zompist bboard

KathTheDragon wrote:Realistically, Uralic need not tell us very much - assuming a shift ejective > implosive > voiced stop, there's no reason to assume that it was still in the 'ejective' phase in PIU.

My own Indo-Uralic material (a draft version is now at academia.edu) does not have very much to tell here. I don't see any special correspondences for the D series (d,ǵ,g,gʷ). In some correspondences the D series seems to behave the same as the T series (e.g. PIE t/d cluster with r/l/n ~ PU d) . In other correspondences it seems to behave the same as the Dh series (e.g. PIE ǵ,g,gʷ, ǵʰ,gʰ,gʷʰ ~ PU x).

I'll reiterate that this paper does contain good examples of PIE *D ~ PU *N that can't be explained as "nasalisation".

Howl wrote:

KathTheDragon wrote:Realistically, Uralic need not tell us very much - assuming a shift ejective > implosive > voiced stop, there's no reason to assume that it was still in the 'ejective' phase in PIU.

My own Indo-Uralic material (a draft version is now at academia.edu) does not have very much to tell here. I don't see any special correspondences for the D series (d,ǵ,g,gʷ). In some correspondences the D series seems to behave the same as the T series (e.g. PIE t/d cluster with r/l/n ~ PU d) . In other correspondences it seems to behave the same as the Dh series (e.g. PIE ǵ,g,gʷ, ǵʰ,gʰ,gʷʰ ~ PU x).

You mean the 'correspondences' basically act exactly how you'd expect borowings from a language with a large inventory into a language with a small inventory to work?

I wonder what could explain correspondences that look exactly like borrowings...

KathTheDragon wrote:I'll reiterate that this paper does contain good examples of PIE *D ~ PU *N that can't be explained as "nasalisation".

I'll put a more detailed evaluation of that document on my big to-do list. It does contain some useful etymologies that should be added to my list. I also should do a more detailed evaluation of Hyllested's work which is really good (even though I disagree with some details).

When it comes to 'd', Kümmel has three reflexes: 'n', 'ń' and 'l' (if I don't include ḱerd 'heart')
- The examples with 'l' might be explained with -l suffixes.
- It is indeed difficult to find good alternatives for näki 'to see' and panï 'to put'. For näki 'to see', it is tempting to link it to nekʷts 'night', but that kind of means the opposite. For panï 'to put' it is tempting to link it to Latin pono, but that one came from earlier po-sino, so that's obviously a chance resemblance.
- And ńüki- ‘to draw, tear’ ~ dewḱ- ‘to draw, lead’ and dn̥gʰweh₁- ‘tongue’ ~ ńaŋk-ćï ‘tongue, gums’ do look like good potential cognates. So this is something that is at least worth considering.

Edit: *ńüki- 'tear, pluck out, tug at' could also be linked to *h₁nek- 'to take, to seize' [LIV2]; *ńaŋkćï 'palate, gill' only means tongue in Saami. And the d in dn̥ǵʰu 'tongue' is only attested in Germanic. Latin and Armenian have 'l'. Celtic has 't'. Oscan has 'f' from earlier dʰ (?). Balto-Slavic has ø. Tocharian has *käntwo (metathesis of *tänkwo)

Salmoneus wrote: You mean the 'correspondences' basically act exactly how you'd expect borrowings from a language with a large inventory into a language with a small inventory to work?

I wonder what could explain correspondences that look exactly like borrowings...

I didn't write it as a 'proof of Indo-Uralic' kind of document. If I did, I would be stressing how much of the basic and most widely attested vocabulary in Uralic these 'borrowings' constitute. And I might even explain how easy it would be to assume the same for Hittite and Tocharian.

Howl wrote:And the d in dn̥ǵʰu 'tongue' is only attested in Germanic.

Not so! Old Latin attests dingua with an initial that can only reflect *d. The l- is adopted from lingere "to lick", and it's similar for the other words seemingly showing *l- (Lith. liežuvìs ~ liẽžti, Arm. lezow ~ lizanem).

An interesting thing about the Tocharian reflex is that only the places of articulation metathesized -- PToch. käntwo must be from pre-PToch. *ǵn̥dʰwéh₂- (rather than **ǵʰn̥dwéh₂-), since PIE *d becomes either PToch. *ts or zero.

The initial of lingua is a Latin-specific innovation -- it's absent in Oscan, which has acc. sg. fangvam... which implies *dʰn̥ǵwéh₂-, doesn't it?

Right, although more specifically a zero-grade suffix. De Vaan thinks the f- is due to influence from some semantically related word with f-, but not having a reasonably complete lexicon for Oscan makes it hard to know what.

Salmoneus wrote:

Zaarin wrote:

Šọ̈́gala wrote:is there anything especially appealing about an unmotivated general ejective > implosive shift, other than that it explains the dearth of /b/? Is this shift typologically common?

It happened in Yucatec Maya (/pʼ/ > /ɓ/ only), and it happened several times independently in Afroasiatic. There doesn't seem to be anything strikingly unusual about the change.

But then, ejective to voiced stop has also happened multiple times.

Has it? The usual example trotted out is Egyptian, where the Proto-Afroasiatic (or at least Proto-Egyptian+Semitic+Berber) *T' and *D series merged into a single series of stops which we transcribe as voiced stops. (These merged again with the voiceless stops in Coptic, which only has one series of stops.)

Curiously, though (the source I'm citing here is an acquaintance at Penn doing a PhD in comparative Afroasiatic), when you look at Semitic transcriptions of Egyptian words and names, that exact series basically never gets transcribed with the Semitic voiced series--both voiceless and emphatic stops are used. So it seems unlikely that the Egyptian "voiced" series actually was...

dhok wrote:

Salmoneus wrote:

Zaarin wrote:

Šọ̈́gala wrote:is there anything especially appealing about an unmotivated general ejective > implosive shift, other than that it explains the dearth of /b/? Is this shift typologically common?

It happened in Yucatec Maya (/pʼ/ > /ɓ/ only), and it happened several times independently in Afroasiatic. There doesn't seem to be anything strikingly unusual about the change.

But then, ejective to voiced stop has also happened multiple times.

Has it? The usual example trotted out is Egyptian, where the Proto-Afroasiatic (or at least Proto-Egyptian+Semitic+Berber) *T' and *D series merged into a single series of stops which we transcribe as voiced stops. (These merged again with the voiceless stops in Coptic, which only has one series of stops.)

Yes, we transcribe them as voiced stops. That doesn't mean that they were voiced stops! The most likely hypothesis would be that they were (voiceless) unaspirated stops, while those transcribed as voiceless stops were aspirated. Such transcriptions are common where no actual voiced stops get in the way; it is simply more economic than fiddling with h-digraphs or superscript ^hs. I don't know what the current common opinion among Egyptologists is, though.

Curiously, though (the source I'm citing here is an acquaintance at Penn doing a PhD in comparative Afroasiatic), when you look at Semitic transcriptions of Egyptian words and names, that exact series basically never gets transcribed with the Semitic voiced series--both voiceless and emphatic stops are used. So it seems unlikely that the Egyptian "voiced" series actually was...

... voiced.

dhok wrote:

Salmoneus wrote:

Zaarin wrote:

Šọ̈́gala wrote:is there anything especially appealing about an unmotivated general ejective > implosive shift, other than that it explains the dearth of /b/? Is this shift typologically common?

It happened in Yucatec Maya (/pʼ/ > /ɓ/ only), and it happened several times independently in Afroasiatic. There doesn't seem to be anything strikingly unusual about the change.

But then, ejective to voiced stop has also happened multiple times.

Has it?

Probably. We can't know for certain, of course, because you can always argue over the reconstructions. But it seems to be likely in many languages in the Caucasus.
For instance, Chechen and Ingush both show 'loss' of voiced stops intervocalically and 'voicing' of ejectives intervocalically, compared to their closest relative, Bats, which 'retains' the 'original' stops. [you could, of course, argue that Bats turned its intervocalic voiced stops into ejectives, etc].
Apparently, something similar probably happened in Tabassaran, although there's more uncertainty over proto-daghestanian reconstruction.
Anatolian dialects of Abaza, likewise, compared to standard Abaza and related languages.
Georgian and Hausa apparently have synchronic voicing of ejectives in many cases.
In dialects of Lezgian, apparently ejectives have deglottalised into unaspirated voiceless, but not yet voiced.

Armenian may also be an example, if you accept the glottalic hypothesis: some dialects have ejectives instead of voiced stops, though the normal theory is voiced > ejective.

Apparently Western Jicaque, but that relies on the reconstruction of Proto-Jicaque; and Subtiaba seems to have voiced stops where eastern (and proto?) Jicaque had ejectives, but apparently that's more controversial.

Final voiced stops in Yucatec correspond to ejectives in Cakchiquel and preglottalised nasals in Pocoman.

Tillamook has dissimilatory voicing of ejectives.

bilabial ejectives in Oromo and colloquial amharic, and maybe velar ejectives in omotic, and sometimes sporadically dental ejectives (eg in soqotri).

Etc etc. It's hard to find any single case where all ejectives became voiced, just because ejectives are rare to begin with, and relatively stable, and the change is only clear where it didn't happen in all daughters, and even then you have to rely on contestable reconstructions. But in general, voiced stops, preglottalised stops, laryngealised stops, unaspirated voiceless stops and ejectives all seem pretty close together in terms of conditioned changes and synchronic changes, so it really seems tendentious to assume that an unconditioned diachronic change is impossible, even if there's no documented example of it.

It's worth noting that changes like 'voiced aspirate to plain voiced' aren't exactly common either!

The usual example trotted out is Egyptian, where the Proto-Afroasiatic (or at least Proto-Egyptian+Semitic+Berber) *T' and *D series merged into a single series of stops which we transcribe as voiced stops. (These merged again with the voiceless stops in Coptic, which only has one series of stops.)

Curiously, though (the source I'm citing here is an acquaintance at Penn doing a PhD in comparative Afroasiatic), when you look at Semitic transcriptions of Egyptian words and names, that exact series basically never gets transcribed with the Semitic voiced series--both voiceless and emphatic stops are used. So it seems unlikely that the Egyptian "voiced" series actually was...[/quote]

WeepingElf wrote:Yes, we transcribe them as voiced stops. That doesn't mean that they were voiced stops! The most likely hypothesis would be that they were (voiceless) unaspirated stops, while those transcribed as voiceless stops were aspirated. Such transcriptions are common where no actual voiced stops get in the way; it is simply more economic than fiddling with h-digraphs or superscript ^hs. I don't know what the current common opinion among Egyptologists is, though.

At least since the ’90s it’s been thought that d, ḏ, g were voiceless, or at least distinguished from t, ṯ, k by something other than voice (ejectivity or non-aspiration) with maybe voiced allophones, already in Egyptian (b remains voiced b into Coptic except when historically final, where it’s written with p).

Salmoneus wrote:

Howl wrote:

KathTheDragon wrote:Realistically, Uralic need not tell us very much - assuming a shift ejective > implosive > voiced stop, there's no reason to assume that it was still in the 'ejective' phase in PIU.

My own Indo-Uralic material (a draft version is now at academia.edu) does not have very much to tell here. I don't see any special correspondences for the D series (d,ǵ,g,gʷ). In some correspondences the D series seems to behave the same as the T series (e.g. PIE t/d cluster with r/l/n ~ PU d) . In other correspondences it seems to behave the same as the Dh series (e.g. PIE ǵ,g,gʷ, ǵʰ,gʰ,gʷʰ ~ PU x).

You mean the 'correspondences' basically act exactly how you'd expect borowings from a language with a large inventory into a language with a small inventory to work?

I wonder what could explain correspondences that look exactly like borrowings...

To be fair, series do drastically reduce in cousin languages, eg. Common IE vs. Anatolian (where the voiced aspirates and voiced merge, but also initial and final voiced-voiceless distinction disappears).

PMP *T *D *NT *ND > PPolynesian *T

and *T *D *Dh > T in Tocharian

Sure, I never said it was impossible. It's just really suspicious. Likewise, dropping half the phonemes in a word is a perfectly possible sound change, between clearly related languages, when it's clear that this is what happened. But when you're doing speculative long-range comparison (and ignoring the possibility of borrowing), allowing suggested sound changes that conveniently drop half the phonemes in a word to produce exactly the desired form makes it much to easy to create 'cognates' that are really chance resemblances or borrowings.

In this case, when your sound changes are:

*t, *d, *r, *l, *n > *d
and
*ǵ,*g, *gʷ, *ǵʰ,*gʰ,*gʷʰ > *x

...you've got two options:
a) the language underwent some massive mergers that cut across all of MOA, phonation and secondary articulations, and that yielded seemingly random outcomes (a voiced stop at one POA, a voiceless fricative at another)
or
b) the language borrowed words that included many phonemes it did not possess, and just replaced them with what seemed like the closest thing available

And sure, a) is possible. If you've got a robust set of soundlaws between related languages in a family, the need to assume a) is not necessarily a disproof. But here we're relying on the probability of a) as our evidence for the existence of the family in the first place, even though b) seems far more probable (and of course massively more likely when priors are taken into consideration, since by default true soundlaws are much less common than borrowings and coincidences).

In my opinion (which I have voiced here before), the Proto-Uralic phonology is probably more conservative the PIE one, because the other Mitian languages are more similar in this regard to the former than the latter. Michael Fortescue has reconstructed, in his 1998 book Langauge Relationships across Bering Strait, a Proto-Uralo-Siberian phonology which is quite similar to those of Uralic and Eskimo-Aleut, and may be a decent model (though not correct in all details) of the Proto-Mitian phonology. It seems that PIE has innovated to a great degree, including the three grades of stops, the three velar series, and the loss of a full palatal series. And of course ablaut!

But that just begs the question of where the three stop series came from. Ablaut's easy to see as an innovation, given how closely tied to the accent it must have originally been, but the stops are less so.

Salmoneus wrote:when your sound changes are:

*t, *d, *r, *l, *n > *d
and
*ǵ,*g, *gʷ, *ǵʰ,*gʰ,*gʷʰ > *x

...you've got two options:
a) the language underwent some massive mergers that cut across all of MOA, phonation and secondary articulations, and that yielded seemingly random outcomes (a voiced stop at one POA, a voiceless fricative at another)

Reminder: Proto-Uralic *d and *x are not [d] and [x], they are, roughly, "something that merges varyingly with *t, *l and *r" and "something that merges with *k almost everywhere, and is lost in Finnic". The traditional reconstruction is [ð ɣ] (*k also lenites to ⁽*⁾ɣ almost everywhere), but there would be many other options like [d g], or [θ x], or [tɬ q]…

Salmoneus wrote:by default true soundlaws are much less common than borrowings and coincidences

I'm pretty sure that there are way more regular sound correspondences than loanwords between e.g. Tocharian and Latin.

Note also that the assumption that PU has lots of PIE-ish loanwords is one of the main arguments why we assume the two of them to have been in contact at all! Theoretically it would be also possible that they were spoken thousands of kilometers away from each other, maximally something like Anatolia vs. Western Siberia.

Tropylium wrote:

Salmoneus wrote:by default true soundlaws are much less common than borrowings and coincidences

I'm pretty sure that there are way more regular sound correspondences than loanwords between e.g. Tocharian and Latin.

But that's a biased sample - because we know Tocharian and Latin are related!

But the odds of two lanugages being (relevantly closely) related or unrelated is not 50/50. It's hundreds to one (and probably thousands to one before the modern era). So any given resemblence between two languages not provably related is far more likely to be coincidence (or borrowing) than cognacy.

The odds that two languages are genetically related also depend on how much they have in common, and how much the geographical distance is between the places where these languages were spoken. For example, the chances of a relationship between Bantu and Quechua (close to zero) are much lower than between Turkic and Tungusic (>50% in my opinion).

Tropylium wrote:

Salmoneus wrote:when your sound changes are:

*t, *d, *r, *l, *n > *d
and
*ǵ,*g, *gʷ, *ǵʰ,*gʰ,*gʷʰ > *x

...you've got two options:
a) the language underwent some massive mergers that cut across all of MOA, phonation and secondary articulations, and that yielded seemingly random outcomes (a voiced stop at one POA, a voiceless fricative at another)

Reminder: Proto-Uralic *d and *x are not [d] and [x], they are, roughly, "something that merges varyingly with *t, *l and *r" and "something that merges with *k almost everywhere, and is lost in Finnic". The traditional reconstruction is [ð ɣ] (*k also lenites to ⁽*⁾ɣ almost everywhere), but there would be many other options like [d g], or [θ x], or [tɬ q]…

The [ð ɣ] reconstruction is assumed by Fortescue in his 1998 reconstruction of Proto-Uralo-Siberian, and some Nostraticists connect these phonemes to PIE breathy-voiced stops. Yet, we don't know whether this reconstruction is correct or not. The establishment of sound correspondences between Uralic and another family, be it IE or whatever, would help elucidating what these phonemes actually were like.

In the sound substitutions found in the PIE-to-PU loanwords, PU *d occurs (besides *t) as reflex of PIE *d and *dh, so it probably was some sort of voiced coronal stop or fricative (or perhaps a flap), ans *x as reflex of *ǵh and of laryngeals, so it may have been a voiced velar fricative. (More on the PIE laryngeals below.) These observations thus support the traditional reconstruction.

Salmoneus wrote:by default true soundlaws are much less common than borrowings and coincidences

I'm pretty sure that there are way more regular sound correspondences than loanwords between e.g. Tocharian and Latin.

Of course! Tocharian and Latin never had direct contact.

Note also that the assumption that PU has lots of PIE-ish loanwords is one of the main arguments why we assume the two of them to have been in contact at all! Theoretically it would be also possible that they were spoken thousands of kilometers away from each other, maximally something like Anatolia vs. Western Siberia.

Then the resemblances would have to be either actual cognates or Wanderwörter, but the trivial nature of the sound correspondences in the largest part (not all) of the resemblances strongly points at direct borrowings from PIE to PU. Actual cognates would involve complex, non-trivial sound correspondences - if IE and Uralic are indeed related, the relationship would be at least about 8000 years deep reckoned from today, or about 3000 years deep reckoned from the two protolanguages - and even Wanderwörter, i.e. borrowings mediated by some intermediate language or languages, would show less trivial sound correspondences because the phonologies of the intermediate languages would have left their marks on them.

Now to my ideas about the PIE laryngeals I mentioned above. I have arrived at the following:

In PIE3 (post-Anatolian PIE), the PIE laryngeals probably were voiced pharyngeal approximants: */ʕʲ ʕ ʕʷ/. These sounds are the most likely to colour vowels, and the fact that the laryngeals could be syllabic points this way too - they must have had a high degree of sonority, higher than fricatives.

In PIE2 (pre-Anatolian PIE), there may have been five sounds that ended up in the three PIE3 laryngeals. There were three velar fricatives */xʲ x xʷ/, which become pharyngeals in PIE3 and are lost in Hittite; and two glottal fricatives */h hʷ/, which also become pharyngeals in PIE3, merging with the plain and labialized velars, and are preserved as /h/ in Hittite. The high frequency of *h explains why *h2 is more frequent than *h1 in PIE3. So we have:

PIE2 *xʲ > PIE3 *ʕʲ = *h1, Hittite zero
PIE2 *x > PIE3 *ʕ = *h2, Hittite zero
PIE2 *xʷ > PIE3 *ʕʷ = *h3, Hittite zero
PIE2 *h > PIE3 *ʕ = *h2, Hittite h
PIE2 *hʷ > PIE3 *ʕʷ = *h3, Hittite h

In PIE1 (pre-ablaut PIE), the three velar fricatives were simply the fricative members of the three velar series, having split from an original */x/ the same way the velar stops had split in connection with the Great Vowel Collapse (the merger of all PIE0 non-high vowels into PIE1 */a/), preserving old [+front] and [+round] features of adjacent vowels (there apparently were no front rounded vowels in PIE0, as we don't have a labialized palatalized series), while *hʷ was a development of earlier *f, closing a gap in the phonology.

In PIE0 (pre-GVC PIE), there were these fricatives: */f s x h/. Maybe also a palatal */ś/ or so, if PIE0 is the same as Proto-Indo-Uralic.

WeepingElf wrote:PIE2 *xʲ > PIE3 *ʕʲ = *h1, Hittite zero
PIE2 *x > PIE3 *ʕ = *h2, Hittite zero
PIE2 *xʷ > PIE3 *ʕʷ = *h3, Hittite zero
PIE2 *h > PIE3 *ʕ = *h2, Hittite h
PIE2 *hʷ > PIE3 *ʕʷ = *h3, Hittite h

I find it implausible that Anatolian *ḫ, which was clearly a velar/uvular fricative, should reflect glottal fricatives, while the original velar fricatives just... disappear. This is completely separate to my other objection that labialised fricatives should be reflected as illabial *ḫ instead of labialised *ḫʷ, by the way.

WeepingElf wrote:[Yes, we transcribe them as voiced stops. That doesn't mean that they were voiced stops! The most likely hypothesis would be that they were (voiceless) unaspirated stops, while those transcribed as voiceless stops were aspirated. Such transcriptions are common where no actual voiced stops get in the way; it is simply more economic than fiddling with h-digraphs or superscript ^hs. I don't know what the current common opinion among Egyptologists is, though.

Currently there’s still ongoing uncertainty between proponents of an ejective series and an optionally aspirated series (e.g. Loprieno, Schenkel, most of the German ‘neuere Komparatistik’ people) and proponents of a plain voiceless series and an aspirated series (e.g. Peust). So pretty much what Astraios said, although there have been suspicions that the opposition wasn’t voiced–voiceless for as long as that hypothesis has existed; the ‘voiced’ stop symbols d, ḏ, g weren’t meant to indicate voiced sounds to begin with when they were introduced. (q also belongs to the same series, most likely. b, as has been noted, seems to have actually been voiced, but fricativized around the time of the New Kingdom when not in final position.)

dhok wrote:The usual example trotted out is Egyptian, where the Proto-Afroasiatic (or at least Proto-Egyptian+Semitic+Berber) *T' and *D series merged into a single series of stops which we transcribe as voiced stops. (These merged again with the voiceless stops in Coptic, which only has one series of stops.)

At least Bohairic Coptic maintained a distinction between the two series of stops in certain positions (before stressed vowels and sonorants followed by a vowel); some authors think the other dialects also did despite not distinguishing them in writing.

WeepingElf wrote:the trivial nature of the sound correspondences in the largest part (not all) of the resemblances strongly points at direct borrowings from PIE to PU.

Most of the trivially loan-looking cases can be also routed through early Indo-Iranian, early Tocharian or even early Balto-Slavic. This is perhaps the most likely for cases where *ḱ, *ǵ, *ǵʰ are reflected as palatals such as *ś or *j in Uralic, since *ḱ *k are likely to have been more like *[k q] (or at least *[k̟ k̠]) than *[kʲ k].

WeepingElf wrote:In PIE0 (pre-GVC PIE), there were these fricatives: */f s x h/. Maybe also a palatal */ś/ or so, if PIE0 is the same as Proto-Indo-Uralic.

PU "*ś" was probably a palatal stop, not a fricative, for various reasons: the traditionally reconstructed contrast with *ć is poorly supported; these both become *ć in Samic; and in loanwords this corresponds with PII *ć, which we by now know to have been an affricate, not a plain sibilant.

KathTheDragon wrote:

WeepingElf wrote:PIE2 *xʲ > PIE3 *ʕʲ = *h1, Hittite zero
PIE2 *x > PIE3 *ʕ = *h2, Hittite zero
PIE2 *xʷ > PIE3 *ʕʷ = *h3, Hittite zero
PIE2 *h > PIE3 *ʕ = *h2, Hittite h
PIE2 *hʷ > PIE3 *ʕʷ = *h3, Hittite h

I find it implausible that Anatolian *ḫ, which was clearly a velar/uvular fricative, should reflect glottal fricatives, while the original velar fricatives just... disappear. This is completely separate to my other objection that labialised fricatives should be reflected as illabial *ḫ instead of labialised *ḫʷ, by the way.

Yes, this is an objection that needs to be considered. One would guess that glottal fricatives get lost easier than velar ones, but English has lost /x/ and preserved /h/ (though some instances of Middle English /x/ did not go to zero but to /f/). The surviving glottal fricatives could have become /x/ later. Note that there are some 2,000 years between PIE2 and Hittite; a lot of things could happen in that time.

Tropylium wrote:

WeepingElf wrote:the trivial nature of the sound correspondences in the largest part (not all) of the resemblances strongly points at direct borrowings from PIE to PU.

Most of the trivially loan-looking cases can be also routed through early Indo-Iranian, early Tocharian or even early Balto-Slavic. This is perhaps the most likely for cases where *ḱ, *ǵ, *ǵʰ are reflected as palatals such as *ś or *j in Uralic, since *ḱ *k are likely to have been more like *[k q] (or at least *[k̟ k̠]) than *[kʲ k].

While I did mark the front velars with a superscript j in my post about the development of the laryngeals from PIE0 to PIE3, that doesn't mean that I firmly think the difference between front and back velars was one of palatalization; they may indeed have been *[k q], with the labiovelars then *[q^w].

WeepingElf wrote:In PIE0 (pre-GVC PIE), there were these fricatives: */f s x h/. Maybe also a palatal */ś/ or so, if PIE0 is the same as Proto-Indo-Uralic.

PU "*ś" was probably a palatal stop, not a fricative, for various reasons: the traditionally reconstructed contrast with *ć is poorly supported; these both become *ć in Samic; and in loanwords this corresponds with PII *ć, which we by now know to have been an affricate, not a plain sibilant.

Interesting, I did not know that.