zompist bboard

WeepingElf wrote:

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

This is possible. PIE *o is reflected as a rounded vowel distinct from *a only in Italic, Celtic, Greek and Armenian. All other branches of IE have merged *o and *a. If PIE *o was not rounded but merely further back than *a, there is no reason to assume that *h3 was rounded.

But Kath's reasoning isn't really evidence for that - it's not uncommon for /u/ to round things while /o/ doesn't. /u/ is often more rounded than /o/ even when both are 'rounded' vowels.

Salmoneus wrote:

WeepingElf wrote:

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

This is possible. PIE *o is reflected as a rounded vowel distinct from *a only in Italic, Celtic, Greek and Armenian. All other branches of IE have merged *o and *a. If PIE *o was not rounded but merely further back than *a, there is no reason to assume that *h3 was rounded.

But Kath's reasoning isn't really evidence for that - it's not uncommon for /u/ to round things while /o/ doesn't. /u/ is often more rounded than /o/ even when both are 'rounded' vowels.

Yes, certainly. Indeed, I am sceptical about this scepticism of roundedness of PIE *o. While I think that the Late PIE vowels *e and *o were pretty open (at least [ɛ] and [ɔ], if not [æ] and [ɒ]), I still think that Late PIE *o probably was rounded. (At least, I see no reason to change the traditional transcription symbols, as such a step would only cause confusion!)

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

And in slavic /i/ palatalizes following velars but /e/ does not. Q.E.D. /e/ must have been non-front *rolls eyes*

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

And in slavic /i/ palatalizes following velars but /e/ does not. Q.E.D. /e/ must have been non-front *rolls eyes*

R.Rusanov wrote:

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

And in slavic /i/ palatalizes following velars but /e/ does not. Q.E.D. /e/ must have been non-front *rolls eyes*

There are even more examples. In Germanic, /i/ triggers fronting of preceding back vowels ("i-umlaut"), but /e/ doesn't; in North Germanic, /u/ triggers labialization of preceding vowels ("u-umlaut"), but /o/ doesn't.

R.Rusanov wrote:

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

And in slavic /i/ palatalizes following velars but /e/ does not. Q.E.D. /e/ must have been non-front *rolls eyes*

/e/ is often less front than /i/.

ObsequiousNewt wrote:

R.Rusanov wrote:

KathAveara wrote:I thought it was very probable that /o/ (and therefore, also /h₃/ ) was not rounded, given that /u/ delabialises velars but /o/ does not.

And in slavic /i/ palatalizes following velars but /e/ does not. Q.E.D. /e/ must have been non-front *rolls eyes*

/e/ is often less front than /i/.

And likewise, /o/ is often less rounded than /u/. It is so in my idiolect, at least.

This isn't my reasoning. As I recall, WeepingElf mentioned it in this thread. But I may be misremembering who said it, and I'm about to go, so I'm not going to look it up.

Ok guys, who want to help improve my lexicon, and what's the best way to share the work? Google Drive? Some kind of version control (probably Git or SVN)?

Mostly the problems are normalizing transcription with data coming from 3 or more sources, and de-duplicating and merging entries. Also, I'd really like to move this over to RDF to make it easier to build in cross references, but I'm not aware of any RDF schemata for lexicography or lexicons, which would have to be developed.

Don't know if I could really contribute anything, but I'd love to try, and I'd probably learn more about PIE.

/i u/ causing palatalization or labialization while /e o/ don't is hardly uncommon (probably more common than the reverse!) Brazilian Portuguese palatalizes /t/ to [tʃ] before /i/ but not /e/, Japanese basically does the same thing (with t -> ts before u but not o, as well), Mandarin has a famous phonology problem with the phonemes /k tɕ ʈʂ/ that occurs before /i/...

I'd like to help, but I'm on a new project job-wise and don't have too much time on my hands. But thanks for providing and improving that ressource!

Mandarin has a famous phonology problem with the phonemes /k tɕ ʈʂ/ that occurs before /i/...

Can you explain further?

Terra wrote:

Mandarin has a famous phonology problem with the phonemes /k tɕ ʈʂ/ that occurs before /i/...

Can you explain further?

If I recall correctly, it's that /ts tʂ k/ before /i~j/ all become [tɕ]. (Similarly with [tɕʰ ɕ].)

(It could be analyzed as either /ts tʂ k/ and /tɕ/ being in a complementary distribution, or the latter being a "collective" allophone of all of the former.)

KathAveara wrote:This isn't my reasoning. As I recall, WeepingElf mentioned it in this thread. But I may be misremembering who said it, and I'm about to go, so I'm not going to look it up.

You're probably thinking of my speculative idea earlier in the thread to reanalyze the *o : *a contrast as *a : *ə.

The absense of known *o / labiovelar interaction is a minor point in favor of the idea, though it certainly doesn't suffice for anything on its own. The basis is instead
1) the quantitative abundance of *o, versus the typologically odd scarcity of *a
2) Brugmann's Law, which suggests that *o was a particularly low vowel in pre-PII
3) the widespread development of *o to *a
4) the lack of explicit evidence for *h₃ having been contrastively rounded

Tropylium wrote:

KathAveara wrote:This isn't my reasoning. As I recall, WeepingElf mentioned it in this thread. But I may be misremembering who said it, and I'm about to go, so I'm not going to look it up.

You're probably thinking of my speculative idea earlier in the thread to reanalyze the *o : *a contrast as *a : *ə.

The absense of known *o / labiovelar interaction is a minor point in favor of the idea, though it certainly doesn't suffice for anything on its own. The basis is instead
1) the quantitative abundance of *o, versus the typologically odd scarcity of *a
2) Brugmann's Law, which suggests that *o was a particularly low vowel in pre-PII
3) the widespread development of *o to *a
4) the lack of explicit evidence for *h₃ having been contrastively rounded

I see.

Why a ə specifically for *o *a? Is there evidence for that and against something like Seri, with ɑ æ?

Then again, if it's a reconstruction that keeps *a distinct from *e...

Nortaneous wrote:Why a ə specifically for *o *a? Is there evidence for that and against something like Seri, with ɑ æ?

Then again, if it's a reconstruction that keeps *a distinct from *e...

In Hittite, *a has merged with *o, so it must have been distinct from *e in Early PIE already.

I am still not convinced of this scepticism against the traditional reconstruction - a very common and unmarked system! - at least for Late PIE; it may be, though, that the non-high vowels huddled more closely together in Early PIE.

I have thought again about the idea that Early PIE had only two non-high vowels, *e ([æ]) and *o ([ɑ] or [ɒ]), while *a split from *e only in Late PIE under the productive vowel-colouring effect of *h2.

My result is a resounding NO.

Why? Quite simple: Where Late PIE has *a, Hittite has not /e/ but /a/, so *a must have been distinct from *e in Early PIE already, and merged with *o in Anatolian. A Late PIE split of *a and *o from a common Early PIE source also makes no sense, as the distribution of Late PIE *a clearly tells that it is a recolouring of *e, and not of *o, as it appears in e-grades and not in o-grades (*h2o gives just *o, not *a).

This means that Early PIE already had three non-high vowels, which can be transcribed with the conventional symbols *e, *a, *o, even if they probably were phonetically closer together to each other, maybe [æ a ɒ], before they drifted apart from each other and became the typologically less marked [ɛ a ɔ] of Late PIE (at least, of the dialects that did not merge *a and *o).

I don't understand. What's your argument for *a?

Salmoneus wrote:I don't understand. What's your argument for *a?

My argument is that:

1. *a cannot have split from *e after the departure of Anatolian because in Anatolian, it merges with *o, and does not "merge" with *e. If *a was a post-Anatolian development from *e next to *h2, one would see /e/ in Hittite cognates of Late PIE forms with *a, but one sees /a/ there, which also reflects PIE *o.

2. *a cannot have split from *o because it occurs in e-grade and not in o-grade forms. If *a developed from *o next to *h2, *h2o would have to develop into *a; but *h2o develops into *o, and *a is the outcome of *h2e.

This means that *a must have split from *e before the departure of Anatolian. Hence, Early PIE must have an *a distinct from *e and *o. Is this so hard to understand?

OK, I get it. But I see two problems:

a) *a may have arisen in both Anatolian and Late PIE after the split. For instance, there may already have been allophonic colouring of *e to *a, with phonemicization only happening later, in both languages

b) does it matter anyway? Even accepting your argument, it doesn't tell us anything about the 'original' vowel system, it would only tell us that by the time anatolian split off there would be three phonemic vowels. It still leaves us knowing nothing about the phonetic values of those vowels, or about their origin. How isn't this just a semantic point about how early you define as "Early PIE" vs "pre-PIE"?

Salmoneus wrote:OK, I get it. But I see two problems:

a) *a may have arisen in both Anatolian and Late PIE after the split. For instance, there may already have been allophonic colouring of *e to *a, with phonemicization only happening later, in both languages

Yes, it could have happened. But the more parsimonious solution is that it had already happened before the split, at least allophonically (a-colouring by *h2 and o-colouring by *h3 apparently were still productive in Late PIE times, as we do not see any exceptions - but then, there is a risk of circularity here because in many forms, the laryngeal is reconstructed by means of its vowel-colouring effect!).

Salmoneus wrote:b) does it matter anyway? Even accepting your argument, it doesn't tell us anything about the 'original' vowel system, it would only tell us that by the time anatolian split off there would be three phonemic vowels. It still leaves us knowing nothing about the phonetic values of those vowels, or about their origin. How isn't this just a semantic point about how early you define as "Early PIE" vs "pre-PIE"?

Sure. It indeed doesn't tell us anything about the phonetic values of the vowels before the split. The symbols *e, *a and *o are just conventional, based on a guess of their values in Late PIE, and may not represent the actual phonetic values of these phonemes. But it is very likely that *e was more fronted than the others, and *o had an additional feature that distinguished it from *a, and given its reflexes in Celtic, Italic and Greek, the best candidate for this feature is [+round]. But it may have been something else, such as pharyngealization or whatever.

What concerns the distinction between Early PIE and Pre-PIE, let me rehearse how I understand these terms (I know that they are somtimes used differently and inconsistently):

1. Late PIE is the latest common ancestor of the non-Anatolian IE langugaes.
2. Early PIE is the latest common ancestor of Late PIE and Anatolian.
3. Pre-PIE is a stage earlier than Early PIE accessible by internal reconstruction.

I admit that "Pre-PIE" is a somewhat hazy notion. Where to stop? It is probably reasonable to distinguish more than one Pre-PIE stage.

WeepingElf wrote:1. Late PIE is the latest common ancestor of the non-Anatolian IE langugaes.
2. Early PIE is the latest common ancestor of Late PIE and Anatolian.
3. Pre-PIE is a stage earlier than Early PIE accessible by internal reconstruction.

I admit that "Pre-PIE" is a somewhat hazy notion. Where to stop? It is probably reasonable to distinguish more than one Pre-PIE stage.

You're certainly free to make up your own terminology if you always can make people remember the definitions. But naming 1) Late PIE will make you being misunderstood a lot, because it's most often used to denotate the immediate pre-break-up stage of PIE, i.e. what you call "Early PIE". Better go for something like Non-Anatolian PIE for 1), at least people won't be confused.

hwhatting wrote:

WeepingElf wrote:1. Late PIE is the latest common ancestor of the non-Anatolian IE langugaes.
2. Early PIE is the latest common ancestor of Late PIE and Anatolian.
3. Pre-PIE is a stage earlier than Early PIE accessible by internal reconstruction.

I admit that "Pre-PIE" is a somewhat hazy notion. Where to stop? It is probably reasonable to distinguish more than one Pre-PIE stage.

You're certainly free to make up your own terminology if you always can make people remember the definitions. But naming 1) Late PIE will make you being misunderstood a lot, because it's most often used to denotate the immediate pre-break-up stage of PIE, i.e. what you call "Early PIE". Better go for something like Non-Anatolian PIE for 1), at least people won't be confused.

OK. I'll change my terminology then. I indeed have noticed that "Late PIE" is usually used for what I called "Early PIE".