zompist bboard

Well we have a Vasco-Caucasian thread, and a Ural-Altaic thread, so why not an Indo-Uralic thread? The Indo-Uralic hypothesis (with or without the wider Nostratic hypothesis) is nothing new (been around for decades) but not accepted by mainstream linguists. However, it's not a crackpot idea either. It's been/is supported by such mainstream linguists as Kortlandt and Pederson. Even splitters like Campbell and Beekes admit that Uralic is the best candidate for a family related to Indo-European (though they remain unconvinced).

My primary purpose in this thread is to discuss my ideas about Indo-Uralic and get some useful feedback, but I'm also very interested other people's ideas. For instance, I recently read an article by Adam Hyllested that caused me to completely revise my opinions about Indo-Uralic laryngeals. My preference is to avoid discussion about the wider question of Proto-Nostratic, but I realize that sometimes Nostratic can shed light upon questions of Indo-Uralic.

In particular one of my goals here is the discovery of regular sound correspondences. I think one of the roadblocks to wider acceptance of Indo-Uralic is that traditionally Proto-Indo-Uralic is reconstructed with a large number of phonemes and the two daughter languages under massive simplifications. For example, PU *k would correspond with PIE *k, *k^, *kw, *g, *g^, *gw, *gh, *g^h, *gwh. Despite the fact that we see basically the same thing in Tocharian and Hittite when compared to PIE, it's no wonder that linguists by and large reject Indo-Uralic. On the flip side PU has 8 vowels in the standard reconstruction which all correspond to PIE *e/o.

As it happens I don't think the situation is as bad as that. I don't think the Proto-Indo-Uralic phoneme inventory was especially complicated. Instead the complicated PIE consonant system is largely the result of secondary developments.

In the next message I'll start talking about some sound correspondences.

OK, then I'll add my own mustard to this thread.

In my opinion, the similarities between Indo-European and Uralic, in terms of morphology, are too obvious to dismiss them as chance resemblance. I am pretty certain that the two families (and maybe others, such as Yukaghir, Chukotko-Kamchatkan and Eskimo-Aleut, which all appear to be related to Uralic, and perhaps Etruscan and other extinct languages on the IE side) are related to each other. I am less certain about Altaic being related to IE and Uralic.

I haven't yet managed to find out much about sound correspondences and the Proto-Indo-Uralic (PIU) antecedents, but there are instances where PIE *s corresponds to PU *t (as in the plural endings), and in the same way, PIE laryngeals (which I assume to have been velar fricatives) correspond to PU *k. PIU had three grades of stops, probably neutral, glottalized and voiced. Neutral stops are preserved in both families. Glottalized stops became PIE voiced stops and PU geminates. Voiced stops became PIE voiced aspirated stops and PU voiced fricatives. However, PU neutralized these distinctions in many positions. All these hypotheses are preliminary.

The threefold IE distinction between plain, palatalized and labialized velars are in my opinion an innovation of IE (actually already Europic, the Neolithic forerunner of IE), wherein the secondary articulations of the velars preserve features of neighbouring vowels. This ties in with another innovation of Europic, which I call the Great Vowel Collapse (GVC). This is a change under which all non-high vowels merged into a single vowel *a, which evolved into the ablauting vowels in PIE. Before the GVC, high vowels were lowered before resonants (this I call Resonant-Conditioned Lowering (RCL). This change is necessary to account for the non-occurence of high vowels (which manifest as *ei~oi~i and *eu~ou~u ablaut sets in PIE) before resonants in PIE. As an example, *kul- 'to turn' became *kol- by RCL, then *kWal- by GVC, and finally *kWel- in PIE.

I look forward to open-minded discussion.

WeepingElf wrote: I haven't yet managed to find out much about sound correspondences and the Proto-Indo-Uralic (PIU) antecedents, but there are instances where PIE *s corresponds to PU *t (as in the plural endings),

Kortland has proposed a law *ti > *si in PIE. This would explain the lack of any PIE roots of the form *tei. There is one regional root *tei which is a locative particle, but it might be a compound *te-i. Starostin lists two other counter-examples but one is clear onomatopoeia and the other is a Baltic-Slavic isogloss. This law also could explain the *s/*t hesitation found in 2nd person verbal suffixes and demonstrative pronouns.

That said, it's not clear to me that the Uralic plural *-t is cognate to the IE plural *-s. The difference between nominative singular and nominative plural is not the addition of *-s, but the addition of *-e- which might be cognate with Uralic plural *-j-. I propose that this plural marker is also found in the genitive plural (but in the o-grade due to the following nasal). The use of *-(e)s as a plural may be a later innovation based on the nominative plural.

and in the same way, PIE laryngeals (which I assume to have been velar fricatives) correspond to PU *k.

I think this correspondence is very important (it's the one I briefly allude to in the OP). Adam Hyllested provides a number of examples,

http://www.nostratic.ru/books/(173)Hyll ... ngeals.pdf

PIU had three grades of stops, probably neutral, glottalized and voiced. Neutral stops are preserved in both families. Glottalized stops became PIE voiced stops and PU geminates.

Do you have any examples? I can think of some examples that contradict this:

PIE *wod "water" , PU *weti "water"
PIE *h2ek "sharp point", PU *kokka "tip, protruding edge, hook"

Voiced stops became PIE voiced aspirated stops and PU voiced fricatives. However, PU neutralized these distinctions in many positions. All these hypotheses are preliminary.

I'll get into this in another message. I think the voiced aspirated stops are, in most cases, derived from plain voiceless stops.

The threefold IE distinction between plain, palatalized and labialized velars are in my opinion an innovation of IE (actually already Europic, the Neolithic forerunner of IE), wherein the secondary articulations of the velars preserve features of neighbouring vowels.

I agree, and this actually ties in with my hypothesis about the development of voiced aspirated stops.

This ties in with another innovation of Europic, which I call the Great Vowel Collapse (GVC). This is a change under which all non-high vowels merged into a single vowel *a, which evolved into the ablauting vowels in PIE. Before the GVC, high vowels were lowered before resonants (this I call Resonant-Conditioned Lowering (RCL). This change is necessary to account for the non-occurence of high vowels (which manifest as *ei~oi~i and *eu~ou~u ablaut sets in PIE) before resonants in PIE. As an example, *kul- 'to turn' became *kol- by RCL, then *kWal- by GVC, and finally *kWel- in PIE.

This is actually quite close to my own view, though I'd merge them into a schwa instead of *a. I think we may have discussed RCL a few years ago on Cybalist.

In this message I'd like to present some evidence that PIE *bh and *dh were, usually, secondarly derived from PIU *p and *t. The specific conditioning environment was the presence of *w, *m, and rounded vowels (unless a palatal consonant is also present). This may seem like a phonetically odd development but I believe that they passed through an intermediate stage of *pw and *tw. This then becomes unified with the development of *kw from PIU *k. Development of PIE voiced aspirate velars (including the labialized and palatalized variants) is quite different, having their origin in intervocalic PIU *k (> *G) and PIU *G. Direct evidence for *G, however, has to come from other Nostratic languages.

After *pw, *tw, *G > *bh, *dh, *g(^,w)h there was a tertiary development of aspiration harmony. So is some supporting evidence.

1) PU *p ~ PIE *bh

PIE *bheg "to split, cut, break apart", PU *pakka "burst, rend, split"
PIE *bher "good, valiant", PU *para "good"
PIE *bhor "to bore, pierce", PU *pura "to bore"

2) PU *p ~ PIE *p

PIE *pel "be afraid, tremble, shake, frightened", PU *peli "to be afraid"
PIE *peri "around", PU *piri "ring, circle, to surround" (Uralic may have a laryngeal in this root, in which case this correspondence is not valid)
PIE *peh2wor "fire", PU *päjwa "heat, warm" (this is one of a number of examples of PU *j corresponding to a laryngeal in PIE)
PIE *peh3l "thumb (Slavic-Italic isogloss), PU * pälka "thumb"
PIE *pelH "many, fill", PU *paljV "thick, many"

3) PU *t ~ PIE *dh

PIE *wedh "lead, carry, bring", PU *weta "take, guide, lead, carry"
PIE *wendh "short hair, fuzz, beard, moustache", PU *wuntV "facial hair" (Uralic reconstruction somewhat uncertain)
PIE *medhu "honey, mead", PU *meti "honey, mead" (suspected Wanderwort)
PIE *lendh "unclutivated land", PU *lamti "low, lowland"
PIE *dheub "deep", PU *toxi "sea, pond"
PIE *dheghom "earth", PU *taxV/*takV "place, site" (Khanty-Hungarian isogloss)

4) PU *t ~ PIE *t

PIE *to demonstrative prn., PU *tä demonstrative prn.
PIE *t- "you", PU *te "you"
PIE *gwet "say, speak, call" (German-Armenian isogloss), PU *kutji "to call" (following Bomhard I interpret Uralic *c' as /t_j/ or perhaps /tj/).
PIE *-nt participle suffix, PU *-nt present participle suffix
PIE *tek "form, fashion, make", PU *te^ki "do, make"

An interesting blog post I found:

http://paleoglot.blogspot.com/2009/10/p ... teppe.html

TaylorS wrote:An interesting blog post I found:

http://paleoglot.blogspot.com/2009/10/p ... teppe.html

If we accept Glen's idea of *t# > *s# in IE then this opens up the possibility of reanalyzing the PIE nominative *-s as *-t instead. This could, in turn, be derived from a demonstrative.

This all looks quite interesting. Is there much agreement on the reconstruction of PU. If so where can I find details, e.g. phonology?

marconatrix wrote:This all looks quite interesting. Is there much agreement on the reconstruction of PU. If so where can I find details, e.g. phonology?

(1) Abondolo, Daniel (ed.): The Uralic Languages. London/NY: Routledge, 1998. [Part of the Routledge Language Family series]

(2) Various things by Björn Collinder, but probably out of print.

(3) Rédei, Károly: Uralisches etymologisches Wörterbuch (5 fascicules, in German) Budapest: Akadémiai kiadó, 1986.

I am not aware of much else available that is not in Finnish or Hungarian.

The Wikipedia may not be a bad place to start on the basics of Uralic phonology. On the whole, most Uraliscists agree on the basics.

Etherman wrote:

WeepingElf wrote: I haven't yet managed to find out much about sound correspondences and the Proto-Indo-Uralic (PIU) antecedents, but there are instances where PIE *s corresponds to PU *t (as in the plural endings),

Kortland has proposed a law *ti > *si in PIE. This would explain the lack of any PIE roots of the form *tei. There is one regional root *tei which is a locative particle, but it might be a compound *te-i. Starostin lists two other counter-examples but one is clear onomatopoeia and the other is a Baltic-Slavic isogloss. This law also could explain the *s/*t hesitation found in 2nd person verbal suffixes and demonstrative pronouns.

That said, it's not clear to me that the Uralic plural *-t is cognate to the IE plural *-s. The difference between nominative singular and nominative plural is not the addition of *-s, but the addition of *-e- which might be cognate with Uralic plural *-j-. I propose that this plural marker is also found in the genitive plural (but in the o-grade due to the following nasal). The use of *-(e)s as a plural may be a later innovation based on the nominative plural.

Maybe. The *ti > *si thing makes sense. Comparisons of plural markers between IE and Uralic are difficult because in IE, plural case markers cannot be broken up into a case morpheme and a plural morpheme.

Etherman wrote:

and in the same way, PIE laryngeals (which I assume to have been velar fricatives) correspond to PU *k.

I think this correspondence is very important (it's the one I briefly allude to in the OP). Adam Hyllested provides a number of examples,

http://www.nostratic.ru/books/(173)Hyll ... ngeals.pdf

Yes.

Etherman wrote:

PIU had three grades of stops, probably neutral, glottalized and voiced. Neutral stops are preserved in both families. Glottalized stops became PIE voiced stops and PU geminates.

Do you have any examples? I can think of some examples that contradict this:

PIE *wod "water" , PU *weti "water"
PIE *h2ek "sharp point", PU *kokka "tip, protruding edge, hook"

Indeed, I am not sure about those correspondences. I currently don't have examples at hand. But Uralic appears to have neutralized voice opposition in initial position but retained them intervocally. I still have to do more research in this matter.

Etherman wrote:

Voiced stops became PIE voiced aspirated stops and PU voiced fricatives. However, PU neutralized these distinctions in many positions. All these hypotheses are preliminary.

I'll get into this in another message. I think the voiced aspirated stops are, in most cases, derived from plain voiceless stops.

The threefold IE distinction between plain, palatalized and labialized velars are in my opinion an innovation of IE (actually already Europic, the Neolithic forerunner of IE), wherein the secondary articulations of the velars preserve features of neighbouring vowels.

I agree, and this actually ties in with my hypothesis about the development of voiced aspirated stops.

This ties in with another innovation of Europic, which I call the Great Vowel Collapse (GVC). This is a change under which all non-high vowels merged into a single vowel *a, which evolved into the ablauting vowels in PIE. Before the GVC, high vowels were lowered before resonants (this I call Resonant-Conditioned Lowering (RCL). This change is necessary to account for the non-occurence of high vowels (which manifest as *ei~oi~i and *eu~ou~u ablaut sets in PIE) before resonants in PIE. As an example, *kul- 'to turn' became *kol- by RCL, then *kWal- by GVC, and finally *kWel- in PIE.

This is actually quite close to my own view, though I'd merge them into a schwa instead of *a. I think we may have discussed RCL a few years ago on Cybalist.

I think I remember such a discussion on Nostratic-L, but I am not sure.

marconatrix wrote:This all looks quite interesting. Is there much agreement on the reconstruction of PU. If so where can I find details, e.g. phonology?

Much literature on Uralic is unfortunately in Uralic languages - apparently, if a Uralicist book is published in Finnish or Hungarian, there is hardly a market for an English translation, as most Uralicists have learned those languages anyway. And what is not in Uralic languages is often in Russian.

But a good summary in English of Proto-Uralic is found in D. Sinor (ed.), The Uralic Languages, published as part of the Handbuch der Orientalistik series. For vocabulary, see K. Redei, Uralisches Etymologisches Wörterbuch, which, however, is in German, and uses a slightly different reconstruction than that mentioned above.

However, PU is not as firmly and detailedly reconstructed as PIE; there are still many controversial matters. Proto-Uralic reconstruction is more difficult than Indo-European because the family, while having about the same time depth, lacks ancient literary languages which in IE bridge about half the time elapsed since the breakup of PIE.

Etherman, do you have any vowel/final/velar/laryngeal correspondences?

marconatrix wrote:This all looks quite interesting. Is there much agreement on the reconstruction of PU. If so where can I find details, e.g. phonology?

For the most part yes. The interpretation of some of the consonants is open however. The biggest difficulty is reconstructing the vowels. Here's a pretty good document with a non-standard approach to the vowels (though it doesn't stray too far from the standard reconstruction).

https://oa.doria.fi/bitstream/handle/10 ... sequence=1

Most of it is in Finnish, but the reconstructions are given in English. Here's a short document on Proto-Uralic cases:

http://www.mv.helsinki.fi/home/jylikosk ... ffixes.pdf

Daquarious P. McFizzle wrote:Etherman, do you have any vowel/final/velar/laryngeal correspondences?

As a general rule all Uralic vowels correspond to *e/o in IE. This means we have to heavily rely on PU for the reconstruction of vowels, and even then Proto-Uralic vowel harmony has blurred most of the distinctions in non-initial syllable vowels. Fortunately, traces have been left behind in IE. Rounded vowels, or at least high rounded vowels, labialized preceeding consonants (see my previous message) in PIE. *u > *o before resonants.

PIE *bhor "to bore", PU *pura "to bore"
PIE *mor "blackberry, mulberry", PU *mura "cloudberry"

Unfortunately it's difficult coming up with solid examples because of IE ablaut. The argument is mostly distributional since PIE doesn't have sequences of *uR, unless R is part of a suffix.

PU *o before non-resonants corresponds to *eu/u in PIE.

PIE *dheub "deep", PU *toxi "sea, pond"
PIE *geup "cave, mine", PU *koppa "hollow, cavity"
PIE *sreu "to flow", PU *c'orV "to flow, drop, overflow" (I assume PIE retained initial clusters so the *o was not originally before a resonant)
PIE *bheug "to run, escape", PU *pukta "to jump, run" (supposedly a derived form from *poki-ta but I don't know the justification for this)

PU *i before non-resonants corresponds to *ei/i in PIE

PIE *neig^h "to bore, stab", PU *nikka "to stick in, push"
These might be derived from a root *ni, c.f.
PIE *neid "to abuse", *ne:ik "to assault"

PIE *leikw "to leave", PU *likki "to move" (note that the second vowel is either *u or *ü because PIE preserved the rounded feature and PU preserve the high feature)
PIE *meiH "to change, exchange", PU *mexi "to give, sell"

I presume PIU *ü > PPIE *i with compensatory labialization, which then undergoes the same evolution as PIU *i.

The evolution of PIU *i/ü in PIE is probably a bit more complicated than that. It appears that in monosyllables it remains *i (c.f. *kwis, *kwim, *kwid). In polysyllabic words it is *ei/i in open syllables not followed by a resonant (the ablaut grade being determined by stress), c.f. *kweies, *keisu. In polysyllabic words it is *e in closed syllables or open syllables followed by a resonant, c.f. *kwesyo, *kwesmi.

In some cases PU *a (and even other vowels) correspnds to PIE *a. The data is pretty inconsistent though so I'm not sure what to make of it.

PIE *(s)kwalo "big fish", PU *kala "fish" (but also *kulta "to fish")
PIE *kant(h) "rim", PU *kanta "edge, shore"
PIE *salHk(^) "willow", PU *s'ala "elm"
PIE *kat "hut, shed" PU *kota "hut, hovel, house"
PIE *kwas "basket" (but Baltic is in e-grade), PU *koc'a "basket"

There are a small number of examples where PU *ä corresponds to PIE *ei, which is counter to the general rule that PU better preserves the vowels. In particular PIU *eiC (or more likely *ejC) > PU *äC, PIE *eiC. PU doesn't have any reliable etymologies with the sequence *ejC.

PIE *peis "to pound, thresh", PU "päc^ka "to cut"
PIE *deik^ "to show", PU *täkkV "to look, observe" (Hungarian-Khanty isogloss)
PIE *leit(h) "to leave, die", PU *läkti "to go away" (here the PU *k corresponds to a PIE laryngeal which causes aspiration on PIE *t)
PIE *weiH "strength, power", PU *wäki "strength, power"

Unfortunately, some of the these PIE-PU correspondences might reflect cross-borrowings (usually from IE into PU) and even (substrate) loanwords from a third party.

For example, IE *salHk(^)- 'willow' (a regional NW word) looks like a substrate loanword (even Matasovic recognizes this!) connected to Tungus-Manchu *dʒali-kta 'hawthorn, willow' and Proto-Adyghe-Kabardian *(p)ttsaɮǝ 'willow'.

This would undermine any Indo-Uralic theory.

Etherman wrote:Well we have a Vasco-Caucasian thread, and a Ural-Altaic thread, so why not an Indo-Uralic thread? The Indo-Uralic hypothesis (with or without the wider Nostratic hypothesis) is nothing new (been around for decades) but not accepted by mainstream linguists. However, it's not a crackpot idea either. It's been/is supported by such mainstream linguists as Kortlandt and Pederson.

You said: 1) it's not accepted by mainstream linguists and 2) it's been/is supported by such mainstream linguists as Kortlandt and Pederson.

A nice contradiction!!!

Octaviano wrote:

Etherman wrote:Well we have a Vasco-Caucasian thread, and a Ural-Altaic thread, so why not an Indo-Uralic thread? The Indo-Uralic hypothesis (with or without the wider Nostratic hypothesis) is nothing new (been around for decades) but not accepted by mainstream linguists. However, it's not a crackpot idea either. It's been/is supported by such mainstream linguists as Kortlandt and Pederson.

You said: 1) it's not accepted by mainstream linguists and 2) it's been/is supported by such mainstream linguists as Kortlandt and Pederson.

A nice contradiction!!!

that's not a contradiction.

'accepted by mainstream linguists' here take them as a sort of collective, and something is accepted by this collective when there's a relatively strong consensus on it.

your crackpottiness is showing through even more, when you're twisting words like that to bring contradictions out of it. plz, stop.

Miekko wrote:your crackpottiness is showing through even more, when you're twisting words like that to bring contradictions out of it. plz, stop.

Should I take this as ad homimem attack?

Octaviano wrote:

Miekko wrote:your crackpottiness is showing through even more, when you're twisting words like that to bring contradictions out of it. plz, stop.

Should I take this as ad homimem attack?

actually, yeah, but I only do that when someone makes a really useless comment based on an overliteral interpretation of what someone says that tries to earn them points by criticizing others for statements that aren't really wrong.

I see you didn't get my irony. Actually, I made an apparently "useless comment" about statements 1) and 2) in order to draw attention to the statement between these two: "However, it's not a crackpot idea either."

Now you know my opinion about the Indo-Uralic theory.

Octaviano wrote:I see you didn't get my irony. Actually, I made an apparently "useless comment" about statements 1) and 2) in order to draw attention to the statement between these two: "However, it's not a crackpot idea either."

oh, I got that as well, and no, it's not a very clever way to present anything, and what you try to say with isn't very clever.

Look, I didn't want to openly say "IMHO, this is a crackpot theory" but this exactly what I think.

Octaviano wrote:Look, I didn't want to openly say "IMHO, this is a crackpot theory" but this exactly what I think.

It's got a lot more evidence for it than your nonsense.

Octaviano wrote:Look, I didn't want to openly say "IMHO, this is a crackpot theory" but this exactly what I think.

So, is this irony, tragedy, or both?

Daquarious P. McFizzle wrote:Etherman, do you have any vowel/final/velar/laryngeal correspondences?

The relationship between velars and laryngeals is complicated. First there is the PIE side which is generally reconstructed with 3 series of velars (velar, palatalized, and labialized). The plain velars are rare and usually occur only in positions where one or both of the other velars don't. In a number of instances the satem languages are inconsistent in regards to whether a plain velar or palatalized velar should occur. Most likely there were only palatalized and labialized velars, with the plain velars being secondarily derived. This may seem typologically unacceptable, but such systems appear in Northwest Caucasian (there's evidence for a substantial amount of contact between IE languages and NWC languages). For PIE 3 or 4 laryngeals are usually reconstructed. Their phonetic realizations are debated and uncertain. *h4 is usually reconstructed to account for those situations in Hittie where *h2 is expected but not found. Also *h4 > h in Albanian, but the evidence is weak. Personally I'm okay without reconstructing *h4 because I'm not married to the theory that all roots are of the form CVC (in fact it's contradicted by massive amounts of evidence) and I think PIE had a *a not derivable from laryngeal effects.

PU has two velars *k and *x. Like PIE, the PU laryngeal *x has uncertain phonetics. IIRC the consensus is that it's a velar glide, but I suppose it could also be a velar fricative. *x has an odd distribution. It occurs only before the second vowel of the root (with a possible intervening consonant), and only if that vowel is [+high]. I think it derives from several different consonants, but most commonly *k. After vowels we find *ka but not *ki (there are some exceptions, for example after *ä <ej> *xji
*psi > *xsi
*jmi > *xmi
*pd'i > *xd'i
*jni > *xni

Note that in all these cases except the last the original cluster is attested before *a but not before *i. The last cluster doesn't appear before either *a or *i, but all attested clusters before *a also occur before *i, so I assume the change based on *jmi > *xmi. I haven't even begun to investigate correspondences between these clusters and PIE clusters. There also appear to be some *xti, *wxi, and *jxi clusters, which I haven't investigated yet. To the best of my recollection *wxi appears only after *o^ and *jxi appears only after *e^.

After that long preliminary discussion we can talk velars and laryngeals in PIU. PIE laryngeals correspond to PU *k (or *x <k> *mk > *mp. PIE velars correspond to PU velars (with the vowel determining whether the velar is palatalized or labialized).

PIE *g(^,w)h appears to be secondarily derived in most cases. It corresponds to PU intervocalic *k (not < *H). I've only found one exception, PIE *tek "to make, fashion", PU *te^ki "to do, make", which curiously is also an exception to the PU rule that *Vki > *Vxi. This PU vowel *e^ (as well as *o^) tends to cause exceptions to rules. If it's/they're some kind of original diphthong that might make all the exceptions regular. Also, this rule does not apply to original *g (> *k in PU). As Weeping Elf mentions intervocalic voiced stops correspond to gemminate stops in PU. There are some exceptions, but I think that means we need to investigate this more. The rule may only apply to *g, which appears to be exceptionless as far as I can tell.

PIE *g(^,w)h also appears to derive from intervocalic *N /N/ > *ng(^,w)h in PIE. In other positions this *N just assimilates in place of articulation to whatever the following consonant is.

There are some examples where in word initial position PIE *g(^,w)h corresponds to PU *k. There are no obvious conditioning environments from which to derive this from PIU *k. From evidence outside of Indo-Uralic (especially Kartvelian) I take these to be examples of an original *G. Something I just noticed is that this initial *G appears only before back PU vowels (note that *k can occur before front and back vowels). Hyllested has proposed a correspondence between PIE *g^h and PU *j. He gives seven examples in initial position and in five of them Uralic has a front vowel. Of the two exceptions one has a very messy Uralic reconstruction and the other (PU *juki "to drink" PIE *g^heu "to pour") may be related to PIE *h2ekw "water" instead.

So I now propose that PIU *G > PIE *g(^,w)h, PU *j before front vowels *k before back vowels. In non-initial position it's not always possible to distinguish PIU *k from *G. This is similar to Hyllested's *gj, for which he proposes special developments after nasals, but his correspondences are inconsistent.

As far as the PIE *H ~ PU *j correspondence goes I assume this goes back to PIU *j. The details aren't entirely clear since this *H can represent any of *h1, *h2, or *h3. It may have to do with the quality of the following vowel. More research is needed here.

It appears that PIU laryngeals are reflected as *k in Uralic, at least in non-initial position. In initial position Bomhard reconstructs them as *0 in PU whereas Hyllested reconstructs them as *k. These two views may not be as far off as it first appears. Bomhard's examples all have either *a, *ä, or *u (I discount *ema "mother" as a nursury word and another which is an exclamation, leaving the only exception as *e demonstrative prn. which may not be directly related to the PIE *h1ei). In Hyllested's list we have 3 roots with *h1 appearing before *i, 7 roots with *h2 before *a, *ä, or *u, and 9 roots with *h3 before *ü, *o, or a consonant (the cluster is broken up in PU). The only time there's an actual clash between Bomhard and Hyllested is when the vowel is *a, *ä, or *u. This seems to require 4 laryngeals. 3 for Bomhard's laryngeals and 1 for Hyllested's laryngeal. It's strange that no laryngeal is found before *e (except perhaps PIE *h1ei ~ PU *e). Bomhard's laryngeals probably have different reflexes before *i, *o, and *ü which need to be investigated. Perhaps they also merge with PU *k.

I'm running short on time so I can't get into explicit details, but I urge everyone to read Hyllested's and Häkkinen's papers which I linked earlier.

Octaviano wrote:Unfortunately, some of the these PIE-PU correspondences might reflect cross-borrowings (usually from IE into PU) and even (substrate) loanwords from a third party.

That's always a danger. That's why it's important to establish regular correspondences (something you've been loathe to do).

For example, IE *salHk(^)- 'willow' (a regional NW word) looks like a substrate loanword (even Matasovic recognizes this!) connected to Tungus-Manchu *dʒali-kta 'hawthorn, willow'

Alternatively this could be a Nostratic root reflected in PIE, PU, and PA.

and Proto-Adyghe-Kabardian *(p)ttsaɮǝ 'willow'.

Maybe a borrowing into PAK.

This would undermine any Indo-Uralic theory.

Given the regular correspondences in lexical roots, grammatical words, and morphology a few borrowings won't hurt anything.