The Great Proto-Indo-European Thread's Sequel

[Richard W]

What bothers me a little is how odd the system looks.
Is *a attested at all, aside from laryngeal influence? Or is it assumed to have existed, but beyond any possible reconstruction?

It makes most sense to see the ablauting short vowel *e as [ɐ] with a contrasting rare [a] somehow developing - possibly from loans, possibly from rare internal occurrences. It's not unlike English vowel of wolf, book and sugar which is assembled from various odd sources. Some people refuse to believe in *a. It's almost marginal.

There seems to be a contrast between *ka and *k̂e (with the first consonant being the plain velar and the second the one that softens), though this could be another case of a conditioned change from *e to *a.

For that matter, why are *i *u syllabic *y *u and not *y *w allophones of the vowel? (For that matter, is there any language described that way?)
Wouldn't it make sense to see zero grade as base forms, say *i, *u with *e added? (Or with *i *u lengthened or something and then breaking into diphthongs?)

There's a contrast between *we as in swépnos and *ew as in *dʰewbʰnós. It's more traditional to write the latter as *dʰeubʰnós. Now, if one takes the zero grade as basic, then one gets the phenomenon called samprasāraṇa in which the *e follows rather than preceding the zero grade vowel. Sometimes both full forms exist - this phenomenon is known as schwebeablaut.

[Talskubilos]

[*] An overview of the modern PIE literature showing that the vast majority of ‘PIE’ words are found only in one branch;

And thus excluding borrowing between different branches? No way!

[*] A demonstration that the various IE branches are not linked through regular sound changes, and thus cannot be descended from a single protolanguage

Sound correspondences might be different depending on the source. For example, some words In Germanic have reduced labiovelars while others do not.

[Talskubilos]

So, since your claim is that reconstructed PIE is not the ancestor of all attested IE langs, I'd want to see a substantiated argument about what makes it impossible to derive all attested IE langs from a singular parent language.

I'd put it that way instead: it's impossible to derive all the attested lexical items in IE langs from a single parent language.

[Talskubilos]

Show us that you are not bluffing, and that your model explains the facts better than the standard model.

I think I'd rather present you the facts and let you draw the conclusions. For example, the 'summer' word I mentioned before has a doublet in Greek hêméra 'day' < *h2ēm-ər- 'heat (of the day)', with correspondences in Semitic *ħamm- 'to be hot; warm', where the voiceless pharyngeal fricative ħ corresponds to the "laryngeal" h2.

I could quote other correspondences if you want.

[Talskubilos]

Bjørn 2017 is an overview recent enough to have an associated blog

Thank you very. Contrarily to WeepingElf, Bjørn proposed a substrate language akin to Semitic as the source of agricultural loanwords.

[Richard W]

So, since your claim is that reconstructed PIE is not the ancestor of all attested IE langs, I'd want to see a substantiated argument about what makes it impossible to derive all attested IE langs from a singular parent language.

I'd put it that way instead: it's impossible to derive all the attested lexical items in IE langs from a single parent language.

What do you mean by an 'attested lexical item'? Does English television count as one? Does seraph? Are you more concerned with words such as inter-Germanic loan English score?

[KathTheDragon]

Greek hêméra 'day' < *h2ēm-ər- 'heat (of the day)', with correspondences in Semitic *ħamm- 'to be hot; warm'

You'll have to do better than this, because there's no evidence for a meaning "heat (of the day)" for the common ancestor of Greek ἦμαρ "day" and Armenian awr "day". It's invented purely to facilitate the comparison with Semitic.

[Nortaneous]

What grounds justify the rejection of the IE-internal etymology for *kʷekʷlos - or do you mean a different word for "wheel"?

Several:

• Striking Sumerian and Semitic look-alikes make one suspect that we have at best the (P)IE representative of a wanderwort - and then we have timing issues at projecting it back to PIE.
• It is a very odd formation from the root *kʷel. Could it be be a folk-etymology?
• It's not attested in Hittite (though I've a feeling we can't declare it as highly unlikely) and the Tocharian form feels quite different - though it would work as a borrowing parallel to core-IE.
• Numerous irregularities within core-IE - they were discussed here two boards ago.
• Is PIE *(s)ker, or its possibly reduced reduplicated apparent derivatives, a doublet? (Through different routes, this gives English ring and circus.)

Since the root is analyzable in PIE, non-analyzable lookalikes elsewhere should be assumed to be loans from PIE. But that depends on whether the lookalikes in Sumerian and Semitic are analyzable, which I don't know.

Let's see what Adams says about the Tocharian forms:

TchA kukäl and B kokale reflect PTch *käuk(ä)le from PIE *kʷukʷlo- from *kʷₑkʷlo-. This *kʷukʷlo- apparently matches Greek kúklos ‘circle, wheel.’ (The semantic development ‘wheel’ > ‘wagon, chariot’ is paralleled by OCS kolo ‘wheel; wagon’ from *kʷolo-.) The Tocharian and Greek forms are closely related to, but phonologically distinct from, the *kʷekʷló- that lies behind Sanskrit cakrá- ‘circle, wheel,’ Old English hwēol ‘wheel,’ Lithuanian kãklas ‘neck’ (< *‘turner’), etc. Both *kʷₑkʷlo- and *kʷekʷló- are reduplicated derivatives of *kʷel- ‘revolve’ (P:639-640; MA:640).

The epenthetic vowel in TB seems a little weird, and I'm not sure where the PToch *äu is supposed to come from. (TB e ~ TA 0 is expected, and the TA epenthetic vowel is regular from there.) I'm not sure why there are two different forms, but the existence of two different forms doesn't necessarily imply a loan - cf. e.g. 'sun'.

Greek-Tocharian lookalikes are expected; Tocharian seems to have been somewhere between Germanic and Greek. Adams again:

Within the Northwestern Group, Tocharian was closest to Germanic of the surviving Indo-European groups, with secondary ties with Baltic and Latin. The Tocharians, however, separated themselves very early from the rest of this group. They moved south and/or east and came into contact with another group of Proto-Indo-European speakers, the Greeks, perhaps in Moldavia or thereabouts, perhaps in the first half of the third millenium-certainly before the Greeks entered the Balkans proper. Subsequently the Tocharians must have continued to drift eastward across the north Pontic steppes and then Central Asian steppes, perhaps in this latter location briefly associating with some group of pre-Indic speakers, ultimately to appear in history two thousand years later in Chinese Turkestan.

And explosive growth doesn't necessarily imply massive substrate influence - Algonquian left American English a few words and not much else.

Whereas pre-IE populations gave rise to a very significant proportion of modern IE populations.

Fair, but cf. postcolonial Africa and India, although maybe it's too early to say.

What bothers me a little is how odd the system looks.
Is *a attested at all, aside from laryngeal influence? Or is it assumed to have existed, but beyond any possible reconstruction?

There's some disagreement about this. *a can be reconstructed, but it's rare and AFAIK it didn't participate in ablaut, so some scholars assume that any word with *a was a loan. I've seen some vague references to *a tending to co-occur with plain velars, which could then mirror *h2 in coloring *e to *a (and which would then probably be uvulars).

In more phonetic-looking reconstructions, *e *o are often rewritten as something like *ə *ɑ.

Sound correspondences might be different depending on the source. For example, some words In Germanic have reduced labiovelars while others do not.

Examples?

I'd put it that way instead: it's impossible to derive all the attested lexical items in IE langs from a single parent language.

That's practically trivial. For all attested lexical items in a language family to be reconstructible to a single common ancestor, you'd need for zero attested lexical items to be loans ("sushi", "television"), ideophonic/sound-symbolic/etc. de novo coinage ("yeet", "chickadee", "barf") or highly irregular alterations ("chunk", "hork"), unamenable to any etymological method ("guy", probably "dog"), etc.

[Richard W]

Since the root is analyzable in PIE, non-analyzable lookalikes elsewhere should be assumed to be loans from PIE. But that depends on whether the lookalikes in Sumerian and Semitic are analyzable, which I don't know.

They're also analysable, and the Semitic stem formation (by reduplication) is fairly frequent. I'm not sure how regular the derivation of Sumerian gigir 'chariot' from a reduplication of giri 'to run' is. I recall suggestions that the Sumerian form was actually pronounced /girgir/.

Sound correspondences might be different depending on the source. For example, some words In Germanic have reduced labiovelars while others do not.

Examples?

ɪ presume he means four, oven, wolf. liver is also added by those who reconstruct an initial *lʲ.

[Travis B.]

[*] An overview of the modern PIE literature showing that the vast majority of ‘PIE’ words are found only in one branch;

And thus excluding borrowing between different branches? No way!

Borrowing between different branches, such as borrowing from Latin into Germanic, practically always has telltale signs such as not obeying the regular sound changes which would have taken place if the words in question were inherited into both branches. This is how we can tell that English plant is from Latin planta, for the noun, and Latin plantō, for the verb (and we know this borrowing into Germanic precedes the High German Consonant Shift, as it is reflected by Standard German Pflanze and pflanzen). Likewise, non-cognates which are superficially similar can be told by that their forms do not obey these rules, this is how we know Germanic *habjaną "to have" is not cognate with Latin habeō "to have, hold" but rather to Latin capiō "to take".

[WeepingElf]

Greek hêméra 'day' < *h2ēm-ər- 'heat (of the day)', with correspondences in Semitic *ħamm- 'to be hot; warm'

You'll have to do better than this, because there's no evidence for a meaning "heat (of the day)" for the common ancestor of Greek ἦμαρ "day" and Armenian awr "day". It's invented purely to facilitate the comparison with Semitic.

Is the word common only to Greek and Armenian? If yes, it is probably a loanword from a pre-IE language of the Balkan Peninsula. That of course doesn't mean Semitic, though a semantic development from 'hot, warm' to 'day' is not impossible. Yet, I doubt that anything Afrasian was ever spoken on the Balkan Peninsula, but it can't be ruled out, so there is a slim chance that Talskubilos is actually right on this one. But even if it is a loanword from a Semitic language, this single word does not really challenge the validity of the standard reconstruction of PIE. Talskubilos must really try harder!

Since the root is analyzable in PIE, non-analyzable lookalikes elsewhere should be assumed to be loans from PIE. But that depends on whether the lookalikes in Sumerian and Semitic are analyzable, which I don't know.

Yes. PIE *k^wek^wklos is clearly analyzable within PIE. It is from the root *k^wel- 'to turn', and appears to be an iterative reduplication - a wheel turns and turns and turns...; also, as I already said, *h₁meh₁msos 'flesh' is formed in a similar way, it seems. I can't say anything about the Sumerian, Semitic or NWC forms, though.

[KathTheDragon]

Is the word common only to Greek and Armenian?

Yes, both etymological dictionaries I'm using (Beekes for Greek and Martirosyan for Armenian) state there are no further cognates.

though a semantic development from 'hot, warm' to 'day' is not impossible.

Obviously not, but as I said, there's no evidence to independently support it. Note in particular that Armenian awr can also mean "time, age" (not 100% clear if that's "an age" as in a long span of time, or "age" as in the number of years lived).

[Talskubilos]

Is the word common only to Greek and Armenian?

Yes, both etymological dictionaries I'm using (Beekes for Greek and Martirosyan for Armenian) state there are no further cognates.

Latin amor 'love', likely a loanword (perhaps from Etruscan) would be related.

Is the word common only to Greek and Armenian? If yes, it is probably a loanword from a pre-IE language of the Balkan Peninsula. That of course doesn't mean Semitic, though a semantic development from 'hot, warm' to 'day' is not impossible. Yet, I doubt that anything Afrasian was ever spoken on the Balkan Peninsula, but it can't be ruled out, so there is a slim chance that Talskubilos is actually right on this one. But even if it is a loanword from a Semitic language, this single word does not really challenge the validity of the standard reconstruction of PIE. Talskubilos must really try harder!

Although in the case the source isn't necessarily Semitic, external comparanda are relevant in order to understand doublets (and even triplets) of seemingly related protoforms among the +2000 lexical items reconstructed for PIE.

[KathTheDragon]

]Latin amor 'love', likely a loanword (perhaps from Etruscan) would be related.

De Vaan cites a Marrucinian cognate amatens "they have received", implying a semantic development "to hold close" > "to be friendly with" or something along those lines. On that semantic basis, we can compare with Vedic amīṣi, amánti "to take hold of; to swear" and further Greek ὀμνῡμι "to swear" which would show the initial laryngeal to be *h₃ anyway. (The Italic vowel *a would be due to *-R̥HV- > *-aRV-)

[Talskubilos]

Yes. PIE *k^wek^wklos is clearly analyzable within PIE. It is from the root *k^wel- 'to turn', and appears to be an iterative reduplication - a wheel turns and turns and turns...;

That's right. My bet is *kʷel- 'to turn' is a loanword, likely from Caucasian. But there's apparently a homonymous verb 'to dwell', attested in Latin colō 'to live, to inhabit', in-cola 'inhabitant', in-quil-īnus 'tenant'.

[WeepingElf]

The root *k^wel- 'to turn' seems to have a pretty good Mitian pedigree, with Turkic *Kul-, Mongolic *kok-ki-, Tungusic *xol/xul-, all with meanings related to 'to turn' (according to An Etymological Dictionary of the Altaic languages by Starostin et al., p. 850; I know this is not all too reliable), and perhaps Uralic *kulke- 'to go'.

[KathTheDragon]

But there's apparently a homonymous verb 'to dwell', attested in Latin colō 'to live, to inhabit', in-cola 'inhabitant', in-quil-īnus 'tenant'.

De Vaan derives colō from *kʷelH- "to turn", but annoyingly doesn't give the semantic motivation. I'd say a development "to turn" > "to go in circles" > "to stay in the same place" > "to inhabit" isn't unreasonable. But note that the Latin verb also means "to take care of", and Latin anculus "man-servant" can be derived from Pre-Latin *ambi-kʷolos, with *kʷelH- as its second element, which is identical to Greek ἀμφίπολος "(female) servant", as well as very close to Sanskrit abhicara- "servant".

[Nortaneous]

English "turn up" [(unexpectedly) end up being at] seems like a reasonable semantic intermediate for both "go" [end up being at] and "live" [habitually be at] - but it's not clear to me how to get from "turn" to "take care of" or "serve"

[Talskubilos]

De Vaan derives colō from *kʷelH- "to turn", but annoyingly doesn't give the semantic motivation. I'd say a development "to turn" > "to go in circles" > "to stay in the same place" > "to inhabit" isn't unreasonable.

Perhaps you're describing a drunk person walking home.

[Richard W]

But there's apparently a homonymous verb 'to dwell', attested in Latin colō 'to live, to inhabit', in-cola 'inhabitant', in-quil-īnus 'tenant'.

De Vaan derives colō from *kʷelH- "to turn", but annoyingly doesn't give the semantic motivation. I'd say a development "to turn" > "to go in circles" > "to stay in the same place" > "to inhabit" isn't unreasonable. But note that the Latin verb also means "to take care of", and Latin anculus "man-servant" can be derived from Pre-Latin *ambi-kʷolos, with *kʷelH- as its second element, which is identical to Greek ἀμφίπολος "(female) servant", as well as very close to Sanskrit abhicara- "servant".

The English compound verb "to revolve around" isn't a bad approximation to the general semantics. So I also think we have polysemy rather than homonymy.