kʷ/ and /ḱw/ just merge to /ḱw/ before front vowels? Two series, not too distinct, collapse together (and to the least-marked POA, if it matters), especially if /k kʷ/ really are uvular, those don't play well with front vowels (though granted, it's usually the vowel that changes). If /ḱw/ is quite a bit rarer than /kʷ/ (I know almost nothing about phoneme distribution) it doesn't work as cleanly, but it probably still works.

EDIT: Dumbness revised to non-dumbness.

[KathTheDragon]

/ḱw/ is a rather rare sequence, while /kʷ/ is incredibly common (relatively speaking, of course). Your solution doesn't really avoid the problem whatsoever - you still require /kʷ/ to front, whilst maintaining that /k/ does not.

[Sleinad Flar]

Nonononono, Kathy.

Post-Mycenaean Greek shows the same weird development as Albanian: Labiovelars are palatized before front vowels, while plain velars aren't. The main POA of Mycenaean would have been the same, wouldn't it (k and k^w). Besides, doesn't k' > k also occur before other sonorants in Albanian? Then the development would have been something like k'R > kR (R including w) and subsequently kw > k^w (> k, s, c).

The facts that the back velars and labiovelars merge and that k'w and k^w have different outcomes in more sensible Satem languages clearly show that the labiovelars had the same main POA as the back velars, not the front velars.

Slightly related: it's funny that the two languages which show the best evidence for 3 dorsal series (Albanian and Luwian) are neighbours of Greek.

[KathTheDragon]

Kathy.

Please don't call me that.

Post-Mycenaean Greek shows the same weird development as Albanian: Labiovelars are palatized before front vowels, while plain velars aren't. The main POA of Mycenaean would have been the same, wouldn't it (k and k^w). Besides, doesn't k' > k also occur before other sonorants in Albanian? Then the development would have been something like k'R > kR (R including w) and subsequently kw > k^w (> k, s, c).

The facts that the back velars and labiovelars merge and that k'w and k^w have different outcomes in more sensible Satem languages clearly show that the labiovelars had the same main POA as the back velars, not the front velars.

Is /kʷ/ > /t/ really palatalisation, or just an unusual phoneme filling a massive hole in the distribution of a typlogically common one? But yes, this is probably the biggest issue with my idea (but it's not a fatal flaw).

Apparently, Albanian failed to palatalise front-velars in words containing sonorants, especially adjacent to sonorants. However, if we view this palatalisation as a wave originating in Indo-Iranian (as it almost certainly was, given the distribution of satemisation) it would make sense for front-velars in some environments to fail to palatalise. Cf Balto-Slavic. However, the key flaw in your argument on this point is that /ḱw/ always palatalises, not just before front vowels, meaning that the merger must have been /kʷ/ > /ḱw/, before the general palatalisation of the front-velars.

It is probable, IMO, that satemisation was in fact an IE-wide fronting of the back-velars from uvular to velar, with the satem languages further fronting the old front-velars from velar to palatal. It's clear that by the end of this shift, the labio-velars were velar, but that could just easily because they were uvular and therefore fronted, or labialised and therefore resisted fronting (except in Albanian before front vowels).

[jal]

Escuse my ignorance on the matter, but what exacty is "ḱ"? Is that just a fancy PIE way of writing "some kind of k which is not k"?


JAL

[WeepingElf]

Escuse my ignorance on the matter, but what exacty is "ḱ"? Is that just a fancy PIE way of writing "some kind of k which is not k"?

The difference is in place of articulation: *ḱ was more forward than *k, something like palatal vs. velar or velar vs. uvular.