The Great Proto-Indo-European Thread

[JounaPyysalo]

@ All:

Tou can follow PIE Lexicon both in FB
https://www.facebook.com/pielexicon
and in Twitter
https://twitter.com/PIELexicon

Also note that I often discuss the matters of the reconstruction with my colleagues in a FB group "Proto-Indo-European" that you can easily join if interested:
https://www.facebook.com/groups/126696954194312/

Atb,

Jouna

[jal]

And if everything else fails, the hat looks nice :)


JAL

[KathTheDragon]

1. The three-laryngeal notation is nothing but writing the PIE vowels /e/ /a/ /o/ as laryngeals, a Semitic feature brought by Möller. This will be detailed in the dissertation as you read forth.

I disagree. Every modern source I have read agrees that PIE possessed at the very least *e and *o as distinct orignal vowel qualities, and some suggest that there may have been an original *a quality too.

I mean the manner of postulation: If you think an equation such as Lat. es- = Hitt. es- = RV. as- "to be" there is no such thing as a "laryngeal" in the beginning of the root. The sole reason for its postulation is that the Semitist Möller believed that the Indo-European and the Semitic languages are genetically sharing the same root-structure, viz. C1C2·(C3).

Since the root √s- ‘to be’ (ablaut *s- *es- *os-) contains only a single consonant, an initial *h1 is added trough the comparison with the root axiom es : C1C2 -> *h1s-.

From a historical point of view the Neogrammarian system had a lot of vowels, but no laryngeal(s) while the laryngeal theory traditionally has had a lot of laryngeals, but very little vowels.

Nowadays there is admittedly an unfortunate trend to combine the worst sides of the both theories into a single one with a lot of vowels and a lot of laryngeals.

In System PIE just the opposite course has been opted: very few vowels and laryngeals.

Again, I disagree. For example, I am re-reading Meier-Brügger's (2003) Indo-European Linguistics, and in the section describing the reconstruction of PIE, they go into detail about how one arrives at the reconstruction with *h₁ for *h₁es- for "to be". The starting point is the Vedic imperfect, which is obviously the augment added to the present and substitution for the secondary endings. However, one would expect a short augment in the plural, where the root vowel is zero-grade, and therefore not available for contraction as in the singular. An initial *h₁ solves this problem perfectly, as well as accounting for the other languages as well. Furthermore, Kloekhorst has published several articles detailing how the three-laryngeal system is reflected in various Anatolian languages, and I see no fault in his argumentation.

[JounaPyysalo]

And if everything else fails, the hat looks nice


JAL

Lol, according to the local wisdom
"Of course the dissertations are written to obtain the hat – and the sword! What else would be the purpose?"

[JounaPyysalo]

1. The three-laryngeal notation is nothing but writing the PIE vowels /e/ /a/ /o/ as laryngeals, a Semitic feature brought by Möller. This will be detailed in the dissertation as you read forth.

I disagree. Every modern source I have read agrees that PIE possessed at the very least *e and *o as distinct orignal vowel qualities, and some suggest that there may have been an original *a quality too.

I mean the manner of postulation: If you think an equation such as Lat. es- = Hitt. es- = RV. as- "to be" there is no such thing as a "laryngeal" in the beginning of the root. The sole reason for its postulation is that the Semitist Möller believed that the Indo-European and the Semitic languages are genetically sharing the same root-structure, viz. C1C2·(C3).

Since the root √s- ‘to be’ (ablaut *s- *es- *os-) contains only a single consonant, an initial *h1 is added trough the comparison with the root axiom es : C1C2 -> *h1s-.

From a historical point of view the Neogrammarian system had a lot of vowels, but no laryngeal(s) while the laryngeal theory traditionally has had a lot of laryngeals, but very little vowels.

Nowadays there is admittedly an unfortunate trend to combine the worst sides of the both theories into a single one with a lot of vowels and a lot of laryngeals.

In System PIE just the opposite course has been opted: very few vowels and laryngeals.

Again, I disagree. For example, I am re-reading Meier-Brügger's (2003) Indo-European Linguistics, and in the section describing the reconstruction of PIE, they go into detail about how one arrives at the reconstruction with *h₁ for *h₁es- for "to be". The starting point is the Vedic imperfect, which is obviously the augment added to the present and substitution for the secondary endings. However, one would expect a short augment in the plural, where the root vowel is zero-grade, and therefore not available for contraction as in the singular. An initial *h₁ solves this problem perfectly, as well as accounting for the other languages as well. Furthermore, Kloekhorst has published several articles detailing how the three-laryngeal system is reflected in various Anatolian languages, and I see no fault in his argumentation.

The laryngeal theory – starting deductively from Möller's Indo-Semitic theory – and the comparative method of reconstruction – starting inductively from the data are two largely different things. Accordingly, so are the reconstructions.

If you want to see faults in Kloekhorst's argumentation, please open the page
http://pielexicon.hum.helsinki.fi/?alpha=12.1

The very first objects you'll encounter there is the root PIE √s- √os- √es- (vb.) ‘(da) sein, existieren’

Check closely how each of the bases *s- *es- and *os- is comparatively confirmed by at least two witnesses (The principle of postulation).

Jouna

[KathTheDragon]

Let us step back for a moment.

We know that the comparative method is the most effective tool we have for reconstruction, as it has been verified on families with attested ancestors. Now, we would expect our reconstructed language to function just as any other language does - coherent single paradigms for each individual verb. That is, each cognate form should go back to a single common ancestor. Yet, you posit all three non-long vocalisms for the 3sg of "to be", and by being more rigorous with your own approach, for the 3pl too, on the basis of Mycenaean e-e-si /ehensi/ and Attic εἰσί. The Doric form can also be traced back to this apparently full-grade form. I do not believe such a state of affairs is acceptable, and therefore, I cannot accept your model as being as superior as it claims to be.

[Pole, the]

From a historical point of view the Neogrammarian system had a lot of vowels, but no laryngeal(s) while the laryngeal theory traditionally has had a lot of laryngeals, but very little vowels.

Nowadays there is admittedly an unfortunate trend to combine the worst sides of the both theories into a single one with a lot of vowels and a lot of laryngeals.

Lel.

[JounaPyysalo]

Let us step back for a moment.

We know that the comparative method is the most effective tool we have for reconstruction, as it has been verified on families with attested ancestors. Now, we would expect our reconstructed language to function just as any other language does - coherent single paradigms for each individual verb. That is, each cognate form should go back to a single common ancestor. Yet, you posit all three non-long vocalisms for the 3sg of "to be", and by being more rigorous with your own approach, for the 3pl too, on the basis of Mycenaean e-e-si /ehensi/ and Attic εἰσί. The Doric form can also be traced back to this apparently full-grade form. I do not believe such a state of affairs is acceptable, and therefore, I cannot accept your model as being as superior as it claims to be.

1. Especially concerning the initial assumption: "Now, we would expect our reconstructed language to function just as any other language does - coherent single paradigms for each individual verb.That is, each cognate form should go back to a single common ancestor."
To this I would say that the comparative method of reconstruction does not make a priori assumptions concerning the protolanguage, but derives its shape on the basis of comparison of the data (the principle of postulation).

2. The structural inferences (such the one based on the vocalisations of the Proto-Indo-Semitic root axiom C1eC2·(C3) : C1C2·eC3 do not necessarily preserve the truth. Thus, while writing
”Yet, you posit all three non-long vocalisms for the 3sg of "to be", and by being more rigorous with your own approach, for the 3pl too, on the basis of Mycenaean e-e-si /ehensi/ and Attic εἰσί. The Doric form can also be traced back to this apparently full-grade form.”
I agree with you that if you mean that the PIE Lexicon approach would mean that the respective stem would have to be reconstructed as PIE *es-, this is indeed the case.

Regarding to this you write: "I do not believe such a state of affairs is acceptable...", but acceptable or not, this is the case in the data:

For the base *es- (with two normal grade vowels, one on the root, other in the suffix) compare the participles of this stem, viz. OLi. esąti- (pt.f.) LinB. ehont- (pt.m.) (Link. e-o-te) both from *esont-.

"... and therefore, I cannot accept your model as being as superior as it claims to be"
- What I mean with this is the following: The reconstruction in PIE Lexicon is completely digitalised. That is, instead of just quoting forms, the platform generates them.

To find out what I mean, please click on any PIE reconstruction (in blue). This is a command for the computer to create the respective IE form on the basis of the digitalised sound laws that you can find at the bottom right corner of the site: Click the full data page in
http://pielexicon.hum.helsinki.fi/?alpha=ALL
Then click "Select rule set", choose a language and press "Show rules".

All the forms in the PIE Lexicon have been created with these rules, and any errors in the generation are shown with the respective phonemes shown in red.

Jouna

[KathTheDragon]

You're missing my point. The one thing we as historical linguists should never lose sight of is the simple fact that PIE was a real language, spoken by real people. This precludes the possibility of PIE somehow having different workings to attested languages by virtue of being reconstructed. The only possible inference from your model is that PIE possessed all of *esti, *osti, and *sti as valid 3sg forms with no clear distribution (indeed, you project all three down into Anatolian itself!) This is in stark contrast to real languages, where there is always a distribution between multiple forms for the same word, be it dialectal, syntactic, a subtle nuance, or even a case of replacement. It is on this basis, and no other, that I reject your reconstruction. That you have a sophisticated database is of credit to your dedication, but it's hardly an independent argument for your model's superiority.

[Cedh]

I have skimmed through the summary chapters of the dissertation and I'd like to summarize what I perceive as Jouna's understanding of PIE phonology in relation to the mainstream laryngealist reconstruction (MLR), which is the version of PIE that I suppose most people here are most familiar with, but which unfortunately is not used for contrast very often within the dissertation.

Vowels: /e a o ē ō/
Resonants: /m n r l u i/
Obstruents: /p t k s/
Laryngeals: /h ɦ/

The resonants are the same as in MLR, existing in both syllabic and nonsyllabic variants. Some additional instances of /i u/ are posited to explain labiovelars and palatovelars.
The vowel inventory is quite similar to the vowel inventory of MLR, but the correspondences are not 1:1. See below.
The obstruents are greatly reduced compared to MLR, with all non-tenuis variants of the stops and all non-plain velars derived from sequences of several phonemes. See below.
The laryngeals basically both correspond to MLR *h₂. The other two MLR laryngeals *h₁ and *h₃ are posited to be unnecessary, the former because it's supposedly an artifact of Möller's Indo-Semitic hypothesis, and the latter by treating /o ō/ as basic. (NB: I'm listing two laryngeals here because Jouna says the condition for when /h/ or /ɦ/ occurs has not yet been found, so even though they're supposed to alternate I think they should best be regarded as two separate phonemes for now.)

For the vowels:
- /a/ is supposed to be lost in most positions when unstressed; where it remains it usually takes the place of the syllabic laryngeal which is reflected as *a in Latin and as *i in Sanskrit, e.g. in MLR *ph₂tḗr > Latin patēr, Skr pitā.
- There is only one ablaut system with five grades, ē : e : Ø : o : ō.

For the obstruents:
- MLR *p *t *k correspond to Jouna's /p t k/.
- MLR *b *d *g are explained as underlying /p t k/ in syllables where there's also a voiced laryngeal /ɦ/.
- MLR *bʰ *dʰ *gʰ are explained as underlying sequences /paɦ taɦ kaɦ/, where the vowel drops out and the voiced laryngeal causes the plosive to become a voiced aspirate.
- Neogrammarian *pʰ *tʰ *kʰ (usually MLR *pH *tH *kH) are explained as underlying sequences /pah tah kah/, where the vowel drops out and the voiceless laryngeal causes the plosive to become a voiceless aspirate (i.e. similar to how MLR explains the Indo-Iranian Tʰ series).
- MLR labiovelars *kʷ *gʷ *gʷʰ (and Neogrammarian *kʷʰ) are explained as sequences of /k/ plus an unstressed prevocalic /u/, with laryngeals added as necessary to create the right phonation (i.e. something like /kuV kuVɦ kuɦV/ (and /kuhV/)).
- Similarly, MLR palatovelars *k̂ *ĝ *ĝʰ (and Neogrammarian *k̂ʰ) are explained as sequences of /k/ plus an unstressed prevocalic /i/ plus laryngeals, i.e. something like /kiV kiVɦ kiɦV/ (and /kihV/).

Is this more or less correct as a summary?

(I must admit I don't understand exactly when a plosive would become voiced and when it would become a voiced aspirate - could you please explain that in some more detail?)

The reconstruction in PIE Lexicon is completely digitalised. That is, instead of just quoting forms, the platform generates them.

Regardless of what to think of your theory as a whole, this approach of testing the validity of linguistic reconstructions through automated simulation of the reconstructed sound changes is something I like quite a lot.

[KathTheDragon]

-snip- Is this more or less correct as a summary?

Yes.

Wrt

(I must admit I don't understand exactly when a plosive would become voiced and when it would become a voiced aspirate - could you please explain that in some more detail?)

the stop becomes breathy voiced when the laryngeal comes in direct contact with the stop (Taɦ and Tɦa), and becomes only voiced otherwise (T...aɦ and T...ɦa; that is, when there is intervening material)

[Sumelic]

- MLR *b *d *g are explained as underlying /p t k/ in syllables where there's also a voiced laryngeal /ɦ/.

Do they have to be in the same syllable? I missed that; the condition I got was that they needed to be in the same morpheme. Or are these equivalent formulations?

[Terra]

Hi, Terra, thank you for your questions, here some answers:

0. Before entering to the details in begin with noting that the answers to the questions concerning the reconstruction can be found in my dissertation, which I recommend you to read perhaps even a couple of times when you have extra hours to spend.

I'll try to find time to read it.

1. In System PIE there is only one "laryngeal" PIE *H, actually a glottal fricative with a voiceless (PIE *h) and avoiced variant (PIE *ɦ).
Traditionally there are not three, but four series of stops reconstructed for Proto-Indo-European by the Neogrammarians, viz.
(a) The series T (= *k p t). This is considered the basic set of items in PIE Lexicon.
(b) The series Th (= *kh *ph *th). This series has been considered as secondary since Saussure and Kurulowicz, who analysed the series as T+h2. PIE Lexicon agrees, but instead of "h2" uses the voiceless glottal fricative PIE *h for this purpose, i.e. Th = T+h.
(c) The series D (= *g b d) is also considered secondary in System PIE in the sense that the voiced stops are considered to gave gained their voice from the voiced glottal fricative PIE *ɦ), i.e. whenever *g b d appears the root also contained PIE *ɦ either before or after the voiced stop D. Thus the Neogrammarian roots with a single voiced stop √D are actually of the form *ɦ—D or *D—ɦ.
(d) The series Dh (= gh bh dh) is actually of the form D+ɦ, i.e. a combination of a voiced stop followed by the voiced glottal fricative PIE *ɦ.

Since (as in the previous case) the voice of the stops is caused by the voice of PIE *ɦ, actually PIE Lexicon only reconstructs a single series of stops, PIE *k p t, not two (as you suggest), three (as in mainstream LT) or four (as in the Neogrammarian system).

However, since the conditions of the alternation PIE *h : *ɦ have not been clarified as of yet and as the data is attested with items implying voiced stops these are preferred in reconstruction instead of voiceless stops. Thus, for instance, for Lat. ago : Av. azaiti etc. PIE Lexicon would reconstruct a stem PIE *ɦɑegi̯e/o- instead of a bit older PIE *ɦɑeki̯e/o- before the shift of the voice from the "laryngeal" to the following stop.

If *d was just *t + *ɦ, and *ɦ didn't leave any other traces, how is reconstructing *t + *ɦ better than just *d, other than wanting to avoid *d because you arbitrarily want to reconstruct a single series of stops? Did *ɦ leave any other remnants, besides affecting stops?

Also, how does your system explain the paucity of roots that have the shape TeDh, DheT, or DeD ?

2. Regarding the relations between PIE and Uralic I must admit that I do not know. I master the 100+ most ancient IE languages, but only speak Finnish as my mother tongue, therefore knowing little or nothing about the bulk of the vocabulary of the Finno-Ugric and Samoyedic. In such circumstances I am unable to compare the groups and therefore stating anything too meaningful about the issue.

Perhaps I could, however, mention that I am not too hopeful with regard to the genetic relation between Finnish and Indo-European: Having compiled an IE etymological dictionary almost 20 years my intuition says that if there were a genetic relation between Finnish and IE I should have noticed that by now.

But as I also said my knowledge of Uralic is very limited, therefore my estimate also potentially fallible.

Okay.

[Sumelic]

If *d was just *t + *ɦ, and *ɦ didn't leave any other traces, how is reconstructing *t + *ɦ better than just *d, other than wanting to avoid *d because you arbitrarily want to reconstruct a single series of stops? Did *ɦ leave any other remnants, besides affecting stops?

Also, how does your system explain the paucity of roots that have the shape TeDh, DheT, or DeD ?

From what I understand, System PIE h/ɦ are supposed to have the same reflexes in all languages aside from the hypothesized effect on the voicing of plosives. (That's why they are analyzed as allophones of one "phoneme" represented by *ḫ). The reflex is ḫ in Hittite when it was not after a plosive (after the deletion of the *a that is always postulated immediately neighboring it), zero when it coalesced with a plosive in Hittite (I think? I'm not sure on the conditions for the loss in Hittite), and zero everywhere in non-Anatolian languages.

You're missing my point. The one thing we as historical linguists should never lose sight of is the simple fact that PIE was a real language, spoken by real people. This precludes the possibility of PIE somehow having different workings to attested languages by virtue of being reconstructed. The only possible inference from your model is that PIE possessed all of *esti, *osti, and *sti as valid 3sg forms with no clear distribution (indeed, you project all three down into Anatolian itself!) This is in stark contrast to real languages, where there is always a distribution between multiple forms for the same word, be it dialectal, syntactic, a subtle nuance, or even a case of replacement. It is on this basis, and no other, that I reject your reconstruction. That you have a sophisticated database is of credit to your dedication, but it's hardly an independent argument for your model's superiority.

I'm confused by some terminological issues here. KathAvarea, would you reject the idea that it is useful to distinguish a linguistic reconstruction of the most recent common ancestral forms of words in a language family, created using the comparative method (the definition I would give for a proto-language) from an attempted description of the most recent actual spoken language/languages ancestral to modern ones that existed at some point in the past? It seems to me that there is no reason to expect these two types of "proto-language" to be identical in fact, although we would hope to be able to get close to the latter through linguistic analysis. One problem I've heard brought up is that the extant features preserved in the family may have diverged at different times, while any description of an actual language must be synchronic in nature.

[KathTheDragon]

You're missing my point. The one thing we as historical linguists should never lose sight of is the simple fact that PIE was a real language, spoken by real people. This precludes the possibility of PIE somehow having different workings to attested languages by virtue of being reconstructed. The only possible inference from your model is that PIE possessed all of *esti, *osti, and *sti as valid 3sg forms with no clear distribution (indeed, you project all three down into Anatolian itself!) This is in stark contrast to real languages, where there is always a distribution between multiple forms for the same word, be it dialectal, syntactic, a subtle nuance, or even a case of replacement. It is on this basis, and no other, that I reject your reconstruction. That you have a sophisticated database is of credit to your dedication, but it's hardly an independent argument for your model's superiority.

I'm confused by some terminological issues here. KathAvarea, would you reject the idea that it is useful to distinguish a linguistic reconstruction of the most recent common ancestral forms of words in a language family, created using the comparative method (the definition I would give for a proto-language) from an attempted description of the most recent actual spoken language/languages ancestral to modern ones that existed at some point in the past? It seems to me that there is no reason to expect these two types of "proto-language" to be identical in fact, although we would hope to be able to get close to the latter through linguistic analysis. One problem I've heard brought up is that the extant features preserved in the family may have diverged at different times, while any description of an actual language must be synchronic in nature.

The former is necessarily anachronistic, but generally unavoidable, while the latter is ideal, but much harder to formulate. However, I do think it's possible to separate out the multiple stages of the development to a fair degree of accuracy, and preforms can be posited to these accordingly. As a case in point, is it not generally agreed nowadays that while *-ō is the latest common form of the masculine n-stem nom.sg, an older form is *-ons, prior to Szemerényi's Law? But if you read what I said carefully, I said that there had to be a distribution between the multiple forms to be acceptable. In the example I gave, *-ō is posterior to *-ons. However, no explanation is offered for how the grades *os- *es- *s- were distributed in the protolang, whether internal to the paradigm, or through the development of the language and its daughters. That is why this reconstruction is not acceptable. Were such a distribution to be offered, I'd happily consider it.

[Sumelic]

The former is necessarily anachronistic, but generally unavoidable, while the latter is ideal, but much harder to formulate. However, I do think it's possible to separate out the multiple stages of the development to a fair degree of accuracy, and preforms can be posited to these accordingly. As a case in point, is it not generally agreed nowadays that while *-ō is the latest common form of the masculine n-stem nom.sg, an older form is *-ons, prior to Szemerényi's Law? But if you read what I said carefully, I said that there had to be a distribution between the multiple forms to be acceptable. In the example I gave, *-ō is posterior to *-ons. However, no explanation is offered for how the grades *os- *es- *s- were distributed in the protolang, whether internal to the paradigm, or through the development of the language and its daughters. That is why this reconstruction is not acceptable. Were such a distribution to be offered, I'd happily consider it.

Thanks for clarifying! I am generally ignorant on the subject of Indo-European and PIE historical linguistics. Are ablaut grades generally understood? I read through the System PIE pdf, but there was very little in it about ablaut; I assumed that its origin was one of the unsolved or contentious problems of Indo-European historical linguistics. Do the competing theories, like the multi-laryngeal theory, offer a better explanation than System PIE for the distribution of the different vocalic reflexes in descendants of this root, or for other words that show up with different ablaut grades in different daughter languages?

[KathTheDragon]

Mainstream theories also posit ablaut in exactly the same way as System PIE does: five grades, *ē *e *Ø *o *ō (although the evidence for *ō being part of the ablaut system is scanty). In the case of *h₁es- "to be", the distribution is e-grade in the singular forms *h₁ésmi *h₁ési *h₁ésti, and zero-grade in the plural forms *h₁smé- *h₁sté- *h₁sénti. All the attested paradigms are understood to derive directly from this. I won't go into details about the individual developments (unless you want me to) because there are lots of daughter languages to cover.

[Sumelic]

I won't go into details about the individual developments (unless you want me to) because there are lots of daughter languages to cover.

I guess I'd be curious to know the mainstream theory about the development of the Hittite plural form with "a" given on the PIE Lexicon site (Hi. a-ša-an-du [3pl]), since JounaPyysalo seems to argue that the Hittite forms in particular of this word are evidence for the absence of a laryngeal in it (and that this form is evidence for an o-grade of the "sein" word). Does positing an initial h₁ allow this form to be explained as a descendent of the zero-grade rather than the o-grade?

[KathTheDragon]

The only satisfying explanation of that initial a I have seen is from Kloekhorst (insomuch as that's the only one I've seen, but I've heard of others that aren't very satisfying) is that the initial a is there because of an unwritten consonant denoted by Kloekhorst as ʔ, which is taken as the reflex of *h₁. His account of the forms is as follows: The starting point is the PIE paradigm with 3sg *h₁ésti, 3pl *h₁sénti. Regular phonetic developments to Proto-Anatolian cause the loss of *h₁ word-initially before an obstruent, leaving (after altering notation) *ʔésti, *sénti. The initial *ʔ is reintroduced into the plural from the singular to avoid this alternation of the root syllable, giving *ʔésti, *ʔsénti (Kloekhorst - I myself would also suppose an epenthesis of *ə at this point to break up the cluster, yielding 3pl *ʔəsénti). Further phonological changes produce Hittite ʔéstˢi, ʔséntˢi (Kloekhorst - I would rather write ʔəsántˢi), spelled e-eš-zi, a-ša-an-zi, where the plene spelling in the singular indicates the initial ʔ, and the a in the plural is to break up the (unwritten) consonant cluster.

[JounaPyysalo]

The only satisfying explanation of that initial a I have seen is from Kloekhorst (insomuch as that's the only one I've seen, but I've heard of others that aren't very satisfying) is that the initial a is there because of an unwritten consonant denoted by Kloekhorst as ʔ, which is taken as the reflex of *h₁. His account of the forms is as follows: The starting point is the PIE paradigm with 3sg *h₁ésti, 3pl *h₁sénti. Regular phonetic developments to Proto-Anatolian cause the loss of *h₁ word-initially before an obstruent, leaving (after altering notation) *ʔésti, *sénti. The initial *ʔ is reintroduced into the plural from the singular to avoid this alternation of the root syllable, giving *ʔésti, *ʔsénti (Kloekhorst - I myself would also suppose an epenthesis of *ə at this point to break up the cluster, yielding 3pl *ʔəsénti). Further phonological changes produce Hittite ʔéstˢi, ʔséntˢi (Kloekhorst - I would rather write ʔəsántˢi), spelled e-eš-zi, a-ša-an-zi, where the plene spelling in the singular indicates the initial ʔ, and the a in the plural is to break up the (unwritten) consonant cluster.

Hitt. as- cannot reflect anything but PIE *o, because only PIE *os- -> Hitt. as- = Pal. as- = CLu. as- = OPr. as- = Northumbr. ar(un) [3pl], etc.

[JounaPyysalo]

You're missing my point. The one thing we as historical linguists should never lose sight of is the simple fact that PIE was a real language, spoken by real people. This precludes the possibility of PIE somehow having different workings to attested languages by virtue of being reconstructed. The only possible inference from your model is that PIE possessed all of *esti, *osti, and *sti as valid 3sg forms with no clear distribution (indeed, you project all three down into Anatolian itself!) This is in stark contrast to real languages, where there is always a distribution between multiple forms for the same word, be it dialectal, syntactic, a subtle nuance, or even a case of replacement. It is on this basis, and no other, that I reject your reconstruction. That you have a sophisticated database is of credit to your dedication, but it's hardly an independent argument for your model's superiority.

To my knowledge PIE *sti is not attested anywhere, therefore it cannot be postulated.

I can see that you don't have a very thorough knowledge on Indo-European languages, which as a rule posses great many devices of forming verbs from a root.

The ancient Indo-European languages are very unlike the modern ones with little variation and pretty steady paradigms.

Regarding to the superiority the proper term is accuracy of predictions, which is ultimately caused by the manner of composition of System PIE: Indo-European linguistics is far more than just the laryngeal theory.

Starting from the beginning of the 19th century, the study can be split into three main periods (add freely +/- 10 years to all periods), viz.
1. The Paleogrammarian era (approximately from 1800 to 1870)
2. The Neogrammarian era (approximately from 1870-1940)
3. The Laryngeal theory (approximately from 1940 to 2010)

System PIE consists of the correctly postulated PIE reconstruction phonemes and IE sound laws from every period, possibly upgraded (especially with regard to PIE *H) when needed to the effect that the obtained system is consistent both internally and with regard to the Indo-European data, which allows the digital version, OS PIE Lexicon, to generate the IE forms with unforeseen accuracy.

With the exception of very few open research problems, especially the PIE accent/tone and minor problems related to individual languages (e.g. Tocharian /a/ vs. /ä/, Germanic /o/ vs. /u/ and so forth) our team is already in possession of the effective solutions to the problems of the PIE reconstruction.

[JounaPyysalo]

I have skimmed through the summary chapters of the dissertation and I'd like to summarize what I perceive as Jouna's understanding of PIE phonology in relation to the mainstream laryngealist reconstruction (MLR), which is the version of PIE that I suppose most people here are most familiar with, but which unfortunately is not used for contrast very often within the dissertation.

Vowels: /e a o ē ō/
Resonants: /m n r l u i/
Obstruents: /p t k s/
Laryngeals: /h ɦ/

The resonants are the same as in MLR, existing in both syllabic and nonsyllabic variants. Some additional instances of /i u/ are posited to explain labiovelars and palatovelars.
The vowel inventory is quite similar to the vowel inventory of MLR, but the correspondences are not 1:1. See below.
The obstruents are greatly reduced compared to MLR, with all non-tenuis variants of the stops and all non-plain velars derived from sequences of several phonemes. See below.
The laryngeals basically both correspond to MLR *h₂. The other two MLR laryngeals *h₁ and *h₃ are posited to be unnecessary, the former because it's supposedly an artifact of Möller's Indo-Semitic hypothesis, and the latter by treating /o ō/ as basic. (NB: I'm listing two laryngeals here because Jouna says the condition for when /h/ or /ɦ/ occurs has not yet been found, so even though they're supposed to alternate I think they should best be regarded as two separate phonemes for now.)

For the vowels:
- /a/ is supposed to be lost in most positions when unstressed; where it remains it usually takes the place of the syllabic laryngeal which is reflected as *a in Latin and as *i in Sanskrit, e.g. in MLR *ph₂tḗr > Latin patēr, Skr pitā.
- There is only one ablaut system with five grades, ē : e : Ø : o : ō.

For the obstruents:
- MLR *p *t *k correspond to Jouna's /p t k/.
- MLR *b *d *g are explained as underlying /p t k/ in syllables where there's also a voiced laryngeal /ɦ/.
- MLR *bʰ *dʰ *gʰ are explained as underlying sequences /paɦ taɦ kaɦ/, where the vowel drops out and the voiced laryngeal causes the plosive to become a voiced aspirate.
- Neogrammarian *pʰ *tʰ *kʰ (usually MLR *pH *tH *kH) are explained as underlying sequences /pah tah kah/, where the vowel drops out and the voiceless laryngeal causes the plosive to become a voiceless aspirate (i.e. similar to how MLR explains the Indo-Iranian Tʰ series).
- MLR labiovelars *kʷ *gʷ *gʷʰ (and Neogrammarian *kʷʰ) are explained as sequences of /k/ plus an unstressed prevocalic /u/, with laryngeals added as necessary to create the right phonation (i.e. something like /kuV kuVɦ kuɦV/ (and /kuhV/)).
- Similarly, MLR palatovelars *k̂ *ĝ *ĝʰ (and Neogrammarian *k̂ʰ) are explained as sequences of /k/ plus an unstressed prevocalic /i/ plus laryngeals, i.e. something like /kiV kiVɦ kiɦV/ (and /kihV/).

Is this more or less correct as a summary?

(I must admit I don't understand exactly when a plosive would become voiced and when it would become a voiced aspirate - could you please explain that in some more detail?)

The reconstruction in PIE Lexicon is completely digitalised. That is, instead of just quoting forms, the platform generates them.

Regardless of what to think of your theory as a whole, this approach of testing the validity of linguistic reconstructions through automated simulation of the reconstructed sound changes is something I like quite a lot.

Hi, Cedh. Your summary is very much correct.

Here some additions and specifying details:

1. On the thing you mentioned not to understand, when a plosive becomes voiced and when not, the problem was with my very short notation. In order to illustrate how the system works (although the conditioning for the alternation h : ɦ is not known), please Ctrl+F PIE √tɑh- √dɑɦ- (vb.) ‘geben, schenken’ in PIE Lexicon full data page.

Under the root PIE √tɑh- √dɑɦ- (vb.) ‘geben, schenken’ you will first encounter the voiceless root

PIE √tɑh- "geben"

PIE *tóɑh- Gr. τό-(pr.) ‘geben’(Grundr2 1:654)(Gr. τότω ‘dato’)
PIE *Π·ski̯ɑhi·toɑh- Hitt. iški·taḫ-(vb2.) ‘ein Zeichen geben ‹Hitt. šagi- ‘Zeichen’›’(HHand. 65)(Hitt. iš-kit9-ta-aḫ-ḫi; or to Gr. δό- (?)

After that the better known voiced root variant follows:

PIE √dɑɦ- (vb.) ‘geben, schenken’

PIE *dṓɑɦ- Gr. δώ-(ao.) ‘geben’(GEW 1:388-9)(Gr. ἔδων [1sg])
PIE *dóɑɦmen- RV. dā́man-(n.) ‘das Geben’(WbRV. 595)(RV. dā́mane [sgD], Brugmann’s Law II
PIE *dēɑɦ- Lat. dā-(ao.) ‘geben, gewähren’(WH 2:360)(Lat. dās [2sg)
PIE *deɑɦ- Lat. da-(√pr.) ‘geben, gewähren’(WH 1:360)(in Lat. damus [1pl.], datum [pp.

Within these forms, after the laryngeal turned into voiced one PIE *tɑɦ- this also turned the initial stop into a voiced one (PIE *dɑɦ-).

In addition, when the vowel PIE *ɑ was lost in the zero grade of the root PIE *dɑɦ-, a voiced aspirate PIE *dɦ- resulted as in
PIE *dɑɦ·Σ- RV. dh·iṣ-(f.) ‘Opfer·lust, Lust zu geben’(WbRV. 683)(RV. dhiṣā́ [sgI])

2. "Neogrammarian *pʰ *tʰ *kʰ (usually MLR *pH *tH *kH) are explained as underlying sequences /pah tah kah/, where the vowel drops out and the voiceless laryngeal causes the plosive to become a voiceless aspirate (i.e. similar to how MLR explains the Indo-Iranian Tʰ series)."
- Also note that in addition to PIE *pɑh *tɑh *kɑh also PIE *phɑ *thɑ *khɑ account for the later Th-series. In the case of the latter the root appears with /a/-vocalism in Indo-European languages preserving this feature.

3. "MLR labiovelars *kʷ *gʷ *gʷʰ (and Neogrammarian *kʷʰ) are explained as sequences of /k/ plus an unstressed prevocalic /u/, with laryngeals added as necessary to create the right phonation (i.e. something like /kuV kuVɦ kuɦV/ (and /kuhV/))."
- Yes, but note that the current notation for aspirated labiovelars is incorrect: Trad. *kʷʰ and **gʷʰ suggest a sequence k-w-h and g-w-h, while in reality these clusters always contained laryngeal component before the labial (i.e. on is to write *kʰʷ and *gʰʷ). In PIE Lexicon the complete segmental analysis for the labiovelar series stands as follows:
Trad. *kʷ = *ku and/or *kw
Trad. *gʷ = *gu and/or *gw (in environment ɦ—gʷ and gʷ—ɦ
Trad. *kʰʷ = *kɑhu- or *khɑu- and/or *kɑhw or *khɑw
Trad. *gʰʷ = *gɑɦu- or *gɦɑu- and/or *gɑɦw or *gɦɑw
(For labiovelars, see System PIE §4.8.3.)

4. "Similarly, MLR palatovelars *k̂ *ĝ *ĝʰ (and Neogrammarian *k̂ʰ) are explained as sequences of /k/ plus an unstressed prevocalic /i/ plus laryngeals, i.e. something like /kiV kiVɦ kiɦV/ (and /kihV/)."

Yes. The exact equations are:

Trad. *k´ = *ki and/or *kj
Trad. *g´ = *gi and/or *gj (in environment ɦ—gi and go—ɦ
Trad. *k´ʰ = *kɑhi or *khɑi- or *kiɑh or *kihɑ
Trad. *g´ʰ = *gɑɦi or *gɦɑi or *giɑɦ or *giɦɑ
(For palatovelars, see System PIE §4.8.4.)

5. "/a/ is supposed to be lost in most positions when unstressed; where it remains it usually takes the place of the syllabic laryngeal which is reflected as *a in Latin and as *i in Sanskrit, e.g. in MLR *ph₂tḗr > Latin patēr, Skr pitā."
Yes. Note especially that PIE *ɑ (written /a/ in System PIE (Pyysalo 2013) due to a font issue) is the equivalent of the Schwa Indogermanicum of the Neogrammarians (i.e. *ǝ) including the rules for that (i.e. Schwa-loss, etc.). In addition do note that instead of equating and replacing Trad. *ǝ with *h2 as was done by Kurylowicz in his "la theorie de la schwa consonantique" (and subsequently MLR) the item Trad. *ǝ = PIE *ɑ is considered a vowel reconstructed as such.

Thus, OS PIE differs from the traditional doctrines with Neogr. *ǝ (e.g. in *pǝtḗr-) and MLR *h₂ (in *ph₂tḗr) in reconstructing BOTH the VOWEL *ɑ and the LARYNGEAL *h, thus positing PIE *pɑhtḗr- for the item in question.

The "a-colouring" and the "syllabic properties" of the "laryngeal are actually properties of the vowel PIE *ɑ, which always appears together with PIE *h in diphonemic PIE *hɑ and *ɑh.

Jouna

[KathTheDragon]

I think I'm done here for now, and will consign this to the pile of "nice-looking theories that I think don't hold water".

[JounaPyysalo]

I think I'm done here for now, and will consign this to the pile of "nice-looking theories that I think don't hold water".

On the contrary KathAveara: PIE Lexicon is the first water-proof theory that exist - just because it comprises of the correct solutions to problems presented during the 200 years of research in Indo-European linguistics.

Just check out the full data page of PIE Lexicon in
http://pielexicon.hum.helsinki.fi/?alpha=ALL
All spots not holding water are marked with red – and of these more than half can be fixed by adding the missing sound laws.

The correctness rate of PIE Lexicon is well above 99 per cents, which is not too bad at all.

Jouna

[KathTheDragon]

I've already told you why I disagree, but if you cannot accept even the fact that I disagree, even more reason for me to stop.