Talk:Proto-Indo-European phonology

Italian PIE

After comprehensive searching and comparing, I finally discovered that:

• Italian PIE phonology article has highest ever number of paradigms, as compared to Wikipedias in other languages.

• Italian PIE phonology article has single PIE reconstruction, without contradictions between various theories of Buck, Beekes ans Ramat.

—The preceding unsigned comment was added by Wikinger (talk • contribs) 21:27, 1 March 2007 (UTC).

Mayrhofer (1986: 170 ff.)

It would be nice to know what book Mayrhofer (1986: 170 ff.) refers to. Tibetologist 23:38, 16 July 2007 (UTC)

What font would I need to see the syumbols next to the 'h's in the Laryngeal section? I switched my browser of the Charis SIL so I can finally see the weird g' palatal velar consonant symbols, but all three laryngeals are just an 'h' followed by a box... making it very hard to determine which is which. —Preceding unsigned comment added by 71.120.201.39 (talk) 20:00, 25 August 2008 (UTC)

Have you tried Arial Unicode? That's what I've got and it works. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 05:35, 26 August 2008 (UTC)

"Standard" theory

I suggest that this article focus maninly on the "standard" theory (aspirated voiced stops, laryngeals, etc.) as presented e.g. in Meier-Brugger, Indo-European linguistics. Competing theories, such as the glotallic theory, should then be presented in a later section. Cheers. Grover cleveland (talk) 21:52, 14 September 2008 (UTC)

Grimm's law, Grassman's law, etc.

I suggest that either

• this article should systematically describe how each phoneme was treated in the descendant languages, or
• it should not do any such description, and leave such material to another article.

I don't mind either way, but right now we are in some kind of half-way house. Grimm's law, Grassman's law, etc. are described. But other, equally important developments (e.g. e,o -> a in Indo-Iranian) are not. Cheers. Grover cleveland (talk) 21:58, 14 September 2008 (UTC)

I agree that those references to reflexes of PIE aspirates and Grimm's Law (I mean in that brief list under the consonant table) are probably unnecessary, though the info on Grassmann's Law might be worth keeping. By the way, we already have this article: Indo-European sound laws.KelilanK (talk) 23:34, 14 September 2008 (UTC)

This article should be on phonological laws that already operated in PIE proper (Szemerenyi lengthening, Siebs' law, Stang's law, Bartholomae's law and all those unnamed ones like "TK > Kþ in post-Anatolian-Tocharian"). Indo-European sound laws should cover reflexes in daughter languages. --Ivan Štambuk (talk) 11:08, 15 September 2008 (UTC)

Phonemes?

Should the consonants really be given between phoneme slashes / / ? The usual representation is just *ḱ etc. On the other hand, the phoneme would use the IPA symbol: /c/ (or probably */c/ because it's a reconstruction). As it is, the table seems to be a strange mixture of API and PIE conventions. - And even if it is correct now, the vowels should be written in the same way. --ἀνυπόδητος (talk) 14:31, 21 October 2008 (UTC)

There's no law that says only strict IPA characters are allowed between slashes. If you read the linguistic literature about Native American languages, you'll see things like "/y/" in reference to the palatal glide and "/š/" in reference to the voiceless postalveolar fricative. Nevertheless, it certainly isn't usual to use slashes in discussions of PIE, and I wouldn't be averse to their being removed. —Angr 14:39, 21 October 2008 (UTC)

I think that it's important to emphasize that it's the phonemes that are being reconstructed. Yes, the vowels should use the same notation, but the section on vowels should be a bit longer, explaining the traditional view that there were only 2 real PIE vowel phonemes: */e/ and */o/, the *i and *u being just the syllabic allophones of the glides (i.e. phonetically vowels, phonologically sonorants), with no real evidence for PIE phoneme */a/ (it was highly-marginal, at best). --Ivan Štambuk (talk) 15:00, 21 October 2008 (UTC)

I agree that the fact that these are phonemes should be stressed, but couldn't we do that in a more... traditional way? The PIE introductions I know mention the phoneme issue, but none of them uses the syntax of this article. People who don't know much about PIE might get the impression that this is the standard way of writing PIE sounds. I think you are right with the vowels; the section could need some cleaning up. --ἀνυπόδητος (talk) 18:00, 21 October 2008 (UTC)

It's not usual to use slashes unless a phonemic-phonetic contrast is being made. The phonemicity of the palatals is debated, so using either slashes or brackets would be taking sides in the debate, which is sidestepped by using neither. The same is true of *h1, which some have suggested might have been two phonemes. kwami (talk) 20:33, 21 October 2008 (UTC)

If you look through the contents of Template:Language phonologies, you'll find that most articles on the phonologies of attested languages do not use slashes in the phoneme charts. I think it's sufficient for us to say in prose at the top of the chart that what follows is phonemes. —Angr 20:45, 21 October 2008 (UTC)

I wasn't aware that the phonemicity of palatovelars is disputable?! Didn't Melchert settled the issue on three series 20 years ago? --Ivan Štambuk (talk) 22:16, 21 October 2008 (UTC)

I'm sure you know more about this than I do. But your link claims Melchert proposed the three velar series, when they're simply centum-satem reflexes. kwami (talk) 23:40, 21 October 2008 (UTC)

People here are confused between how phonemes are represented in attested languages versus protolanguages. While it's true that, for example, a proto-sound like *p is a "phoneme" and that it may be represented as */p/ when pronounced as such, it's not true that these symbols are simply interchangeable with IPA notation. The phoneme *h2 **CANNOT** be represented in IPA notation as */h2/ whatsoever because "h2" is not a valid IPA symbol! Depending on the theories of the IEist in question regarding the pronunciation of *h2, however, the sound may be represented as some universally recognized IPA symbol (such as /x/, for example, a velar fricative). Mixing up these notations belies vast ignorance of the topic. Less editing, more reading. --24.78.208.183 (talk) 20:17, 21 October 2008 (UTC)

• Dear IP that calls me a "plain ignorant of basic comparative linguistics, which is all too typical on the children-maintained Wikipedia.": Have you actually read the sentece in the lead, saying: The reconstruction of abstract units of PIE phonological systems (i.e. segments, or phonemes in traditional phonology) is much less controversial than their phonetical interpretation. This especially pertains to the phonetic interpretation of PIE vowels, laryngeals and voiced stops. ? IPA has nothing to do with it. Symbols employed by the IEist denote reconstructible phonemes, nothing else. Even the terminology is completely conventional [so-called "laryngeals" are anachronistic misnomer, and "palatovelars" are most likely not real palatals]. The exact phonetic status of dorsals and laryngeals is not 100% reconstructible (though it's more or less debatable). Actually, the only English-language PIE introductory that I've read actually uses the slashes throughout fairly consistently. We should make a difference between e.g. real and the only PIE sibilant */s/ and it's voiced allophone in assimilated clusters (e.g. *nisdos "nest") which later became phonemicized as */z/ in some daughter languages, *y and *i [all the same phoneme], *þ resulting from "thorn clusters" etc. Glottalic theory, which should be discussed in separate section, is just another phonetical interpretation of the same phonological system. We could be using A, B, C, D...instead, but we use traditional symbols that are as close to what most people think are the actual PIE sounds, in Late PIE. --Ivan Štambuk (talk) 20:19, 21 October 2008 (UTC)

I've told him I will block him if he continues edit warring and personal attacks. kwami (talk) 20:33, 21 October 2008 (UTC)

I see you already did block. I was just about to do so myself. —Angr 20:39, 21 October 2008 (UTC)

As nasty and uninformed as he was, there is a valid question of whether we should be using slashes when it is not uniformly agreed that all of these distinctions were phonemic. It's also common to omit slashes in inventory tables of living languages, even when they're used in the text. kwami (talk) 20:47, 21 October 2008 (UTC)

Yes; see my comments above. Just because the anon was rude, arrogant, and mistaken about the reason we shouldn't be using the slashes, that doesn't mean we should keep them. —Angr 20:55, 21 October 2008 (UTC)

I see Ivan Štambuk has already removed the slashes in the table. Should't they be removed in the text as well? --ἀνυπόδητος (talk) 10:51, 22 October 2008 (UTC)

I have no problem with that, but only after the allophony and the disputed phonemic status of some of them is discussed. I'm about to rewrite the section on three dorsals series dispute and on the vowels, could you please wait an hour or so.. --Ivan Štambuk (talk) 12:02, 22 October 2008 (UTC)

Take your time! --ἀνυπόδητος (talk) 13:13, 22 October 2008 (UTC)

By the way, I'm sorry for the hubbub I obviously raised by starting this section. I had no idea my simple question would cause some IP to insult you and start an edit war. Cheers --ἀνυπόδητος (talk) 13:16, 22 October 2008 (UTC)

The IP started causing trouble before you started this section, though. —Angr 13:56, 22 October 2008 (UTC)

I've finished with what I had in mind writing, now it's up to you guys to fix bad English and possible NPOV.. --Ivan Štambuk (talk) 16:36, 22 October 2008 (UTC)

three series of velars

I completely rewrote this section to make it more theoretically sound and less POV towards the three-series view. The previous version had major issues in that most of the arguments used aren't valid in historical linguistics. Benwing (talk) 06:27, 27 October 2008 (UTC)

Can you be more specific on "not valid in historical linguistics" ? ^_^ AFAIK, there are no known well-defined rules for regularly deriving the "secondary" palatovelars in all Satem languages (esp. in PIE words for dog and tooth which you removed), other then some ad-hoc devised exceptions (anti-rules). This way the section concludes as if the three-way contrast is some remnant from Brugmannian times, being held just because people are too conservative. --Ivan Štambuk (talk) 06:59, 27 October 2008 (UTC)

The essence of the two-series argument is that the "plain velar/palatovelar" distinction never existed in PIE, nor in any of the Centum languages derived from PIE. Rather, it was an originally allophonic distinction that developed in the Satem languages as a result of a secondary change of simple velars into palatovelars in most positions, and was turned into a phonemic distinction by loss of the labiovelars. That is: all early IE languages were heavily ablauting, and thus following the change just described, would have had lots of alternations between plain and palatovelars in different forms of he same root. No problem as long as it's all allophonic: cf. the complex allophony of phoneme /t/ in English. When labiovelars merged with plain velars, however, the allophony got obscured -- some roots alternated e.g. /k/ and /c/, others only had /k/. Analogy would quickly have generalized one or the other across each root, not necessarily in a predictable fashion. Furthermore, in such a sound change that works through areal spread rather than strict genetic derivation, you'd expect differences both in the allophonic environments involved in the original split and in the results of the process of analogy, and we do see both of these.

The problem with your arguments is, as I see:

• • • "This theory, although possibly responsible for some pre-PIE developments which have left discernible traces in Late PIE, cannot explain all the reflexes of velars in attested Indo-European languages". Necessarily, an analogical smoothing out of an unpredictable alternation makes it impossible to provide strict sound laws explaining the results. But that doesn't require us to reconstruct more phonemes in the parent language. A simple example from the history of English: Around 400 AD, /k/ became palatalized to /tʃ/ in the vicinity of certain front vowels; likewise for /g/. Very quickly, further vowel changes obscured the original distribution, and analogy soon smoothed out paradigms, leading to even more disturbances; furthermore, certain very common words (e.g. 'give', 'get') borrowed their form from nonpalatalizing dialects (Danish and Northern English). By Chaucher's time, less than 1000 years later, it's completely impossible to produce any set of sound changes to handle all the variants between 'ch' and 'k'/'c' even if we ignore loan words. The result is a three-way velar, labiovelar, and palato-alveolar distinction in English. But nonetheless no one insists on reconstructing a separate series of palato-alveolar sounds in Proto-Germanic, much less in PIE, based on this evidence.
    • "The usual argument is that allegedly no Indo-European language actually distinguishes the reflexes of all three rows, and that for Centum languages the formula *ḱ = *k ≠ *kʷ is valid, and for Satem *ḱ ≠ *k = *kʷ" Whether Albanian and Armenian show evidence of three rows is orthogonal to the question of whether the "plain velar/palatovelar" distinction is original or secondary, since Albanian and Armenian are clearly Satem languages. This is not at all the same for Luwian, since it shows no other Satem influences. Even so, for Luwian we need to show that the attested three-way distinction is not due to a later process (c.f. what happened in English), which is hard due to how common palatalization processes are, the small number of roots in question, the questionable etymologies of some of the roots and the small and late attestation of Luwian compared to other Anatolian langs.

Benwing (talk) 08:45, 27 October 2008 (UTC)

btw I don't mean to imply that belief in parent language with three series is unreasonably; if you get that POV then see if you can NPOV the conclusion. however, I don't think the pro/con arguments have been correctly stated in most places. esp. the *huge* disparity between the clear evidence for labiovelar phonemes in pre-Satem and total lack of such evidence for palatovelar phonemes in pre-Centum, and the likewise the total lack of any dialect differences in the supposed Centum innovation eliminating the palato/plain distinction in velars, despite the fact that the innovation had to have occurred totally independently in at least two or three separate places (Tocharian, Hittite, elsewhere) due to geographic separation. Benwing (talk) 09:04, 27 October 2008 (UTC)

A few comments to the above: I'm not sure what you mean by "cf. what happened in English"; are you talking about the k → č change seen in "cool/chill", "break/breach", etc.? Or are you talking about the actually Romance change of k → ts → s in French cent "hundred", which is mirrored in Romance loanwords in English like "century"? Also, you seem to be implying that Hittite in particular or Anatolian in general was a centum island in a satem sea, just like Tocharian was. But Anatolian was spoken immediately adjacent to Greek, so it was no island. —Angr 09:11, 27 October 2008 (UTC)

Thanks, you again open my eyes ^_^ Reading relevant parts from Sihler 1995 on b.g.c. [1] got me thinking. His explicitely stated conclusion: However, this is an artifact of the method, not a picture of the early history of PIE: there never was a variety of PIE with three dorsal stops, strikes me as *shocking*. I've never seen this expressed so succinctly and explicitly. Basically Sihler argues for a genetic classification of "Proto-Satem", with Centum group reflecting original pre-Late-PIE state (which would make Satem a genetic node, but Centum not - just being archaic). One smaller weakness in this is that sometimes the relative chronology of RUKI and Satemization is not reconstructed in the same order in e.g. Indo-Iranian and Balto-Slavic, which would not make it a strictly common innovations but parallel spread of some areal feature.

I don't know about Indo-Iranian, but there is no 100% clear evidence for labiovelar traces in Balto-Slavic - the development of prothetic *i/*u before resonants in BSl. is still unsolved (this would be the only known trace of labiovelars in BSl., and there are alternative theories for the distribution of *i/*u that presuppose the merger of labiovelars and velars).

I'm also a bit perplexed with Sihler's usage of the term palatal - for him "palatovelars" were indeed palatal stops, not palatalized (fronted) velars?! No wonder that the shift of palatals to velars in Centum group would for him be "nothing short of phenomenal.

Sihler mentiones that the crucial evidence in Armenian and Anatolian seems to hinge upon especially difficult or vague or otherwise dubious etymologies, which is not surprising considering the languages evolved and the fundamental improbability of the proposition. The Albanian evidence is of better quality, but can be accounted for (and easily) without recourse to a threefold system of dorsals. - Can this be further substantiated? Craig Melchert is enough of an authority not to be dismissed with some generalized statement such as this. Luwian material is generally held as crown evidence in favour of threeway contrast of PIE dorsals; if it is indeed so unreliable, it should deserve a bit more.

I think that the necessity of the reconstruction three-way PIE dorsal contrast from the viewpoint of the pure historical linguistics must be further clarified, by examples of roots in which there is no obvious motivation for palatalization, and by examples of three-way contrast being reflected in identical environments.

Also I think that the traditional "depalatalizations" (which are not mentioned in this version of the section) in Balto-Slavic and Albanian of palatovelar series need to be contrasted to what appear (at least to me) random environments where in Satem the pre-Satem plain velars would "remain intact". This provides two different perspectives on the "Centum" reflexes in "Satem" languages, which is not so trivial scholarly conundrum. --Ivan Štambuk (talk) 13:55, 27 October 2008 (UTC)

Angr, I tried to respond awhile ago to your comments but the response seems to have disappeared. I was indeed referring to the "cool/chill", "break/breach" change in English. As for Hittite, agreed that it is not a "centum island in a satem sea" given its location in Anatolia. (Although AFAIK it's attested as early as 1800 BC, and I'm not sure whether Greek was in this area so early.) But the point is rather that Anatolian split off from PIE thousands of years before it ended up in Anatolia. If Hittite only lost its three-way distinction due to Greek influence rather than independently, then it seems to me that (1) it's not so likely that such an areal influence across languages so divergent would completely affect all lexical items; (2) it's quite likely that traces of the previous phoneme would remain (e.g. influences on adjoining vowels, esp. given the already-existing complex and irregular vowel system, which would protect against analogical effects that might wipe out such a change; or occasional lexical items with /s/ or some other palatal-ish phoneme); (3) it's quite likely that there would be other evidence of influence from the same language, esp. borrowed words.

Ivan, I think the analogy between the English palatalization I just mentioned and the two-velar theory should answer your comment in the final two paragraphs. Starting from Modern English and Modern German, a pure application of the comparative method would require us to construct a three-way distinction between e.g. */y/ (young, jung; year, Jahre), */g/ (for-get, ver-gessen; give, geben; be-gin, be-ginnen), and */gy/ (yesterday, gestern; yellow, gelbe). But in fact, no such three-way distinction ever existed; this is just an artifact of the comparative method combined with complex changes in English, involving a fair amount of inter-dialect borrowing and analogizing (which the comparative method cannot handle).

Note also: What you call "pure historical linguistics" is a misnomer -- you really mean "using only the comparative method". The problem is, when the issue of the PIE velar correspondences was being investigated, historical linguistics was in its infancy and the comparative method had only just been formulated (aka the Neogrammarian Hypothesis). At that point it was basically the only tool available -- the method of internal reconstruction hadn't yet been discovered and even the concept of "phoneme" was decades away. As a result, many of the arguments made at that time in favor of one side or the other are completely obsolete now but they're still being rehashed. I suspect that the consensus would be quite different today had the debate occurred recently rather than almost 150 years ago.

Finally, it's not obvious to me that Sihler is really arguing for a Proto-Satem clade, in the strict genetic sense, as much as just a dialect area that was in close areal contact at the relevant point in history. He also, for example, argues that Greek and Sanskrit were at one point a dialect area, on the basis of shared vocabulary, morphological developments, etc. Benwing (talk) 01:52, 16 November 2008 (UTC)

Ivan, thanks for the link to Melchert's online article. I looked through it, and in the conclusion he proposes either (1) palatal k' always becomes z, while plain velar k becomes k; or (2) palatal k' becomes z only before front vowels (and w), and merges with velar k elsewhere as k. the reason for this hesitancy is that all examples presented where k' > z are before front vowels (or in paradigms which include both front and back vowels). if we assume that (2) is correct, then proof of the three-way split in Luvian hinges on evidence that velar k does NOT also become z before front vowels. problematically, only one single piece of evidence appears to exist for this (the word for "comb") -- not much to base a theory on. Benwing (talk) 03:36, 23 November 2008 (UTC)

Yeah, the proof would depenend on the evidence of PIE *k not becoming /z/ before front vowels (or alternatively—on evidence of PIE *ḱ becoming /z/ provably before non-front vowels, which would be much harder to find). The real problem is that what appears one mere counter-example is one third of all non-tentative examples of Luvian *k < PIE *k! (at least of those linked in Melchert's paper). I've looked for more in Anatolian Historical Phonology on books.google.com but it doesn't allow preview for the page 251 which deals with reflexes of PIE *k in Luvian :( [Though I can imagine what it would claim as the on the page 53 Melchert reconstructs 3-way dorsal contrast for Common Anatolian]. Newer and abundant evidence would certainly be desirable. --Ivan Štambuk (talk) 05:04, 23 November 2008 (UTC)

Sihler appears to indeed advocate some kind of "Proto-Satem" clade, as he uses that exact phrase on p. 93 and p. 154 (footnote c.) Interestingly, another source of mine (non-English PIE handbook) remarks that the scepticism towards the 3-way dorsal contrast usually comes from "Western linguists" ^_^. If someone could redo the "shifting" scheme in SVG (or something) that Sihler presents on p. 153, it would be highly-illustrative for the article (or can it be done in Wiki markup? I doubt that those lines can be drawn..). --Ivan Štambuk (talk) 07:19, 16 February 2009 (UTC)

Notation for nonsyllabic glides

It appears that lots of WP articles uses just 'i', and 'u', instead of 'y' and 'w', not even bothering to use that combining diacritic underneath to indicate non-syllabicity (as in i̯, *u̯). This could be confusing as some people reconstruct genuine PIE */i/ (and moreover */ī/). Since 'y' and 'w' are both easier to read and write than i̯ and *u̯, can we agree to use them as opposed to the alternatives? --Ivan Štambuk (talk) 12:52, 9 November 2008 (UTC)

There should be no problem with glides after consonants or word-initially. They are mostly transcribed as *y *w on WP and I agree that this is the easiest solution. Of course, a strict distinction has to be made between glides and vowels (non-syllabic *y, *w vs. syllabic *i, *u).

The question is what to do with glides after vowels (i. e. what many scholars interpret as diphthongs): *ei or *ey? The first is more common as far as I know, and it is by far the most common transcription on WP. The only book I can think of at the moment writing glides in such positions is LIV. On the other hand, I don't know any scientific reason for writing *ei. The *i is definitely nonsyllabic, and to make things more complicated, some books write *i after vowels but *y between vowels, i. e. *ei vs. *eye. These problems could be easily avoided if we write glides throughout, but it will be some work to find and correct all of these. --ἀνυπόδητος (talk) 14:55, 9 November 2008 (UTC)

I think I've read somewhere that 'y'/'w' is more favored by American, and i̯/u̯) by European scholars, so favoring any should be "taking sides" but this shouldn't be much of a controverse. I think that the 1) consistency 2) clearness are the most important, i.e. if 'y' and 'w' are used word-initally and after the consonants here on WP, they should be used consistently throughout, and if 'i and 'u' are used, they should be used with non-syllabicity marker and consistently in all the other positions. I mildly favour y/w because it's easier to type and read.. --Ivan Štambuk (talk) 16:40, 9 November 2008 (UTC)

What do you mean by "all the other positions"? Do you mean if we use *i̯, *u̯ for glides (word-initially and after consonants), we should also use *ei̯, *eu̯ for diphthongs (like the LIV), or that we should use *ei, *eu (like, for example, Fortson)? --ἀνυπόδητος (talk) 19:11, 9 November 2008 (UTC)

The first one - that we should use it for diphthongs too. The same sound - the same symbol in all positions. --Ivan Štambuk (talk) 19:48, 9 November 2008 (UTC)

I fully agree. And if we use the *y, *w notation, the diphs would be *ey, *ew. I suggest waiting a bit to see if someone objects, or if people would prefer *i̯, *u̯, and then start changing the relevant pages. --ἀνυπόδητος (talk) 19:55, 9 November 2008 (UTC)

For historical as well as for practical reasons, "diphthongs" like ei, eu, etc. should not be written *ey, *ew. First of all, no major scholarly work I know of uses those spellings consistently. Besides, in most daughter languages those diphthongs have evolved separately from the vowels and often also from VyV/VwV sequences. If you change their notation, then for coherence's sake you will have to add explanations like "the sequence of a monophthong *e plus the glide *w became /u/ in Greek" to all relevant Wikipedia articles, while it's so much easier and more concise to write this change as *eu > u. Of course you can have something like *ew# > u (a similar notation is traditionally used for the "diphthongs" of Proto-Austronesian, although they are fewer in number and phonotactically more restricted), but that will go against all major references this article quotes - not to mention against tradition - and could be interpreted as original research.KelilanK (talk) 21:26, 9 November 2008 (UTC)

If we use *y and *w, it should only be for the sake of typographic convenience. In my experience, using y and w in place of *i̯, *u̯ is not an American habit but rather a custom of works intended for more general audiences. In their scholarly writing, Americans like Jasanoff, Watkins, and Cowgill use *i̯, *u̯, but in works like Watkins's appendix to the American Heritage Dictionary, which is intended for a broader audience, he uses *y and *w. As for the diphthongs, there's no reason not to use the conventional *ei *eu, etc., since they're typographically easy, and *ei̯, *eu̯ are just as uncommon as *ey *ew. —Angr 22:08, 9 November 2008 (UTC)

Can you please state those historical reasons for not writing [j] and [w] as 'y' and 'w', but using 'i' and 'u' signs that are reserved for their syllabic allophones?

I'm proposing this just for the reason of consistency of articles on WP. The fact that there is no wide-spread consistency on this notation in all PIE works in no way legitimizes the misleading [IMHO] "chaos" that is currently practiced.

Diphthongs have "evolved" differently, except when they did not. What makes the former case more relevant?

I'm not sure I comprehend this; hasn't Ancient Greek preserved *ew as ευ? Isn't "ew#" supposed to denote "word-final sequence of ew" ?

It's certainly more concise, but writing e.g. *deywos as *deiwos and *dyēws as *dyēus would arguably be much more confusing, using twofold notation for the same sound.

I'm not that sure on "original research" argument - I assure you that there are books that use *y and *w fairly consistently for both diphthongal and non-diphthongal non-syllabic PIE glides. --Ivan Štambuk (talk) 01:38, 10 November 2008 (UTC)

As for *deiwos vs. *dyēus, it doesn't look especially confusing to me because I've seen similar-looking reconstructions many times. I could agree that a new user might find this confusing, but on the other hand they might not. Most of the time the diphthongs do evolve differently, e.g. Greek *euC becomes ευC but *ewV becomes εV. Similarly Latin *euC becomes u:C but *ewV becomes ovV; Sanskrit *euC becomes oC but *ewV becomes avV; etc. Benwing (talk) 02:47, 1 December 2008 (UTC)

Well, I've seen it too (just as I've seen the notation I'm proposing many times used consistently), but that doesn't mean it's a good practice to follow. I cannot imagine why PIE sound *[j] should be marked as *i in one case, and as *y in another, and PIE sound *[w] as *u in one case, and *w in another. That looks awfully confusing to me. *i in *deiwos is the same sound as *y in *dyēus, and *u in *dyēus is the same sound as *w in *deiwos. The alternative notation looks highly counter-intuitive to me and I cannot imagine how someone unfamiliar with the intricate details PIE phonology would come to realize that by pure mental logic.

As for the "dipthongs evolve differently" - I thought that the user who said that was referring to something else. What you write as *euC sequence (and where I say it should be *ewC) is a diphthong by definition, a tautosyllabic (closed syllable) VR sequence. VR in VRV sequence is not diphthong by definition, being in heterosyllabic position, so it (understandably) has different reflexes in daughter languages (i.e. as a normal sequence of a vowel followed by a sonorant). I also don't see how this particular issue of is/is not in favour of my proposal.. --Ivan Štambuk (talk) 03:54, 1 December 2008 (UTC)

I don't have any problem with your proposal and I agree it's more consistent. I was just suggesting that there might be arguments for doing it the other way. Benwing (talk) 07:16, 2 December 2008 (UTC)

2¢ from s.o. who hasn't contributed anything to this article: I agree that we should be consistant; I also think that we should avoid "obvious" typographic shortcuts, since we have a mostly novice readership that they won't be obvious to. As for which symbols to use, we should go by whatever's most used in serious publications. We have no need for typographic convenience—unless our readers have font-support issues. kwami (talk) 08:20, 2 December 2008 (UTC)

That's sort of the problem - serious publications aren't consistent, and it isn't clear what counts as a "serious" publication. Watkins's appendix of IE roots to the AHD is certainly serious, but it's intended for a general audience, not a specialist audience. I think the majority of works for general audiences (and there aren't many of those) use y and w for the glides, while the majority of works for specialist audiences use i̯, u̯; the majority of works for both types of audiences use ei/eu/oi/ou etc. for the diphthongs. So we come back to the perennial question at Wikipedia - are we writing for a general audience or a specialist one? On the one hand, Wikipedia as a whole is certainly intended for a general audience; on the other hand, how likely is it that anyone who isn't already at least a little familiar with Indo-European linguistics is going to be reading this page in the first place? —Angr 12:40, 2 December 2008 (UTC)

Include audio?

For someone like me - who is not a linguist - it is difficult to know exactly how these consonants and vowels are pronounced. E.g. what is the difference between h1, h2 and h3? Is it possible to add audio - perhaps mp3 files? Or could you, if these already exist somewhere else, provide a link? —Preceding unsigned comment added by Gabelsberger (talk • contribs) 22:01, 31 January 2009 (UTC)

Even linguists don't know how h1, h2, and h3 were pronounced. Just think of them as algebraic variables like x, y, and z. —Angr 22:32, 31 January 2009 (UTC)

Discussion of accent?

Unless I've missed something, I don't see any discussion of the pitch accent of PIE (although the acute-accent notation is used in some places, it is not explained). It seems as if this ought to belong in the article. Grover cleveland (talk) 02:57, 17 February 2009 (UTC)

Well, Proto-Indo-European accent is a separate article, but you're right, it should be briefly summarized here. —Angr 07:35, 17 February 2009 (UTC)

Vowels section

While this part of the article is provocative and stimulating, it has a few issues:

• seems a lot like WP:OR (not least for the almost complete lack of citations)
• it is rather opaquely presented
• it presents a viewpoint that is not in the mainstream, but doesn't make that clear to the reader.

For example, if I understand this section correctly, it suggests analysing long "o" in *pṓds as a sequence of two short vowels */peods/. I haven't seen this idea presented in any of the main introductions to PIE. Where does it come from? Grover cleveland (talk) 03:08, 17 February 2009 (UTC)

I think you misunderstood that sentence, it does not claim that */o:/ can be interpreted as dimoric */eo/, but that */o:/ as */oo/ and */e:/ as */ee/.

The lack of citations doesn't make it OR. I think that that section makes a rather good compromise on explaining the various viewpoints on PIE vocalism you can find the usual handbooks.. we cannot just e.g. assume that there was PIE */a/ or */ā/ just become some reconstruct them, as some others do not: we must present a coherent overview of all viewpoints per NPOV policy. Hence the various "systems" of PIE vocalism depending on the perspective. Same goes for the phonetic vs. phonological *i and *u, or the examples of long vowels that are not explainable as morphologically-conditioned lengthened grades. And there are some other general issues that section doesn't touch yet.. If you think that the text is confusing and misleading, feel free to rewrite it.. --Ivan Štambuk (talk) 06:47, 17 February 2009 (UTC)

I've removed the bit about /o:/ being "dimoric o" (the word incidentally is "bimoraic") because it's trivial. Long vowels in most languages are analyzed as being bimoraic, and the fact isn't relevant to any particular discussion here. —Angr 07:32, 17 February 2009 (UTC)

OK, I though it was worth adding that PIE had only one time of long vowels (bimoric), as some of its immediate daughters (Proto-Germanic and Proto-Balto-Slavic) distinguished two kinds of prosodically relevant "long vowels": bimoric (long) and trimoric (hyperlong). There was no such distinction in PIE, where all long vowels were bimoric, and with ictus on the second mora (when accented). --Ivan Štambuk (talk) 11:54, 18 February 2009 (UTC)

Well, as long as the comparison is being made directly it's okay. But just saying that PIE long vowels are bimoraic without comparing it to cases of trimoraic long vowels seems trivial. It's probably also a little OR-ish (unless someone's said in print); I'd rather go with something like "PIE is reconstructed to have two degrees of vowel length, short and long, unlike Proto-Germanic and Proto-Balto-Slavic, which have been reconstructed to have three degeres: short, long, and superlong." After all, for all we know the PIE long vowels were as long as the PGmc and PBS superlong vowels. —Angr 13:44, 18 February 2009 (UTC)

The lack of citations doesn't make it OR. I think that that section makes a rather good compromise on explaining the various viewpoints on PIE vocalism you can find the usual handbooks.

Maybe -- but it would be interesting to see which authors make the suggestions that e.g. /o/ was phonetically [ə], or that there is a typological similarity between the *PIE vowel system and those of some of the Cacucasian languages. I just haven't found those ideas so far in my (pitifully meager) reading on the subject. Cheers, Grover cleveland (talk) 07:29, 18 February 2009 (UTC)

There are zillion theories on the interpretation of PIE vocalism, and if the article covered them all there the whole section would grow immensely and distract the reader from several crucial points. Those are (IMHO): 1) status of phonemic */a/ 2) phonetic/phonological status of the glides *y/i and *w/u and their syllabification rules 3) status of long vowels and their relationship to synchronic PIE rules (lengthened grades and various contractions as results of PIE phonological rules vs. phonologically-unjustified long vowels like in *pṓds) 4) allowed diphthongs. Anything more detailed would prob. be an overkill.

Typologically PIE is commonly compared to Nortwest Caucasian language (like Abkhaz which has 2 real vowels just like PIE) in minimal vocalism and rich consonantism for about half a century now. This is taken to extreme in glottalic, Pontic and Nostratic frameworks which analyse the correspondences from various viewpoints. However, now that you mention it, methinks that that innocent remark might be out of place and should prob. be removed as misleading/confusing, as it insinuates pre-PIE genetic relationship in theories that are generally held as still unproven, and which are discussed on other WP articles. --Ivan Štambuk (talk) 11:54, 18 February 2009 (UTC)

Labiovelars vs. velar-labial sequence

Are there any minimal pairs for *kʷ vs. *kw, *gʷ vs. *gw or *gʷʰ vs. *gʰw? If not, what is the justification for seeing the labiovelars as separate phonemes, rather than sequences of two phonemes? After all, English has words like "queen" and "what", deriving directly from the PIE *labiovelars via Grimm's Law, yet no one suggests that English has unit phonemes *kʷ or *hʷ. Grover cleveland (talk) 16:36, 23 February 2010 (UTC)

There's certainly a difference between *kʷ and *ḱw, because they developed differently in the satem languages. I don't know if there are literal minimal pairs, though. I'm not sure if it can be shown that there's a difference between *kʷ and *kw, but there are several reasons for interpreting *kʷ and its brothers as being single phonemes. For example, they can occur before consonants without developing syllabic [u]: the zero grade of a root like *twer- would be *tur- because *w between two consonants syllabifies to *u, but the zero grade of a root like *kʷer- would be *kʷr-, not **kur-, showing that *kʷer- doesn't have the *w phoneme in it. +Angr 17:27, 23 February 2010 (UTC)

Ah -- that's a good argument -- thanks. Btw, are there any PIE words containing a velar followed by *w (let alone minimal pairs?). I haven't come across any. Another interesting possibility would be the existence of any undisputed roots containing both velars and labiovelars (since PIE roots cannot contain two stops from the same place of articulation). Ringe mentions one dubious one as possibily post-PIE, but I can't remember what it is. I think all other possible pairs of places of articulation are attested -- am I right? Grover cleveland (talk) 21:48, 23 February 2010 (UTC)

I don't know of any words with a true velar + w cluster. I guess the first place to look would be Sanskrit words containing kv. I think the root you're thinking of is *kʷek- 'sleep', but I'm not 100% sure. You seem to be questioning the validity of the labiovelars, but if anything is in doubt, it's the distinction between the palatals and the true velars; those are the two series that may well have started out as a single series before getting differentiated in a way that's no longer easy to figure out. In other words, there is not much chance that pre-PIE didn't have *kʷ, but there is a fair chance that it didn't have both *ḱ and *k. +Angr 13:44, 24 February 2010 (UTC)

"inside"

All of the reconstructed roots with */b/ inside are usually confined to a few Indo-European branches ...

"inside"? Is this an ESL attempt to say "roots containing */b/", or what? — And does "All ... usually" mean all, or most? —Tamfang (talk) 03:58, 8 December 2011 (UTC)

Proto-Indo-European initial /*b/

How many roots with PIE initial /*b/ are known?? This question is important because it relates to a mention of it at the P article in the Usage section. Georgia guy (talk) 15:15, 28 May 2013 (UTC)

Very few indeed. Only a handful of the roots listed at wikt:Index:Proto-Indo-European/b start with regular b rather than bʰ (sometimes spelled bh on that page, but it's the same thing), and many of the roots listed on the page that do start with regular b are kind of questionable since they're not attested in very many languages and may not really be Proto-Indo-European at all. It's true that p was extremely rare word-initially in Proto-Germanic; the words listed at wikt:Index:Proto-Germanic/p are mostly loanwords or of obscure origin. I don't know if any of them has an impeccable PIE pedigree. But p isn't that rare in Germanic outside word-initial position; modern English words like sheep, ship, sleep, lip, harp, help, slip, keep, hope, etc., all had noninitial p in Proto-Germanic. Angr (talk) 18:51, 28 May 2013 (UTC)

Those are words that end in P. What pattern does their etymology follow?? I know that there's the sp combination, and it is preserved in English; the p doesn't turn into f in these words. Georgia guy (talk) 18:56, 28 May 2013 (UTC)

Kluge's law might account for some of them. CodeCat (talk) 19:10, 28 May 2013 (UTC)

Voiceless aspirated stops

Joseph Wright wrote grammars of ancient languages describing the sound changes they underwent in developing from PIE. His phonological system includes a series of voiceless aspirated stops (mediae aspiratae). (See the page from his Greek grammar on Google Books.) I'm curious, does anyone know what happens to this series in the modern PIE phonological system? Are they subsumed under "voiced aspirated", perhaps with the (presumably) distinct development explained by a set of environments? Perhaps this question is only interesting to those like me who read ancient language grammars, but it's relevant to describing the historical development of PIE theory. — Eru·tuon 20:44, 28 December 2014 (UTC)

I think they're plain voiceless in modern theory. CodeCat (talk) 21:00, 28 December 2014 (UTC)

Most or all of the voiceless aspirated stops of Indo-Iranian ( > voiceless fricatives in Iranian) come from clusters of plain voiceless stops with *h₂ or possibly *h₁. David Marjanović (talk) 23:17, 22 November 2015 (UTC)

Refs for claims about velars

Does anyone have any sources for the "citation needed" statements in the PIE phonology#Dorsals section, about alternative realizations for the palatals and palatovelars (and in fact for the whole of the argumentation that follows them)? W. P. Uzer (talk) 22:35, 19 November 2015 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified one external link on Proto-Indo-European phonology. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive http://web.archive.org/web/20130517043355/https://openaccess.leidenuniv.nl/dspace/bitstream/1887/2662/1/299_013.pdf to https://openaccess.leidenuniv.nl/dspace/bitstream/1887/2662/1/299_013.pdf

When you have finished reviewing my changes, please set the checked parameter below to true or failed to let others know (documentation at {{Sourcecheck}}).

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 18 January 2022).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—^{cyberbot II}_{Talk to my owner:Online} 03:47, 23 April 2016 (UTC)

No PIE /a/ phoneme?

What nonsense is this, /a/ is the most common sound on this earth. There is no way that PIE did not have some variety of it, whether that was ə, ɐ, æ, ɑ, or ʌ. To say otherwise can only be foolish. Why is a notion like this one even included in this atricle, which is suppose to be using scholarly sources from well trained linguists, not pseudo scientific hogwash. — Preceding unsigned comment added by Idielive (talk • contribs) 23:18, 10 March 2017 (UTC)

Ignorant. The sound [a] existed as an allophone of /e/, so for all purposes they are treated as the same phoneme. And it's believed by scholars that before laryngeal colouration, [a] was the unique phonation of /e/. Read a little bit more before insulting the most respectable experts in a field of science you know nothing about. –Tom 144 (talk) 14:36, 10 March 2019 (UTC)

Liquids?

The *l and *r are given as “Liquids” on the chart. Is the *r known to be a tap, approximant, or a trill? 66.87.125.206 (talk) 17:55, 8 March 2019 (UTC)

My guess is that it was an alveolar trill, with possible tap (which is a trill with one contact) and approximant (which is an *undershoot* tap, so it's unlike English [ɹ] in that it's alveolar rather than postalveolar and lacks labialization) allophones. The syllabic [r̩] (transcribed with ⟨*r̥⟩ in the article) was likely a trill. Kbb2 (ex. Mr KEBAB) (talk) 11:32, 13 March 2019 (UTC)

/pibh/

We say, In PIE *ph₃ the *p regularly gives *b; for example, the reduplicated present stem of *peh₃- 'to drink' > *pi-ph₃- > Sanskrit píbati. Is it just literally this one word? I have never seen any other word used as an example. I dont think we should be calling it regular if it's only known to have happened once. —Soap— 22:53, 18 February 2020 (UTC)

Ape is said to possibly come from *h₂éph₃ō -> *h₂ébō, which also gave Celtic *abū, but it's the only other example I'm seeing right now... Does seem hard to call it a regular rule --Ngfsmg (talk) 02:36, 13 October 2020 (UTC)

Vowel chart

In this article, there is a vowel chart with four vowels, long and short, as reconstructed by the Leiden school. There follows a paragraph about how the number of vowels is disputed. I propose that we either a) add vowel charts for the other common reconstructions or b) remove the vowel chart that is currently there; as it stands, this is highly misleading for people not reading very carefully, because it would seem to favor a two-vowel reconstruction. It might even be a borderline case of violating WP:NPOV. JeanLackE (talk) 20:48, 21 July 2020 (UTC)

Bruh

Bruh where is the vowel chart and why are you guys using the weird unofficial pseudo-IPA EWW use the frekin IPA — Preceding unsigned comment added by 103.215.54.107 (talk) 00:55, 18 October 2020 (UTC)

In historical linguistics, we customarily do not solely use IPA, but rather a blend of IPA and transcription conventions. There are several vowel charts in this article—see my post above taking issue with one of them. JeanLackE (talk) 15:31, 2 November 2020 (UTC)

Superscripts and subscripts

Hey everyone, I'm updating MOS:SUPERSCRIPT to more correctly document best practices for superscripts and subscripts. In most cases (mostly science and math notations), the Manual of Style prefers HTML markup like ¹, but for the International Phonetic Alphabet we prefer Unicode characters. It looks like for Proto-Indo-European we prefer Unicode? So for example, we would want to write *h₁ and *gʷʰ, not *h₁ or *g^wh? -- Beland (talk) 16:29, 16 March 2021 (UTC)

Acute accent

What does the acute accent in words mean? I can't find where/if this is described.  Nixinova   T   C   04:55, 22 January 2022 (UTC)

@Nixinova: It indicates the location of the pitch accent. —Mahāgaja · talk 17:48, 9 December 2022 (UTC)

Syllabic Consonants

why are the syllabic consonants in the vowel section marked with the "devoiced" diacritic (r̥, l̥, etc.) and not with the "syllabic" diacritic (r̩, l̩, etc.)? 2A02:6680:110A:6DB9:D4EA:9C3F:8881:E082 (talk) 10:09, 26 June 2022 (UTC)

It is standard notation in Indo-European linguistics. Confusing, isn't it?  Tewdar  11:01, 26 June 2022 (UTC)

*a and *ā in PIE

Currently, under the "Vowels" section, the table gives a four-vowel reconstruction of PIE, with *e, *ē, *o, and *ō. The following text (third paragraph) also states the following:

Furthermore, all the daughter languages have a segment *a, and those with long vowels generally have long /aː/, /iː/, /uː/. Until the mid-20th century, PIE was reconstructed with all of those vowels. Modern versions incorporating the laryngeal theory, however, tend to view the vowels as later developments of sounds that should be reconstructed in PIE as laryngeals *h₁, *h₂, h₃. For example, what used to be reconstructed as PIE *ā is now reconstructed as *eh₂; *ī, *ū are now reconstructed as *iH, *uH, *H representing any laryngeal; and *a has various origins, among which are a "syllabic" [H̥] (any laryngeal not adjacent to a vowel) or an *e next to the "a-coloring" laryngeal *h₂e. (Though they may have phonetically contained the vowel [a] in spoken PIE, it would be an allophone of *e, not an independent phoneme.) Some researchers, however, have argued that an independent phoneme *a must be reconstructed, and it cannot be traced back to any laryngeal.

Not only is this completely unsourced, it also seems to give undue weight to a non-mainstream opinion, as far as I can tell. In the Handbook of Comparative and Historical Indo-European Linguistics, Byrd writes:

While most present-day Indo-Europeanists reconstruct *a as a phoneme for late PIE, some (most famously the "Leiden School" [LS]; see Beekes 2011) do not, eschewing typical reconstructions such as */sals/ 'salt' (Ved. salilá- 'salty', Gk. ἅλς, Lat. sal-code: lat promoted to code: la , etc.) in favor of laryngealistic reconstructions: *sh₂als (← /*sh₂els/). Thus, for the LS, *a was always a surface allophone of */e/, colored by an adjacent */h₂/. It is, however, very difficult to avoid the reconstruction of certain forms with *a vocalism. For example, Hitt. apa, Gk. ἀπό, and Lat. abcode: lat promoted to code: la 'away, off' may only be traced back to */apó/, not */h₂epó/, and it is quite difficult to derive Skt. nas-, OCS nosь 'nose' from */nh₂es-/, as one would expect the syllabification *n̥h₂es-... (pp. 2057-2058)

There are certain isolated forms which may have possessed */ī/ and */ū/: *u̯īs- 'poison' (Av. vīšcode: ave promoted to code: ae , Lat. vīruscode: lat promoted to code: la ) beside *u̯is (Ved. viṣá-, Av. viša-code: ave promoted to code: ae ), PIE *mūs 'mouse' (OE mūs) beside *mus- (Lat. musculuscode: lat promoted to code: la 'muscle'). Were one to reconstruct *u̯ihₓs- and *muhₓs, the short vowel variants could not be explained. (Such instances of long high vowels are likely due to monosyllabic lengthening; see [ω] below.) Likewise, while *nās- 'nose' (Lat. nārēscode: lat promoted to code: la 'nostrils') may be mechanically derived from */neh₂s-/, it would be difficult to connect this form with the short-vowel variant *nas- cited above. Additional instances of *ā were also derived by (η): */-eh₂m/ > *-ām (Skt. sénām 'army'). (p. 2058)

(You can access the relevant pages in Beekes 2011 on the Internet Archive here.)

Mayrhofer 1986 also writes (DeepL Translate):

The existence of */a/ in Indo-European (without the cases *eh₂e → ah₂a, § 5.2.2) is not in doubt. Admittedly, the distribution is conspicuously weaker than */e/ and apparently also restricted: "in more than half of the cases [a] appears before and after guttural," writes J. Schindler (Wurzelnomen 5, § 2.2.2), and a related view is held by T. V. Gamkrelidze; according to him, roots of the structure C₁VC₂ tend to /V/ = /a/, "if the consonant sequence in the root is accessive or extrovert, with a velar phoneme in its initial consonant." (pp. 169-170)

IE */ā/ was present in the proto-language but not frequent, similar to IE */a/ (§ 7.3.7.3). Cf. */mātér-/ 'mother' (Ved. mātár- etc.), with primary */ā/ and as a 'nursery word' not to be connected with a more or less suitable verbal root */meh₂-/. (p. 172)

Based on this, I think it's clear that we should probably fix up the "Vowels" section of the article to reflect mainstream scholarship, as well as change the vowel chart. MeasureWell (talk) 03:54, 8 December 2022 (UTC)

References