The Great Proto-Indo-European Thread

[KathTheDragon]

Hittite wehan- / wahan-. I do not know anything about Hittite derivational morphonology, but it does not seem immediately impossible to segment this as weh-an- instead of weha-n-?

I don't think any judgements about this word can be made until it can be identified. The only word I could find resembling this is a verb with stem weh-^zi~wah-, weh-^ari. Not that this cannot be taken back to a root of the supposed form, but must instead reflect *weih₂-, the active paradigm being secondary to the middle.

[KathTheDragon]

As far as I have been able to gather (and stop me if I'm wrong), in most roots only a subset of grades are attested, most often the *e-grade, *o-grade and zero grade; and some noun roots or bound morphemes show no alternation at all. And whenever the *ē-grade and *ō-grade are found, they generally occur alongside their short counterparts. That is, there are no roots that would be known only in e.g. *ē-grade versus *o-grade, or only in *ō-grade versus zero grade. The conclusion to draw from this would seem to be that ablaut is not an alternation between five entirely independent grades — instead there are two interwoven alternation patterns, one of which involves an *e : ∅ : *o alternation, and the other involves an *V : *V̄ alternation, building on the previous one; both of them morphologically conditioned rather than . (I would also find it an interesting question to investigate whether "core" ablaut could be further broken into two patterns, an *e : *o alternation and a *V : ∅ alternation.)

Alternatively, both could be taken back to an older original alternation *V : *V̄ in a most ancient prestage of PIE.

[marconatrix]

Another would be to reconstruct a distinct coloring element *A and a distinct phonation-adjusting element *h, of which either both or only one may be present:
• coloring and non-plain phonation implies *ChVA
• coloring and plain phonation implies *CVA
• no coloring and non-plain phonation implies *ChV
• no coloring and plain phonation implies *CV
Etc.

Exactly, demanding the /ha~ah/ conjunction simply generates a host of unnecessary entities. The above is much simpler. We might then reformulate VA -> AV, followed by hA -> X, giving :

CXV = Cha; CAV = Ca; CHV = Che/o; CV = Ce/o;
That would be a 3-laryngeal theory (A, H, X) that doesn't account for voicing and e/o alternation. To load those extra features on to the laryngeals you would I think need to have six of them.

I don't see that there is any obviously 'correct' way of arranging the required features across the possible segments. You can have several original vowels, or several different laryngeals or other colouring agents, or a reduced set of elements and rely as Jouna does on playing with the different sequential permutations. "Ye pays yer money ..."

[JounaPyysalo]

The vowel /a/ attached to PIE *h/ɦ is usually provable (see the example above).

No, an attested /a/ cannot "prove" an *a — unless, of course, you circularly assume that your own theory already is true and that *a is the sole source of later /a/.

- What I mean with proof in this case is the fact that usually the "a-colouring" is somewhere attested, which in turn is taken to reflect PIE *ɑ (i.e. *h2 of the mainstream theory in which PIE *ɑ is replaced with h2).

I previously gave an example from (Brittonic) showing that a 'laryngeal' could, according to circumstances, have various derivatives, including both vocalic and consonantal segments. I.e. there is no theoretical motivation for attaching an original /a/ to (almost?) every /ɦ~h/.
- The theoretical motive comes on the one hand from the issue that already the Neogrammarians proved the existence of a vocalic photo-phoneme *ə (= PIE *). On the other hand the connection between between the traditionally reconstructed PIE stops and the "a-colouring" of the roots needs to be simultaneously solved. For an example of the latter Ctrl+F the root √dɦɑn- √dɦɑon- √dɦɑen- ‘Tot’ in PIE Lexicon, representing the general ablaut class Ion. θνητό-: Gr. θάνατο-.

Having got a bit bogged down in Jouna's thesis, I wandered over to looking a bit at Uralic and found this really nice and interesting paper dealing with 'laryngeals' which as it happens gives another case of /x/ ([γ]?) > /ə/ :
http://www.sgr.fi/sust/sust253/sust253_janhunen.pdf
– Yes, I know Janhunen's paper – and Prof. Janhunen – very well as Juha happens to be the Principal Investigator of the PIE Lexicon project:
http://pielexicon.hum.helsinki.fi/about.html
I and Juha very much agree on a single laryngeal in PIE as do all partners of PIE Lexicon project. The mono laryngealism has always been there, not only in the research history (Zgusta, Szemerényi, Burrow, Laroche, Tischler, etc.) but also among the practicing researchers. What has been missing is an inductive theory of PIE "laryngeal" and the "vocalism" which has been now published – and soon digitaised.

This btw is my idea of 'useful speculation'.

[jal]

Jouna, could you please stop quoting whole swaths of text (mostly quoting entire posts verbatim), sometimes posting without even replying? Thanks.


JAL

[KathTheDragon]

Actually, Jouna's responses are embedded inside the quotes headed by "-"s.

[Cedh]

Jouna, could you please stop quoting whole swaths of text (mostly quoting entire posts verbatim), sometimes posting without even replying? Thanks.

As Kath has remarked, Jouna usually answers within the quoted posts, which is fairly hard to see. I've been marking Jouna's words in italics there for more clarity; however, I only read this thread about once a day so you got to see the posts before my edit. But you're right, proper quote tags would be better. Jouna, could you please make an effort to use them?

[JounaPyysalo]

offer a couple of more examples of ablaut PIE *ē : e : Ø : o : ō within a single root in §2.5.2 of my dissertation (note that the examples are independent of the "laryngeals" thus proving the existence of the alternation. It is not difficult not find more examples, one of these contained in the data of the root PIE √tekw- ‘run, flow, etc.’ This is not a "hypothetical construct" but the ablaut pattern you also meet as the bases of the root PIE *dēɑɦ- : deɑɦ- : dɑɦ- : doɑɦ- : dōɑɦ- “Gene, schenken”.

I am not talking about the fact that some words show multiple ablaut grades; but the assumption that therefore any other words could occur in any ablaut grade as well, even when no particular facet of alternation is directly attested (i.e. found within a single language).

"As far as I have been able to gather (and stop me if I'm wrong), in most roots only a subset of grades are attested, most often the *e-grade, *o-grade and zero grade; and some noun roots or bound morphemes show no alternation at all. And whenever the *ē-grade and *ō-grade are found, they generally occur alongside their short counterparts. That is, there are no roots that would be known only in e.g. *ē-grade versus *o-grade, or only in *ō-grade versus zero grade."
- This is not accurate: Roots can be preserved with any of the PIE *ē : e : Ø : o : ō (or their combinations) depending on which grades were originally formed – and which were preserved all the way to us.

The conclusion to draw from this would seem to be that ablaut is not an alternation between five entirely independent grades — instead there are two interwoven alternation patterns, one of which involves an *e : ∅ : *o alternation, and the other involves an *V : *V̄ alternation, building on the previous one; both of them morphologically conditioned rather than . (I would also find it an interesting question to investigate whether "core" ablaut could be further broken into two patterns, an *e : *o alternation and a *V : ∅ alternation.)
- This is admittedly one way of understanding the matter, except that there are no consistent rules that could generate, say, PIE *ē from PIE *e, i.e. the entire set PIE *ē : e : Ø : o : ō is original and its oppositions distinctive also in the Pre-PIE assuming that it would make any sense to speak of such an entity.

This is actually an entirely general question I have about the reconstruction of PIE, although Jouna's system seems to move to the exact opposite direction from what seems to me like the sensible direction — i.e. he's assuming ablaut to be a more general system than has been thought, rather than trying to find the specifics of how and when it applies.
— Regarding the "sensible direction" I would like to put the issue to its proper context in the research history of the entire study.
0. The study was initially set forth by Sir William Jones who in his famous lecture in 1786 announced the genetic relationship between several IE languages, hence setting forth the pure paradigm without any attachments to the schools of thought (or mainstreams) to be discussed next:
1. The Paleogrammarians, the first mainstream of the IE linguistics, approximately from 1800 to 1870 – adopted Sanskrito-centrict typology as their paradigm of the phoneme inventory of the protolanguage. With regard to the vowel system this meant monovocalism with two quantities for PIE, viz. Paleogr. *a (= Skt. a) and Paleogr. *ā.
2. This paradigm was abandoned by the Neogrammarians who insisted that the qualities and quantities attested in the "European" languages, i.e. Neogr. *a *e *o *ā *ē *ō were actually original (and hence the Paleogrammarian paradigm was wrong). With the addition of two other vowels required by the correspondence sets, viz. Neogr. *å and *ǝ they ended up with a system of eight distinct vowels for the photo-language, widely used between 1870-1940.
3. The Neogrammarian period lasted some 70 years until the Paleogrammarian "monovocalistic school re-entered the study, this time in the form of Møller's Indo-Semitic theory in which the six vowels Neogr. *a *e *o *ā *ē *ō were again interpreted on the basis of a single proto-vowel and the colouring and compensatory lengthening rules in a well-known manner: LT *h2e h1e h3e eh2 eh1 eh3, a tendency to dominate the field between 1940-2010.
4. In parallel to this development the Neogrammarian school re-organized itself as monolaryngealism, first proposed by Zgusta (1951) with support from prominent Anatolian linguists such as Laroche and Tischler and from Indo-Europeanists such as Szemerényi and Burrow. Yet again in this theory the PIE vowels are understood as original in the sense that no attempt is made to reduce them through a Semito-centric paradigm (with fundamental vowel /ä/) is made. Although the formulation of this theory was initially a simple outline, approximately containing PIE *H (for Hittite h) and the Neogrammarian vowels system, I've detailed the theory in an explicit manner in System PIE with the result that I foresee the tradition once again seizing the initiative in Indo-European linguistics.

Generally in this axis there are thus two opposite trends going beyond the historical mainstream movements, viz.
1+3. The Paleogrammarian+Laryngeal theory. Typical for this trend is to pick an typological model aimed at reducing the vowel system by means of deduction. Between 1800-1870 this goal was advanced by the “Sanskrit-as-the-mother-of-all-languages” paradigm (with *a and *ā) and when Møller's Indo-Semitic theory made its breakthrough through Kurylowicz and Benveniste (camouflaging it as an Indo-European theory) through LT *h2e h1e h3e eh2 eh1 eh3.
2+4. The Neogrammarian+Monolaryngealism. Typical for this trend is to avoid external typology (or paradigm) and define the Indo-European vowels by means of inductive comparison of the Indo-European languages. Between 1870-1940 a system of eight phonemes for vowels Neogr. *a *e *o *ā *ē *ō *å *ǝ was established and despite the domination of the laryngeal theory the correct solution to the laryngeal problem was established: Hitt. h = PIE h.

In this context what is "sensible" depends on our angle of approach towards the science in a very general level. From the point of view of the "2+4"-tradition the "sensible" approach of the "1+3"-tradition is hardly nothing else than deductive forcing of the data to an (unfounded) typology, a procrustean bed (to quote a commonly used metaphor) with fruits resembling the products of the Amazonian headshrinkers http://en.wikipedia.org/wiki/Shrunken_head

Any idea of an common (P)IE loss of unstressed vowels is impossible, resulting immediately in inconsistency (I've tried this and it was a catastrophe).

Perhaps. Such a development would probably have to be pre-PIE, in which case new loanword/affective vocabulary acquired after the initial introduction of zero grade alternations would have to be somehow first sieved apart in order to be able to determine the original conditioning.
– It makes little sense to distinguish between PIE and Pre-PIE, because everything that happened in "Pre-PIE" would have affected PIE turning it into quite different entity than it is on the basis of the attested languages.

[JounaPyysalo]

the original phonemes and features of PIE are exclusively decided on the basis of the data.

And regardless other people have ended with different reconstructions from you, which ought to demonstrate that multiple decisions are possible.

in the actual reconstruction we are capable of determining the original nature of the phonemes taken that the data is sufficient. This is quite hypothetical and distant from the process of reconstruction.

The data is generally not sufficient, that's exactly my point.

Please show me how that would actually generate the full data PIE √uehɑ- √uohɑ- √uhɑ- (vb.) ‘wenden, drehen, bewegen’ – and I will tell you whether that is possible or not.

This seems like it would remain *weha-/*woha-/*uha- in all three initial possible adjustments I suggested (with non-coloring *h₁ if we go along that route), although apparently you posit stem-final *a solely on the grounds of Hittite wehan- / wahan-. I do not know anything about Hittite derivational morphonology, but it does not seem immediately impossible to segment this as weh-an- instead of weha-n-?
— No, the stem final *·ɑ- is posited due to the ‘a-vocalism’ in Greek, implying that PIE *ɑ had to be present (it being the sole cause of ‘a-vocalism’). Since the vowel *ɑ could not precede PIE *h due to the absence of ‘a-colouring’ in Hitt. ueh- = Umbr. ue- it had to follow PIE *h, thus implying PIE *uehɑ-. Hitt. u̯eḫan- is taken from PIE *u̯ehɑen- but also PIE *u̯ehɑon- would be possible Trad *a and *o colliding in Hitt. a.

I'm however somewhat concerned that you only cite data for this root from Hittite, Greek and Umbrian, which sounds like that there is a risk that rather than common PIE inheritance, it is a loanword from Anatolian to Greek (and perhaps thence to Umbrian, or perhaps that word is not related at all). It's easy to argue for whatever reconstruction at all if you're allowed to hand-pick the data to examine.
- The comparative method of reconstruction follows the principle of postulation ("Durch zweier Zeugen Mund wird alle Wahrheit kund" — August Fick) and there is nothing wrong in the etymology of Goetze and Pedersen bringing the Greek forms here (for ‘a-vocalism’ cf. also Lat. uannus).
Also note that the data of PIE Lexicon is not hand-picked but genuinely represents much of the entire PIE data. In our "Summer Challenge" we have committed to reconstruct any Indo-European correspondence that we are asked to by anyone.

the set PIE *hɑ : ɑh (voiceless) and PIE *ɦɑ : ɑɦ (voiced) suffices to the generation of the Indo-European data.

I do not doubt that they suffice to generate the data, but what I do not believe is that that makes it the best possible solution.
— You can always point me an alternative and we check it out.

Your reconstruction in fact leaves completely unaddressed the known connection between e.g. a-coloring and laryngeals, since you trace both phenomena back into two different segments; but do not address at all the question of why would they be forced to so often co-occur! Your "diphoneme" seems like an unnatural construct that has no parallels in natural languages. And iff PIE had truly independent phonemes *a and *h, we'd expect sequences like *tak or *ehe to be also found.

Your *ha/*ah also seems to be often an unwarranted postulation. E.g. you first claim a connection between *h and voicing, or *a and a-coloring, and then only use your assumption about the two of them always co-occurring to suppose that any root with a voiced consonant should also have an *a next to the *h, or any root with an a-colored reflex should also have an *h next to the *a. I fail to see what the benefit of this is. I would expect single cases of a-coloring could only be used to reconstruct your *a, and positing the presence of *h should always be argued separately.
— There are several points related, all of which, however are more or less "structural" and therefore not necessarily preserving the truth so be cautious with this:
(a) Ultimately the whole idea of the laryngeal theory, viz. explaining the vowel qualities with laryngeals is totally unrealistic and phonetically unfounded. Especially the vowel qualities relevant here, e a o are produced in mouth cavity, not in the throat (or "larynx" in the broad sense). That this explanation become possible in the first place happened only because Möller needed a lot of "laryngeals" to PIE in order to match the real entities of the Semitic with something – to which his only option was to re-write the Indo-European *e a o (vowels) as *E A O (consonants).
In order to fix this error PIE Lexicon restores the vowels PIE *e o ɑ preferring no longer to write these as "h1 h3 h2".
(b) The idea of the co-occurence of PIE *ɑ and PIE *h is hardly nothing but a way to explain the "colouring effect" erroneously attached to the "laryngeal" in the laryngeal theory. Although the proponents of LT are often ignorant of the history of the field, it should be underlined that Neogr. *ə is a comparatively well-defined phoneme and a vowel. The a mistake Kurylowicz made in equating that vowel (= PIE *ɑ) with a consonant Hitt. h (in 1927), is now corrected in OS PIE Lexicon. The issue is that whenever the Old Anatolian suggests a laryngeal (Hitt. h) the rest of the group seems to respond with ‘a-vocalism’ at least in some forms – and vice versa. Therefore both two items are reconstructed.
(c) Finally the general structural features of PIE, in particular the consequences of the existence of the zero grade in the schema PIE *ē : e : Ø : o : ō should be taken into account: Although this is structural argumentation, it is clear that had PIE *h not been accompanied with the vowel PIE *ɑ this would have yielded a huge number of unpronounceable root shapes in hCC ChC CCh.

This moreover seems to mean that e.g. your idea about a connection between *h and voiced stops also becomes dubious. Say, on page 373 you re-reconstruct *deḱm̥ '10' as *deɦaḱeah-, on the grounds of an a-colored reflex in Armenian. But this does not need to imply anything about a laryngeal being also present.
– In comparative method TWO WITNESSES are required for postulation. The second witness in this case are all the languages belonging to the correspondence, each one of them pointing to PIE *d, which in turn implies PIE *ɦ—d or d—ɦ. Since there is no ‘a-vocalism’ in the Greek initial the only place where the "laryngeal" can be posited is *de(ɦ)ɑḱeɑh-.

[JounaPyysalo]

Jouna, could you please stop quoting whole swaths of text (mostly quoting entire posts verbatim), sometimes posting without even replying? Thanks.

As Kath has remarked, Jouna usually answers within the quoted posts, which is fairly hard to see. I've been marking Jouna's words in italics there for more clarity; however, I only read this thread about once a day so you got to see the posts before my edit. But you're right, proper quote tags would be better. Jouna, could you please make an effort to use them?

Thank you Cedh, I'm in a hurry so I've used a short cut, but I'll try to get used to that. (Note that I try to begin the reply with "—" to indicate that my response is to follow.) You may have to try to last this today as I've got a deadline, but I'll learn to use the tags as soon as possible.

Thank you very much for the reminder, it's certainly best for all to use the common procedure.

Jouna

[KathTheDragon]

For a quick fix, Jouna, just preceed your responses with the closing tag [/ quote] and follow it with the opening tag [ quote] (obviously without those spaces I put in to prevent them being parsed)

[JounaPyysalo]

Hittite wehan- / wahan-. I do not know anything about Hittite derivational morphonology, but it does not seem immediately impossible to segment this as weh-an- instead of weha-n-?

I don't think any judgements about this word can be made until it can be identified. The only word I could find resembling this is a verb with stem weh-^zi~wah-, weh-^ari. Not that this cannot be taken back to a root of the supposed form, but must instead reflect *weih₂-, the active paradigm being secondary to the middle.

I know that the mainstream LT attempts to posit weih2, but this makes no sense in the face of the ablaut alternative Hitt. uah- (from PIE *woha-) incompatible with the assumed PIE *i in the root.

If having problems with the data why not follow the links to PIE Lexicon from which you'll find references to the dictionaries from which the forms are taken? The full entry for Hittite in PIE Lexicon answers to your questions providing all this information:
PIE *u̯ehɑ- Hitt. u̯eḫ-(1A.) ‘sich wenden, drehen, (hin und her)bewegen’)(HHand. 200)(Hitt. ú-e-eḫ-zi [3sg.], Kimball 1999:210, HIL. 1149f)

Jouna

[JounaPyysalo]

For a quick fix, Jouna, just preceed your responses with the closing tag [/ quote] and follow it with the opening tag [ quote] (obviously without those spaces I put in to prevent them being parsed)

[/ quote] Thank you, KathAveara, I'll try that immediately now, let's see whether I got it right (it sound funny to begin with a closing tag and end it with the opening tag, but I'll try anyway [ quote]

[Tropylium]

1. The Paleogrammarians, the first mainstream of the IE linguistics, approximately from 1800 to 1870 – adopted Sanskrito-centrict typology as their paradigm of the phoneme inventory of the protolanguage. With regard to the vowel system this meant monovocalism with two quantities for PIE, viz. Paleogr. *a (= Skt. a) and Paleogr. *ā.
2. This paradigm was abandoned by the Neogrammarians who insisted that the qualities and quantities attested in the "European" languages, i.e. Neogr. *a *e *o *ā *ē *ō were actually original (and hence the Paleogrammarian paradigm was wrong). With the addition of two other vowels required by the correspondence sets, viz. Neogr. *å and *ǝ they ended up with a system of eight distinct vowels for the photo-language, widely used between 1870-1940.
3. The Neogrammarian period lasted some 70 years until the Paleogrammarian "monovocalistic school re-entered the study, this time in the form of Møller's Indo-Semitic theory in which the six vowels Neogr. *a *e *o *ā *ē *ō were again interpreted on the basis of a single proto-vowel and the colouring and compensatory lengthening rules in a well-known manner: LT *h2e h1e h3e eh2 eh1 eh3, a tendency to dominate the field between 1940-2010.
4. In parallel to this development the Neogrammarian school re-organized itself as monolaryngealism, first proposed by Zgusta (1951) with support from prominent Anatolian linguists such as Laroche and Tischler and from Indo-Europeanists such as Szemerényi and Burrow. Yet again in this theory the PIE vowels are understood as original in the sense that no attempt is made to reduce them through a Semito-centric paradigm (with fundamental vowel /ä/) is made. Although the formulation of this theory was initially a simple outline, approximately containing PIE *H (for Hittite h) and the Neogrammarian vowels system, I've detailed the theory in an explicit manner in System PIE with the result that I foresee the tradition once again seizing the initiative in Indo-European linguistics.

Generally in this axis there are thus two opposite trends going beyond the historical mainstream movements, viz.
1+3. The Paleogrammarian+Laryngeal theory. Typical for this trend is to pick an typological model aimed at reducing the vowel system by means of deduction. Between 1800-1870 this goal was advanced by the “Sanskrit-as-the-mother-of-all-languages” paradigm (with *a and *ā) and when Møller's Indo-Semitic theory made its breakthrough through Kurylowicz and Benveniste (camouflaging it as an Indo-European theory) through LT *h2e h1e h3e eh2 eh1 eh3.
2+4. The Neogrammarian+Monolaryngealism. Typical for this trend is to avoid external typology (or paradigm) and define the Indo-European vowels by means of inductive comparison of the Indo-European languages. Between 1870-1940 a system of eight phonemes for vowels Neogr. *a *e *o *ā *ē *ō *å *ǝ was established and despite the domination of the laryngeal theory the correct solution to the laryngeal problem was established: Hitt. h = PIE h.

This is not the whole picture. The modern trilaryngealism I'm familiar with admits separate *o and according to some researchers even *a (which I am sympathetic to). Only those instances where a laryngeal is either demonstrable by Anatolian cognates, or those where non-standard ablaut occurs, should be taken back to *e plus laryngeal coloring.

The reason for this is that as far as I can tell, laryngeal coloring cannot be comparatively reconstructed. Neogrammarian *a is valid, at least as an allophone, for PIE proper. The laryngealistic reanalysis as *e colored by *h₂ is a pre-PIE development, though one that can be internally reconstructed. Any remaining cases however could be explained thru any of the following:
• loanwords
• onomatopoeia and other coinages
• the existence of a separate *a already before laryngeal coloring
• other conditional sound changes generating *a

One thing though that seems to muddle the waters a lot is that too often laryngeal coloring is rendered as a simplistic algebraic correspondence "*h₂e > *a", showing only the endpoints: Neogrammarian non-laryngealist late PIE and laryngealist pre-PIE. To me it however appears to be the case that there are a great number of processes here that need to be distinguished, e.g.

1) Pre-IE: laryngeal coloring, e.g. h₂e > h₂a, eh₂ > ah₂
2) Pre-IE?: coda laryngeal merger, Vhₓ > VH / _C _# (where probably H = [ʔ])
3) Possibly pre-IE: loss of word-initial *h₁ (I am aware of one or two loanwords in Uralic that would argue against this, but I do not think we can conclusively rule out that they could be instead from a para-IE language)
4) Non-Anatolian PIE: reduction of remaining *h₂ to /h/
5) einzelsprachlig: loss of word-initial *h₂ (this could normally be considered Non-Anatolian PIE, but there are the known Albanian/Armenian cases and the recently mentioned in this thread Western Iranian cases of possible retention)
6) einzelsprachlig: compensatory lengthening, VH > Vː (postdates e.g. Winter's Law in Balto-Slavic, Lachmann's Law in Latin)
7) einzelsprachlig: loss of postconsonantal H, CHV > CV (postdates rise of voiceless aspirates in Indo-Iranian and partly elsewhere; I do not know if the "no-aspirate" branches i.e. Celtic, Italic, Germanic, Baltic, Albanian offer any evidence for the dating of this)

This is just a quick outline. I'm not sure where exactly e.g. the developments of *h₃ or the loss of intervocalic laryngeals would fit in. "Laryngeal vocalization" would also have to be included, and broken in parts. Neogrammarian *ə seems to be indeed reconstructible already for PIE proper, though according to current analyses it's these days considered an epenthetic vowel, not a separate phoneme.

---

Similar issues as in mainstream laryngeal theory would appear to be true in Jouna's system as well. In particular, the presumed contractions *ea > *a, *ae > *a, *ēa > *ā, *aē > *ā, and the loss *a > ∅ in most positions appear to be reflected in every single IE language, and hence they should probably be reconstructed for PIE proper.

So this in mind, I consider it quite possible that only a single laryngeal occurred in PIE proper. The monolaryngealist vs. trilaryngealist question seems to boil down to whether it should be considered to reflect the merger of several pre-PIE laryngeals, or not. Here though the selection of data becomes of great importance. All roots with a narrow distribution should be excluded from initial analysis; so should all roots with plausible loanword etymologies (e.g. *tawros 'bull' from Semitic), since these risk being innovations that are newer than the establishment of the "tainted" ablaut system.

It makes little sense to distinguish between PIE and Pre-PIE, because everything that happened in "Pre-PIE" would have affected PIE turning it into quite different entity than it is on the basis of the attested languages.

No offense, but this is nonsense. "Everything that happened in pre-PIE" is exactly those developments that has made PIE into what it is, not into what it isn't.

All natural languages have ancestors, from which they have evolved. If PIE was a natural language, it too had ancestors, and it too had undergone earlier development. The only reasonable way to deny this is to deny that PIE was a natural language (e.g. to treat it as some kind of an artificial construct created by [insert crackpot historical conspiracy theory here].)

[JounaPyysalo]

1. The Paleogrammarians, the first mainstream of the IE linguistics, approximately from 1800 to 1870 – adopted Sanskrito-centrict typology as their paradigm of the phoneme inventory of the protolanguage. With regard to the vowel system this meant monovocalism with two quantities for PIE, viz. Paleogr. *a (= Skt. a) and Paleogr. *ā.
2. This paradigm was abandoned by the Neogrammarians who insisted that the qualities and quantities attested in the "European" languages, i.e. Neogr. *a *e *o *ā *ē *ō were actually original (and hence the Paleogrammarian paradigm was wrong). With the addition of two other vowels required by the correspondence sets, viz. Neogr. *å and *ǝ they ended up with a system of eight distinct vowels for the photo-language, widely used between 1870-1940.
3. The Neogrammarian period lasted some 70 years until the Paleogrammarian "monovocalistic school re-entered the study, this time in the form of Møller's Indo-Semitic theory in which the six vowels Neogr. *a *e *o *ā *ē *ō were again interpreted on the basis of a single proto-vowel and the colouring and compensatory lengthening rules in a well-known manner: LT *h2e h1e h3e eh2 eh1 eh3, a tendency to dominate the field between 1940-2010.
4. In parallel to this development the Neogrammarian school re-organized itself as monolaryngealism, first proposed by Zgusta (1951) with support from prominent Anatolian linguists such as Laroche and Tischler and from Indo-Europeanists such as Szemerényi and Burrow. Yet again in this theory the PIE vowels are understood as original in the sense that no attempt is made to reduce them through a Semito-centric paradigm (with fundamental vowel /ä/) is made. Although the formulation of this theory was initially a simple outline, approximately containing PIE *H (for Hittite h) and the Neogrammarian vowels system, I've detailed the theory in an explicit manner in System PIE with the result that I foresee the tradition once again seizing the initiative in Indo-European linguistics.

Generally in this axis there are thus two opposite trends going beyond the historical mainstream movements, viz.
1+3. The Paleogrammarian+Laryngeal theory. Typical for this trend is to pick an typological model aimed at reducing the vowel system by means of deduction. Between 1800-1870 this goal was advanced by the “Sanskrit-as-the-mother-of-all-languages” paradigm (with *a and *ā) and when Møller's Indo-Semitic theory made its breakthrough through Kurylowicz and Benveniste (camouflaging it as an Indo-European theory) through LT *h2e h1e h3e eh2 eh1 eh3.
2+4. The Neogrammarian+Monolaryngealism. Typical for this trend is to avoid external typology (or paradigm) and define the Indo-European vowels by means of inductive comparison of the Indo-European languages. Between 1870-1940 a system of eight phonemes for vowels Neogr. *a *e *o *ā *ē *ō *å *ǝ was established and despite the domination of the laryngeal theory the correct solution to the laryngeal problem was established: Hitt. h = PIE h.

This is not the whole picture. The modern trilaryngealism I'm familiar with admits separate *o and according to some researchers even *a (which I am sympathetic to).

JOUNA: Yes, the "modern" trilaryngealism uses both the Neogrammarian vowels and the three laryngeals, thus being the weakest and the most incoherent of all theories in existence.

Only those instances where a laryngeal is either demonstrable by Anatolian cognates, or those where non-standard ablaut occurs, should be taken back to *e plus laryngeal coloring.

JOUNA: In that case welcome to the monolaryngealism: the terms "standard" and "non-standard" mean little these days as the broader Neogrammarian vowel system has been degraded into the monovocalism of the LT

The reason for this is that as far as I can tell, laryngeal coloring cannot be comparatively reconstructed.

JOUNA: The issue is not that "laryngeal colouring" could not be comparatively reconstructed. The issue is that it does not exist phonetically. Vowel qualities are produced in the mouth cavity, not in the throat.

Neogrammarian *a is valid, at least as an allophone, for PIE proper.

JOUNA: The cover symbol Neogr. *a3 = Neogr. a has been correctly postulated by the Neogrammarians for the correspondence type Lat. a = Skt. a = Lith. a = Arm. a = Gr. a etc. It was also correctly noted by Møller that Ae [= PIE *ɑe in PIE *Hɑe] yields Neogr. a. In System PIE the overstated compensatory lengthening rule of the LT has been abandoned with the result that the second half of the cover symbol Neogr. *a = PIE eɑH. In this manner no independent /a/ is ever required in the reconstruction of PIE. It is sufficient to postulate PIE *ɑ (i.e. Schwa indogermanicum).

The laryngealistic reanalysis as *e colored by *h₂ is a pre-PIE development, though one that can be internally reconstructed. Any remaining cases however could be explained thru any of the following:
• loanwords
• onomatopoeia and other coinages
• the existence of a separate *a already before laryngeal coloring
• other conditional sound changes generating *a

JOUNA: The colouring of PIE *ɑe, *eɑ -> IE a is by no means Pre-PIE development. Quite the contrary we've already coded them as the first post-PIE changes in PIE Lexicon, look these yourself, say for Hittite, in http://pielexicon.hum.helsinki.fi/?showrule=21

One thing though that seems to muddle the waters a lot is that too often laryngeal coloring is rendered as a simplistic algebraic correspondence "*h₂e > *a", showing only the endpoints: Neogrammarian non-laryngealist late PIE and laryngealist pre-PIE. To me it however appears to be the case that there are a great number of processes here that need to be distinguished, e.g.

1) Pre-IE: laryngeal coloring, e.g. h₂e > h₂a, eh₂ > ah₂
2) Pre-IE?: coda laryngeal merger, Vhₓ > VH / _C _# (where probably H = [ʔ])
3) Possibly pre-IE: loss of word-initial *h₁ (I am aware of one or two loanwords in Uralic that would argue against this, but I do not think we can conclusively rule out that they could be instead from a para-IE language)
4) Non-Anatolian PIE: reduction of remaining *h₂ to /h/
5) einzelsprachlig: loss of word-initial *h₂ (this could normally be considered Non-Anatolian PIE, but there are the known Albanian/Armenian cases and the recently mentioned in this thread Western Iranian cases of possible retention)
6) einzelsprachlig: compensatory lengthening, VH > Vː (postdates e.g. Winter's Law in Balto-Slavic, Lachmann's Law in Latin)
7) einzelsprachlig: loss of postconsonantal H, CHV > CV (postdates rise of voiceless aspirates in Indo-Iranian and partly elsewhere; I do not know if the "no-aspirate" branches i.e. Celtic, Italic, Germanic, Baltic, Albanian offer any evidence for the dating of this)

This is just a quick outline.

I'm not sure where exactly e.g. the developments of *h₃ or the loss of intervocalic laryngeals would fit in.

JOUNA: LT *h₃ is a modern way of writing PIE *o, hence the developments of *h₃ coincide with PIE *o

"Laryngeal vocalization" would also have to be included, and broken in parts.

JOUNA: Of course consonants (obstruents/fricatives) are not vowels and therefore have no vocalisations.

Neogrammarian *ə seems to be indeed reconstructible already for PIE proper, though according to current analyses it's these days considered an epenthetic vowel, not a separate phoneme.

JOUNA : Yes. The correspondence set of the neogrammarians defining the sound as Lat. a (Gr. a, etc.) : OInd. i is provable especially in cases which the absence of 2nd palatalization in Sanskrit implies an /i/ of non-palatal origin.

---

Similar issues as in mainstream laryngeal theory would appear to be true in Jouna's system as well. In particular, the presumed contractions *ea > *a, *ae > *a, *ēa > *ā, *aē > *ā, and the loss *a > ∅ in most positions appear to be reflected in every single IE language, and hence they should probably be reconstructed for PIE proper.

So this in mind, I consider it quite possible that only a single laryngeal occurred in PIE proper. The monolaryngealist vs. trilaryngealist question seems to boil down to whether it should be considered to reflect the merger of several pre-PIE laryngeals, or not. Here though the selection of data becomes of great importance. All roots with a narrow distribution should be excluded from initial analysis; so should all roots with plausible loanword etymologies (e.g. *tawros 'bull' from Semitic), since these risk being innovations that are newer than the establishment of the "tainted" ablaut system.

It makes little sense to distinguish between PIE and Pre-PIE, because everything that happened in "Pre-PIE" would have affected PIE turning it into quite different entity than it is on the basis of the attested languages.

No offense, but this is nonsense. "Everything that happened in pre-PIE" is exactly those developments that has made PIE into what it is, not into what it isn't.

All natural languages have ancestors, from which they have evolved. If PIE was a natural language, it too had ancestors, and it too had undergone earlier development. The only reasonable way to deny this is to deny that PIE was a natural language (e.g. to treat it as some kind of an artificial construct created by [insert crackpot historical conspiracy theory here].)

Jouna: I tried to use the tags, but just in case added also my name, let's see whether my commenting has improved after this or not,

[Tropylium]

I'm however somewhat concerned that you only cite data for this root from Hittite, Greek and Umbrian, which sounds like that there is a risk that rather than common PIE inheritance, it is a loanword from Anatolian to Greek (and perhaps thence to Umbrian, or perhaps that word is not related at all). It's easy to argue for whatever reconstruction at all if you're allowed to hand-pick the data to examine.

The comparative method of reconstruction follows the principle of postulation ("Durch zweier Zeugen Mund wird alle Wahrheit kund" — August Fick) and there is nothing wrong in the etymology of Goetze and Pedersen bringing the Greek forms here

No, this does not hold. An etymology does not need to have anything formally wrong in it to be regardless unreliable. I believe the classic example is that "gun" and "whiskey" are formally reconstructible for Proto-Plains Algonquian. Also e.g. "bullet" is formally reconstructible for Proto-Samic, "book" is formally reconstructible for Proto-Finnic, and "radar" is formally reconstructible for Old Finnish — all regardless flagrant violations of known history.

"Explaining all the data as inherited" strikes me as a sign of a poor theory, not the sign of a good one. Not all etymological comparisons that have been made are a priori correct, especially ones that have been proposed before the formulation of a stable, generally-agreed-on system of reconstruction.

(a) Ultimately the whole idea of the laryngeal theory, viz. explaining the vowel qualities with laryngeals is totally unrealistic and phonetically unfounded. Especially the vowel qualities relevant here, e a o are produced in mouth cavity, not in the throat (or "larynx" in the broad sense).

This is false. Vowel coloring by laryngeal consonants is well attested in languages such as Arabic (where /a/ has a back allophone [ɑ] next to emphatic consonants and pharyngeals, and elsewhere a front allophone that I've seen varyingly described as [ɛ] or [ɜ] or [æ]) or Northern Haida (where /a/ has a back allophone [ʌ] next to uvulars and pharyngeals, elsewhere a front allophone [ə]).

The phonetical founding is straightforward too. Pharyngealization or uvularization imparts on vowels a similar acoustic coloring as retraction and lowering, and the latter is easier to pronounce; therefore we expect a laryngeal-colored allophone such as [əˤ] to soon be replaced by the corresponding backed and lowered vowel [ʌ]. Not all sound changes need to have an articulatory basis.

(b) The idea of the co-occurence of PIE *ɑ and PIE *h is hardly nothing but a way to explain the "colouring effect" erroneously attached to the "laryngeal" in the laryngeal theory. Although the proponents of LT are often ignorant of the history of the field, it should be underlined that Neogr. *ə is a comparatively well-defined phoneme and a vowel. The a mistake Kurylowicz made in equating that vowel (= PIE *ɑ) with a consonant Hitt. h (in 1927), is now corrected in OS PIE Lexicon. The issue is that whenever the Old Anatolian suggests a laryngeal (Hitt. h) the rest of the group seems to respond with ‘a-vocalism’ at least in some forms – and vice versa. Therefore both two items are reconstructed.

I follow this part of the logic well enough, but I think you go off the rails when you begin postulating laryngeals and after that their attached *a's on the basis of non-direct reflexes such as stop voicing or aspiration or what you call Fortunatov's Law II. Having to posit, within your framework, laryngeals and *a co-occurring in some words does not imply that they would need to co-occur always.

(c) Finally the general structural features of PIE, in particular the consequences of the existence of the zero grade in the schema PIE *ē : e : Ø : o : ō should be taken into account: Although this is structural argumentation, it is clear that had PIE *h not been accompanied with the vowel PIE *ɑ this would have yielded a huge number of unpronounceable root shapes in hCC ChC CCh.

This is internal reconstruction, not a comparative one; and the problem can be easily circumvented by positing epenthetic *ə in awkward zero-grade forms.

In comparative method TWO WITNESSES are required for postulation.

You keep repeating this quote, but I don't think it is valid as a general principle. It is merely a rule of thumb that in some cases follows as a corollary of Occam's Razor. But it cannot be applied generally, since it is possible for two languages have undergone a similar innovation.

Otherwise we'd run into conflicing issues such as:
• Armenian and Germanic provide two witnesses that the traditional series of mediae should be reconstructed as voiceless;
• but, meanwhile, Greek and Indic provide two witnesses that the series should be reconstructed as plain voiced;
• and Celtic and Balto-Slavic provide two witnesses that we should not reconstruct any distinction between the mediae and the voiced aspirates in the first place.
Hence when it comes to e.g. setting up proto-language phonology, two witnesses appears to be sometimes an insufficient criterion.

The second witness in this case are all the languages belonging to the correspondence, each one of them pointing to PIE *d, which in turn implies PIE *ɦ—d or d—ɦ.

Circular logic: the part of your thesis I was quoting here is exactly the point where you attempt to argue in the first place that PIE *d implies a laryngeal at all.

Your method in this and various other places appears to be:
1) list some examples where a laryngeal and phonetic feature X co-occur
2) claim that this allows you to infer a laryngeal from every single instance of X
3) reconstruct "invisible" laryngeals that disappear from the data without a trace, save for triggering feature X.
Which only creates an unfalsifiable non-theory.

[Tropylium]

Yes, the "modern" trilaryngealism uses both the Neogrammarian vowels and the three laryngeals, thus being the weakest and the most incoherent of all theories in existence.

This sounds like you are optimizing the wrong thing. The measure of a reconstruction's strength is not merely the number of proto-phonemes that it assumes, but the sum total of assumptions it makes altogether. If the elimination of a phoneme turns the proto-language into something typologically unlikely, this is not progress; if the elimination of a phoneme requires assuming new sound changes than then recreate this phoneme, this is not progress; if the elimination of a phoneme requires assuming typologically unlikely sound changes, this is also not progress.

The colouring of PIE *ɑe, *eɑ -> IE a is by no means Pre-PIE development. Quite the contrary we've already coded them as the first post-PIE changes in PIE Lexicon, look these yourself, say for Hittite, in http://pielexicon.hum.helsinki.fi/?showrule=21

PIE is by definition the last common ancestor of the Indo-European languages, not the "most regular" common ancestor stage. If a rule exists in identical shape in every descendant, and is not preceded by any developments that only apply to some branches, it is necessary to consider it to have already taken place in PIE — even if the rule introduces morphophonological complications.

Of course consonants (obstruents/fricatives) are not vowels and therefore have no vocalisations.

This critique seems to be quite common, but I think it confuses synchronic and diachronic sound changes. Obstruent vocalization is well attested, usually with an intermediate approximant stage (e.g. Latin noctem > Old French noit /nojt/ > French nuit /nɥi/). As I see it, the sensible analysis is that a "zero-grade laryngeal" develops into a vowel in late PIE, not that it immediately alternated with a "vocalic laryngeal".

(Although, take a look at Berber or Salishan languages or the Ōgami language of the Ryukyuan family sometimes: they do some quite wild stuff with phonotactics, and "syllabic fricatives" could be quite well posited for them.)

This argument also does nothing to rule out epenthetic vowels. Have you followed recent work by e.g. Andrew Miles Byrd on this matter?

[JounaPyysalo]

Yes, the "modern" trilaryngealism uses both the Neogrammarian vowels and the three laryngeals, thus being the weakest and the most incoherent of all theories in existence.

This sounds like you are optimizing the wrong thing. The measure of a reconstruction's strength is not merely the number of proto-phonemes that it assumes, but the sum total of assumptions it makes altogether. If the elimination of a phoneme turns the proto-language into something typologically unlikely, this is not progress;

JOUNA: PIE Lexicon uses the smallest number of assumptions in existence: Only PIE *ē e ō o ɑ H; no compensatory lengthening, no Proto-Indo-Semitic root axiom, no Neogr. *a, etc. Typology is typology, and I understand the general point, so why not consider it this way: Which is typologically more likely system: {ē e ō o ɑ h} or {h1 h2 h3 e} regardless of what values are attached to h1 h2 h3?

"if the elimination of a phoneme requires assuming new sound changes than then recreate this phoneme, this is not progress; if the elimination of a phoneme requires assuming typologically unlikely sound changes, this is also not progress."

JOUNA: I disagree. Möller's one and only positive contribution to the Indo-European linguistics is his analysis of the second part of the cover symbol Neogr. *a into *Ae. A new sound change, viz. a colouring rule, is required, but I see this as a positive progress bringing regularity into former irregularity. Consequently the rule is accepted.

The colouring of PIE *ɑe, *eɑ -> IE a is by no means Pre-PIE development. Quite the contrary we've already coded them as the first post-PIE changes in PIE Lexicon, look these yourself, say for Hittite, in http://pielexicon.hum.helsinki.fi/?showrule=21

PIE is by definition the last common ancestor of the Indo-European languages, not the "most regular" common ancestor stage. If a rule exists in identical shape in every descendant, and is not preceded by any developments that only apply to some branches, it is necessary to consider it to have already taken place in PIE — even if the rule introduces morphophonological complications.

JOUNA: No. PIE is defined on the basis of two factors, viz. the (primary) phoneme inventory and the sound changes. We define the point when both are intact and unchanged as the Time T+0. However, when the first sound law applying to all languages, viz. the colouring rule
define RColouring1a e -> a || ɑ _ .o. RepeatAll; # PIE *ɑe → ɑa | Colouring rule for *ɑe | (Pyysalo 2013: 2.2.5.3)
took place there also emerged a new phoneme PIE *a previously non-existing phoneme *a to the language. This clearly indicates that PIE *T+1 is no longer PIE *T+0 i.e. this change cannot be considered to have taken place in PIE, but to have created a new language form from PIE.

Of course consonants (obstruents/fricatives) are not vowels and therefore have no vocalisations.

This critique seems to be quite common, but I think it confuses synchronic and diachronic sound changes. Obstruent vocalization is well attested, usually with an intermediate approximant stage (e.g. Latin noctem > Old French noit /nojt/ > French nuit /nɥi/). As I see it, the sensible analysis is that a "zero-grade laryngeal" develops into a vowel in late PIE, not that it immediately alternated with a "vocalic laryngeal".

(Although, take a look at Berber or Salishan languages or the Ōgami language of the Ryukyuan family sometimes: they do some quite wild stuff with phonotactics, and "syllabic fricatives" could be quite well posited for them.)

This argument also does nothing to rule out epenthetic vowels. Have you followed recent work by e.g. Andrew Miles Byrd on this matter?

JOUNA: I don't think the problem is in the sector you are discussing of: Already the Neogrammarians proved the existence of the cover symbol *ǝ, a vowel. Once done it is strictly speaking illegitimate to remove (or in this case replace) it with a consonants, because consonants and vowels are fundamental oppositions. Since we must postulate *ǝ (= PIE *ɑ) on this basis anyway, it is pointless to derive its vocalic (syllabic, colouring) properties from a consonant of any kind.

[JounaPyysalo]

I'm however somewhat concerned that you only cite data for this root from Hittite, Greek and Umbrian, which sounds like that there is a risk that rather than common PIE inheritance, it is a loanword from Anatolian to Greek (and perhaps thence to Umbrian, or perhaps that word is not related at all). It's easy to argue for whatever reconstruction at all if you're allowed to hand-pick the data to examine.

The comparative method of reconstruction follows the principle of postulation ("Durch zweier Zeugen Mund wird alle Wahrheit kund" — August Fick) and there is nothing wrong in the etymology of Goetze and Pedersen bringing the Greek forms here

No, this does not hold. An etymology does not need to have anything formally wrong in it to be regardless unreliable. I believe the classic example is that "gun" and "whiskey" are formally reconstructible for Proto-Plains Algonquian. Also e.g. "bullet" is formally reconstructible for Proto-Samic, "book" is formally reconstructible for Proto-Finnic, and "radar" is formally reconstructible for Old Finnish — all regardless flagrant violations of known history.

JOUNA: As I mentioned earlier I am strictly committed to the reconstruction of PIE here, hence not making general statements on Uralian or anything of the sort. Simply indicating that the standard procedure in the Comparative Method has been and is the same: Attestation in two Indo-European branches suffices for the reconstruction with the obvious reservation of loan words.

"Explaining all the data as inherited" strikes me as a sign of a poor theory, not the sign of a good one. Not all etymological comparisons that have been made are a priori correct, especially ones that have been proposed before the formulation of a stable, generally-agreed-on system of reconstruction.

JOUNA: I am keenly aware of this. One of the reasons for the digitalisation of OS PIE was the desire to set forth a unitary means of testing for all etymologies so far (inclung those proposed myself as well.

(a) Ultimately the whole idea of the laryngeal theory, viz. explaining the vowel qualities with laryngeals is totally unrealistic and phonetically unfounded. Especially the vowel qualities relevant here, e a o are produced in mouth cavity, not in the throat (or "larynx" in the broad sense).

This is false. Vowel coloring by laryngeal consonants is well attested in languages such as Arabic (where /a/ has a back allophone [ɑ] next to emphatic consonants and pharyngeals, and elsewhere a front allophone that I've seen varyingly described as [ɛ] or [ɜ] or [æ]) or Northern Haida (where /a/ has a back allophone [ʌ] next to uvulars and pharyngeals, elsewhere a front allophone [ə]).

Yes, i understand that, but what happens in the examples above (similarly also, e.g., in Hindi) is a coarticulation. In the laryngeal theory there is only /e/ (i.e. no distinctive articulation) whereas the distinctions of the vowels /a/ /e/ /o/ are assumedly emerging from the throat. I think this is a very bad idea: Have you ever tried so pronounce a h1 h2 h3 in your throat in a such manner that it turns an /e/ in your mouth cavity into /a/ or /o/?

The phonetical founding is straightforward too. Pharyngealization or uvularization imparts on vowels a similar acoustic coloring as retraction and lowering, and the latter is easier to pronounce; therefore we expect a laryngeal-colored allophone such as [əˤ] to soon be replaced by the corresponding backed and lowered vowel [ʌ]. Not all sound changes need to have an articulatory basis.

(b) The idea of the co-occurence of PIE *ɑ and PIE *h is hardly nothing but a way to explain the "colouring effect" erroneously attached to the "laryngeal" in the laryngeal theory. Although the proponents of LT are often ignorant of the history of the field, it should be underlined that Neogr. *ə is a comparatively well-defined phoneme and a vowel. The a mistake Kurylowicz made in equating that vowel (= PIE *ɑ) with a consonant Hitt. h (in 1927), is now corrected in OS PIE Lexicon. The issue is that whenever the Old Anatolian suggests a laryngeal (Hitt. h) the rest of the group seems to respond with ‘a-vocalism’ at least in some forms – and vice versa. Therefore both two items are reconstructed.

I follow this part of the logic well enough, but I think you go off the rails when you begin postulating laryngeals and after that their attached *a's on the basis of non-direct reflexes such as stop voicing or aspiration or what you call Fortunatov's Law II. Having to posit, within your framework, laryngeals and *a co-occurring in some words does not imply that they would need to co-occur always.

JOUNA: The proof for this is sought in PIE Lexicon through an extensive study of the vocabulary itself by means of complete induction – just as for other induction hypotheses presented in OS PIE Lexicon. If an error is detected the rule in question is corrected or abandoned, if beyond repair.

(c) Finally the general structural features of PIE, in particular the consequences of the existence of the zero grade in the schema PIE *ē : e : Ø : o : ō should be taken into account: Although this is structural argumentation, it is clear that had PIE *h not been accompanied with the vowel PIE *ɑ this would have yielded a huge number of unpronounceable root shapes in hCC ChC CCh.

This is internal reconstruction, not a comparative one; and the problem can be easily circumvented by positing epenthetic *ə in awkward zero-grade forms.

JOUNA: That sort of solution is not available in Comparative Method, because you can only posit "epenthetic *ə" if at least two branches imply that – and be able to formulate the conditions for it appearance in positions əhCC CəhC CCəh or əhCC CəhC CCəh which is impossible: First the IE languages only allow the postulation of PIE *ɑ rather than PIE *ə and secondly its position cannot be predicted, but it must be comparatively inferred on the basis of the data.

In comparative method TWO WITNESSES are required for postulation.

You keep repeating this quote, but I don't think it is valid as a general principle. It is merely a rule of thumb that in some cases follows as a corollary of Occam's Razor. But it cannot be applied generally, since it is possible for two languages have undergone a similar innovation.

JOUNA: It is not a corollary of Occam's Razor but the fundamental principle of Comparative Method, explained by Pedersen (1962:274) to mean that:
“If a word [or an object of any level]is found in the two branches, then it was also to be found in the original language which divided into these branches.”
(I discuss the principle of postulation and its relation to reconstruction (i.e. everything that is postulated) in §1.5.5 of my dissertation "Reconstruction and the principle of postulation).

Otherwise we'd run into conflicing issues such as:
• Armenian and Germanic provide two witnesses that the traditional series of mediae should be reconstructed as voiceless;
• but, meanwhile, Greek and Indic provide two witnesses that the series should be reconstructed as plain voiced;
• and Celtic and Balto-Slavic provide two witnesses that we should not reconstruct any distinction between the mediae and the voiced aspirates in the first place.
Hence when it comes to e.g. setting up proto-language phonology, two witnesses appears to be sometimes an insufficient criterion.

JOUNA: Yes, we could also postulate PIE *a for RV. a = Lith. a – and so forth, but there is also another principle that has to be followed, viz. the completeness of the data is to be observed. In this sense the principle of postulation is recursive, meaning that first two, then another two and then another two languages are to be compared until the stock of data has been exhausted. This note was only omitted by FIck, Pedersen and myself due to its evident character.

The second witness in this case are all the languages belonging to the correspondence, each one of them pointing to PIE *d, which in turn implies PIE *ɦ—d or d—ɦ.

Circular logic: the part of your thesis I was quoting here is exactly the point where you attempt to argue in the first place that PIE *d implies a laryngeal at all.

JOUNA: Not so, because both PIE *ɦ and PIE *ɑ (in diphonemic PIE *ɦɑ) imply the very same position (PIE *deɦɑ·), and together explain all the features of the data: The voice of the stop PIE *d, the expected ‘a-vocalism’ in Armenian still allowing for the /e/ in the majority of the languages. It is not circular to explain the entire data with an unified reconstruction.

Your method in this and various other places appears to be:
1) list some examples where a laryngeal and phonetic feature X co-occur
2) claim that this allows you to infer a laryngeal from every single instance of X
3) reconstruct "invisible" laryngeals that disappear from the data without a trace, save for triggering feature X.
Which only creates an unfalsifiable non-theory.

JOUNA: I fail to follow this: With regard to each of the points I'd say that:
1) It has taken 15 years to compile the database of PIE Lexicon and test the applicability of the *h/ɦ and *ɑ and so far there has been no problems created with the hypothesis.
2) A laryngeal h2 is inferred from each occurrence of ‘a-vocalism’ in the mainstream laryngeal theory as well – to which I subscribe with the reservation of tiny improvements to the actual rules: Instead of just two rules, one for *h2e and another for *eh2 there is a more comprehensive set of rules applied in OS PIE Lexicon viz.
Neogr. *ā = PIE *Hɑē or PIE *ēɑH Neogr. *a = PIE *Hɑe or PIE *eɑH Neogr. *ǝ = PIE *Hɑ or PIE *ɑH (where H = PIE *h or PIE *ɦ)
3) "reconstruct "invisible" laryngeals that disappear from the data without a trace, save for triggering feature X". This is true for both System PIE and the mainstream laryngeal theory. You should, however note that the disappearance of the "laryngeals" (or rather the glottal fricative) is caused by the sound law PIE *h/ɦ -› IE Ø (except in Old Anatolian), i.e. the loss of PIE *h/ɦ, which cannot be prevented from taking place due to the regularity of the sound changes.

[Pole, the]

JOUNA: PIE Lexicon uses the smallest number of assumptions in existence: Only PIE *ē e ō o ɑ H; no compensatory lengthening, no Proto-Indo-Semitic root axiom, no Neogr. *a, etc. Typology is typology, and I understand the general point, so why not consider it this way: Which is typologically more likely system: {ē e ō o ɑ h} or {h1 h2 h3 e} regardless of what values are attached to h1 h2 h3?

It is a dual strawman.

First: You're not postulating /(i u) e ē o ō ɑ h/. You are postulating /(i u) e ē o ō ɑh hɑ ɑɦ ɦɑ/.

Second: The mainstream postulates at least /(i u) e o h₁ h₂ h₃/, with the values of the last three commonly assumed to be [h x ɣ] or similar.

The latter inventory does not imply anything typologically unusual — abundant velar-or-further-back fricatives are a common phenomenon, even in European languages (see German [h x ʁ] or Dutch [ɦ x ɣ] — both languages also have [ʔ]).

The former inventory implies existence of such unsplittable sequences as [ɑh] or [hɑ], with no separate [ɑ] or [h] being considered. It is indeed not a common occurrence. I also do not know of any parallel of than in a natural language — if you do, please let us know.

[marconatrix]

The theoretical motive comes on the one hand from the issue that already the Neogrammarians proved the existence of a vocalic photo-phoneme *ə (= PIE *). On the other hand the connection between between the traditionally reconstructed PIE stops and the "a-colouring" of the roots needs to be simultaneously solved. For an example of the latter Ctrl+F the root √dɦɑn- √dɦɑon- √dɦɑen- ‘Tot’ in PIE Lexicon, representing the general ablaut class Ion. θνητό-: Gr. θάνατο-.[/i]

A truly deadly argument. However, thick as I am, I cannot see why you can't postulate an "a-coloured laryngeal" which can (i) colour adjacent vowels, and (ii) be itself vocalised to /a/. Note that some PIE theorists have characterised 'H1-3' as 'ə1-3' which implies entities that can manifest as either vowels or consonants, as well as modifying (adding features [+low], [+voice], [+aspiration]) adjacent segments. In your example, I would have /dHn/ > /d^ɦHn/ > /d^ɦan/ where there was no original vowel (zero grade). That is, in this particular case the H modifies the /d/ to /d^ɦ/ before being vocalised to /a/ or perhaps colouring an epenthetic /ə/ > /a/. That is both (i) and (ii) above.

Btw. If all Greek /a/'s derive from laryngeal colouring, how do you account for the second /a/ in θανατος ?

I and Juha very much agree on a single laryngeal in PIE as do all partners of PIE Lexicon project.

Then as a minimum you must specify the environments which distinguish your [h] from [ɦ], otherwise you are fudging the issue of at least two laryngeals.

[Artaxes]

In my opinion there were a three laryngeal fricatives: h₁ = /h/, h₂ = /ħ/, h₃ = /ħʷ/.

[JounaPyysalo]

JOUNA: PIE Lexicon uses the smallest number of assumptions in existence: Only PIE *ē e ō o ɑ H; no compensatory lengthening, no Proto-Indo-Semitic root axiom, no Neogr. *a, etc. Typology is typology, and I understand the general point, so why not consider it this way: Which is typologically more likely system: {ē e ō o ɑ h} or {h1 h2 h3 e} regardless of what values are attached to h1 h2 h3?

It is a dual strawman.

First: You're not postulating /(i u) e ē o ō ɑ h/. You are postulating /(i u) e ē o ō ɑh hɑ ɑɦ ɦɑ/.

JOUNA: Exactly so: which one you prefer, depends of whether a phoneme- or functional-based description is chosen.

Second: The mainstream postulates at least /(i u) e o h₁ h₂ h₃/, with the values of the last three commonly assumed to be [h x ɣ] or similar.

JOUNA: It is very difficult to discuss meaningfully what is the mainstream, since there are dozens of versions these days: The theories accepting Lex Eichner also postulate long vowels – and often the schwa is added from nowhere.

The latter inventory does not imply anything typologically unusual — abundant velar-or-further-back fricatives are a common phenomenon, even in European languages (see German [h x ʁ] or Dutch [ɦ x ɣ] — both languages also have [ʔ]).

The former inventory implies existence of such unsplittable sequences as [ɑh] or [hɑ], with no separate [ɑ] or [h] being considered. It is indeed not a common occurrence. I also do not know of any parallel of than in a natural language — if you do, please let us know.

JOUNA: This is a very good point, Pole, and I share the concern: There is a similar observation in my dissertation where I discuss the typoology (see p. 461-2); My expertise of the languages of the world are limited to the Indo-European languages – and besides I am not too hopeful in finding a typological parallel: the sequences *e+ɑ and *ɑ+e tend not to be long-living in the languages. Since *h can disappear with relative ease it is by no means obvious that such a system would have made it through the time thus preserving a parallel. But I am glad that you took this issue up, perhaps someone knowing better can find out something.

[marconatrix]

2+4. The Neogrammarian+Monolaryngealism. Typical for this trend is to avoid external typology (or paradigm) and define the Indo-European vowels by means of inductive comparison of the Indo-European languages. Between 1870-1940 a system of eight phonemes for vowels Neogr. *a *e *o *ā *ē *ō *å *ǝ was established and despite the domination of the laryngeal theory the correct solution to the laryngeal problem was established: Hitt. h = PIE h.

The trouble is that this sort of inclusive 'induction' method from all attested sources tends to produce extra spurious phonemes in the reconstructed protolanguage. This is because regular conditioned changes become blurred in the daughter languages by analogy and random variation.

No, the stem final *·ɑ- is posited due to the ‘a-vocalism’ in Greek, implying that PIE *ɑ had to be present (it being the sole cause of ‘a-vocalism’). Since the vowel *ɑ could not precede PIE *h due to the absence of ‘a-colouring’ in Hitt. ueh- = Umbr. ue- it had to follow PIE *h, thus implying PIE *uehɑ-. Hitt. u̯eḫan- is taken from PIE *u̯ehɑen- but also PIE *u̯ehɑon- would be possible Trad *a and *o colliding in Hitt. a.

And here you provide a perfect example. There is an observed mismatch between the expected developments in Hittite/Umbrian vs those in Greek, and your solution is to 'multiply entities' rather than accept that most probably either (i) there was some conditioned process at work, the causes of which may no longer be apparent, or (ii) an analogical or random change took place in one of the branches. As I've already remarked it's amazing and rather useful that most of the time sound changes are regular, they "run on rails". Sometimes however a word will "jump the tracks". Accidents happen!

[JounaPyysalo]

The theoretical motive comes on the one hand from the issue that already the Neogrammarians proved the existence of a vocalic photo-phoneme *ə (= PIE *). On the other hand the connection between between the traditionally reconstructed PIE stops and the "a-colouring" of the roots needs to be simultaneously solved. For an example of the latter Ctrl+F the root √dɦɑn- √dɦɑon- √dɦɑen- ‘Tot’ in PIE Lexicon, representing the general ablaut class Ion. θνητό-: Gr. θάνατο-.[/i]

A truly deadly argument. However, thick as I am, I cannot see why you can't postulate an "a-coloured laryngeal" which can (i) colour adjacent vowels, and (ii) be itself vocalised to /a/. Note that some PIE theorists have characterised 'H1-3' as 'ə1-3' which implies entities that can manifest as either vowels or consonants, as well as modifying (adding features [+low], [+voice], [+aspiration]) adjacent segments. In your example, I would have /dHn/ > /d^ɦHn/ > /d^ɦan/ where there was no original vowel (zero grade). That is, in this particular case the H modifies the /d/ to /d^ɦ/ before being vocalised to /a/ or perhaps colouring an epenthetic /ə/ > /a/. That is both (i) and (ii) above.

JOUNA: I have already discussed about this with Tropylium above (i) ‘colouring laryngeals' are not phonetically realistic (ii) neither is a "vocalised laryngeal" (consult
https://helda.helsinki.fi/handle/10138/41760
for the details).
Finally, as directly related to the data, you should note that the LT rules break down with this example: There is also Gr. τέθνα- with short final /α/ – and no vowel emerges in the first syllable. In order to achieve completeness in this and similar cases a broader set of rules is necessary.

Btw. If all Greek /a/'s derive from laryngeal colouring, how do you account for the second /a/ in θανατος ?

JOUNA: Please Ctrl+F the items in PIE Lexicon to see yourself. When referring to that I mean that the reconstruction is already available.

I and Juha very much agree on a single laryngeal in PIE as do all partners of PIE Lexicon project.

Then as a minimum you must specify the environments which distinguish your [h] from [ɦ], otherwise you are fudging the issue of at least two laryngeals.

JOUNA: Yes, indeed: Actually Juha and I have discussed this several times. We both agreed that we could do with a single PIE *h thus sticking to the traditional theory and notation due to which the issue is somewhat "cosmetic". On the other hand there are several factors why PIE *h : ɦ was ultimately opted, viz.
1. The traditional notation for the series bh dh gh is wrong (still so in IPA). The proper notation is actually *bɦ dɦ gɦ (with a voiced glottal fricative) which makes them in terms of PIE as clusters b+ɦ, etc. with *ɦ.
2. More importantly I wanted to bring forth the distinction PIE *h : ɦ because the traditional roots with a single inspirited media D are of the form *ɦ—D or D—ɦ, i.e. the feature ‘voice’ implies a laryngeal PIE *ɦ, which is critical for the reconstruction because with the help of this we are able to identify numerous items otherwise lost.
3. Finally the distinction between PIE *h : ɦ is important in particular in PIE Lexicon because with the explicit distinction of the voiceless and voiced items we can make accomplish comprehensive digital searches in PIE Lexicon gathering all instances of the voiced PIE *ɦ together, thus enabling the study of its causes and the possible formulation of the voicing rule, if possible at all.