The Great Proto-Indo-European Thread's Sequel

[Ketsuban]

For what it's worth I find myself equally unconvinced by arguments of the form "X is common, which argues against its truth because it implies X doesn't change easily due to sound change and we know Proto-Indo-European broke up into daughter languages so we should expect it to show unstable structures" and arguments of the form "X is uncommon, which argues against its truth because we can be reasonably certain Proto-Indo-European was a language spoken by people not substantially different than us, so it probably didn't show typologically unusual features".

And on that note, IIRC I have seen people who have posited that the laryngeals were something like *x *xʷ *χ.

This is certainly appealing in its symmetry, although as I understand it there's some evidence (e.g. Sanskrit pibati from a reduplicated *peh₃- > *pi-ph₃-) to suggest that *h₃ was voiced.

[Travis B.]

For what it's worth I find myself equally unconvinced by arguments of the form "X is common, which argues against its truth because it implies X doesn't change easily due to sound change and we know Proto-Indo-European broke up into daughter languages so we should expect it to show unstable structures" and arguments of the form "X is uncommon, which argues against its truth because we can be reasonably certain Proto-Indo-European was a language spoken by people not substantially different than us, so it probably didn't show typologically unusual features".

I see how both of these arguments can be unconvincing. I personally am of the view, though, that one should treat reconstructed languages as languages spoken by real people and not merely reconstruction artifacts, so I tend to favor reconstructions that are plausible over deliberately disfavoring plausible reconstructions because "well if it were plausible it would have never changed and we would see it in daughters".

With that in mind, I would say that there is no reason why *k could not have palatalized in some daughters, leading to the pull-chain *k *kʷ *q > *c *kʷ *q > *c *kʷ *k and so on (or conversely a push-chain where *q started getting fronted, resulting in *k fronting as well to keep the two distinct). There is no reason to deliberately choose a less plausible reconstruction indefinitely back in time just because "if it were stable it would have survived to date".

And on that note, IIRC I have seen people who have posited that the laryngeals were something like *x *xʷ *χ.

This is certainly appealing in its symmetry, although as I understand it there's some evidence (e.g. Sanskrit pibati from a reduplicated *peh₃- > *pi-ph₃-) to suggest that *h₃ was voiced.

There is no reason that *h₃ could not have become voiced in the daughter that spawned Indo-Aryan. For instance, I have heard it posited that ME /x/ may have become voiced in the daughter that spawned StdE in positions where it did not merge with /f/ prior to disappearing altogether. (Of course, that is based largely on orthographic evidence, i.e. why it was written <gh>, parallel to, say, how /ɣ/ was commonly written <gh> in MD.)

[abahot]

There is no reason that *h₃ could not have become voiced in the daughter that spawned Indo-Aryan. For instance, I have heard it posited that ME /x/ may have become voiced in the daughter that spawned StdE in positions where it did not merge with /f/ prior to disappearing altogether. (Of course, that is based largely on orthographic evidence, i.e. why it was written <gh>, parallel to, say, how /ɣ/ was commonly written <gh> in MD.)

I never thought of this, but it actually makes an awful lot of sense.

However, I find it hard to believe that the laryngeals patterned exactly with the three velar series. Because then *h₂ and *h₃ would have merged in satem branches, etc etc etc. Maybe it still makes sense, and in any case this inventory looks pretty typologically plausible to me, more in the sense of PIE being internally consistent than consistent with changes we see in the daughters.

I've also heard the theory advanced that in PIE, there were only two velar series, and the front velar series arose originally as allophones of the plain velar series in a broader trend of palatalization. To me, this does seem to smack of prioritizing evidence from Western IE branches, which do not show distinct reflexes at all. (On the other hand, I am also inclined to view reconstruction of the "voiced aspirates" having breathy-voiced pronunciation in the opposite sort of way, as prioritizing evidence from Indo-Aryan.)

[Ketsuban]

I found myself on the Wikipedia page for s-mobile and noted a number of roots with *a, which is generally not considered solidly reconstructible for Proto-Indo-European, so I went through not just the roots on that page but the roots on Wiktionary's list containing *a; I noted which ones are mentioned on Wiktionary, what Pronk (2019) has to say about them if anything and the general vibe I get from them. (This is very much not comprehensive; Wiktionary has an awkward distinction between "roots" and "lemmas", so I didn't look at for example *albʰós, for which Pronk notes the unstable ground it's resting on since Hittite al-pa-aš "cloud" is primarily associated with storms and rain rather than the kind of white fluffy clouds implied by the semantic association.)

• *bak- "club, peg, stick" is a classic; it seems nobody can ever fully dismiss it despite it stinking like week-old fish because there's no other conclusive explanation for the things it ties together. The semantics are terrible: Germanic *pagilaz (English pail) is just about defensible as rod > measuring rod > liquid measure > liquid container; Celtic *bakkos "hook" strains credibility; Slavic *bokъ "flank, side" feels nonsensical. Latin baculum < *bakklom is based on no non-Latin evidence and seems to match up dubiously with Greek βάκτρον.
• *h₁yaǵ- "sacrifice, worship" is geographically restricted; with the exception of one Anatolian (which I can't find) and one Tocharian word (of which this dictionary says "uncertain") all the examples come from Greek, Latin and Sanskrit.
• *kagʰ- is largely limited to Italic and Celtic. It's also an instance of Wiktionary disagreeing with itself: the etymology of Armenian cʻank is given as "uncertain"; the page for Sanskrit kákṣa gives PII *káćšas < *koḱs-o-s from the PIE root *koḱs- "joint", which also feels like a root on dubious ground; and the page for Welsh cael says it's a syncopated form of caffel, ultimately from the zero-grade of...
• *kap- is an instance of Wiktionary just plain being wrong; Pronk is quite decisive in showing the correct reconstruction is *kh₂ep-. (Yes, this is out of alphabetical order.)
• *kakka- is clearly a kind of sound-symbolism; it stands out to me that the only family within Indo-European with a huge consonant shift (Germanic) retains it as a "loanword" to hang onto the velar sound rather than continuing something like **hahha. Pronk dismisses *atta "dad" along similar lines.
• *skabʰ- is addressed by Pronk directly; while he doesn't come down on the side of a specific reconstruction, he presents alternatives which are more parsimonious in not needing *a to function.
• *(s)kand- is listed on the Wikipedia s-mobile page; while it has no Wiktionary link, other words reference it in their etymology section with a red link. It's another geographically restricted one: there's no Celtic, Germanic or Balto-Slavic lexemes. κάνδαρος "charcoal" is a little strange semantically, but sparkle > embers > charcoal seems... okay? The main problem is just the fact it has *a when so many of those are coming up snake eyes.
• *(s)mal- is a joke; it lists *(s)mel- on the page itself as an "alternative form".
• *(s)kap- is glossed as "tool"; we're given Greek σκεπάρνον "carpenter's axe/adze/tool for hewing and smoothing the trunks of trees" and Latin cāpus "???". I'm not sure what a castrated rooster has to do with "tool", but both of these words are listed as probably of substrate origin on Wiktionary so I feel reasonably confident calling this a phantom.
• *(s)wagʰ- is glossed as "resound"; we're given Greek ἠχή "noise, sound" and English sough. Both words' Wiktionary pages list *(s)weh₂gʰ- as the PIE root, making this the easiest murder mystery of all time.

This leaves the interesting one: *(s)kʷal-o-, glossed as "big fish". We have Avestan kara "kind of fish" (I hate this kind of gloss; sure, not every family can be like the Polynesian languages where animal words can be tidily reconstructed as referring to a specific genus, but it always reminds me of Vovin's criticism of the Altaic hypothesis), Germanic *hwalaz, Greek ἄσπαλος (glossed by Hesychius as ἰχθύς "fish"), Latin squalus "shark" and Prussian kalis "sheatfish". This isn't the most geographically diverse root in the world, but the vibes I get from it are more solid than something like *káput, the Germanic *hwalaz can refer to catfish (e.g. German Wels) as well as whales, and the sheatfish species Silurus glanis has the distinct advantage of being a popular sport fish native to the Black and Caspian Seas, not far from the Indo-European Urheimat in what's now Ukraine.

[Zju]

What are the s-mobile variants of *h₁yaǵ-? Wiktionary lists none.

[Ketsuban]

There are none, that one came from Wiktionary itself rather than the s-mobile page.

[abahot]

*bak- "club, peg, stick" is a classic; it seems nobody can ever fully dismiss it despite it stinking like week-old fish because there's no other conclusive explanation for the things it ties together. The semantics are terrible: Germanic *pagilaz (English pail) is just about defensible as rod > measuring rod > liquid measure > liquid container; Celtic *bakkos "hook" strains credibility; Slavic *bokъ "flank, side" feels nonsensical. Latin baculum < *bakklom is based on no non-Latin evidence and seems to match up dubiously with Greek βάκτρον.

I feel that we have three options here:

• Explain it as "onomatopoeic" and ignore all the questions that brings up.
• Embrace the null hypothesis and say that these roots are unrelated. Sure, they look vaguely similar, but they also don't have meanings which are very related. Perhaps Latin and Germanic are related (because another meaning of *pagilaz given is "peg"), but maybe Celtic (and especially Slavic) are just unrelated. If we looked at English over time, we might see certain words as essentially coming out of thin air. For example, if we didn't preserve phrases like "what's up", "'sup", "'sup", and so forth, the use of "suh" /sʌ/ as a greeting appears to have originated from nowhere. (Or, for a more direct example, take "simp".) It may well just be the case that we are missing intermediate forms of this sort that would hypothetically have existed in Pre-Proto-Slavic, -Celtic, and so forth tying these words to other roots than *bak, or even arising in irregular ways from other words.
• Accept that "terrible" semantics can and do happen in real languages. Consider the evolution of the word "boomer". The actual etymology is from the root "boom-" along the lines of ("make a loud and deep noise" > "do very well" > "time when lots of children were being born" > "one born during said time" > "older person with outdated political views"). If Proto-Germanic *bōmarijaz were proposed to descend from PIE *b^hoh₂m- along these same lines, there would be ridicule because this etymology is absurd! And yet it happened in real life.

[WeepingElf]

You must not forget that PIE was no pristine isolated language, but was in contact with other languages around it, and thus probably contained loanwords that were borrowed into the language at a late stage when /a/ and /b/ had become phonemic. Such loanwords need not comply with the reconstructed word structure constraints - the example of /f/ in Slavic shows that even phonemes can be borrowed.

[Man in Space]

the example of /f/ in Slavic shows that even phonemes can be borrowed.

And English, when French loans phonemicized voicing distinctions in fricatives.

[Sol717]

Likewise, ME /x/ either disappeared or turned into [f], outside Early Modern Scots, without a corresponding change of /k/.

To be annoyingly late and pedantic, /x/ was also retained in the Yorkshire-Lancashire border area until very recently; the 1950-1961 Survey of English Dialects recorded light night [lɪçt nɪçt] at Muker in the Yorkshire Dales, while in Alexander Ellis's 1889 The Existing Phonology of English Dialects, forms of this sort are recorded alongside bought sigh tough [bɔxt sɪç tɔʊ̯x] at the Colne Valley further south along the border; for more discussion, see The Dialect of Dentdale in the West Riding of Yorkshire, § 8:46. Less surprisingly, Ellis also records retention of /x/ in areas of Northumberland adjacent to Scotland.