The Great Proto-Indo-European Thread

[Soap]

Holzer formulates six sound changes characteristic of Temematic; most notably the PIE voiceless stops merging into the voiced stops, and the PIE breathy-voiced stops becoming (plain) voiceless stops. (The language is named for theses changes: tenuis > media, media aspirata > tenuis.) These changes look a bit weird: one would expect a neutralization of a voicing oppositio in stops to result in voiceless stops, not voiced ones, and the shift from breathy-voiced to plain voiceless doesn't really look as if it could happen in one move. However, I have found a different solution: 1. The feature [+voice] is neutralized, resulting in the voiced (unaspirated) stops merging into the voiceless stops, and the breathy-voiced stops becoming (voiceless) aspirated stops. 2. The resulting unaspirated/aspirated system then shifts to a voiced/voiceless one.

I agree, that sounds like a good explanation.

Still, though, could it just be that Balto-Slavic simply lacked a distinction that Temematic had, and didnt pick it up during loanwords? e.g. maybe the Tememian system was b/p/pʰ, with only a single change from PIE, but the proto-Balto-Slavic people heard /p/ as ⁅b⁆ the way English speakers loaned "goomba" from southern Italian cumpari and prefer to romanize Chinese cities iwth names like "Beijing" even though both stops are voiceless. This is because in English, voiceless stops are aspirated ... strongly at the beginning of a word, but still somewhat in the middle of a word. (voice onset time)

In fact, now that I think of it, if we hypothesize that Greek plain voiceless stops were similarly heard as voiced stops by speakers of Balto-Slavic, maybe Temematic is actually Greek itself, or a closely related macro-Hellenic language that died out.

[KathTheDragon]

A more interesting idea is that rounded labials mostly merged with rounded velars, but occasionally just delabialised. Miguel Carrasquer Vidal brought the idea to my attention. It's one possible explanation of *p ~ *kʷ alternations in IE, such as English oven, wolf and liver. (The initial of the latter is seen as a rare development of *lʲ, with conventionally reconstructed PIE *y being the usual reflex.) He backed up the latter with AA *lib 'heart'. I'm not sure if the curious Egyptian i̓b supports initial *lʲ.

To my knowledge, Germanic is the only branch that shows a labial instead of a labiovelar in 'wolf', which makes it a purely Germanic phenomenon, and thus we should look for a purely Germanic solution (which has been proposed: labiovelars regularly become labials after *w). There's no reason to believe 'liver' to be cognate with *Hyekʷr̥, and I believe that even within Germanic there's a variant *uhnaz for 'oven', showing the reflex of a labiovelar. So while it's a nice idea, it's not necessary.

[Soap]

Swedish has ugn for "oven", which Im told is pronounced with [ŋ]. This could be evidence for a pre-existing plain velar consonant or for the loss of labialization next to /u/.

[KathTheDragon]

It's well known that Germanic delabialised labiovelars early next to *u

[KathTheDragon]

According to Kroonen, Swedish ugn < OSw. ughn, oghn, ofn, omn < PGmc *ufnaz by dissimilation of *f from *u. He claims the same for Gothic auhns, citing auhuma < *ufumô in support. Icelandic has both ofn and ónn, which removes inner-Germanic evidence for old *uhnaz.

The cognates from across IE don't inspire confidence: Hittite ḫapp(e)n-, Greek ἰπνός, Old Prussian wumpnis. This word is most probably an outer-IE loan.

[WeepingElf]

The cognates from across IE don't inspire confidence: Hittite ḫapp(e)n-, Greek ἰπνός, Old Prussian wumpnis. This word is most probably an outer-IE loan.

What do you mean by "outer-IE"? A language belonging to an "outer", i.e. early diverging, now extinct branch of IE (as opposed to "Inner IE", the known non-Anatolian IE languages), or, which is little else than a matter of definition, a language closely related to IE, such as my hypothetical Aquan language? In my work on a dictionary of substratum loanwords in western IE languages, I have met with some items which look somewhat similar to words in other IE languages such as Hittite, Sanskrit or Greek, but not plainly cognate - either the sounds don't really match, or the meaning is too far off, or both. Such cases can IMHO perhaps be explained by borrowing from a related but not ancestral language. The word discussed here may be another example of this.

[hwhatting]

According to Kroonen, Swedish ugn < OSw. ughn, oghn, ofn, omn < PGmc *ufnaz by dissimilation of *f from *u. He claims the same for Gothic auhns, citing auhuma < *ufumô in support. Icelandic has both ofn and ónn, which removes inner-Germanic evidence for old *uhnaz.

The cognates from across IE don't inspire confidence: Hittite ḫapp(e)n-, Greek ἰπνός, Old Prussian wumpnis. This word is most probably an outer-IE loan.

The only word that looks related to the Germanic if we reconstruct a PIE predecessor like *up-(V)n- without outlandish contortions is is the Old Prussian - consonant stems regularly become i-stems in Balto-Slavic, prosthetic /w/ isn't unheard of, and the /m/ may go back to an infixion of the n-suffix like in Lithuanian vanduo "water". Beekes shows evidence that the Greek word was aspirated earlier and quotes a proposal deriving it from the root *sep- with the /i/ coming from "a secondary zero grade" (I'm not sold on that). According to Kloekhorst, the Hittite word goes back to *H3ep-(V)n- and is related to Greek optao: "bake". So it could be that the Greek and Anatolian words are just chance similarities, while the Germanic and Old Prussian could really be related.

[KathTheDragon]

What do you mean by "outer-IE"?

Er, sorry, non-IE.

The only word that looks related to the Germanic if we reconstruct a PIE predecessor like *up-(V)n- without outlandish contortions is is the Old Prussian - consonant stems regularly become i-stems in Balto-Slavic, prosthetic /w/ isn't unheard of, and the /m/ may go back to an infixion of the n-suffix like in Lithuanian vanduo "water".

Yes, I agree with this. But this still leaves the correspondence with a limited distribution, and so still probably a non-IE word.

[Šọ̈́gala]

Still, though, could it just be that Balto-Slavic simply lacked a distinction that Temematic had, and didnt pick it up during loanwords? e.g. maybe the Tememian system was b/p/pʰ, with only a single change from PIE

That sounds plausible. Or maybe some from column A and some from column B. That is, Temematic appears to require three sound changes:

1. dʰ > tʰ ... resulting in t/d/tʰ
2. t > d ... resulting in d/tʰ
3. tʰ (original dʰ per step 1) > t ... resulting in d/t

But the third step is actually unnecessary since there's no reason to think early Balto-Slavic speakers would treated *tʰ as anything other than PBS *t. The second step may or may not be necessary, depending on how we expect they would have treated tenuis stops (although, by default, one assumes PBS *t was tenuis, so *t in a loan would be equated with that).

In fact, now that I think of it, if we hypothesize that Greek plain voiceless stops were similarly heard as voiced stops by speakers of Balto-Slavic, maybe Temematic is actually Greek itself, or a closely related macro-Hellenic language that died out.

Very interesting. Step 1 above is what Temematic has in common with Greek. If that's all that's required for Temematic, then, yeah, maybe that is Macro-Hellenic. This scenario would require that Proto-Balto-Slavic had switched their voiced/voiceless contrast to an aspirated/unaspirated contrast (with voicedness unspecified on unaspirated stops) just like in modern English. I'm not aware of evidence against this off the top of my head, but there might be other loanword data that implies otherwise. Presumably the unaspirateds deriving from PIE "mediae" would have to still be pre-glottalised.

Maybe Proto-Balto-Slavic didn't have tenuis at all. Maybe it was just *tʰ, *d. In that case, perhaps they adopted Temematic *t as a new phoneme, which only later merged with *d. This requires the same sound change as in Step 2 above, but it locates it in the history of Balto-Slavic rather than than Temematic. For all I know, PBS didn't have any "plain" voiceless or voiced stops. Maybe it was exclusively *tʰ, *ˀd, *dʱ. They could potentially do all kinds of crazy things with *t and *d in loans.

[hwhatting]

The only word that looks related to the Germanic if we reconstruct a PIE predecessor like *up-(V)n- without outlandish contortions is is the Old Prussian - consonant stems regularly become i-stems in Balto-Slavic, prosthetic /w/ isn't unheard of, and the /m/ may go back to an infixion of the n-suffix like in Lithuanian vanduo "water".

Yes, I agree with this. But this still leaves the correspondence with a limited distribution, and so still probably a non-IE word.

Or it could be a regional formation from the root *wep- Pok. 1149 "throw, scatter", for the sematics s. Sancr. vapra: fire place. It would't be the only Balto-Slavic-Germanic isogloss.

[Šọ̈́gala]

Eng. "peg" is an interesting word. It's the only inherited (i.e. direct from Proto-Germanic) word in English that I know of which begins with p- and appears to have a solid Indo-European etymology. That is, it appears to continue the elesuive PIE *b-. Wiktionary gives

... from Proto-Germanic *pig-, *pag- (“peg, stake”), from Proto-Indo-European *bak-, *baḱ- (“club, pointed stick, peg”). Cognate with ... Irish bac (“stick, crook”), Latin baculum (“staff”), Latvian bakstît (“to poke”), Ancient Greek βάκτρον (báktron, “staff, walking stick”). Related to beak.

So, great. That's Germanic, Celtic, Italic, Balto-Slavic, and Greek. All the biggest European branches of IE are represented.

Except ... what's with the -g in peg? If it was really *bak-, *baḱ- in PIE, shouldn't Grimm's Law have made it *pah- in Proto-Germanic, resulting in something along the lines of paugh in modern English?

I don't know enough about the development of Germanic vowels to know if the vocalism presents any additional problems. Except, you don't have to be an expert to suspect that it's weird for PIE *a to result in both *i and *a in Proto-Germanic, per Wiktionary's "... from Proto-Germanic *pig-, *pag- ..." We might potentially be left with just p- < PIE *b- as the workable part of the comparison, plus the obvious semantic match. That could just be a coincidence.

Pin (in the sense of a rolling pin) and pen (in the sense of a pig pen, i.e. something fenced in with posts) are an interesting comparison. They apparently go back to Proto-Germanic, and Wiktionary gives a PIE source, but I've never seen much evidence that this root really exists outside of Germanic. "pick" has a similar origin, apparently originally a verb before it was a noun. One might even want to look at "prick", again of uncertain origin before Proto-Germanic. Or, really spitballing here, maybe even "pig", with the semantic connection being that the pig is the "penned-in one".

The distribution of this etymon in Europe only could readily suggest that *bak-, *baḱ- doesn't really go back to PIE per se, but is a loan from a pre-PIE language in Europe. That scenario is highly conducive to handwaving to the effect that any odd alternations can be explained as dialectal variation or opaque morphology in the forgotten source language.

[Znex]

Except ... what's with the -g in peg? If it was really *bak-, *baḱ- in PIE, shouldn't Grimm's Law have made it *pah- in Proto-Germanic, resulting in something along the lines of paugh in modern English?

Verner's Law is the easy answer in my mind. The tonal accent was presumably originally on the case suffix instead of the root, and hence Verner's Law applied.

[Richard W]

Except ... what's with the -g in peg? If it was really *bak-, *baḱ- in PIE, shouldn't Grimm's Law have made it *pah- in Proto-Germanic, resulting in something along the lines of paugh in modern English?

Verner's Law is the easy answer in my mind. The tonal accent was presumably originally on the case suffix instead of the root, and hence Verner's Law applied.

That again would result in something like *pough, cf. lough.

The attestation is not ancient - it appears to be associated with Dutch peg and Middle Dutch pegge.

[hwhatting]

That again would result in something like *pough, cf. lough.

Are you sure? As far as I know, English "gh" goes back to Old English "h", so to un-Vernered PGmc. /x/, not Vernered /ǥ/. I'd expect something like **pay.
As far as I can see, any English word of the shape CVg is suspect of being a loan or some kind of expressive / onomatopoeic formation, as OE *CVǥ is continued either by CVy or by CVw

The attestation is not ancient - it appears to be associated with Dutch peg and Middle Dutch pegge.

You could derive MDu pegge from PIE *bakyó- > PGmc. *paxjá- > WGmc. paǥja-. This again would mean peg is a loan in English, as the expected outcome of paǥja- would be *pedge.

[Sumelic]

That again would result in something like *pough, cf. lough.

Are you sure? As far as I know, English "gh" goes back to Old English "h", so to un-Vernered PGmc. /x/, not Vernered /ǥ/. I'd expect something like **pay.

*ǥ was devoiced word-finally to [h~x] in Old English (like like *ƀ was devoiced word-finally to [f]), except for when it had developed to a palatal approximant [j] (or the geminate palatal [ddʒ]). E.g. Modern English "dwarf" from PG *dwergaz developed through an intermediate form with final /x/. However, I think the devoicing of *ǥ was less consistently represented in writing than the devoicing of *ƀ. It seems the "yogh" symbol (insular g) was often used to represent both voiced and unvoiced sounds. Additionally, many Old English words that ended in devoiced [h~x] had inflected forms with [ɣ], later [w], that are thought to have contributed to their modern English forms. An example of a doublet is the word "enough" and its archaic plural/adverbial form "enow".

Palatalization of velars depended on the identity of the surrounding vowels, and I don't remember the conditions for a-brightening (as in dæġ). I think it can depend on the identity of the vowel in the word's suffix.

[Soap]

Native English words like dog, pig, beg, frog, dig, etc seem by wiktionary to be always from verbs or diminutives. I wouldnt be surprised if the word "big" started out as a diminutive, even, since diminutives don't always indicate things that are small. Wiktionary seems not to believe the theory that spring and frog are from the same root, one with a nasal infix, and the other without.


(edited a bit to touch up the mistypes I made on my phone. "kluge love"? I dont revemember what that was)

[Šọ̈́gala]

Thanks for pointing out Verner's law, I hadn't thought of that.

For purposes of early Germanic and Indo-European comparisons, it's the same difference if the word is inherited per se or is an intra-Germanic loan, e.g. Dutch to English.

[Sumelic]

Native English words like dog, pig, beg, frog, dig, etc seem by wiktionary to be always from verbs or diminutives. I wouldnt be surprised if the word "big" started out as a diminutive, even, since diminutives don't always indicate things that are small. Wiktionary seems not to believe the theory that spring and frog are from the same root, one with a nasal infix, and the other without.

(edited a bit to touch up the mistypes I made on my phone. "kluge love"? I dont revemember what that was)

You might have been meaning to mention "Kluge's law", but it seems that process is only a possible source of voiceless geminates in PG, not voiced ones.

[hwhatting]

@Sumelic: thanks. I had just looked at several cases of English "gh" and seen that they all went back to OE "h"; I had forgotten about the hardening.

[KathTheDragon]

Native English words like dog, pig, beg, frog, dig, etc seem by wiktionary to be always from verbs or diminutives. I wouldnt be surprised if the word "big" started out as a diminutive, even, since diminutives don't always indicate things that are small. Wiktionary seems not to believe the theory that spring and frog are from the same root, one with a nasal infix, and the other without.

(edited a bit to touch up the mistypes I made on my phone. "kluge love"? I dont revemember what that was)

You might have been meaning to mention "Kluge's law", but it seems that process is only a possible source of voiceless geminates in PG, not voiced ones.

However, intra-paradigmatic analogy does seem to have produced geminate fricatives and voiced geminate stops

[Šọ̈́gala]

re: the oft-mooted problem of PIE's apparent stop system (tenuis-voiced-voiced aspirate), which is highly unusual if not just plain unattested, has anyone suggested the possibility that PIE actually had a voiceless aspirate series that systematically spirantised shortly before the spread and diversification of Indo-European? Personally, I'm pretty comfortable with the Cao Bang/Kümmel model (t, ɗ, d > t, d, dʱ), but I've been toying with this alternative idea.

If voiceless aspirates spirantised, what was the result? The obvious thing to consider would be the laryngeals. Also possible would be that they merged with other existing non-stop phonemes or that they went to ∅.

pʰ > ɸ > ʍ > w (or, alternatively, ɸ > h~ʔ = h₁; or ɸ > ∅)

tʰ > sʰ > h~ʔ = h₁

For h₂ and h₃, some combination of dorsals. I could easily imagine kʰ > x > X = h₂, with secondary articulations neutralising after the sound is no longer a stop, so that *k̂ʰ and *kʷʰ merge with *kʰ. The sticky wicket is h₃. I mean, *kʷʰ is the obvious option if you believe that h₃ was a rounded fricative/approximant. But I tend to think that h₃ was distinguished by backness and/or voicing compared to h₂. I have a vague impression (but no data) that, in post-velar positions, voicing can tend to become unspecified. So, maybe if so-called *k was really a uvular, spirantisation started as qʰ > X, and then the fricative subsequently began to be realised as voiced. There may or may not have been a chain shift pushing *x (early h₂) and *X~*ʁ (early h₃) both into articulations further back; or there may or may not have been leveling of both to uvular articulation with voicing as the only distinction.

It would be tempting to get *b involved in this, since we're looking for a source for a voiced laryngeal. But I struggle with both a) why *b alone among the plain voiced sounds (and alone among voiced bilabial stops) would spirantise (remember that, in this model, the so-called plain voiced series was in fact plain voiced from early on); and b) how *b could fit in with what we think the laryngeals might have sounded like. Regarding the latter, maybe a debuccalisation like b > β > β̞ > ɦ? I'd like this better if there were precedent for ɦ spontaneously fortiting to ʁ~ʕ~ʢ; or precedent for ɦ strongly coloring adjacent vowels.

I guess you could combine this approach with the Cao Bang model. After all, the actual Cao Bang language does have both voiceless and voiced aspirates. So you could imagine Early PIE with tenuis-implosive-voiced, same as Proto-Zhuang-Tai; and then mid-PIE with the full four-way contrast, same as Sanskrit and Cao Bang; and then late PIE with the classic weird tenuis-voiced-mediae aspiratae system. This is not very satisfying, though, because this "late PIE" would still have to be before Hittite and Tocharian split off. We're really talking something more like "Early Early PIE", "Mid Early PIE", "Late Early PIE".\

P.S. Why won't chi render properly? I had to substitute capital X for chi above.

[KathTheDragon]

h₃ was distinguished by backness and/or voicing compared to h₂.

There's no evidence for voicing in *h₃ beside *pibh₃eti, which can be alternately explained anyway (i.e. initial devoicing of *b)

[Šọ̈́gala]

The only case that I've found where there are different grades of vowel coloring from different consonants (leaving aside palatal and rounding assimilation) is in modern Danish, where the backmost coloring results from /ʁ/ (orthographic "r"). However, there's no /X/ in Danish to compare it to as far as what coloring that would result in. That's why I say backness and/or voicing. It's not much to go on, just a guess from very limited data. In the Danish example, voiced velars have the same effect that voiceless velars do, so voicing by itself doesn't seem to be decisive. It might require postvelar place and voicing together for all I know.

Fair enough, though. If this tenuous Danish analogy plus pibh₃eti is all there is to go on, then maybe h₂ is a uvular fricative (of uncertain voicing) and h₃ is a pharyngeal or epiglottal fricative (of uncertain voicing).

[Nortaneous]

aren't *-h2- *-h3- sometimes reflected in Anatolian as -hh- and -h-?

[hwhatting]

aren't *-h2- *-h3- sometimes reflected in Anatolian as -hh- and -h-?

Not according to Kloekhorst (Hittite Etym. Dict., pp 94 - 97); H3,depending on its position, either merges with H1 (as glottal stop or zero) or with H2. Also, intervocalic "h" and "hh" always go back to H2 (H1 and H3 > glottal stop), with "hh" the regular outcome and "h" due to lenition. I don't know what the current thinking of other Anatolian scholars (e.g Melchert) is on this matter.