Talk:Glottalic theory

Ejectives in Armenian?

I'm not a professional linguist, but doesn't Armenian have ejectives? And isn't it an indo-european language?

Yes, it most certainly is Indoeuropean, and it might have ejectives, but if it does, they are not primary (ie, inherited directly from Indoeuropean), so it isn't of much relevance here, I believe. Or did you have something else in mind? --Twid 17:15, 19 Mar 2005 (UTC)

Some dialects of armenian look like other indo european languages like greek and latin (having voiced stops where they do) and others have glottalic stops similar to IE in the proposed glottalic theory. I like this theory. It makes Germanic sound changes less anomalous. — Preceding unsigned comment added by 79.236.121.9 (talk) 17:49, 7 March 2016 (UTC)

Addition: Classical Armenian (and that's the decisive stage of the language) certainly had ejectives. There's a long treatment of this (in German) in Jan Henrik Holst: Armenische Studien, 2009, pp. 23 - 44. As to Modern Armenian, the other writer is right: it depends on which dialect you mean; some dialects still have the ejectives, some don't. — Preceding unsigned comment added by 93.202.199.92 (talk) 17:17, 7 April 2016 (UTC)

Note that in romanizations of armenian, the apostrophe (p', t', k') marks aspiration, not ejectives Canaron (talk) 09:17, 6 August 2019 (CET)

Chain shifted all three??

The text segment on noting that Germanic chain shifted all three is inaccurate. As the above poster noted, the Germanic sound changes are less anomalous under Glottalic theory. The text as is reads wrong, as under glottalic, PIE(G) [PIE with glotallic] b, d, g and gw are unchanged into Germanic! So, really it should state that only two shift. The unaspirated stops pass from p->f, t->þ, k->h and kw->hw. (I suspect that Germanic perhaps had these stops first pass quickly through an aspiration phase that led to the further shift, and recall seeing an article somewhere on that...) The remaining Germanic-identifying change is that the ejectives are reduced to stops with or without aspiration, i.e. t` becomes simple t.

It makes a good contrast too with Italic, Greek, etc. choosing a different series of *two* chain shifts, where Italic, et al. keep p, t, k and kw intact, instead turning b, d, g and gw into fricatives, thus allowing the ejectives to migrate into voiced stops. --Sturmde 29 June 2005 13:48 (UTC)

Actually the reference to chain-shifting all three sounds is accurate. But the sentence it is in is prone to misunderstanding; indeed I misread it the same way on first reading. The sentence is subject to the condition, “If the parent language had a typologically unstable system (the traditional p/b/bh system).” All of the examples (Indo-Iranian, Greek and Italic, Balto-Slavic, Germanic, Armenian) then describe the changes from the traditional reconstruction. This is a little confusing in an article on the glotalic theory, but the point is that the typological instability of the traditional reconstruction supposedly explains the many different reflexes of the voiced (aspirates?) in the daughter languages.

The glotalic theory certainly fits Germanic very well. But if the voiced (asperates?) were really unaspirated, why did they evolve in so many different ways. Was it to make phonetic space for the evolution of ejectives into voiced stops? Then why didn’t the ejectives evolve in many ways (or just remain ejectives)? --teb728 22:40, 15 July 2005 (UTC)

More objections

Proto-Indo-European did not permit a root to begin and end with a voiced stop, i.e., there are no such roots as *deg or *ged in the reconstruction above. I wonder if this argument should be treated seriously. At least according to Adolf Erhart's book Indoevropské jazyky. Srovnávací fonologie a morfologie, not only the roots as *deg but also the roots as *tegh, *dhek are prohibited! Why? It seems to be a similar phenomenon, caused by a similar reason, but yet it cannot be explained the same way as the absence of the *deg-type roots. The roots as *tek, *teg, *dek, *dhegh are all possible and frequent. And finally, the roots as *degh, *dheg are also possible but rare.

Moreover, Gamkrelidze and Ivanov divided IE languages into Taihun and Decem groups (it's a pity that not a word on this here; the names are analogical to Centum and Satem). According to these scholars, the Taihun group (mainly Armenian and Germanic, maybe also Anatolian and Tocharic) represents the older stage of the IE consonantism, while the Decem group (Indo-Iranian, Balto-Slavic, Albanian, Greek, Italic, Celtic) is more progressive with voiced stops on the place of supposed previous ejectives. There is a huge problem in the argumentation, however. Gamkrelidze and Ivanov reconstruct voiceless, ejective and voiced rows basing on both Germanic and Armenian facts - but without sufficient base for such a reconstruction.

In Old Armenian we find voiceless and voiced unchanged, but there are non-aspirated consonants on the place of postulated ejectives. At the same time, Germanic has voiceless fricatives (f, þ, x) on the place of voiceless stops, aspirated voiceless stops (p, t, k) on the place of postulated ejectives and voiced fricatives (β, ð, γ) on the place of postulated voiced stops.

Two facts should be emphasized. Voiced stops in Germanic are the result of further development of previous fricatives (and not in all positions). If the Glottalic theory were right, we could observe the process of the type d > ð > d here. It is not impossible but less probable than the change dh > ð > d like in the classic (non-glottalic) attempt.

The second significant observation now. De facto aspirated (except in some positions) voiceless stops in Germanic (p, t, k) come from the same source (ejectives) as non-aspirated voiceless stops in Old Armenian. In the same time, Armenian has also aspirated voiceless stops (p`, t`, k`) but they come from a different source (voiceless) and they correspond to Germanic voiceless fricatives.

Example: Gothic taihun "10" has the aspirated t- while Armenian tasn "t.s." has the non-aspirated t-. Maybe it has been the conventional transliteration used for Armenian which has confused the authors of the Glottalic theory: it is Old Armenian aspirated t` (and not t) which has (at least in the most instances) the same phonetic value like Germanic (aspirated) t.

Yet another observation. According to the traditional point of view, there was a consonantal change in Proto-Germanic (ex. *dhog^wh- > *daγ(w)). Next, there was the second consonantal change in some Germanic dialects (ex. *daγ(w) > Tag in modern German). Similarly in Armenian: after the first common Armenian consonantal shift there was the second consonantal shift in some Armenian dialects. Isn't it striking?

It looks like if languages, once put out of the steady state, tend to subsequent changes next. So, the simplest solution (in accordance to Occam's razor) seems to be to presume that Common Germanic and Old Armenian consonantal systems are more changed in comparison to the PIE state than consonantal systems of most other IE languages.

And finally. The classic view (such as Illych-Svitych presented) among followers of the Nostratic hypothesis is as follows:

Nostratic		Classic IE		Glottalic IE
ejective	*t’	voiceless	*t	voiceless	*t
voiceless	*t	voiced	*d	ejective	*t`
voiced	*d	voiced aspirated	*dh	voiced	*d

Note that ejective sounds are changed in placed. So, the glottalic hypothesis gains no support even from this. So, in my opinion, it has too many weak points and it lacks real arguments.

Perhaps part of these remarks should be noticed in the "objective chapter" of the article.

--Grzegorj 13:21, 15 August 2005 (UTC)

The problem I see with some of your arguments is that they are not universally accepted. For example, there is no consensus that voiced stops developed from fricatives in Germanic. Glottalic theory types (nb not me) are among those that assert the opposite development. Also, the argument based on aspiration is not sound because (a) there is no way to tell whether Gothic or other old Germanic languages had aspiration; (b) even in the modern languages, aspiration is only initial, exactly where you'd expect it given a strong initial stress (which is not reconstructed for early Proto-Germanic!). Hence aspiration could very easily have developed *within* Germanic, after the development of dynamic stress -- quite possibly even after the split into daughter langs. Benwing 03:58, 17 August 2005 (UTC)

You are fully right. There are some "but's" however. It seems to be obvious that reconstructions based on two or more languages are much more probable than those based on only one language (are the latter permissible at all in science?). The followers of the glottalic theory usually claim that their view is based on similarities between Armenian and Germanic (Gmc). But they do not emphasize (it would have been very strange if they had emphasized) that their Germanic reconstructions are not generally acceptable. So, their colossus stands on two legs, one of them is made of clay.

In fact, voiced fricatives (on the place of classic IE voiced aspirated = glottalic IE voiced stops) are present in most Gmc languages. E.g. IE (classic) *gh > Old English gh [γ] > ZERO or y in Modern English; similarly *bh > v except some positions. Even if we find *b on the place of *bh in German, there is foundation to assume existance of *β in the past, namely, the old *w seems to have mixed with *β and developed into b like in gelb < *γelw-, cf. English yellow.

In all Gmc languages *f, *þ, *x, *s > *β, *ð, *γ, *z after an unstressed syllable ("Verner's law"). Indeed, some scholars do propose *f > *b etc. instead, which is however very doubtful for some reasons:

• the change "a voiceless fricative" > "a voiced fricative" is just more probable than "a voiceless fricative" > "a voiced stop",
• the parallel change *s > *z is undoubtful (so, a voiceless fricative to a voiced fricative, not to a voiced stop),
• the voiceless stops did not undergo to the change,
• if *f had really changed into *b immediately, two fonologic features (fricativeness and voicing) should have been claimed to be changed at the same time which is much less probable than a simple change.

Anyway, the results of "Verner's law" fully mixed with the descendants of (classic) *bh, *dh, *gh. So, e.g. if we assumed the immediate shift *x > *g, we would have *x > *g > γ at least in some positions (fricative > stop > fricative).

For all those reasons, the model with Gmc *β, *ð, *γ is simpler and more probable than that with *b, *d, *g. As I said, "it is not impossible but less probable". I would even say that the idea of Gmc *b, *d, *g would violate Occam's razor. And I would say that there is no reason for assuming Gmc *b, *d, *g, except the glottalic theory. But then this assumption cannot be invoked as an argument for it (because of a logic error of vicious circle).

And about aspiration: the most important is that Old Armenian aspirated stops never correspond to Germanic aspirated stops but rather to voiceless fricatives. Besides, it is not true that aspiration in Germanic languages is only initial. For example, the German orthoepic norm tells about aspiration of [p t k] also in other positions, including the final position after a stressed vowel (like in Galopp, Betrieb, Gewalt, Kontrast, Musik, weg) and the position before [l r n] like in April. Icelandic also knows aspiration in many positions. So, it is more likely that aspiration disappeared in some positions because it can be explained with the help of one reason (reduction in an unstressed syllable in some position). If we assumed that aspiration is secondary in Gmc, there would be hard to find one reason for explaining all instances where aspiration can happen.

Of course I do not say that these arguments are "better" than those which have already been presented in the article. These are only addings.

--Grzegorj 00:09, 18 August 2005 (UTC)

Two random comments. One: surely Nostratic is even more controversial & less solid a theory than this, and wouldn't thus make a very good objection? And two: eg. *bʰ > *β is a change in both voicing and MOA just as well as eg. *ɸ > *b is, so the implausibility of such changes doesn't seem like a plausible objection either.

--Tropylium 15:40, 21 April 2007 (UTC)

Nostratic is not more controversial and less solid because it is based on an outer reconstruction which is always more plausible than an internal reconstruction. This is why it is a good objection.

You might be right, as I'm not sure how good Nostratic's standing exactly is these days. But I do believe I could think of an exception to that "always" (something involving kooky theories about isolates, frex...)

You are also wrong in the point 2. *bʰ > *β is not a change in both voicing and MOA because both bʰ and β are voiced (please study the meaning of both symbols before you use your argument). No voicing change is supposed to happen here then, the only change is "aspirated occlusive" > "fricative". There is nothing special in this change, examples are numerous. Thus the discussed objection is plausible, despite of what you say.

Grzegorj 08:55, 14 June 2007 (UTC)

"Voiced aspirated" is just another way of saying "brethy voice" AFAIK. Still voiced, yes, but not modal (full) voice - and that is what voiced fricativs have, 99% of the time. I'll also grant that it's a smaller change in voicing than unvoiced > fully voiced, but I believe the point does still stand.

Do any of these numerous spirantization examples of yours happen to involve voiced aspirates in non-IE langs, BTW? I'm only aware of that occuring in Germanic and Iranic. --Tropylium 10:32, 14 June 2007 (UTC)

(reindenting) It suddendly occurs to me that it is possible, especially within the glottalic framework, that Verner's law predates Grimm's. This minimizes the voicing/MOA flip-flops. And checking the VL article, hey presto, that idea is presented there alreddy in an entirely non-glottal context. :) --Trɔpʏliʊm • blah 00:15, 24 January 2009 (UTC)

In actually examining samples of the so-called "Taihun" languages, Tocharian for example does not even have an initial "t" for its word for 'ten' - in fact, the "sam" word would imply it is probably Satem anyway and so outside of the Decem/Taihun split. So I think that the statement about what the decem and taihun languages are needs to be rewritten or deleted. --50.67.247.221 (talk) 00:40, 6 November 2015 (UTC)

Ejectives

Were there ejectives in PIE? Read the following quotation: "The central tenet of the Glottalic Theory of the Proto-Indo-European obstruent system is that what has been traditionally reconstructed as a plain voiced stop series can be more convincingly reconstructed as a series of ejectives, which are inherently voiceless. However, the abundant evidence that this series was voiced has led to outright rejection by many of the Glottalic Theory. In an apparent attempt to obviate these objections, Frederick Kortlandt claims to have found ample comparative evidence that this series underwent a change from ejectives to implosives to voiced stops in several branches of Indo-European. In this address evidence will be presented to demonstrate that Kortlandt's implosive model is fraught with implausibility and inconsistencies and should also be rejected." (Simpson Humanities Center)

So, the evidence that Germanic (aspirated or not) *p *t *k, as well as Armenian non-aspirated *p *t *k are secondary and stand for original voiced stops is abundant. Charles Barrack demonstrated that even Kortland's attempt (ejective *t’ > implosive *d’ > Germanic *t^(h) = Indian d) causes objections.

--Grzegorj 07:39, 18 August 2005 (UTC)

So what are these arguments? Your link (which I fixed BTW) provides only a summary of the conclusions.

I however offer this paper: How Old Is The English Glottal Stop? I may be only an amateur in linguists, but to me those seem like good arguments about glottalization being an old feature of at least Germanic unvoiced stops. And according to another paper I've redd, the 1st laryngeal, commonly (AIUI) interpreted to having been a glottal stop, had similar effects on vowels in Slavic as the traditionally *voiced series (before its merging with *brethy.) (This could probably be added to the section mentioning Winter's Law by someone who knoes a bit more about it than me.) --Tropylium 15:25, 21 April 2007 (UTC)

This paper looks highly suspect to me. Kortlandt is really stretching in his attempts to find evidence for pre-glottalization and ignoring lots of other, equally plausible (or more plausible) explanations -- e.g. concomitant glottalization is hardly unusual in voiceless stops at the ends of words, and such stops often become simple glottal stops (e.g. in lots of Chinese dialects). Benwing (talk) 03:15, 1 December 2008 (UTC)

It's really unfortunate that Barrack's arguments are apparently not provided anywhere easily accessible (are there proceedings?). Talking with Martin Kümmel and others at a conference last year, I was surprised to find that they (as mainstream Indo-Europeanists) had warmed to at least the possibility to phonetically interpret the traditional media (voiced stops) as glottalised, more specifically implosives, in view of certain well-known root constraints (even if they are only relative and not absolute, meaning that certain root shapes are strikingly rare, statistically speaking). That does not mean that mainstream Indo-Europeanists necessarily accept other arguments in favour of the glottalic "theory" (recte: hypothesis), though, such as typology, pre-glottalisation and pre-aspiration in modern Germanic languages (or even the High German sound shift), the Armenian evidence or Lachmann's law, all of which I'd consider at least problematic (some variant of Winter's law seems to be accepted by many – as it is in the LIV –, but the precise formulation of the law and its phonetic interpretation are still not quite clear). The "pre-glottalised voiced stops" posited by the Leiden school at some point (as it had become clear to me some time before that) would seem to be nothing else but implosive stops, phonetically. It would make sense to assume implosives for the latest PIE stage but since it is *b and not *g that is rare (not quite absent, though – the LIV has a number of verbal roots ending in it but none starting with it) this cannot have been the original state and we have to assume ejectives for some older stage.

To sum it up, it would be premature to declare the glottalic "theory" flat-out dead – even if it may have seemed like that not long ago (Schumacher demolished the arguments in class quite cogently, or at least almost all of them), when even a controversial scholar such as Vennemann had abandoned it. (His reason, however, seems to have been the incompatibility of the hypothesis with his Semitic speculations, a reason unlikely to turn out relevant for mainstream linguists. No offence to him, who is a pleasant enough and charming gentleman in person, a skilled rhetorician and far from unimaginative, unintelligent, unoriginal or not bold, he does have numerous admirable qualities, I just seem to disagree with his methods, reasoning, conclusions – and especially his suggestion that the opposition to his Semitic etymologies is somehow politically motivated – on a regular basis, and I've learned from experience to automatically distrust any view that Vennemann endorses, and funny enough it seems that he's being consistently out of step with the Indo-Europeanist mainstream yet again.) It's one of those hypotheses that die surprisingly hard and keep giving you the feeling that there may be something to them; I mean, it's not exactly a crazy hypothesis, if anything, it's just somewhat unorthodox, in more than one sense.

One phonetic development that is especially striking is the fact that the only coronal (non-nasal stop that appears at the end of words in at least Italic, if not even PIE (as assumed by many investigators), is the voiced stop, while phonetically, this is totally unexpected: Usually, word-final stops are neutralised to the voiceless counterpart (in Armenian, interestingly, it's the voiceless aspirate). What you, Benwing, say, would rather support an interpretation of the word-final d as an originally glottalised stop. --Florian Blaschke (talk) 16:16, 12 May 2011 (UTC)

The t-d-dh exemplification

The Polish linguist Witold Mańczak in his book Wieża Babel (in Polish) gives the information what the example is that contradicts the assertion that no language is known which has voiced aspirates unless it also has voiceless aspirates. I have moved the observation on disputable character of this assertion into the "objections" chapter, and added the additional information taken from W. Mańczak's book, as the information is really hard to be found elsewhere. --Grzegorj 09:20, 18 August 2005 (UTC)

About "Objections"

Please don't cut off quite large parts of the article without discussing it first, OK? -- Grzegorj 01:38, 5 January 2006 (UTC)

Sorry about that. It seemed to require a lot of work to save. The stuff about "typologically unstable systems", for example, needs support. Yes, you'd expect ejectives to remain in some branches (unless of course they weren't actually ejectives), but it's bizarre to state that the branches wouldn't go in different directions phonologically if they had a "stable" system. As for empirical evidence, that's precisely what the vowel lengthening laws are supposed to be. I'll try giving it another passover with a little more finesse; please let me know if I take out anything you think can be justified. kwami 02:10, 5 January 2006 (UTC)

Oh, some fragments of the article need additions but it is the attraction of Wikipedia that no subject is ultimately finished and you can put more and more additions. And, which is the most important, we should report various views here rather than force our own ones. (BTW. the "unstability" stuff is not mine but Benwing's, see "history".) Despite of what you write, the Glottalic theory does not seem to be generally accepted, at least in its original form (and even if the glottalic - not necessarily ejective - character of the D series is accepted in fact, the old t/d/dh symbols are in use in the prevailing part of publications), and, as a consequence, it seems to be very useful to inform the reader about various objections against the theory which can be found in the literature (regardless we agree with them or not). Counterarguments, i.e. arguments which shake the original arguments of the followers of the glottalic theory, are yet more interesting. I mean the statement that there are no languages which have breathy voice consonants but no voiceless aspirates. Such a statement was formulated some years ago as one of the most important arguments but now we know that it is false. That is why I changed "no" into "few" in the article, and added the fragment on Kelabit. Note that this piece of information is rather hard to find elsewhere and that is why I have protested against removing it. Of course, some of the arguments of the anti-glottalicists can also be "counterized". I would suggest you to add your objections against "typologically unstability" instead of removing the thing, as it would be more useful for the reader than presenting the problem in the light of one point of view only. --13:03, 8 January 2006 (UTC)Grzegorj

Generally accepted?

"Although controversial at first, some variant of the glottalic theory is generally accepted today." Well, I can't pretend to know enough of the literature to be sure, but it seems to me that "generally accepted" is a considerable overstatment. Vennemann, for example, has changed his mind and no longer accepts the theory - see the article in the 2006 Transactions of the Philological Society. In the most recent handbook (2004), Fortson's Indo-European Language and Culture, it says "The glottalic theory enjoyed a not insignificant following for a time, and still has adherents; but it has been rejected by most Indo-Europeanists."

At best, one might say "still controversial", it would seem. --Pfold 11:29, 11 December 2006 (UTC)

The article itself seems to present a controversy. I wouldn't say that it's "generally accepted" enough to use that wording in an encyclopedia. --Puellanivis 21:10, 5 February 2007 (UTC)

Cross-linguistic tendencies & internal evolution of PIE

Concerning the part tagged for verification:

One objection to this reconstruction is that the voiced consonants are frequently voiceless in the daughter languages; aspirates in Greek and voiceless fricatives in Latin, for example. While it is common for aspirates to become tenuis and then voiced, as pʰ → p → b (lenition), the reverse is rare.

Okay, what? From what little I've redd, changes like b → pʰ happen all the time - eg. in Old Chinese and various other languages of East Asia - while I'm aware of NO examples of changes in the reverse direction. Additionally, changes of the former sort are usually postulated to happen in this sequence:

• The voicing - which is harder to sustain in stops than in continuants - becomes brethy
• Brethy voice is re-interpreted as aspiration
• Aspiration is then re-interpreted as the distinguishing feature
• The now-redundant voicing then disappears, leaving a stable system

Note especially how this is in line with the view of glottalic PIE being a precedessor of traditional PIE. And on another note, the article claims that this would require uniform internal evolution of PIE; I'm not sure how that folloes, since in this view b → bʰ (→ pʰ) would be limited to Italic, Greek, and Indo-Aryan, while Germanic and Armenian would have pʼ → p directly. Doesn't seem very uniform to me. I also don't understand why needing to postulate a period of internal evolution for PIE would be a bad thing in the first place? Doesn't the fact that PIE existed as a single language downright imply that? --Tropylium 16:27, 21 April 2007 (UTC)

I completely agree with you. I thought that reading that was funny after I just read up on breathy voice in it's respective Wikipedia article.

"In some...Bantu languages, 'breathy voiced' stops have been phonetically altered to devoiced stops..."

Proof positive right there and with citations too! :D --Redroot 19:27, 08 July 2007

User:Kwamikagami claims that the article for lenition has the sorces required. I think he's missing the point; yes, b → p → pʰ is rare (altho that's exactly what has gone on in most of Germanic if you agree with non-glottalic PIE reconstruction). We're not interested in the intermediate [p], tho; to repeat, the usual intermediate on the reverse path is the traditionally reconstructed bʰ! Also, while each step of pʰ → p → b is common, the chain as a whole isn't - and it's impossible here anyway due to the PIE tenuis stops not being dragged along too. --Tropylium 07:58, 4 October 2007 (UTC)

Yet another amateur here, but what Tropylium says here seems right on the money. Boldly getting rid of the 'pʰ -> p -> b' stuff and reorganising the 'internal evolution' stuff. 4pq1injbok (talk) 20:58, 19 January 2009 (UTC)

Vocalic lengthening before voiced stop

I puzzled as to why this should be considered a puzzling phenomenon - it is common enough in English!

It's puzzling because only occurs before voiced consonants that were originally glottalic in Balto-Slavic and Latin (when they are made to be voiceless.) So a root like wegh would become weg in Proto-Balto-Slavic but one with weg would become we:g, inexplicably. In Latin, this only happens if the final consonant was devoiced: wegh-to- would become vectu- but weg-to- would become ve:ctu-. Someone should definitely flesh this section out with real world examples! —Preceding unsigned comment added by 75.172.35.143 (talk) 09:36, 6 September 2007 (UTC)

"Inexplicably"? Doesn't the aspiration take a bit of time, too, so that [wegʰ] and [weːg] have the same length? That wouldn't surprise me at all. In my native southern German, a stressed syllable can either have a long vowel or a lengthened consonant (or a consonant cluster), so that in all cases it has a similar length. Lenes can't be lengthened, and that means that the few words like Ebbe, Egge, Bagger – at least some of which are imported from Low German – sound overly hasty. David Marjanović (talk) 18:47, 2 February 2008 (UTC)

Lengthening before consonant clusters consisting of voiced stop + (voiceless) stop is also found in Dutch dialects. Phonetically, this is really trivial. If Lachmann's law really has a phonetic basis, which is controversial, I doubt it has anything to do with the glottalic theory, even if it is correct in some form (especially the implosive or preglottalised variant; for Winter's law, this is more plausible). That this lengthening does not happen in front of old aspirates is unsurprising if you keep in mind that the Proto-Italic reflexes are probably voiceless fricatives, which then go back to voiceless aspirates as in Greek as an intermediate step from breathy-voiced stops. I think Lachmann's law is probably irrelevant and one of the weakest arguments considering that its interpretation is so uncertain. In fact, the common Italic/Celtic (Proto-Italo-Celtic?) rule that inserts /a/ after the first consonant in CDC clusters just like in CCCC clusters (for the /a/-epenthesis in clusters of four consonants, see Schrijver 1991: 416, 488ff.), which suggests the reinterpretation of CDC as CʔDC, is far more suggestive, and it is clearly different from and precedes Lachmann's law: frāctus < *frāgtos < Proto-Italic *fragtos < Pre-Proto-Italic *bʰra(ʔ)gtos < PIE *bʰr(ʔ)gtos was affected by both. It is exactly this example which I find striking and which makes me find the idea of a pre-glottalised or implosive realisation of the PIE voiced stops attractive. --Florian Blaschke (talk) 22:00, 30 January 2015 (UTC)

Is it really true that /p'/ tends to be absent in languages that have ejectives?

I wonder if whoever first came up with this factoid looked too hard at Na-Dené or other North American languages that generally have very few labials. Several have /w/ as the only labial consonant, others have /w/ and /m/ and/or /b/ (voiceless lenis) – but, AFAIK, /pʰ/ is just as rare as /p'/. Is this different in Afro-Asiatic, Kartvelian and Caucasian languages? I thought not…

Incidentally, the classical Proto-Nostratic reconstruction could also explain why classic PIE /b/ was so rare: classic PIE corresponds to classic PN /p/, and if that one was aspirated (as its direct descendant is in the Kartvelian languages, but admittedly nowhere else), it is the one most expected to be absent…

While I am at it, I will invert the "is now generally accepted" part. It is plain wrong. It makes the article look like it was written 20 years ago. David Marjanović (talk) 18:59, 2 February 2008 (UTC)

Yes, it's true, and independent of the lack of labials in some North American families. You get the opposite pattern in implosives. Restricting ourselves to the more common labial-alveolar-velar places, if a language has one implosive, it will be ɓ. ɗ is rather uncommon, and ɠ is rare. If a language has one ejective, it will be k’. t’ is less common, and p’ uncommon. It is assumed this pattern is due to the volume of air necessary to make these sounds easily: implosives work better with a large air pocket above the glottis, whereas ejectives work better with a small air pocket. Aspirate pʰ, on the other hand, is extremely common: if you have aspirates and labials, you will almost always have pʰ.

There is a regional lack of p (vs. b) around the Sahara, including Arabic, where p may have shifted to f. However, this areal feature does not extend to any of the posited homelands of p-IE, and some think it may be due to the more recent prestige of Arabic. A summary of these patterns can be found in the World Atlas of Language Structures, (Dryer, Haspelmath, Gil, & Comrie, eds). kwami (talk) 19:49, 2 February 2008 (UTC)

I just ran thru a bunch of language phonologies... Some examples:

• Hausa language#Phonology - Lacks both p and p', like many Chadic languages
• Hadza language#Sounds - Lacks p' and t' (but not further back ejectivs like ts')
• Gǀwi language#Other consonants - Ditto, and I don't think Khoisan heritage applies as an explanation!
• Proto-Mayan#Phonology - b' replaces p'

--Tropylium (talk) 00:42, 3 February 2008 (UTC)

I should add, the general tendency still appeared to be to have no holes whatsoever in the inventory, and I didn't spot a single language with just k' (maybe that's a theoretical universal similar to "n comes before m"?). But if there was a hole, it was always p'. --Tropylium (talk) —Preceding comment was added at 00:46, 3 February 2008 (UTC)

As far as I'm aware, this is completely parallel to the situation in nonglottalised stops: If a language lacks one of the voiceless stops, it's usually /p/ (the "frontmost" stop). If a language lacks one of the voiced stops, it's usually /g/ (the "backmost" stop among the common voiced stops). This seems to have a phonetic explanation, but I can't think of it anymore. If ejectives are interpreted as glottalised voiceless stops and implosives as glottalised voiced stops, this absolutely makes sense. --Florian Blaschke (talk) 14:27, 12 May 2011 (UTC)

Check out WALS.[1] Lack of /p/ is a circum-Saharan areal feature. Lack of /g/, however, crops up sporadically all over the world. So I don't think the two are directly comparable. — kwami (talk) 08:49, 16 September 2011 (UTC)

My Austrian German dialect has kʰ but neither pʰ nor tʰ and I actually have the most trouble consistently aspirating this pʰ when trying to speak Standard-German or even English. For me and other speakers it often ends up as a voiceless unaspirated p instead. So I was surprised to learn here that pʰ supposedly come so easily if you already have other aspirates. 213.47.74.51 (talk) 03:34, 25 September 2011 (UTC)

Relative rareness of p and g has an easy phonetic explanation. See here:

[2]

Basically, maintenance of voicing requires a minimum pressure drop across the glottis. This requires more lung force when the space behind the tongue closure is small (/g/) than large (/b/). Insufficient lung force and the sound devoices. Contrarily, to get an audible release in a voiceless stop, you need a minimum pressure drop across the tongue closure, which is harder to achieve when the space behind the closure is big and requires a bigger volume of air to build up the pressure (as in /p/). This is not really a problem with /b/ because voiced sounds are much louder (that's why we don't usually have voiceless vowels or approximants). For the same reason it's easy to see why both ejective and aspirated p's are relatively rare -- in fact, both require even more air than plain /p/.

BTW there's no reason why the (relative) absence of PIE /b/ needs to be typological. It has in fact been suggested that the word-initial change /b/ > /w/ occurred in pre-PIE. Benwing (talk) 07:27, 14 February 2013 (UTC)

I also doubt that the lack of /p/ in various African languages is due entirely to Arabic influence. This may be the case in Berber and Hausa but it seems rather less likely in Yoruba and extremely unlikely in the case of Ewondo and other languages in Cameroon / Southeast Nigeria. Benwing (talk) 08:03, 14 February 2013 (UTC)

Thanks, that's the explanation I was thinking of. Also, when languages have implosives, front implosives such as [ɓ] or [ɗ] are much more frequent than [ʄ] or even [ɠ], let alone [ʛ]; I've actually noticed that in East Asia. Actually, this makes me think that if the traditional PIE voiced stops were really implosives, the b-gap is completely contrary to what is to be expected, what with [ɓ] actually being the most common implosive stop. Either there was a change like /b/ or /ɓ/ > /w/ (this has been proposed by Michael Weiss, actually), or the voiced stops were originally (in pre-PIE) ejectives after all (which would be more satisfying from the point of view of areal typology and the phonotactic constraints), and they changed into PIE implosives as an intermediate step to the attested voiced stops. --Florian Blaschke (talk) 21:16, 30 January 2015 (UTC)

Another possibility Kümmel mentions is that (pre-)PIE /b/ or /ɓ/ became /m/ instead (first proposed by Jochem Schindler, apparently), or that, vice versa, the PIE voiced stop row goes back to denasalised nasals or nasal clusters (prenasalised stops?), but /m/ remained (he mentions some possible cognates in Uralic which contain nasals where PIE has voiced stops). --Florian Blaschke (talk) 21:24, 30 January 2015 (UTC)

Typology and universals

Hello everyone! I admit this is a bit off-topic, but it is an extremely interesting subject (to me, at least :D). Could we possible create an article devoted to these typological matters? I am especially interested in phonetics and phonology, but other subsystems of language would be useful as well, I suppose. --Pet'usek ^{[petrdothrubisatgmaildotcom]} 07:22, 9 May 2008 (UTC)

Linguistic typology, Implicational hierarchy and Linguistic universal exist, but they probably could use work. --Tropylium (talk) 22:00, 9 May 2008 (UTC)

"Unfalsifiable"

I don't really understand the addition of this part. Unfalsifiability, AFAIU, is a very strong objection; something unfalsifiable "isn't even wrong". But that's not the case here, as far as I can see. The theory has not been falsified yet, but that is not what is required; the requirement is that it would have to be impossible, even in theory, to present any evidence to the contrary. But just the possibility of Nostratic eventually being accepted, and within that scenario the possibility of the decem-series not being the ejectiv series, should be sufficient to constitute falsifiability. (The "More objections" discussion some ways up here indeed has a table to exactly that effect.)

The original version just stated "unprovable" - I thought that was about the lack of direct evidence, which is cover'd in a previous paragraph. Am I not seeing something here?

Or, if some critics have simply call'd the theory unfalsifiable, that's a different case then - it just needs a citation and possibly a small discussion to the short-lived footnote's effect - but is that really a consensus view? --Trɔpʏliʊm • blah 19:32, 26 January 2009 (UTC)

See JIEES 22/1994, pp. 133-149, "Proto-Indo-European *b and the Glottalic Theory" for one quotation of GT as pre-PIE being unprovable due to the lack of comparative evidence, and some intelligent search could probably yield more as that is something as obvious as 2+2=4. It is unprovable (or unfalsifiable, if you will), within the current generally accepted framework of PIE, namely due t the status of PIE (or Indo-Hittite) not having any obvious genetic relationship to other proto-families. Nostratic is supported by a tiny fraction of linguists today, even less then the GT itself, so you cannot possibly legitimise one with the other.

Comparative method can only reconstruct the latest stage of a language (some diverging dialect continuum), but in order to reconstruct earlier stage of a language you need comparisons with cognate families or some highly consistent internal reconstruction. Of those little supporters of glottalic theory today, namely the folks at the Uni of Leiden, AFAIK none of them treats the glottal stops (or "preglottalized stops" or whatever the fancy current term is) as some "pre-PIE", but Late PIE directly reflectable in the daughters. --Ivan Štambuk (talk) 20:12, 26 January 2009 (UTC)

Genetic evidence has come to support the Glottalic theory, in more recent years.

It is relevant to point out that recent genetic studies (by Nasidze, 2000 and Underhill 2009) have finally come to support the hypothesis that Haplogroup R (and R1), and it's descendant haplotypes R1a* (R1a1), originated nearer Northwest Iran. While the former marker, has been linked to post-glacial (and perhaps Neolithic), Caucasoid expansion, the distribution of the latter marker is widely believed to represent (at least one wave) of Indo-European expansion.

How on earth does this support glottalic theory of PIE? --Ivan Štambuk (talk) 05:02, 5 December 2009 (UTC)

And I for one distrust genetic "proofs". Some 10 years ago we had no genes whatsoever in common with the neanderthals, now we have some 4% in common. And languages aren't genetically transmitted. Rursus dixit. (^mbork³!) 22:53, 20 September 2010 (UTC)

My god. How does it not support Glottalic theory? — Preceding unsigned comment added by 2601:882:100:D7B0:5D18:C1DE:2A83:245E (talk) 14:02, 1 July 2015 (UTC)

Korean

It might be noted that the structure of the revised glottalic theory (barring some differences in location of articulation) is exactly the same as that of modern Korean: a voiceless series with allophonic, weak aspiration when word-initial, a glottalised series, and a strongly aspirated voiceless series. The first series shows weak voicing intervocalicly. Jakob37 (talk) 01:42, 21 November 2010 (UTC)

The tense(d) or faucalised consonants of Korean cannot be described as glottalised at all, neither even as implosive, see Korean phonology. They're actually rather like voiceless geminates, judging from the description. Also, the series which shows weak voicing intervocalically is the unaspirated (lax(ed)) series. Can't see any resemblance here. --Florian Blaschke (talk) 13:26, 8 August 2012 (UTC)

potential garbage in typological claims

Kortlandt's work in Slavic accent is spot-on, but the stuff I've seen in support of the glottalic theory looks like a lot of garbage. Kortlandt obviously knows his stuff when it comes to careful and complex ordered sound changes but seems out of his league with typological arguments or arguments requiring comparing lots of ancient and modern varieties to work out what's original and what's not. As a simple example, Kortlandt was quoted in this article as "pointing out" that the change bh->b is not found, hence typologically suspect. But this is utter garbage and trivially falsifiable. This exact change is found independently numerous times in the Indic languages -- the first and most obvious place to look if you want to check out the truth of this statement. Examples are the far northwestern Dardic languages (but not all, sometimes developing into tone); Sinhala in Sri Lanka, in the far south; some dialects in Bangladesh, in the northeast; etc.

Interesting notes, but these examples cannot really be separated from the glottalic theory itself. I have not seen anyone lay out the model in full, extended detail, but my impression is, in the glottalic view these would be treated as retentions in marginal areas rather than innovativ deaspiration, just like how groups like Iranian or Celtic would have undergone a merger B' → B rather than Bʰ → B. The claims on Sindhi imply that glottalic theory would involve reconstructing the main difference between the two voiced series as glottalication even during the disintegration of Indo-Aryan.

Romany meanwhile has bh -> ph; places in Bangladesh have bh -> β; etc.

Bʰ → Pʰ is obviously a lot more plausible than B → Pʰ. However, since glottalic theory asserts that B → Bʰ is a plausible change, this doesn't seem like an issue. --Trɔpʏliʊm • blah 22:59, 15 February 2013 (UTC)

All these changes are exactly parallel with postulated IE changes. You'd think people making typological assertions against the traditional PIE model would at least bother to check the Indic languages -- the only area generally preserving bh dh gh -- to see how these have actually developed, to see what change is really reasonable or not.

In fact, nearly all of the typological claims of the glottalic theory have been disputed. I think we need to be much more careful including any unchallenged typological assertions by pro-glottalic authors. Benwing (talk) 07:06, 14 February 2013 (UTC)

re Adjarian's Law

In case I do not find time to go into this in the near future: the fact that Adjarian's Law is a counterargument against Glottalic Theory has nothing to do with the original value of the traditional voiced aspirate series, per se (which the GT is not even about). It is instead about how to reconstruct the Proto-Armenian unaspirated voiceless series. The problem is the existence of a dialect group (Middle Armenian) where the Old Armenian Pʰ, P, B system evolves into [Pʰ, B, P]. This development is impossible if we only allow shifts in ±voicing and ±aspiration - something should have merged into something else.

Kortlandt's proposed solution is to reconstruct a glottalized *Pʼ series, which becomes voiceless in standard East Armenian, but voiced in the problem dialects. Adjarian's law allows a different detour: the B series first develops into voiced aspirates, which then break into voiced stop + breathy voiced vowel; the P series also becomes voiced; and then, the voiced series is devoiced specifically before these breathy voiced vowels, "undoing" the merger of the two non-Pʰ series. --Trɔpʏliʊm • blah 12:59, 5 November 2014 (UTC)

Can you add it to the article? --Ivan Štambuk (talk) 18:45, 7 November 2014 (UTC)

This would probably take some reorganization to add a section treating Armenian in partcular. --Trɔpʏliʊm • blah 20:09, 7 November 2014 (UTC)

Hmm, my understanding is that the B series is now thought to have been breathy-voiced in Old Armenian (like in Indo-Aryan), as it is still so realised in several dialects. This obliterates the need for the reconstruction of ejectives. Ejective realisations do exist in Eastern Armenian, but they are mostly weak (barely audible) in Armenia (own experience) and only become more noticeable as one travels north (or so I heard), and they are particularly conspicuous in Tiflis. Of course, the Armenians there are bilingual in Georgian, so the obvious explanation for the ejective realisation is that it is simply Kartvelian influence. Also, a few very old Iranian loanwords such as arcatʿ "silver" have taken part in the Armenian sound shift as well, showing that it is real and that ejectives cannot be reconstructed for Proto-Armenian (nor for Old Armenian, probably). --Florian Blaschke (talk) 19:51, 29 January 2015 (UTC)

The link to "Middle Armenian" is to diachronically middle, not geographically. Do we cover these dialects somewhere? — kwami (talk) 21:36, 29 January 2015 (UTC)

Not really, all we have is Classification des dialectes arméniens. However, see the table at Armenian language#Consonants at the end, which confirms what I said about breathy-voiced stops. --Florian Blaschke (talk) 01:46, 30 January 2015 (UTC)

Implosives instead of ejectives

Details in chapter 2 (this is in German, but probably largely understandable even for non-German-readers). Very useful rundown on the relevant facts and arguments, but it would have been nice to mention that the development from ejectives to voiced stops is actually attested in Chechen and some other Northeast Caucasian languages, see Nakh languages#The voicing of ejective consonants, even if perhaps on a global scale the change may be comparatively rare (Kümmel 2008 probably goes into this, however). --Florian Blaschke (talk) 20:59, 30 January 2015 (UTC)

Broken link. :-( David Marjanović (talk) 23:58, 25 November 2015 (UTC)

Fixed. --Florian Blaschke (talk) 17:07, 13 September 2020 (UTC)

Caucasian substrate

New (2015) publication by Bomhard; see [3]. Joshua Jonathan -Let's talk! 04:17, 18 July 2016 (UTC)

Meaning of "voiced aspirated stops" were the only voiced stops

I really don't understand this, early in article I read According to them, the non-aspirated forms occurred in roots where two non-ejectives were present because of a rule that prohibited more than one aspirate in the same root. And also, Thus, an Indo-European *DʰeDʰ (where *Dʰ represents any non-ejective stop) might be realized as *DeDʰ (attested in Indic and Greek) or as *DʰeD (attested in Latin). In contrast, traditional theory would trace a form attested as both *DeDʰ and *DʰeD to an Indo-European *DʰeDʰ.

Later however, I come across this statement, If the "plain voiced stops" were not voiced, the "voiced aspirated stops" were the only voiced stops. The second constraint can accordingly be reformulated as two nonglottalic stops must agree in voicing.

These cant both be true. If only one aspirate was tolerated in a root, then this means that PIE breathy voiced stops were not the only voiced stops. The PIE root bheydh (to trust) would have been phonetically realized as either bʱɛjd or bɛjdʱ. Is there something I'm not getting. Since Dʰ represents any non-ejective stop, that includes dʱ and tʰ. Idielive (talk) 17:51, 9 April 2017 (UTC)

I think the idea is that the voiced aspirate stops were the only voiced stop phonemes, and unaspirated voiced stops were merely allophones of the aspirated voiced stop phonemes. AJD (talk) 20:39, 9 April 2017 (UTC)

The Germanic borrowing argument discredited

The term '*rik-' in Proto-Germanic was borrowed from Proto-Celtic '*rixs-' or '*riks-', not '*rig-'. For references see A Comprehensive Introduction to Nostratic Comparative Linguistics (Bomhard 2023) and Wiktionary (https://en.wiktionary.org/wiki/Reconstruction:Proto-Germanic/r%C4%ABks). Pcg111 (talk) 13:18, 13 April 2024 (UTC)