PIE *g’enH1 and *gʷenH2 as cognates ("king" and "queen")

« previous post | next post »

[This is a guest post by German Dziebel, commenting on "PIE *gene- *gwen-" (8/10/23).]

I will strike a dissenting note here. The two roots in question – *g’enH1 and *gʷenH2 are likely cognates. There seems to be a non-random distribution of palatalized and labialized velars in IE stems with nasals – palatovelars are favored in stems with m, while labiovelars are favored in stems with n. E.g.,

nGʷ roots: *nogʷno- 'naked', *nogʷt- 'night', *snoigʷho- 'snow', *h₂ongʷo- 'anoint', *h1ngwni- 'fire', *negʷhro- 'kidney', *gʷenh₂ 'wife', *kʷoino- 'price', *penkʷe- '5', *h₁lengʷʰ- 'light', *gʷʰen- 'slay, strike', *sengʷh- 'sing', *neigʷ- 'wash' 

vs.

mG'-roots: *H3moiǵhlo- (assimilated to njegull(ë) in Gheg Alb), *meǵh₂s 'great', *meh₂ǵ- 'smear, anoint', *ǵheyōm 'winter', *dheǵhōm 'earth', *ḱoimo- 'household, family', *mreǵh-, *mosgho- 'brain', *h₂melǵ- 'milk', *smeḱur 'chin, beard', *deḱm̥ '10', *h1ḱm̥tóm '100' *h₂émǵʰu- 'narrow' (Hitt hamenk- 'tie, bind').

Although there are seeming exceptions (e.g., PIE *gʷher- ‘hot’ yields -mo-derivatives in Gk  θερμός, Alb zjarm, Arm jerm, in all those branches the labiovelar is found in a palatalized state), those exceptions are limited in number and can be explained as later assimilations. This is likely what happened with PIE *g’enH1 and PIE *gʷenH2 where only *gʷenH2 is “legal”, while *g’enH1 is likely assimilated from either *g’emH1 or *gʷenH1. As a supporting proof for this inference one can cite Baltic *gmti ‘beget, give birth’ (Lith gimti, Latv dzimt, OPruss gemton) that must be going back to *gʷem- (no connection to PIE *gʷen- ‘come, step’ (Lat venio:, Gk baino:, etc., with assimilation creating stems such as Germ *kwemaną (comp. *faima 'foam' < PIE *spoineh₂), PToch *kum (comp. mekwa ' nails' < *nogʷho-) and InIr *ǰámati (comp. Skrt ūrmí, Avest varəmi 'wave' but Lith vilnis, Slav *vъlna 'wave')). PIE *gʷem- went through assimilation and generalized labiality across the stem in exactly the opposite way from PIE *g’enH1 that generalized palatality. As a sum total, it’s most likely that the PIE word for ‘beget, give birth’ was * gʷen(H1)- and hence it can hardly be separated from *gʷenH2 ‘woman, wife’. Germ *kʷēniz 'wife' was likely applied to ‘queen’, too, as in Old English, and was a cognate counterpart to *kuninga- ‘king’. It’s to be expected that the words for ‘king’ and ‘queen’ were derived from a single root as they do in so many IE languages – living and dead – from Hitt hassu ‘king’, hassusara ‘queen’ onward.

Selected readings

The modern English word gender comes from the Middle English gender, gendre, a loanword from Anglo-Norman and Middle French gendre. This, in turn, came from Latin genus. Both words mean "kind", "type", or "sort". They derive ultimately from a Proto-Indo-European (PIE) root *ǵénh₁- 'to beget', which is also the source of kin, kind, king, and many other English words, with cognates widely attested in many Indo-European languages. It appears in Modern French in the word genre (type, kind, also genre sexuel) and is related to the Greek root gen- (to produce), appearing in gene, genesis, and oxygen. The Oxford Etymological Dictionary of the English Language of 1882 defined gender as kind, breed, sex, derived from the Latin ablative case of genus, like genere natus, which refers to birth. The first edition of the Oxford English Dictionary (OED1, Volume 4, 1900) notes the original meaning of gender as "kind" had already become obsolete.

 



105 Comments

  1. Dwight Williams said,

    October 7, 2024 @ 5:10 pm

    Yeah, my old browser/OS combo is leaving off some of the glyphs here…

  2. Laura Morland said,

    October 7, 2024 @ 6:36 pm

    It's been a few years since I studied Indo-European linguistics, and so I want to make sure that I've understood correctly, that:

    only *gʷenH2 is “legal”, while *g’enH1 is likely assimilated from either *g’emH1 or *gʷenH1 ….

    It’s to be expected that the words for ‘king’ and ‘queen’ were derived from a single root as they do in so many IE languages – living and dead….

    It is therefore possible that the original IE word for Queen came first, and the word for King was derived from it? If so, moving from linguistics to sociolinguistics, what does that imply about a possible matrilineal culture of our IE ancestors?

    (See: the situation in ancient Crete, and even in the Odyssey. Why so many suitors, over 20 years, for Penelope's hand, unless marrying the Queen is the only way to become the King?)

  3. German Dziebel said,

    October 7, 2024 @ 8:18 pm

    @Laura. It's an interesting question. PIE *gʷenH2 is very stable and attested in nearly all the branches. (Michael Weiss (in the original post) mentioned that Italic does not have a reflex of *gʷenH2. I think Lat mulier 'woman' might constitutes such a reflex via *gʷNyes- > *bNyer- (rhotacism and a Sabellic gʷ > b) > *bLyer- (comp. glo:ra *mLyer- > *mulier (capitals indicate IE syllabic resonants)). Phonetically, as I argued, *gʷenH2 is more conservative than *g'enH1-. Gk βᾰσῐλεύς 'chieftain, overlord' (really the Greek kunungaz) can be interpreted as based on *gʷNti- 'wife, woman' reflecting Margalit Finkelberg's theory (https://www.jstor.org/stable/638900?seq=1) that Bronze Age Hittite and Greek chiefs acquired their power from marrying the daughters of local, pre-Hittite and pre-Greek chiefs. Genealogical descent was patrilineal, but control over the land was matrilineal. Sons-in-law succeeded their fathers-in-law but sons could not inherit princely power from their fathers. However morphologically most IE names for 'queen' tend to be marked and derived from the IE names for 'king' (Lat rēgīna vs. rēx, Hitt hassusara vs. hassu, etc.)

  4. Chris Button said,

    October 7, 2024 @ 8:25 pm

    I'm certainly no IE expert either, so I look forward to some comments from the experts. But I do have a few (possibly naive) questions:

    As a supporting proof for this inference one can cite Baltic *gmti ‘beget, give birth’

    But isn't this usually derived from an entirely separate root? Could you perhaps be conflating unrelated things as evidence?

    Although there are seeming exceptions

    A cursory look at the reconstructed lexicon suggests there might be at least a fair few though. Can these all really be explained as "later assimilations"?

    It’s to be expected that the words for ‘king’ and ‘queen’ were derived from a single root as they do in so many IE languages – living and dead – from Hitt hassu ‘king’, hassusara ‘queen’ onward.

    I like the attempt to connect them. But even if we explain away the possible gʲ- versus gʷ- alternation, the "laryngeals" remain different. Should they just be swept under the rug?

  5. German Dziebel said,

    October 7, 2024 @ 8:52 pm

    @Chris. "But isn't this usually derived from an entirely separate root?" It is true that Balt *gem-/gM- are usually interpreted as belonging with PIE *gʷem 'come, step'. But this is only because the Baltic forms never had their "home" in word families with the same meaning (comp. Lith gỹmis 'birth, nature', gimine 'family', etc. next to Gk γένος, Lat genus). It's also because what's reconstructed as *gʷem- 'come, step' was, I believe, a "forbidden" root shape in PIE and we need to follow Gk βαίνω and Lat venio: in reconstructing *gʷen- and explain forms such as *kwemana as assimilated ones.

    "But even if we explain away the possible gʲ- versus gʷ- alternation, the "laryngeals" remain different. Should they just be swept under the rug?" It's tru – again- that the laryngeals are different. However H2 in *gʷenH2 is a feminine marker. The function of H1 in *g'enH1 is unknown. The Baltic forms don't have a laryngeal at all, which may point to a bare stem *gʷen- as original.

    "Can these all really be explained as "later assimilations". There are indeed some well-attested ones that I can't fully explain (e.g., *g'enu- 'knee' or *g'enH2- 'chin') but the lists above seem compelling enough to me and later assimilations are certain for some of the items (e.g., Toch mekwa 'nails' < *H3nogwho- above).

  6. David Marjanović said,

    October 8, 2024 @ 5:37 am

    An interesting idea, but it needs more work, and of course all the usual dangers of internal reconstruction apply.

    However H2 in *gʷenH2 is a feminine marker. The function of H1 in *g'enH1 is unknown.

    That's an important point; by current understanding the comparison is really between the roots *gʷen- and *ǵenh₁-. That means we don't have to figure out how *h₁ and *h₂ could be related, but it replaces that problem with the new problem of how *h₁ could appear or disappear; I'm not aware of any evidence that *ǵenh₁- might be derived from a hypothetical **ǵen-.

  7. German Dziebel said,

    October 8, 2024 @ 8:19 am

    @David. "I'm not aware of any evidence that *ǵenh₁- might be derived from a hypothetical **ǵen-."

    PIE *g'em-/*gm-ro 'son-in-law' (Gk γαμβρός, Skrt (jā́mātṛ, Avest zāmātar, possibly Celt *damos with a surprising link to Toch B täm 'beget, be born') is not reconstructed with H1 but the overall agreement with Lat gener, Alb dhender, BS źénˀtis, -as is too strong to ignore. If it were from *g'mH- we would have obtained **gameros in Greek. Now that we have a hypothesis of why m/n alternates in roots with different velars the connection between *g'em- and *g'enH1 is becoming even more plausible suggesting that H1 in *g'enH1 is not part of the root. Anat *hams- > Hitt hass- 'beget, be born', with a totally unexpected onset but a familiar -m- and the same semantic extension hassu- 'king' as we see in Germ *kuninga-, does not have H1 either. Finally Balt *gmti 'beget, be born' is not reconstructed with a laryngeal either.

  8. TR said,

    October 8, 2024 @ 12:24 pm

    A priori the idea seems phonologically unlikely. A labiovelar losing its labial component due to an /n/ in the root isn't "assimilation" in any sense I'm familiar with, and seems phonetically unmotivated. I'd like to see parallels for such a development.

  9. TR said,

    October 8, 2024 @ 12:28 pm

    As for "Gk βᾰσῐλεύς 'chieftain, overlord' (really the Greek kunungaz) can be interpreted as based on *gʷNti- 'wife, woman'" — what is this *gʷNti- and how do you account for the -λ-?

  10. Chris Button said,

    October 8, 2024 @ 12:55 pm

    The labiovelar component of the onset not being distinctive in *gʷem- is eminently reasonable though.

  11. TR said,

    October 8, 2024 @ 1:53 pm

    Maybe, but we're talking about a supposed change *gʷen- to *gen-. Are there any parallels for such a change? What would be the phonetic motivation?

  12. German Dziebel said,

    October 8, 2024 @ 2:23 pm

    @TR. *gʷNti- – an extended stem based on *gʷN-, comp. Luwian wanatti 'woman'. No idea what -λ- stands for.

    "A priori the idea seems phonologically unlikely…I'd like to see parallels for such a development." The larger data cited above seems to suggest that there's co-dependency between a labiovelar vs palatovelar choice depending on the quality of the nasal. There's one other case in addition to *gwenH2 ~ g'enH1 where the choice is ostensibly (and it's a new etymological claim) being made within in a cognate set. PIE *meh₂ǵ- 'smear, anoint' (e.g., Slav *mazati 'smear, anoint', *maz-slo 'butter', Arm macanim 'glue to, adhere)' ~ *h₂ongʷo- 'anoint' (Lat unguo: 'anoint', OPruss anctan 'butter', Arm awcanem 'anoint' wg'), while -n- dropped. But the labiovelar does not simply drop labiality. It acquires palatality seemingly compensatorily. This development is known as awcanem-rule and it's known to affect just a handful of forms. It's currently believed that a shift from a labiovelar to a palatovelar in Armenian happened because just like u and i are both short, high vowels, ʷ and j are also articulatorily similar. Just by looking at PIE *meh₂ǵ- vs. *h₂ongʷo-, it's impossible to say which velar-nasal arrangement was original. It's also possible that originally there was a different, let's say uvular stop in place there that got fronted to a velar and depending on the nasal the choice was either palatalized or labialized variety. In the case of *gʷenH2 ~ *g'enH1 my claim that assimilation affected *g'enH1 is based on the stability and legality of the *gʷenH2 form.

  13. German Dziebel said,

    October 8, 2024 @ 2:31 pm

    Oops, my comment came out garbled.

    Instead of "wg'), while -n- dropped" please read the following: Arm awcanem shifted from an inherited labiovelar to a palatovelar either by shifting -w- and dropping -n- (ngʷ > wg') or by fusing w and n into m and then shifting m to w, so that ngʷ > mg' > wc. I prefer the latter because of an attested macanim.

  14. German Dziebel said,

    October 8, 2024 @ 3:18 pm

    @Chris. "The labiovelar component of the onset not being distinctive in *gʷem- is eminently reasonable though." Would you say the same about the palatovelar component in *g'en-?

  15. TR said,

    October 8, 2024 @ 3:40 pm

    As far as I can tell the awcanem rule involves labialization getting reassigned from a labiovelar to an immediately preceding nasal. That's phonetically understandable, and not at all parallel to something like *gʷen- > *gen-.

    In *meh₂ǵ- vs. *h₂ongʷo- there's an errant laryngeal once again, so you'd have to assume an additional metathesis. More likely the roots are unrelated.

    I just looked through the LIV index and counted 24 verb roots containing a palatal and *n. It seems hardly supportable to say that "labiovelars are favored in stems with n".

  16. German Dziebel said,

    October 8, 2024 @ 4:04 pm

    @TR. "In *meh₂ǵ- vs. *h₂ongʷo- there's an errant laryngeal once again, so you'd have to assume an additional metathesis." Yes. Both forms are metathetical to each other on all counts.

    "More likely the roots are unrelated." It depends on how we frame cognation. We have two PIE root shapes with identical meaning and derivations, both root shapes are distributed between closely related branches (OPruss anctan vs. Slav *maz-slo < possibly from *maz-sno with the same secondary nasal affix), having the same phonetic components (H2, nasal, velar) but different order both on the phonemic (laryngeal) and feature (labiality, palatality) levels. Only the velar(ity) is anchored firmly in the middle of the string. Typically different order does not preclude cognation as long as the sounds are in correspondence (e.g., Toch kantwo 'tongue' is a metathesized cognate of Germ *tungǭ 'tongue').

    "I just looked through the LIV index and counted 24 verb roots containing a palatal and *n. " I did some of it , too (and need to do more work here for sure), and had issues with many instances mostly because of limited distribution of forms not necessarily suggesting the PIE status. Did you check for any verbs that have a labiovelar and m? I reinterpreted *gwem 'come, step' as *gwen- based on Lat venio: and Gk baino:. See above.

  17. German Dziebel said,

    October 8, 2024 @ 4:22 pm

    @TR. Also Armenian favors the cognation of *meh₂ǵ- vs. *h₂ongʷo because it attests to minimally divergent forms macanim and awcanem (hypothetically < *amcanem).

    "the awcanem rule involves labialization getting reassigned from a labiovelar to an immediately preceding nasal." I don't equate *g'enH1 ~ *gʷenH2 with the awcanem rule directly but the variation that we find between Balt *gmti (< PIE *gʷm-), PIE *g'em- 'beget, son-in-law' and PIE *g'enH1 'beget, son-in-law' involves the same reassignment of palatality and/or labiality as the awcanem rule. The targets are not contiguous but this is secondary as we have secure cases of distant labial assimilation in Toch mekwa, Luw tammuga 'nail' < *H3ngʷho- and Arm merk 'naked' < *negʷ-ro, PIE *nogʷno-,

  18. TR said,

    October 8, 2024 @ 5:12 pm

    We've had this discussion before, but where formal cognacy is concerned it seems wise to apply what Michael Weiss calls the "rule of multiple funninesses": if your etymology requires more than one irregular or ad hoc phonological change, it's unlikely to be correct. Allowing yourself too many degrees of freedom makes it so easy to equate unrelated forms that any true cognates, if such exist, will likely be swamped by the many false positives the method inevitably produces.

    Verbs with a labiovelar and *m — there is *h₂meygʷ- which seems secure from ἀμείβω, Lat. migrō, but appears to be confined to those branches. Not sure if there are other good examples other than the common root *gʷem- which you reject.

    The supposed change *gʷen- &glt; *ǵen- isn't an assimilation since *n is not a palatovelar, so cases of distant labial assimilation don't seem relevant to me. It's such a weird change that it's hard to accept without a phonetic explanation and some clear parallels.

  19. Chris Button said,

    October 8, 2024 @ 5:38 pm

    @ German Dziebel

    gwen and gjen vs gem and gjem would work, although im talking in genealities rather than in terms of the structure of PIE. The w and m are both labial, so gwem might merge with gem.

  20. Chris Button said,

    October 8, 2024 @ 5:39 pm

    *I'm talking in generalities.

  21. German Dziebel said,

    October 8, 2024 @ 5:59 pm

    @TR. Weiss's definition of cognation is very exquisite but can hardly be taken literally and be useful practically. If we take it literally, why would there be no more than 1 "irregular" change that's allowed, why not keep it to "no exceptions at all" and how do we determine likelihood? If we are talking in practical terms, Arm awcanem and macanim will fit Weiss's definition of cognation, while awcanem and Lat unguo: won't. (The nature of the laryngeal is ambiguous, there are exceptions to the awcanem rule such as anjuk 'narrow' and the velars don't match). Meanwhile, no scholar, including Weiss, doubts their cognation.

    "Verbs with a labiovelar and *m — there is *h₂meygʷ- which seems secure from ἀμείβω, Lat. migrō, but appears to be confined to those branches." It's an interesting example but I agree it won't fit the PIE horizon as other branches have reflexes of the more basic root *H2mey- without labiovelar extensions.

    "gwen > g'en…It's such a weird change." That's important to know.

  22. TR said,

    October 8, 2024 @ 6:26 pm

    It's not a definition, but a rule of thumb: the more funninesses are required, the most likely it is the partial similarity is accidental. Of course there are funninesses and funninesses, but on the whole you'll find very few accepted etymologies that rely on multiple ad hoc formal changes. That's almost the definition of the comparative method. (I'm not an Armenianist but I don't think unguō : awcanem is an example; "we're not sure which laryngeal it was" isn't an ad hoc change, and for the velar there seems to be a regular palatalization rule after *w/u.)

    I think you're being sarcastic with "That's important to know", but of course phonetic explainability and the existence of phonological parallels are important when proposing a new sound change.

    BTW the semantic derivation doesn't seem straightforward either: are there languages where the basic word for "woman" comes from "one who gives birth"? If this is the case for the PIE word shouldn't we expect an agent suffix of some kind?

  23. German Dziebel said,

    October 8, 2024 @ 7:04 pm

    @TR. No I wasn't sarcastic. I concur with what you wrote.

    "Of course there are funninesses and funninesses, but on the whole you'll find very few accepted etymologies that rely on multiple ad hoc formal changes."

    In such discussions I like to use IE words for 'tongue' as a guiding stick. tungo, lingua, tengae, fangvam, jihva, hizwa, kantwo, lezu, liezuwis, jenzykъ – all branch-specific forms are irregular in the strong position (and this excluding even more difficult cases of Gk glo:ssa < *dlo:gwhya and Alb gluhe). yet all of them share the same meaning and the same polysemy "tongue, language" and are universally accepted as cognates. In addition, rephrasing your assertion above, PIE phonology is based on comparanda that involves identical semantics and very often highly variable (including irregularities) form. Hence semantics (in addition to fundamental formal compatibility) are two ingredients of cognation and sources for new rule formulations.

    "are there languages where the basic word for "woman" comes from "one who gives birth"? If this is the case for the PIE word shouldn't we expect an agent suffix of some kind?" It's a legitimate question to ask. The problem is that we don't know exactly if the proposed cognation between 'beget, be born' and 'woman, wife' should be modeled as "woman = one who gives birth." There's no agentive affix in the woman forms. But if we look at the 'son-in-law, bridegroom' forms *g'em- or gwem- and *g'enH1- we may expect them to be based on the word for 'woman' (e.g., *ženixъ 'bridegroom' is an agent noun implying 'wife-getter') and some of the forms carry what might be interpreted as an agentive affix *-(t)er (jama:tar, dhender, etc.) In fact Gk gamos 'marriage' and gambros are clearly linked. But PIE *g'em- or gwem- and *g'enH1- all mean 'beget, give birth' (e.g., Avest zāmi 'giving birth') and not 'woman'. Unless – which is a new thought – the original stem for 'woman' was *g'en-, *g'em- and then *gwem emerged from *g'em and normalized as *gwen-H2.

  24. German Dziebel said,

    October 8, 2024 @ 7:15 pm

    As a side remark for all, Greek has a separate rule whereby labiovelars lose their labial component in stems with another labial. E.g., κᾰρπός 'wrist' but Germ *hwerbaną 'turn'. This means that γαμβρός and γᾰ́μος can easily be from *gʷem-.

  25. TR said,

    October 8, 2024 @ 8:28 pm

    You're right that the "tongue" word is notoriously irregular. I think it's just the Indo-Iranian forms that may contain multiple funninesses (the relation of the Greek and Albanian forms seems very doubtful, and with the others it's either metathesis or contamination affecting the first consonant). In this case the semantic identity makes it easier to accept.

    I don't understand your last paragraph — what can the semantic relation be if not "woman = one who gives birth"?

  26. German Dziebel said,

    October 8, 2024 @ 8:56 pm

    @TR. "what can the semantic relation be if not "woman = one who gives birth"?" I was just saying that there's no evidence for such an understanding if by evidence we mean an agentive affix of sorts. Forms such as Lat genitor 'begetter, father' and genitrīx 'begetter, mother' are widely attested. But *gwenH2 'woman, wife' has a different relationship to *g'enH1. What I was trying to point out is that forms for 'son-in-law, bridegroom' that have an agentive affix must be built on a form for 'woman, wife' but as a component of those forms (and not as a standalone lexeme) 'woman, wife' is coded as *g'enH1, *g'em and not as *gwenH2 (we have Lat gener instead of expected **vener or **bener, etc.) This again, from a different angle, raises the question of how *g'enH1 and *gwenH2 are connected formally. In other words why do g' and gw appear to be allophonic variations on the basic phoneme g in the context of n/m? Back to semantics: it's possible that PIE *gwenH2 meant 'female, marriageable member of the group of blood relatives called *g'enH1."

    "contamination". I reject the possibility of those. So -l- in lezu, liezuwis, lingua, glo:ssa and gluhe is a systemic feature of this set and not an accident.

  27. TR said,

    October 9, 2024 @ 3:59 pm

    "Greek has a separate rule whereby labiovelars lose their labial component in stems with another labial" — I think there are only two good examples of this, καρπός and κόλπος "curve"; both have very similar phonological environments *kʷ(V)[l/r]p-, so it seems unsafe to extend the rule further. (The environment is quite similar to that seen in the dissimilation Lat. quinque > It. cinque — does the presence of an intervening sonorant promote this kind of dissimilation?) But maybe there are more cases I'm unaware of.

    It doesn't seem obvious to assume that "forms for 'son-in-law, bridegroom' that have an agentive affix must be built on a form for 'woman, wife'" — on the contrary, the presence of an agentive suffix implies they're built on a verb form.

  28. German Dziebel said,

    October 9, 2024 @ 6:02 pm

    @TR. There's also καπνός 'smoke' < *kʷep-. Kroonen doubts that δελφῡ́ς 'womb' comes from *gʷelbho-, which I found interesting because anecdotally cross-linguistically dental reflexes of labiovelars are said to be uncommon.

    " the presence of an agentive suffix implies they're built on a verb form." yes, good point. But those verbs are likely derived from 'wife' as in Slav *ženixъ 'bridegroom' from žénja se 'get a wife'. Alternatively son-in-law, bridegroom as a begetter is implausible.

  29. TR said,

    October 9, 2024 @ 6:37 pm

    καπνός is more doubtful (and seems to be often reconstructed with *kw- rather than *kʷ-), but maybe. Kroonen does seem to assume the dissimilation should apply to the voiced labiovelar too, but I'm not sure what the basis for that is as he doesn't cite any examples.

  30. German Dziebel said,

    October 9, 2024 @ 8:13 pm

    @TR. "I'm not sure what the basis for that is as he doesn't cite any examples." He uses the dissimilation in the voiceless series as the basis for his belief and then reacts to the lack of a labiovelar in Germ *kalbi- 'calf' to cast doubt on δελφῡ́ς.

    I tend to suspect that PIE *gʷelbho- is cognate with PIE *g'elt- (Germ *kelþa- 'child, womb', Skrt jártu 'vulva') and *kalbi- is from *g'elbho-. But I don't see any motivating factor for the labio-palayal alternation here.

  31. German Dziebel said,

    October 10, 2024 @ 5:16 am

    @Chris, @TR. Would it be phonetically plausible for a (palato)velar to assimilate to a non-contiguous labial stop and become a labiovelar. Hypothetically, for the sake of an argument, PIE *gʷelbho- 'womb, fetus, calf' < *g'el- with the addition of -bho-affix also seen in Gl elaphos 'deer', Germ *lamba- 'lamb'. Or PIE *gʷrebh- (Gk brephos 'kid', Slav *žrěbę 'foal', etc.) p, b, ph in IE centum languages (Skrt asva ~ Lat equus) – cna this pattern be phonetically plausibly extended to non-contiguous palatovelar + labial stop situations?

  32. German Dziebel said,

    October 10, 2024 @ 5:22 am

    Again, some of it came out mixed up. Below is a corrected segment.

    "Or PIE *gʷrebh- (Gk brephos 'kid', Slav *žrěbę 'foal', etc.) p, b, ph in IE centum languages (Skrt asva ~ Lat equus) – can this pattern be phonetically plausibly extended to non-contiguous palatovelar + labial stop situations?"

  33. German Dziebel said,

    October 10, 2024 @ 5:23 am

    It's still not pasting correctly…

  34. Chris Button said,

    October 10, 2024 @ 7:13 am

    @ German Dziebel

    Unlike with Old Chinese, and more broadly Proto-Tibeto-Burman when covering language groups I am personally familiar with (Burmish, Kuki-Chin), I do not have the knowledge to mess with individual PIE reconstructions without toppling the house of cards.

    Just a few more general comments I can perhaps offer up if at all helpful:

    – It's great that you're asking questions about phonological and phonetic plausibility. Sometimes historical linguists can posit algebraic reconstructions without properly considering linguistic reality (in fact, you could even say the PIE two-vowel e/o system suffers from that–surely it would just be schwa/a in terms of any viable system).

    – Given the right conditioning environment, theoretically any sound change is possible. The question is does that conditioning environment really exist and how plausible is it (Pamela's question in the comments here hits the nail on the head in that regard: https://languagelog.ldc.upenn.edu/nll/?p=66278).

    – If you have identified the right environment (as you claim, albeit shakily, above), then perhaps you can challenge some of the accepted orthodoxy to address why a seemingly untoward sound change might regularly (unless discrepant as say in a loanword) occur. For example, would the uvular hypothesis better account for your proposal by knocking palatovelars out of the system?

  35. German Dziebel said,

    October 10, 2024 @ 7:56 am

    @Chris. Thank you for your comment. Could you clarify what you have in mind when you say: "or example, would the uvular hypothesis better account for your proposal by knocking palatovelars out of the system?"

  36. German Dziebel said,

    October 10, 2024 @ 7:59 am

    @TR. I looked at the data for the delabalization of voiceless labiovelars before a labial stop again and I agree with you that's it's insufficient. In addition to what you wrote, kolpos 'lap' does not have a strong semantic fit to Germ *hwalfa- 'arch, vault' (the latter likely belonging with *kwel- 'wheel, circle') and for me that's a deal breaker as I don't accept semantically loose formal equations.

  37. Yves Rehbein said,

    October 10, 2024 @ 2:18 pm

    > As a supporting proof for this inference one can cite Baltic *gmti ‘beget, give birth’ (Lith gimti, Latv dzimt, OPruss gemton) …

    One might think of … Old English nominative wilcuma, chiefly West-Germanic. That is in a patriarchal, patrilineal and patrilocal society (w.r.t. to Laura Morland, above).

    > … that must be going back to *gʷem- (no connection to PIE *gʷen- ‘come, step’ (Lat venio:, Gk baino:, etc., …

    Interesting. It's a disengenious gambit. *gʷem- "come, step" is widely cited. To sepparate *gmti from it does not exactly support the argument towards **gʷen- unless there is more behind it. To be clear, if Balto-Slavic is unique evidence of **gʷem-, it would be difficult to argue that this must be Proto-Indo-European rather than innovation, as in the following

    > … with assimilation creating stems such as Germ *kwemaną [Dziebel]

    I missed that part of the argument at first. As a "forbidden" root shape as you say (or "phonological conspiracy" in terms of Andrew Byrd) a systematic methodical approach to the argument would benefit from a separate root which shows palatalization and a similar meaning.

    The discussion so far has suggested that much.

    That would be *ǵʰengʰ- or *ǵʰeh₁- (not palatal in LIV: *gʰeh₁-), but I'm very uncomfortable with this because it is involved in the suppletive paradigm of Germanic "go", which is suppletive with *h2ey-.

    One might also think of *ǵneh₃- "know". I mean something like **ǵnéh₃t (perfective) "to know; to get to know, to meet": aorist ἔγνων, अज्ञात् (ájñāt), kñā- https://en.wiktionary.org/wiki/Reconstruction:Proto-Indo-European/ǵnéh₃t I do not know how they know the second sense. But it does include "to know carnally".

    I hope that's sufficiently clear vis-a-vis semantics. I understand when somebody speaks of "an entirely separate root" they usually mean something like race versus race. I am agnostic and would totally compare that, beginning with Etruscan rasna, Latin gēns. Here be dragons.

    For a problematic counter-example consider *ǵem[H]- as in monogamy [PS: as indeed you have]. It is fairly close to *dʰéǵʰōm (cp. groom). On the other hand it looks like *ḱóm to me (cp. German Gemahl "groom", Mahl "meal, similarly companion, which I reject; cp. Latin comes).

    As for the labial versus nasal, consider climax, κλίνω, *ḱley-, *ḱel- (potentiall in the name of Hallstadt) versus hang, Hang ("slope, hill, hill side"), *ḱenk-. We might also add campus, Campania respectively Rome. More to the point, one might wonder if to hail from may be comparable with Tocharian B keṃ "earth, ground", keñiye "thereof" (WT) or "of a country, land":

    śaumo kañiye [sic!] rīne śem

    "a man came from the country" (Adams, A Dictionary of Tocharian B)

    [man.SG country.NISBA city.RECTUS(?) come.PAST]

    Incidentally, śem is the past tense of käm-. śaumo is not related to any of these. rīne is found under rīye "city", but Adams and Wiktionary give conflicting opinions, the latter refering to 里 /*rɯʔ/ (Zhengzhang) or /*(mə.)rəʔ/ (B&S)! So I feel reminded of merx, confirming my slight Etruscan bias.

    [PS: TL;DR: The pattern is not quite regular, since it has no clear starting point. I show four odd examples that don't exactly fit the pattern: *gʷem-, *ǵʰengʰ- , *ǵem[H]-, *ḱenk-. For one, semantics is subjective, whether Balto-Slavic *gmti- is "isolated". I would like to concede the point, but it remains questionable. Secondly, PIE morphology is heavily biased by the Grecko-Aryan model, which does not agree so much with Anatolian, whether Germanic is closer to the one or the other. Thirdly, I share doubts about *ǵem[H]- but we do not come to agree on a solution. Fourth, *ḱenk- takes precence because it is attested in Hittite. Finally, Tocharian and Etruscan serve as sanity check, which must fail because … I would have to be insane to go there. I just hope the text mark-up checks out this time.]

  38. Yves Rehbein said,

    October 10, 2024 @ 3:46 pm

    The previous comment was written on the commute and delayed by some time. In the meantime I have written another response.

    > As for "Gk βᾰσῐλεύς 'chieftain, overlord' (really the Greek kunungaz) can be interpreted as based on *gʷNti- 'wife, woman'" — what is this *gʷNti- and how do you account for the -λ-? [TR]

    Beekes considers -leus pre-Greek also in Achileus.

    A Lycian qλdãns is discussed by Romain Garnier (2022, Lord Qλdãns and related forms), who suggests PIE *ḱuh₁-er-i-yó-on- > Proto-Anatolian *ḱuʔ-er-iyṓ(n) (p. 196) and investigates the suffix and potential cognates, compare κῡ́ρῐος, church. Does this match Myc. qa-si-re-u, basilica? https://www.jstor.org/stable/27257699?seq=15 The sibilant is problematic for this comparison.

    It may be notable, at any rate, that Linear B ⟨re⟩ is rather similar to Cypriot syllabary ⟨sa⟩ (as in wa-na-sa "queen"), a distant congener to Linear B, and variants of this shape elsewhere (debatable).

    As a parallel to *-Nti consider Proto-Anatolian *Hant-

    "Especially the semantics of the ‘front’ gave rise to denominative verbs such as ‘to place to the front’, ‘to order’ (Oettinger 1979a:367, HEG A-H:149-153, contra EDHIL:290-291), ‘to be king, to rule’ via ‘having the status of the front’ (see Rieken & Sasseville 2014a)." https://www.ediana.gwi.uni-muenchen.de/dictionary.php?lemma=1065

    Compare, however, ἀντί, ἀμφί. Sigmatism is not obvious to me in this case. Αζα- from Scythian doesn't really help, nor Celtic Ambaxtos.

    Then compare Greek γυγαί and Gyges versus Lydian kuka- (Gyges, the tyrant(!) of Gordion), Luwian huha- "grandfather" (so memorable, thanks to Al Pacino!), Lycian xuga (admittedly not a perfect match on qλdans), Carian (šr)quq (much better), Hittite ḫuḫḫā, PA *Héu̯h-eh2-/*HuH-éh2- https://www.ediana.gwi.uni-muenchen.de/dictionary.php?lemma=1500

    See also Kloekhorst on *h2 respectively Akkadian ⟨ḫ⟩ (I am not up to date with his research, though). [PS: See especially Akkadian numerals "one", "first"!]

    NB: Haftbefehl, "Chabos wissen wer der Babo ist". This is supposed to be Romani or there abouts. I guess it could be akin to baba "(Hinduism, Islam, Sikhism) A holy man, a spiritual leader." https://en.wiktionary.org/wiki/baba A lull word.

    TL;DR: It is implied that *h2 hardened in some way. I didn't read Margalit Finkelberg's theory.

  39. German Dziebel said,

    October 10, 2024 @ 4:56 pm

    @Yves. Thanks for extended comments.

    AFAICS, *ḱenk is reconstructed with a palatovelar only because of Skrt śaṅk but śaṅk means 'to doubt' and its belonging with Hitt kānki ~ kankanzi 'hang' and Germ *hangana 'hang' can be questioned.

  40. David Marjanović said,

    October 10, 2024 @ 7:08 pm

    BTW the semantic derivation doesn't seem straightforward either: are there languages where the basic word for "woman" comes from "one who gives birth"?

    Well, the basic word was *sōr (*|sor-s|) before *gwēn (*|gwen-h₂|) ousted it.

    It's still not pasting correctly…

    Replace every < by &lt; and (just to be sure) every > by &gt; so the text doesn't get parsed as nonexistent HTML tags and deleted.

    kolpos 'lap' does not have a strong semantic fit to Germ *hwalfa- 'arch, vault'

    The accent of kólpos doesn't fit either: it's *hwalba- with Verner (German wölben, Wölbung, Gewölbe).

    climax, κλίνω, *ḱley-, *ḱel- (potentiall in the name of Hallstadt)

    No. The etymology of Hallstatt*, Hallein, Bad Hall, Bad Reichenhall etc. etc. etc. is well known to specialists (even though it hasn't made it to the general public that still thinks it's basically Welsh).

    * Yes, with tt. None of that Middle Dutch dt phenomenon here.

    It is implied that *h2 hardened in some way.

    Yes, that's apparently a western/late Anatolian phenomenon.

    BTW, I'm not linking to Kloekhorst's long Academia page because there are already two links in this comment and I want to keep it out of moderation.

    can be questioned

    Literally everything can be questioned; that's not interesting. The interesting question is: which hypothesis is the most parsimonious?

  41. David Marjanović said,

    October 11, 2024 @ 4:20 am

    Heh, it got into moderation anyway – two links are enough to trigger that. Anyway, direct link to Kloekhorst in all his glory.

  42. German Dziebel said,

    October 11, 2024 @ 7:33 am

    @David. "Well, the basic word was *sōr (*|sor-s|) before *gwēn (*|gwen-h₂|) ousted it." There's not enough evidence to postulate PIE *sōr. Luw hassu-sara is the only true piece of evidence that there was any *sōr anywhere. The rest of the evidence (e,g., Lat uk-sor 'wife') comes from misanalysis. Lat uxor, Arm amusin, Germ *wi:ba are all from *woyk'- 'house, etc.'

    On the other hand, Lyc lada 'wife' is probably from *gwlndeH2 (with a regular loss of gw- and with n…n > l/r…n dissimilation similar to Toch kuli 'woman', Aln grua 'wife' and Lat mulier 'woman'). It's an extended stem, to be fair, but probably the same as -tti- in Luw wanatti 'wife.

    "Literally everything can be questioned; that's not interesting. The interesting question is: which hypothesis is the most parsimonious?"

    Superficial similarity between forms without a firm semantic alignment is probably not a good entry into a parsimony contest in the first place.

  43. German Dziebel said,

    October 11, 2024 @ 7:44 am

    @David. "The accent of kólpos doesn't fit either: it's *hwalba-". Could that be a case of Greek retraction? Typically Verner is identified through Sanskrit accent.

    @TR, David. The only way to reconcile the semantics of *hwalba 'vault, arch' and *kolpos 'lap' if, correspondingly, PIE *kwel- 'wheel', circle' and Balto-Slav *kwel- 'knee, lap' (Slav koleno) are related under the seme 'turn (of objects), bend (of body parts).' Welsh olwyn 'wheel' < PIE *HoHl- 'elbow' would support this.

  44. David Marjanović said,

    October 11, 2024 @ 10:50 am

    Luw hassu-sara is the only true piece of evidence that there was any *sōr anywhere.

    Define "true"…? Evidence – I didn't say proof, which is for mathematics rather than science, I said evidence – is all over the place: sister > *swé-sōr "woman from one's own clan", the feminine forms *tísres & *kʷétesres, the list of Indo-Iranian words and suffixes listed on pp. 133–134 of this paper, and the Germanic phenomena the paper is about.

    Could that be a case of Greek retraction?

    Do you mean the regularization of verb accent that is not supposed to have happened in nouns? Or could kólpos be a tómos-type noun derived from an end-stressed adjective?

  45. Jonathan Smith said,

    October 11, 2024 @ 11:47 am

    IDK about IE, but similar non-random cooccurrence restrictions~limitations attributable to phonotactic constraints~preferences must be everywhere. But IMO it's tough to claim that such-and-such a disprefered structure is necessarily illegal, secondary, or what have you.

    E.g., /(C)wVw/ and /(C)jVj/ syllables are completely (?) illegal in Mandarin… but some people/places say (pinyin) yai2 'cliff' ("standard" ya2). In Cantonese (which retains consonantal codas), /PVP/ has been totally (?) removed inclusive of /(K/G)wVP/… but there seem to be (jyutping) jai, jeoi. Broadly parallel phenomenon exist in English — "wow" "yay" "mom" etc. are often ideophonic/nursery words and the like — less good as "normal" words/syllables. Clusters may amplify this tendency in statistical terms (e.g. no "spumt" or something like that — probably not even as nonce ideophone or the like.)

    But these are all just stastical tendencies.

  46. German Dziebel said,

    October 11, 2024 @ 2:19 pm

    @David. Thanks for the link. This idea has become popular in the past decade with a couple more papers out. It seems that people really want to reconstruct PIE *so:r and they see it where I don't necessarily see it (e.g., *swesor can just be an old -s- stem as in snus-os, glo:s-) or they push InIr data too far into PIE. I think Skrt *stri: may come from *sestri:, comp. Slav *sestra 'sister, without -w-, with the generalization of meaning from sister to woman. Baltic did it with PIE *meH2ter 'mother' > Lith moterys 'woman'. PIE *gwenH2 is just so robust as a PIE etymon that compared to it *so:r is all over the place a bit.

  47. xiesong said,

    October 11, 2024 @ 8:47 pm

    From an angle of Chinese, King and Queen would be easily viewed as 乾 qián and 坤 kūn,the top two of the Eight Diagrams (八卦). Chinese classic I Ching (易經) interprets 乾 as being sky (天)、king (君)、father (父), male (男), etc.. Modern pronunciation of 乾 qián is the result of palatalization of traditional kian sound ( from k to q, similar to English tʃ ). 坤 kūn in I Ching means earth (地), queen (后), mother (母), female (女), etc.. Its traditional pronunciations range from kun, kwan, to khuen among different dilect areas but overall very close to modern kūn.

  48. German Dziebel said,

    October 12, 2024 @ 5:02 am

    @xiesong. Fascinating! qián and kūn are not cognates, are they?

  49. German Dziebel said,

    October 12, 2024 @ 5:25 am

    @Jonathan. "IDK about IE, but similar non-random cooccurrence restrictions~limitations attributable to phonotactic constraints~preferences must be everywhere. But IMO it's tough to claim that such-and-such a disprefered structure is necessarily illegal, secondary, or what have you."

    There can be confounding factors that interfere with the smooth operation of a constraint. Or maybe there are rules that override other rules in some situations. But I tend to favor a position where constraints are real. In a similar to the original post's IE context, IE n and m maintain their polarity observed in their interactions with different velars in the contrasting ways in which they assimilate to labials vs dentals. Across branches, nP(h) > mP(h) and mT(h) > nT(h) where P and T stand for any labial or dental stop. But in Baltic languages the second rule does not operate. So we get Balt *śímta '100' next to Lat centum and Germ *hundą (with assimilation) < PIE *ḱm̥tóm.

    So the linguistic system continues to reinforce in the articulatory polarity of n and m. To my knowledge, it does not do it to w/y or l/r.

  50. German Dziebel said,

    October 12, 2024 @ 5:30 am

    Next to Skrt śaṅk 'doubt', Hitt kānki ~ kankanzi and Germ *hangana 'hang' there's a case of Skrt śram 'tired', Gk κρεμάννῡμῐ 'hang' where the proposed attraction of m to palatovelar is supported. But ironically there's a semantic weakness of this comparison, too.

  51. xiesong said,

    October 12, 2024 @ 11:00 pm

    @German Dziebel, I don't know if I understand the word "cognate" correctly. If cognate means "genetically related," 乾坤 could be cognates for they are phonically close and semantically opposite. But the best cognate for 乾 is 健 jian,meaning "strong, healthy and brave." I Ching has manifested the hexagram of 乾 as "天行健,君子以自强不息." Google translate it as "Heaven is in good health, and a gentleman strives for self-improvement," which is not a perfect one. Someone translate it as " Just as the celestial bodies run energetically without stop, a gentlem should strive for self-improvement without rest." Other words like 坚(堅) jiān, meaning "hard;solid;firm" and 劲(剄) jìng, meaning " strong; powerful; forceful" are definitely cognate with 乾.

  52. German Dziebel said,

    October 14, 2024 @ 10:34 am

    @TR. On Gk kolpos and karpos 'wrist' once again.

    It's quite possible that those two forms are just an instance of boukolos-rule whereby two labiovelars are avoided/dissimilated. In Germanic where the first labiovelar has been ascertained (*hwelfan and *hwerbana) the second labial can be from a labiovelar as well as the root-type (2 labials and a resonant) fits the type of root conducive to the labialization of labiovelars (Germ *fimfe '5' &lt *penkʷe-, *wulfa- 'wolf' &lt *wlkʷo- and *twalib- ‘12’ (Goth twalif, twalib-) &lt *-likʷ- (cf. Lith. dvylika). This would make *hwelfan (&lt *kʷelkʷ-) and *hwerban (&lt *kʷerkʷ-) reduplicated roots. They are of a different type from *kʷekʷlo- 'wheel, neck' but their existence is nevertheless semantically understandable since the referents are the same notions of "turning", "curving", etc.

  53. TR said,

    October 14, 2024 @ 11:18 am

    It's an interesting idea but I don't think there are any known PIE reduplicated forms of the type C1VC2-C1-, and AFAIK that type of reduplication pattern is cross-linguistically either very rare or nonexistent.

  54. German Dziebel said,

    October 14, 2024 @ 12:54 pm

    @TR. Such forms are well documented in IE languages – they do reduplicate segments but for some reason they are not counted as reduplicatives as a morphoogical type.

    Comp., e.g.,
    Attested.
    Gk κίρκος, κρῐ́κος 'circle, ring'.
    PIE *kᵂl(H)kᵂ, Lat calx 'heel' (with delabialization and coloring as in canis 'dog' caseus 'cheese'), SCr kuk 'thigh, hip', Bulg кълка 'hip', Latv kulksnis 'hock', Lith kulksnis 'ankle', OPruss culczi 'hip'. (next to normally reduplicated Lat pople:s 'knee', Lith kaklas 'neck' and unreduplicated BS *k(w)el-/k(w)ol- 'knee' and metathesized *olkᵂ- 'elbow' (Slav *olkъtь).
    Inferred.
    Germ *halsa- 'neck' (*hᵂolhᵂso-) (inference based on Lith kaklas 'neck'), *hursa 'horse' (*hᵂrhᵂso-), Lat curro:, Toch B kwarsar, etc. The Latin reflexes may be questions as typically RKS sequences turn into RS sequences without -RS- shifting to -RR-.

    When I worked on kinship universals it was interesting to observe that languages typically lexicalize all the structurally possible slots (at different frequencies in a worldwide sample) and those structures tend to be asymmetrical with one another. In our case here, unreduplicated stems CVR- (common) alternate with metathesized stems RVC (rare). Then fully reduplicated stems CVRCVR (e.g., Hitt hulhuliya 'fight'), sometimes broken by dissimilation or metathesis (Lat cancer < *karkino-, Gk dendreon < *der-drewon) "split" into 2 possible asymmetrical types: CVCR (e.g., PIE *kwekwlo- 'wheel') vs. CVRCV (Gk κίρκος).

  55. TR said,

    October 14, 2024 @ 1:17 pm

    As far as I can tell those etymologies are all quite doubtful (either your proto-forms themselves and/or the forms as instances of reduplication). It would take a lot more robust evidence to convince me that such a pattern exists at all, let alone being well-documented.

  56. German Dziebel said,

    October 14, 2024 @ 1:57 pm

    @TR. First, we may have digressed. The presence of 2 labiovelars in a stem does not mean reduplication (boukolos is not a reduplicated form but a compound). I may have opened a can of worms by referring to *kʷelkʷ-and *kʷerkʷn as reduplicatives. But they do like ones and they come from a semantic domain where reduplication is known. The main point was that kolpos and karpos may be instances of a known law (i.e. boukolos) and don't require another one (loss of labiality by a labiovelar next to a labial stop). Second, forms such as κίρκος, κρῐ́κος and Germ *hringa- or calx and kulksnis "reduplicate" the same consonant, so technically they are just as reduplicative as κύκλος, pople:s or kaklas. Some of the etymologies aside, the facts above are descriptive in nature and won't change regardless of an etymological solution.

    Otherwise my previous comment was a mixed bag of novel and established etymologies, which made it confusing, I agree.

  57. David Marjanović said,

    October 15, 2024 @ 10:01 am

    (e.g., *swesor can just be an old -s- stem as in snus-os, glo:s-)

    Then where does the extra *-r come from?

    I think Skrt *stri: may come from *sestri:

    That would be irregular. The zero-grade of *ses- "sleep" in Skrt and in Hittite is sas-, not just s-. The zero-grade of *ped- "foot" is *-bd- only after prefixes; word-initially, it would probably have become *d-, which was replaced by *ped- wholesale.

    Slav *sestra 'sister, without -w-

    I don't know how that happened, BTW.

    PIE *gwenH2 is just so robust as a PIE etymon that compared to it *so:r is all over the place a bit.

    Sure; the idea is that it – well, phonologically regular *gʷēn, remodeled to morphologically regular *gʷenh₂ or *gʷeneh₂ in various branches – was well established in PIE but had only recently replaced *sōr in that role.

    I may have opened a can of worms by referring to *kʷelkʷ-and *kʷerkʷn as reduplicatives.

    Yes :-)

    BTW, just to make sure: the "two labiovelars" in boukólos < *gʷow-kʷólos are not the *gʷ and the *kʷ, but the *w and the *kʷ: *|wkʷ| surfaced as /wk/, *|kʷw| surfaced as */kw/ (including *[ku]).

  58. German Dziebel said,

    October 15, 2024 @ 6:16 pm

    @David.

    "Then where does the extra *-r come from?"

    It's all over IE kin terms. E.g., among the non-ter terms Lat uxor from *woyk'sr or (thematized) PIE *swekuro- 'husband's father'.

    "I think Skrt *stri: may come from *sestri. That would be irregular."

    Even pre-accentually?

    "Slav *sestra 'sister, without -w-…I don't know how that happened, BTW."

    Neither do I but distributionally it's broader than presently thought, I think. Hitt sas-, Skrt ses- sleep' next to PIE *swe-p-no next to Germ *s(w)le:pan 'sleep'. All maybe related to *swe- 'self'.

    "the "two labiovelars" in boukólos < *gʷow-kʷólos are not the *gʷ and the *kʷ, but the *w and the *kʷ:"

    How do we know this, BTW? I tend to think that dissimilation comes from the second labiovelar not the u. boukolos, kuklos. There are other phonetic processes that are sometimes lumped under boukolos-rule, e.g., Germ *kwemana yielding *kumana- 'come' (with syllabic M turning um and delabializing preceding kw) but that seems different. This occurs even without a labiovelar. Lat uxor from *wu:ksor from *woyk's-r.

  59. German Dziebel said,

    October 15, 2024 @ 6:19 pm

    In Germanic labials, including labiovelars, often develop -u- before them: *augan 'eye', *haubuda ' head'. A poorly understood phenomenon.

  60. German Dziebel said,

    October 16, 2024 @ 5:53 am

    @David. Re: *sweso:r. If we look at Anatolian, it's contrasted with Nuclear IE by having Hitt neka, Luw nanasri (from *na(g)nasri ?), Lyc A nere/i (from *ne(g)nehri from *ne(g)nesri). I interpret it as derived from Anat *H2en-/*H2em- 'give birth, beget' (as in *H2ne(g)n-, which could be a reduplication with dissimilation from *HneHn-). if we use it as an etymological model, then it would make sense to interpret *swe- in *sweso:r as identical with PIE *suH- 'give birth' that yielded *suHnu-, *suHyu-, *sukter 'son, brother's son'. Luw -sri- and *-so:r indeed must be identical but I would see them as two different affixes -s- and -r-.

  61. David Marjanović said,

    October 16, 2024 @ 1:49 pm

    Lat uxor from *woyk'sr

    Oh, look, there's *sr right there…

    Even pre-accentually?

    Oh yes. "They're sleeping" is sasánti (Skt), sasanzi (Ht).

    Hitt sas-, Skrt ses- sleep' next to PIE *swe-p-no next to Germ *s(w)le:pan 'sleep'. All maybe related to *swe- 'self'.

    Do you think that's currently testable? You're going back well before PIE if you take the coda off a content-word root.

    There is the idea that *ses-, the only root with two identical consonants, is the baby-language version of *swep-, though.

    There are other phonetic processes that are sometimes lumped under boukolos-rule, e.g., Germ *kwemana yielding *kumana- 'come' (with syllabic M turning um and delabializing preceding kw) but that seems different.

    That is the boukólos rule. Given that it already explains boukólos and kúklos, what do we need your extra dissimilation for?

    In Germanic labials, including labiovelars, often develop -u- before them: *augan 'eye', *haubuda ' head'. A poorly understood phenomenon.

    If you have any more examples, I'm all ears – ears (*aus/zan-) being currently blamed for the *u in *augan-.

    This occurs even without a labiovelar. Lat uxor from *wu:ksor

    Bans on the sequences [ji] and [wu] are extremely common worldwide; they don't need any further explanations. Especially not in early IE where [w] and [u] were a single phoneme and long consonants were forbidden.

    it would make sense to interpret *swe- in *sweso:r as identical with PIE *suH- 'give birth' that yielded *suHnu-, *suHyu-, *sukter 'son, brother's son'.

    Yes, that would make sense, but it would leave the *é entirely unexplained.

    two different affixes -s- and -r-

    …of which at least *-s- would be unexplained.

  62. TR said,

    October 16, 2024 @ 5:47 pm

    I tend to think that dissimilation comes from the second labiovelar not the u

    There are other boukolos examples without two labiovelars: ὑγιής, εὔχομαι, etc.

    There's no direct evidence that uxor ever had *w-, BTW. If it's from *woik-, wouldn't we expect **vīxor, like vīcus : οἶκος? You'd also need a double o-grade to account for the second vowel.

  63. German Dziebel said,

    October 16, 2024 @ 8:12 pm

    @David.

    "Oh, look, there's *sr right there…"

    And? There's also one in *h₂ews-ro 'dawn' and elsewhere. Are all those sr's names for 'woman'? Germ *wi:ba 'wife' from *woyk'wo and Arm amusin 'spouse' from *sm-woyk'ino don't have -s-. THis suggests, albeit indirectly, that -s- was a morpheme just like -ino and -wo- and not a lexeme meaning 'woman'. Gk oar 'wife' could also go to *woy(H)sr (syllabic R is identical in oar and uxor), with a loss of -k- for which there's another rule but PIE *wiHro- 'husband' *woyk's- > *woyk'sr.

  64. German Dziebel said,

    October 16, 2024 @ 8:15 pm

    It didn't come out right.

    Trying again.

    "Oh, look, there's *sr right there…"

    And? There's also one in *h₂ews-ro 'dawn' and elsewhere. Are all those sr's names for 'woman'? Germ *wi:ba 'wife' from *woyk'wo and Arm amusin 'spouse' from *sm-woyk'ino don't have -s-. THis suggests, albeit indirectly, that -s- was a morpheme just like -ino and -wo- and not a lexeme meaning 'woman'. Gk oar 'wife' could also go to *woy(H)sr (syllabic R is identical in oar and uxor), with a loss of -k- for which there's another rule but PIE *wiHro- 'husband' < *woyHro clearly doesn't have -s- but still has -r- and hence no 'woman' formally or semantically.

    "That is the boukólos rule." Yeah, but I think people attribute delabialization to a wrong phoneme. The examples are right though. And when we add kolpos and karpos, we consolidate 2 rules under one. There are also potential examples for labiovelar dissimilation in Germanic.

    "if you have any more examples, I'm all ears." *raubana 'rob' ~ Lat rapere 'seize'. You can look up "A new sound law for Proto-Germanic?" by Dimitri Mikhailovich Pisarev on Academia.edu.

    "Yes, that would make sense, but it would leave the *é entirely unexplained."

    Why? u is zero-grade for *we. Its the long vowel in *suHnu- that does not have an easy parallel in *swe.

    "…of which at least *-s- would be unexplained."

    -s- is found across nominals (kin terms or not). It's an old s-stem marker, the current thinking goes. So *woyk'os to *woyk's- to *woyk'sr.

  65. German Dziebel said,

    October 16, 2024 @ 8:34 pm

    @David. In the syncope in Indo-Iranian see Kummel 2020. https://www.italian-journal-linguistics.com/app/uploads/2021/05/9_Kuemmel.pdf

    I may need to modify my proposal to *sustri > *stri as he only describes it for high vowels.

  66. German Dziebel said,

    October 17, 2024 @ 7:39 am

    @David.

    "Do you think that's currently testable? You're going back well before PIE if you take the coda off a content-word root."

    I had to think about it. It's a tradeoff between leaving intact small isoglosses such as *ses- or *sle:pana that will remain limited to 1-2 branches and connecting them together into a word family with a coda-level variation. The first approach is a blind alley. If ses- and sle:pana are innovations, then why are they etymologically obscure? If they are older than *swep-, then why do they still fit the same syllable structure as *swep-? And why neither of them are semantically different from each other. PIE *drem- 'light sleep' (Slav *dremati 'nap') and *swep- 'deep sleep' (Lat sopor) with further post-PIE meaning generalization are semantically differentiated. But *ses- and *sle:pana semantically are just like later, post-PIE *swep-.

    Will Pre-PIE etymologies be ever as testable or compelling as PIE etymologies? Maybe… with time. But that's a subjective thing. Objectively the material is decomposable into more granular units when semantics is maintained the same. E.g., Anat *h2ans-/*h2ams- 'beget, give birth' can be compared to Hitt hartu 'descendant' with a resulting H2en- ~ H2er- variation. Same for PIE *g'el- 'fetus, child' (Skrt jartu, Germ *kilth-) next to PIE *g'en- 'beget'. Also PIE *gwel- 'fetus, womb' (Gk delphus, Germ *kalba-, with *g' and not *gw) and *gwen- 'wife, woman'. The 4 PIE sets will be different across 2 dimensions – resonant type and labialized vs. palatalized velar. Should we leave them alone forever or should we try to model their diversification from a single pre-coda etymon with rules governing the alternations? It may be a job for AI. :)

    In addition to assessing phonetic variation between roots, one would need to assess if there's an excess of roots with the same meaning but lopsided distributions at the PIE level.

  67. David Marjanović said,

    October 19, 2024 @ 6:56 am

    Oh, look, there's *sr right there…

    And? There's also one in *h₂ews-ro 'dawn' and elsewhere.

    *-ró- is a well-known suffix. Why would it suddenly lose its stress and its vowel and not be followed by an ending?

    Germ *wi:ba 'wife' from *woyk'wo and Arm amusin 'spouse' from *sm-woyk'ino don't have -s-. THis suggests, albeit indirectly, that -s- was a morpheme just like -ino and -wo- and not a lexeme meaning 'woman'.

    No, it suggests different branches constructed words for "wife" from "household" separately, in three different ways, and one of them was compounding with *sor- while the other two were derivations with adjective-forming suffixes.

    …if uxor really is from *wojkʲ-, on which I defer to TR.

    PIE *wiHro- 'husband' < *woyHro

    Tocharian wir means "young". By being an adjective, it explains why *wiH-ró- is stressed like an adjective and has an adjective-forming suffix. The Indo-Actually-European reflexes don't seem to mean "husband", but to be one of those patriarchal "man ~ warrior ~ hero" words; late zero-derivation from "young" ("a youth") makes sense for that, less so for "husband".

    *raubana 'rob' ~ Lat rapere 'seize'. You can look up "A new sound law for Proto-Germanic?" by Dimitri Mikhailovich Pisarev on Academia.edu.

    Thank you, I will!

    In the meantime, please check out section 3.11 of this open-access book chapter. It's on the Germanic and Italic "head" word and the mess that surrounds them all the way to India. The extra *-w- is optional even within Germanic, but doesn't seem to occur anywhere else, so Pisarev may well be right.

    Yes, that would make sense, but it would leave the *é entirely unexplained.

    Why? u is zero-grade for *we. Its the long vowel in *suHnu- that does not have an easy parallel in *swe.

    That way you'd have a verb root *swe- which wouldn't end in a consonant. That's not supposed to be possible. Postulating *sweh₁- to fix that would account for *suH-nú- and its root cognates, but it couldn't account for *swésor-; neither the Saussure effect nor any other sound law I can think of would delete *h₁ in that position.

    …actually, the Wetter-Regel could if it's real: *sweHsr- should become *swesr-. But unambiguous examples have been hard to find, and the semantics of this word would be difficult as well.

    I may need to modify my proposal to *sustri > *stri as he only describes it for high vowels.

    Interesting; but a double-zero-grade *sustri should be quite a bit younger than PIE, and a root *swes- with two identical consonants would need some explaining as well, a verb root *swe- even more so as mentioned above.

    If ses- and sle:pana are innovations, then why are they etymologically obscure?

    *ses- makes sense as a baby-language version of *swep-, as mentioned; that would in particular explain why it so blatantly violates the root constraint I just brought up. The Germanic word, already suspect because of its *p, forms part of a Kluge mess (German schlapp and schlaff "low-energy"), and then there's Slavic slab- "weak".

    Should we leave them alone forever or should we try to model their diversification from a single pre-coda etymon with rules governing the alternations?

    I'm not saying it can't be done, and I'm not saying the apparent patterns can't be real; I'm saying the time when any such process was productive must have been well before PIE, so within PIE we're dealing with the observed situation of similar but synchronically unrelated roots.

  68. David Marjanović said,

    October 19, 2024 @ 9:42 am

    Pisarev's presentation is here. A few of the examples are better explained in other ways, but the rest holds, as far as I can see. I'm going to call it Pisarev's law.

    (One quibble about the data: German Rahm is claimed to be used mostly in compounds, but in reality that varies regionally. Where I'm from, it says Sauerrahm on the cups sold in the supermarkets, but we only ever say Rahm; cream in general has other names. However, I have no idea how old the geographic distribution is or how it developed.)

  69. German Dziebel said,

    October 19, 2024 @ 11:37 am

    @David. "*-ró- is a well-known suffix. Why would it suddenly lose its stress and its vowel and not be followed by an ending?" I'm not saying it's the same -r- in uxor, oar as in *ausra. I'm just saying that there's no need to interpret -sor as a noun meaning woman when both -s- and -r- are known as separate morphemes (e.g., Gk ane:r, Arm ayr, Gk damar 'wife' or the other ones I mentioned earlier).

    "Tocharian wir means "young"." There's no indication of it being a cognate of Lat vir 'man, husband', Lith vyras 'husband' and Skrt vi:ra 'man, husband' outside of it being their look-alike. PIE *weyHro- is a term for (adult, married) males and as such it can hardly be a derivative from "young".

    "but a double-zero-grade *sustri should be quite a bit younger than PIE, and a root *swes-". That fits the model of it being derived from *sweso:r and restricted to Indo-Iranian.

    "The Germanic word, already suspect because of its *p, forms part of a Kluge mess (German schlapp and schlaff "low-energy"), and then there's Slavic slab- "weak"."

    "No, it suggests different branches constructed words for "wife" from "household" separately, in three different ways, and one of them was compounding with *sor- while the other two were derivations with adjective-forming suffixes."

    Technically yes, but since other IE branches time and again used morphemes to derive 'wife' from *woyk'- it's less likely that one of them settled on a lexeme. To be fair, Gk oar could be from *woysr.

    Germ *sle:pan is probably part of several isoglosses including Slav *svoboda 'freedom, village', *sloboda 'same' (*swloboda), Germ *selban 'self' ( "weak" – all built around *swe- 'self'.

    "if uxor really is from *wojkʲ-, on which I defer to TR." It's a new etymology but it's defensible (unlike a dozen old ones). There's a precedent of deriving Arm amusin from *woyk-'.

    "Postulating *sweh₁- to fix that would account for *suH-nú- and its root cognates, but it couldn't account for *swésor-" Another option is *suHnu- < *sHunu- with a well-documented laryngeal metathesis. But that would change the whole *swe- etymon.

  70. TR said,

    October 19, 2024 @ 12:21 pm

    German, how do you account for the different vowels of uxor and vi:cus (sorry, can't do macrons on my phone) if you derive both from *woik'-?

  71. German Dziebel said,

    October 19, 2024 @ 1:42 pm

    @TR.

    "how do you account for the different vowels of uxor and vi:cus."

    Must be from *woyk' and *weyk'- respectively.

  72. German Dziebel said,

    October 19, 2024 @ 1:49 pm

    vi:cus is identical to Goth weihs 'village, settlement'. uxor corresponds to Germ *wi:ba 'wife' < *woyk'wo-.

  73. German Dziebel said,

    October 19, 2024 @ 2:02 pm

    Also Toch B īke < *weiks (Adams 67).

  74. German Dziebel said,

    October 19, 2024 @ 2:12 pm

    Also, see Lat līber < *łeiβero- < *leuβero- < *leuðero- as in Gk ἐλεύθερος where e in eu diphthong is attested.

  75. TR said,

    October 19, 2024 @ 3:44 pm

    That seems to work phonologically (at least, I don't think there's any strong counter-evidence to a change *woi- > Lat. ū-; vīdī would be a counterexample if from *woid-, but it may not be). But as far as I can see you still need a second member -sor- to account for the stem ending in -ōr-, and a double o-grade seems very anomalous.

  76. German Dziebel said,

    October 19, 2024 @ 7:48 pm

    @TR. "I don't think there's any strong counter-evidence to a change *woi- > Lat. ū-."

    In fact I would argue Lat oi to u: only happens after a labial.

    sūdor < *swoyd-, uxor < *ūxor < *woyḱsr̥, mūnus, mūnia, com-mūnis < *moyn-, fūcus < *bhoiko, pūnire < *kʷoyn-, spūma < *spoHin-, prūdēns Adj. 'prudent, wise' (< *pro-woid- next to prōvidēns Partic. 'foreseeing' < *pro-weid-), cūra < *kʷoys-, ūva Lat or regularly. Can the oblique stem -ōr- be explained otherwise?

  77. German Dziebel said,

    October 19, 2024 @ 7:51 pm

    Came out corrupted again.

    After kʷoys read

    "Even Lat u:nus can be matched with Balt *vienas '1'.

    "But as far as I can see you still need a second member -sor- to account for the stem ending in -ōr-."

    Based on Gk oar, uxor has a syllabic R (R leads to Lat or regularly). Can -ōr- in oblique cases be explained otherwise?

  78. TR said,

    October 20, 2024 @ 12:12 am

    The ō of uxor is long (in the nom. sg. it was shortened in Classical Latin, but the long form still appears in Plautus); syllabic R yields Lat. -ŏr- / -ŭr-, but I don't see how it can yield -ōr-.

  79. German Dziebel said,

    October 20, 2024 @ 7:56 am

    @TR. It's good that we agree on the *woyk'- part as etymology gets out of the way. The rest is a stalemate. I understand your point. soror and uxor have indeed identical paradigms, and long o in the 'sister' forms is attested in other branches. However external evidence (damar, ear) show a syllabic R in the words for 'wife' and ear and uxor were likely derived from the same *woyk'-. Internally, there's no direct evidence for so:r 'woman' in Italic (or any other branch really with a possible exception of InIr) hence it's hard to imagine it got added to *woyk'-. So I don't have any other option but to postulate analogy from soror remaking *uxR into uxor into uxo:r and then back to uxor in Classical Latin.

  80. TR said,

    October 20, 2024 @ 12:51 pm

    I think there is one other option, namely that uxor and ὄαρ, which share nothing beyond the final /-r/, are unrelated.

  81. Yves Rehbein said,

    October 20, 2024 @ 3:55 pm

    To paraphrase E. R. Luján Martínez, Sobre la etimologia de UXOR (1996), the majority lf proposals accept an element *sór going back to Meillet. Van Windekens disagrees. De Vaan later rejects all prior proposals.

    IMHO the X of matrix has to be taken into account. Since I previously meantioned the graphematics of Phrygian ψ, see also Etruscan Ψ versus Χ, Cretan ΧΣ and more to the point Sumerian SAL. See likewise the Germanic -tree suffix, Old High German -ter conceivably with Verners law, yet not so in EWAhd vol. 1 who cite Grimm, who has not figured out Verner's. The same grapheme turned upside down stands significantly for runic Tyr. Only in Younger foothark does it appear as Yr ⟨R⟩ < *z, yr interpreted as a tree name. Further evidence comes from fotksamal in place of futskamal (cf. Anthonson) and one South Germanic Futharz instead of Futhark (cf. Beuchte Fibula in Dübel et al.). Yr would be cognate Yves, I believe. This is personal!

    I support PIE *weyḱ- in wife but have been told off before, so I shall say nothing of it. The root is problematic per se, cp. PIE *wi- "far, apart". It is everywhere, well supported by Sanskrit, yet different in Hittite u-.

    For the lulz I must note German Wichser "tosser, wanker" could be a Weichei "softy", Landei "country folk" (cf. vīcīnus "neigbour"); compare wax (WT: wichsen), which brings us back to mulier if from mollis "soft" (pace mulgeo "to milk"), as also Greek μαλάκας and, incidentally, Sumerian sal "fine, delicate" (logogram for munus "female").

    Sin cera, Yours truly

  82. Yves Rehbein said,

    October 20, 2024 @ 4:35 pm

    Correction: In Anglo-Saxon runes it appears randomly as ⟨x⟩ to render Latin. ⟨z⟩ is well attested in the Elder Futhark, assumed predating the change to ⟨R⟩.

    Beware that I evidently did not double check the references:

    * Elmer Antonsen. 2002. Runes and Germanic Linguistics (Trends in Linguistics. Studies and Monographs. 140). Ch. 5 "The fifteenth rune".

    * Klaus Düwel, Robert Nedoma & Sigmund Oehrl. 2020. Die südgermanischen Runeninschriften (Ergänzungsbände zum Reallexikon der Germanischen Altertumskunde 119).

  83. German Dziebel said,

    October 20, 2024 @ 5:29 pm

    TR. "uxor and ὄαρ, which share nothing beyond the final /-r/, are unrelated."

    They also share -o-. -o- is important because it's the ablaut grade only found in this word family in oikos and *woyk'sor. -y- is also regularly lost in both branches (in two different ways). And the meaning is the same. So all is aligned. The interesting issue is -k'- which I explain as a laryngeal hardening similar to vi:vo: but vi:xi, but it's possible that the root is *woy- and -k'- is an affix.

    Even if oar and uxor are unrelated, the lack of **so:r 'woman' in Italic does not make the proposition that Italic is the only branch that used -so:r to add to *woyk'- believable.

    Finally it's non-parsimonious to leave oar unrelated to any other form as it becomes an ultimate instance of the obscurum per obscurius fallacy. I understand if there was a solid alternative but there's none.

  84. German Dziebel said,

    October 20, 2024 @ 5:45 pm

    @Yves,

    "The root is problematic per se, cp. PIE *wi- "far, apart".

    I think this root gets in the way of people seeing *wey-/*weyk'- in a broader range for etyma, e.g., PIE *widheweH2 'widow' which should be derived from *wey- and not from wi- 'far, apart' as it is often proposed. Comp. semantically PIE *ḱei-wo-, -ro 'orphan, widow' (Skrt śayú 'orphan', Lith šeire 'widow', šeirys 'widower', Slav *sȋrъ 'orphan') next to PIE *ḱei-wo- 'husband, wife' (Lat cīvis 'citizen', Germ *hīwô 'married couple, household', Latv sieva 'wife', *ḱoi-mo- (Germ *haima- 'home, house, village', Lith šeimà 'family', Latv saime 'family, household', Slav *sěmь 'household member, serf', *sěmьjà 'family').

  85. TR said,

    October 20, 2024 @ 5:50 pm

    They don't share -o-. The idea that the u- of uxor goes back to *oi is part of your proposal, not part of the evidence. And there's nothing unparsimonious about leaving a word unetymologized — if we don't have a good etymology, that is the only parsimonious thing to do. Certainly more so than positing theories that have to add an unexplained consonant here, subtract one there, and appeal to analogy to sort out the vowel.

  86. German Dziebel said,

    October 20, 2024 @ 6:21 pm

    @TR. "They don't share -o-. The idea that the u- of uxor goes back to *oi is part of your proposal, not part of the evidence." Evidence of what? That the two forms are presently divergent? That's an expected part of evolution. they are supposed to diverge. An etymological proposal involves establishing more identities with every step back in time toward an etymon.

    w is regularly lost in Greek
    y is regularly lost in Greek
    s is regularly lost in Greek
    u: is a regular outcome of (w)oy in Latin
    s is attested in Latin

    These 4 regularities attested across the corpus create a shared *woy-..sr between oar and uxor. Now that's evidence of the coalescence of two forms in the past. And it's part of my proposal only because it's already attested on a broader etymological material, which constitutes evidence for a reconstruction.

  87. German Dziebel said,

    October 21, 2024 @ 6:58 am

    @TR. There's a Hesychian gloss οἰχμή = δούλη ('slave girl / or 'woman-slave') that can also be matched with uxor via *woyksme:. Both semantic developments – 'house' to 'wife' and 'house' to 'slave' – are common in IE.

    This gloss shows well that -s- in uxor is a morpheme. In Greek KsR > khR (where R is n, m.m l, r) is regular

  88. German Dziebel said,

    October 21, 2024 @ 8:58 am

    Comp. δμώς 'slave' (from *dmons?) ~ δάμαρ 'wife'. δούλoς , -η 'slave' from *domslo-. Next to these forms there's of course Gk δεσπότης (with morpheme -s- as in οἰχμή and uxor), dámpati Du. 'husband and wife', Lat dominus, domina 'master (of the house, of slaves') or from *woyk'- OPruss waispattin 'wife, lady of the house', Lith viešpats 'lord, governor'.

  89. TR said,

    October 21, 2024 @ 3:37 pm

    Thanks, that's interesting — I wasn't aware of οἰχμή. I don't see that it tells us anything about uxor, though, nor that it shows a morpheme -s-; it seems to contain the fairly productive nominal suffix -σμή (όσμή, δυσμαί, ὁρμή etc.).

  90. German Dziebel said,

    October 21, 2024 @ 3:52 pm

    @TR. -σμή is a deverbal affix and *woyk' is a noun. So when uxor and οἰχμή. are lined up against each other it both seem to segment into *woyk'-s-meH2 vs. *woyk'-s-o:r.

  91. German Dziebel said,

    October 21, 2024 @ 3:53 pm

    With -s- being a Nominative marker from *woyk'(o)s.

  92. German Dziebel said,

    October 21, 2024 @ 4:00 pm

    Re: ὀδμή, ὀσμή, Frisk writes, "zu σμ aus δμ unten", so so apparently no σμή- affix here.

  93. TR said,

    October 21, 2024 @ 4:34 pm

    It's true that most -σμή nouns are based on verbal roots, but the same goes for most -μή nouns, and neither suffix tends to form words for people in any case. (Not to mention that *weiḱ- is also a verbal root.) The idea of a nominative case ending somehow finding its way into the middle of a derived form is another ad-hoc non-starter.

  94. German Dziebel said,

    October 21, 2024 @ 4:59 pm

    @TR. *weyḱos 'house' may or may not derive from *weyḱ- 'enter'. But οἰχμή. and uxor are both derived from the noun (just like damar and other forms above).

    μή obviously has a Feminine -eH2, so there's no μή-affix in οἰχμή but only -m-. if we look at Germ *wi:ba 'wife' < *woik'-wo, it was a neuter noun with -wo- standing for a collective notion likely similar to the use of semja 'family' in Old Russian in the sense 'wife'.

    "The idea of a nominative case ending somehow finding its way into the middle of a derived form is another ad-hoc non-starter." In other similar etymologies this is how people explain -s-.

  95. TR said,

    October 21, 2024 @ 5:09 pm

    I don't see that it matters much whether you call the suffix -μή or -μ-. The fact that words in -μή/-σμή/-θμή seem never to refer to people makes me doubtful that that's what we're looking at here, but if you had to choose one of the suffixes, -σμή obviously leaves you with one less thing to explain.

    What commonly accepted etymologies involve a nominative-marked stem plus a derivational suffix like -μή?

  96. German Dziebel said,

    October 21, 2024 @ 5:43 pm

    @ТР. "What commonly accepted etymologies involve a nominative-marked stem plus a derivational suffix like -μή?"

    Take any mobile s-stems such as *H3eH1es-, *H2eH1s-os 'mouth'. You have -s- marker internally. There's no problem with it right? Then you see it internally with a derivational affix (e.g., Germ *wi:hsilo 'sour cherry' < *weik'-s-). What else do you think that is but not the Nominative -s?

    "-σμή obviously leaves you with one less thing to explain." But -s- in οἰχμή and uxor is the same but the derivational affixes are different. So we don't need to explain much. *woyk'-s-mo-, *woyk'-s-r that became uxor by analogy with soror. What needs an explanation is the loss of -k- in oar.

  97. TR said,

    October 21, 2024 @ 5:49 pm

    I don't understand how s-stems like "mouth" are relevant. The -s- there isn't a nominative ending.

  98. German Dziebel said,

    October 21, 2024 @ 5:57 pm

    TR. Lat ōs 'mouth' < *h3oh1s, with -s as a Nominative ending. Let's take a derivative from this root, namely ōscillium 'cavity' with a double diminutive derivational affix. ōscillium is based on the Nominative ōs. The same internal placement of Nominative -s is found in Genitive ōris from *ōzis from *ōsis. What am I missing?

  99. TR said,

    October 21, 2024 @ 6:03 pm

    The fact that the -s- is part of the stem. It appears before all case endings, including the zero-marked nominative.

  100. German Dziebel said,

    October 21, 2024 @ 6:06 pm

    @TR. Sure, you can put it this way. *weyk'os > *weyk's (Nominative -s became part of the stem) > uxor, *oiks-m-H2.

  101. German Dziebel said,

    October 21, 2024 @ 6:08 pm

    But it's not part of the affix because it's paired with two different derivational affixes, -o:r- and -m-H2.

  102. German Dziebel said,

    October 21, 2024 @ 6:14 pm

    It's interesting that in uxor -ks- did not metathesize into -sk- as in aesculus 'oak' from *H2eig'-s or vespa 'wasp' < *webh-s.

  103. German Dziebel said,

    October 22, 2024 @ 7:53 am

    @TR. If the pattern "nasal – labiovelar" ~ "labionasal – palatovelar" was a true PIE phenomenon then -μή in οἰχμή goes back to *n due to the palatovelar in *woyk'-. In this case we end up with a more clear *woyk'sr > *uksr *> *uxor > uxo:r (by analogy with soror) > Nom/Voc. uxor and *woyk'snH2 > οἰχμή.

  104. David Marjanović said,

    October 27, 2024 @ 6:55 am

    "Tocharian wir means "young"." There's no indication of it being a cognate of Lat vir 'man, husband', Lith vyras 'husband' and Skrt vi:ra 'man, husband' outside of it being their look-alike. PIE *weyHro- is a term for (adult, married) males and as such it can hardly be a derivative from "young".

    "Man" > "husband" is such a common extension of meaning (German, Ukrainian…) it makes it impossible to assume "husband" is primary. This is a patriarchal "man ~ warrior ~ hero" words that eventually settled down. :-)

    Klaus Düwel, Robert Nedoma & Sigmund Oehrl. 2020. Die südgermanischen Runeninschriften (Ergänzungsbände zum Reallexikon der Germanischen Altertumskunde 119).

    That's on academia.edu; I read it a while ago.

    "The idea of a nominative case ending somehow finding its way into the middle of a derived form is another ad-hoc non-starter." In other similar etymologies this is how people explain -s-.

    Not with the nominative.

    The standard explanation of δεσπότης is a phrase, *déms pótis, "the house's lord" – the first *s is the genitive ending. A nominative wouldn't make any sense.

    And if you explain all *s-stems as nominatives whose *-s was reinterpreted as part of the stem, how do you explain the neuter *s-stems?

  105. Yves Rehbein said,

    October 31, 2024 @ 1:18 am

    Very good point!

    Nota bene: Melchert, "He reconstructs the word for ‘mouth’ as a neuter acrostatic s-stem or root noun (Melchert 2010b:59): nom./acc. sg. *(h1)óh1(-)s (Hitt. /ā́i̯s/ or /ái̯is/, Lat. ōs), gen. sg. *(h1)éh1(-)s-e/os." (eDiaAna #443).

    Note how similar this is to -ššara -a-aš-ša-ra, *h1es-r- ‘woman (vel sim.)’, "The -r- is explained as the remnant of a common gender r-stem, i.e. *h1s-r-´ > -ššar- (+ thematization) > -ššara-."

    "Lat. uxōr- is ambiguous and no reliable testimony for a PIE stem *(h1)sor- ‘woman’. It could continue a preform *h1uks-sor- ‘housewife’ (formed to *h1uk-s-, the zero grade of PIE *h1éu̯k-os/es- n. ‘house’ [> Ved. ókas-] + *sor- ‘woman’, cf. Harðarson 2014a:35). Alternatively, Pinault 2013a:250f. explains the Latin noun from a preform *h1uk-s-or- ‘the one being in the house’, a delocative formation derived from a loc. sg. *h1uk-s-ér ‘in the house’." (eDiAna #744)

    Since this root was already mentioned, you recognize the difference. I do not recognize it, but it matches what I said about *wi- and Hittite u- (cf. PIE *h2ou-, HIL: u-; *h1i̯éh1-/h1ih1-´, eDiAna #1669).

    See also *h1ḗs-/h1és "to be, exist; sit": "In the case of Indo-Iranian and Greek, the strong stem *h1ḗs- may theoretically go back to PIE *h1éh1s- (as per LIV²:232)." (eDiAna #1455)

    This was interpreted as a reduplicated form, though it seems now they do not want to rely on it.

RSS feed for comments on this post