The Great Proto-Indo-European Thread

[WeepingElf]

Tropylium has written an interesting blog post about Anatolian and PIE vowels. The bottom line: The Anatolian data points at a system like *i *e *ɜ *a *u instead of the traditionally reconstructed *i *e *a *o *u for Proto-Anatolian, and he points out that the former may be more archaic than the latter. Now, if Early PIE had the "quincunx" system (as I call it: four vowels in the corners of the vowel space, and one in the middle) reconstructed by Tropylium, the feature that distinguished *h3 from *h2 wouldn't have been labialization, and the whole idea that the three laryngeals were counterparts of the three velar stop series would go out of the window. However, Hyllested has found hints at a connection between PIE laryngeals and Proto-Uralic vowel qualities that supports the velar series idea, and therefore indirectly points at the traditional system in Early PIE.

[Zju]

Was extending verbs with prefixes from adverbs or preposition the state of affiars in wide or narrow PIE? Seeing how this feature is present in Slavic, Germanic, Romance languages and Greek, but in different forms I've been wondering for a while. Or maybe it was a feature of the substrate.

[KathTheDragon]

It's also a feature of Anatolian and Sanskrit. However, there are clear signs that univerbation was not present in PIE already.

[KathTheDragon]

Tropylium has written an interesting blog post about Anatolian and PIE vowels. The bottom line: The Anatolian data points at a system like *i *e *ɜ *a *u instead of the traditionally reconstructed *i *e *a *o *u for Proto-Anatolian, and he points out that the former may be more archaic than the latter.

This has interesting consequences for the phonetics of the laryngeals. It's been traditionally assumed that *h₂ should be uvular or pharyngeal, due to its a-colouring effect, and *h₃ similar but labialised due to its o-colouring effect. Under this new interpretation, it should instead be *h₃ that is uvular/pharyngeal, since now it is the a-colouring laryngeal. This fits with *h₃ being somewhat rarer than *h₂ (but how many of the cases of *h₂ are due to reconstruction bias?) I'm not sure how to interpret *h₂, though. Perhaps just plain velar.

[√mak·a√g-láwar·ē]

Tropylium has written an interesting blog post about Anatolian and PIE vowels. The bottom line: The Anatolian data points at a system like *i *e *ɜ *a *u instead of the traditionally reconstructed *i *e *a *o *u for Proto-Anatolian, and he points out that the former may be more archaic than the latter.

This has interesting consequences for the phonetics of the laryngeals. It's been traditionally assumed that *h₂ should be uvular or pharyngeal, due to its a-colouring effect, and *h₃ similar but labialised due to its o-colouring effect. Under this new interpretation, it should instead be *h₃ that is uvular/pharyngeal, since now it is the a-colouring laryngeal. This fits with *h₃ being somewhat rarer than *h₂ (but how many of the cases of *h₂ are due to reconstruction bias?) I'm not sure how to interpret *h₂, though. Perhaps just plain velar.

This makes me think of the common reinterpretation of traditional palato-velar series *Ḱ as plain velar *[k, g, gʱ] and traditional plain velar series *K as uvular *[q, ɢ, ɢʱ], especially if one aligns laryngeals *h₂ and *h₃ with these stops. The relative rarity of both the traditional plain series *K (a-coloring uvulars *[q, ɢ, ɢʱ]) and *h₃ (a-coloring uvular *[ʁ ~ ʁ̞]?) when compared to the frequency of both traditional palato-velar series *Ḱ (*[k, g, gʱ]) and *h₂ (*[x]?) seems less typologically marked if this were the case.

[KathTheDragon]

The non-existence of labial and labio-velar fricatives could then be attributed to a merger of the labio-velar into the labial fricative (an abundantly attested change) followed by debuccalisation (also attested). The resulting /h/ may be identifiable as *h₁, or some subset of *h₁'s occurances, if there is merit in the idea that *h₁ actually represents two phonemes that can no longer be distinguished, a glottal stop and a glottal fricative.

[√mak·a√g-láwar·ē]

Is there any indication of correspondences with Uralic or Afro-Asiatic that could support a proto-phoneme *ɸ or *β, for example sets with some labial sounds in other languages and *h₁ in PIE? I’ve yet to see anyone reconstruct any labial fricatives for Nostratic.

Other idea semi-related to this: Maybe this is what happened to those instances of *b that didn’t become *w or *m. Perhaps there was some phonological conditioning that split *b into a labial fricative *β > *w, *m in some instances and merged it with *ɣʷ > *h₁ in others.

[KathTheDragon]

Actually, on the putative split of *b > *w, *m, I believe I've found a conditioning for the different reflexation, which leads me to posit *b > *w as the basic change, with *w > *m being a subsequent conditional development (though it obviously can't be excluded that we had concurrent *b > *w and *b > *m). The obvious place to start is the only root structure where we can reliably (assuming, that is, that this hypothesis is even correct) identify original *b - in *wReC- and *mReC- roots. Now, we can exclude those roots that seem to be derivatives of other roots, and hence aren't relevant to the discussion at hand. In the LIV (I'll cite in the original German), these are *mneh₂ "denken an" < *men "einen Gedanken fassen" and *myeḱs- "sich festsetzen" < *meyḱ- "mischen". The remaining roots are *mlewh₂- "sprechen", *myewh₁- "bewegen", wleykʷ- "befeuchten", wred- "weich werden/machen", *wreg- "einer Spur folgen", *wreh₁- "finden", *wreh₁ǵ- "brechen, reißen", *wrep- "sich neigen", *wremb- "sich neigen", *wreng- "verdrehen, (ver)biegen", *wrengʰ- "winden, zusammendrehen", *wreyḱ- "drehen, einhüllen", *wreyt- "drehen, winden", *wr/lesk(ʷ)- "abhacken", *wr/leyH- "zusammendrücken", *wyeh₁- "umwickeln, umhüllen", *wyekʷ- "unwinden, umspannen", *wyeth₁- "wanken". There are a number of roots restricted to a single branch, and possibly a few extended roots, but that doesn't hurt my conclusion, which is this: both roots with *mR- have coda *w, and no roots with *wR- do. This is comparable to the dissimilation *w > m adjacent to *u found in Hittite (e.g. tumeni with the ending -weni, which conceivably began in PIE times already.

Also, I'm not saying that there had to have been an original *f/ɸ -, just offering an explanation of where it could have gone if there had been one.

[2+3 clusivity]

I am interested to see that the proposed dissimilation also does not appear to be triggered by the labiovelars.

[KathTheDragon]

Again, Hittite corroborates this - consider akueni to the root eku-/aku-

[Tropylium]

The obvious place to start is the only root structure where we can reliably (assuming, that is, that this hypothesis is even correct) identify original *b - in *wReC- and *mReC- roots.
(…) The remaining roots are *mlewh₂- "sprechen", *myewh₁- "bewegen", wleykʷ- "befeuchten", *wred- "weich werden/machen", *wreg- "einer Spur folgen", *wreh₁- "finden", *wreh₁ǵ- "brechen, reißen", *wrep- "sich neigen", *wremb- "sich neigen", *wreng- "verdrehen, (ver)biegen", *wrengʰ- "winden, zusammendrehen", *wreyḱ- "drehen, einhüllen", *wreyt- "drehen, winden", *wr/lesk(ʷ)- "abhacken", *wr/leyH- "zusammendrücken", *wyeh₁- "umwickeln, umhüllen", *wyekʷ- "unwinden, umspannen", *wyeth₁- "wanken".

You might want to take a look at the proportion of pre-PIE roots with two *B series stops you end up hypothesizing here.

The most natural explanation for *wR- and *mR- surely is that these come by syncope from earlier bisyllabic word roots (which may well be the case for many of the other word-initial clusters as well). Perhaps the "correct" question to ask would be why do **nR-, **lR-, **yR- not occur? Some of these could have simply metathesized (e.g. *yw- > *wy-?) or simplified (*ly- > *l- or *y-? *nr- > *dr-?), and there may not have been any initial **r- to potentially end up as cluster-initial. But that still leaves open questions.

The non-existence of labial and labio-velar fricatives could then be attributed to a merger of the labio-velar into the labial fricative (an abundantly attested change) followed by debuccalisation (also attested).

Another option is *Hʷ > *Hw, a la Germanic. PIE has interestingly many cases of *h₂w in roots.

Moreover, apparently the current understanding is that Proto-Anatolian in fact had *xʷ for PIE *h₂w and *h₃w. What if this is an archaism rather than an innovation?

[√mak·a√g-láwar·ē]

Also, I'm not saying that there had to have been an original *f/ɸ -, just offering an explanation of where it could have gone if there had been one.

I figured you were; I was genuinely asking (non-rhetorically) whether anyone had noticed Nostratic (or at least Indo-Uralic) cognates with labials where PIE has a laryngeal.

You might want to take a look at the proportion of pre-PIE roots with two *B series stops you end up hypothesizing here.

Isn’t this exact issue a major part of the support for the glottalic theory? If there were some rule that disallowed 2 plain lenis (or ‘glottalized’) stops, this is probably what triggered the *b > *w/m in these cases in the first place, and the resulting rarity of remaining *bR- may have been subject to hypercorrection to align with the rules KathTheDragon outlined above.

The most natural explanation for *wR- and *mR- surely is that these come by syncope from earlier bisyllabic word roots...

Now that I look at all of the roots together, some of them seem oddly similar to other roots with similar syllable shapes without an initial *w-. For example, alongside *wrengʰ- ‘winden, zusammendrehen’ there is *strengʰ- ‘zusammendrehen’ and *grenǵʰ- ‘drehen,’ and a slightly less phonologically similar *stregʷʰ- *drehen, wenden’; all of these may have derived from a root **reg(ʷ)ʰ-, with prefixing and nasal-infixing to alter the meaning. In fact, there is a root *(h₂)regʰ- meaning ‘sich aufrichten,’ which seems to be almost the opposite semantically (“straighten out” as opposed to “turn, twist”). Maybe whatever root *w- (?< *b) could have had a negating or contradicting meaning.

A similar opposition may be found between roots *h₃reǵ- ‘gerade richten, ausstrecken’ vs. *wreng- ‘verdrehen, (verbiegen)’ and *wreg- ‘einer Spur folgen’; *rek- ‘ordnen, festlegen, bestimmen’ vs. *wrek- ‘drehen, einhüllen’ (and possibly *bʰrekʷ- ‘zusammendrängen’). There might also be some etymological relationship between *wreh₁- ‘finden’ and *reh₁t- ‘treffen, finden.’

There may also be additional evidence for this in the form of potential zero grades of the “non-prefixed” roots. For example, for *(h₂)regʰ- vs. *wrengʰ- & *strengʰ- & *grenǵʰ-, there is also *werǵʰ- ‘(zu)binden’; for ‘ *h₃reǵ- vs. *wreng- & *wreg-, there is also *h₂werg- ‘sich umdrehen, sich wenden’; and for *rek- vs. *wrek- (?& *bʰrekʷ-), there is also *terkʷ- ‘sich drehen.’

This could also just be a huge set of coincidences, but it’s interesting to think about.

[jal]

For example, for *(h₂)regʰ- vs. *wrengʰ- & *strengʰ- & *grenǵʰ-, there is also *werǵʰ- ‘(zu)binden’; for ‘ *h₃reǵ- vs. *wreng- & *wreg-, there is also *h₂werg- ‘sich umdrehen, sich wenden’; and for *rek- vs. *wrek- (?& *bʰrekʷ-), there is also *terkʷ- ‘sich drehen.’

There also tʷerk - 'den Arsch schütteln' which might be related.


JAL

[√mak·a√g-láwar·ē]

For example, for *(h₂)regʰ- vs. *wrengʰ- & *strengʰ- & *grenǵʰ-, there is also *werǵʰ- ‘(zu)binden’; for ‘ *h₃reǵ- vs. *wreng- & *wreg-, there is also *h₂werg- ‘sich umdrehen, sich wenden’; and for *rek- vs. *wrek- (?& *bʰrekʷ-), there is also *terkʷ- ‘sich drehen.’

There also tʷerk - 'den Arsch schütteln' which might be related.


JAL

AHAHAHAHA.

[Zju]

Yeah I dunno about that. Why should all PIE words ultimately derive from CVC roots or CVC+CVC compounds? It makes it look like a conlang. Natural languages have all sorts of shapes for 'primary', 'underived' roots. Their exact shape and number depends on the language, but e.g. they may be CV, VC, CVC, CCVC, VCCV, VCCVC, etc. Why wouldn't PIE just have a similar array of different 'primary' roots?

Consider the English words house, home, human, habitat - each of them from unrelated PIE word. Ditto for spike and spear. Infact, there is a word spoon, which seems to be almost the opposite semantically. Maybe whatever suffix was added -e/-i could have had a negating or contradicting meaning.

PU reconstruction seems to have the same fixed idea, to derive (nearly) all words from CVCCV words, which then should be CVC- roots with -CV suffixes. (or am I missing something?)

[√mak·a√g-láwar·ē]

Why should all PIE words ultimately derive from CVC roots or CVC+CVC compounds? It makes it look like a conlang.

I’m not proposing this at all (I’m not really proposing anything, just noticing patterns & speculating).

I was just rolling with Tropylium’s idea that the verbal roots in *wr- and *mr- “come by syncope from earlier bisyllabic word roots” (which I don't necessarily agree with), probably of a shape CVCVC, by saying that MAYBE the segment CV- could be a prefix, because the -CVC part of this root might be mirrored in other attested roots with similar or identical meaning, and a plain CVC of opposite meaning. This isn’t “all PIE words” or even all verbs, just this suspicious set of roots all in *wr_K. Even if this weren't the case, having two 'primary' CVCVC roots such as **wVregʰ- and **stVregʰ- both with exactly identical meaning would at least make me consider looking into the possibility of an original morpheme **regʰ-...

However, at this point I’m not even entirely convinced by my specific examples, especially with the unexplained alternation between the G series (*g vs. *ǵ vs. *gʰ vs. *gʷʰ). I think there might be something behind *wreh₁- ~ *reh₁t- and behind *wrengʰ- ~ *strengʰ- ~ *(h₂)regʰ-, but overall it was all a little far-fetched from the get-go and I’m aware of that, I just figured this would be the place to throw out the idea.

Consider the English words house, home, human, habitat - each of them from unrelated PIE word. Ditto for spike and spear. Infact, there is a word spoon…

I know this is supposed to be tongue in cheek, but the difference is that unlike English, PIE does have a system of ablaut that yields derivations like CVC (full grade) > CV-CC (zero grade with prefixing), for example *sed- > *ni-sd-. I don't see why some earlier stage of Pre-PIE couldn't have had a development of CV+CVC > C-CVC, obscuring the original prefix.

[Tropylium]

You might want to take a look at the proportion of pre-PIE roots with two *B series stops you end up hypothesizing here.

Isn’t this exact issue a major part of the support for the glottalic theory? If there were some rule that disallowed 2 plain lenis (or ‘glottalized’) stops, this is probably what triggered the *b > *w/m in these cases in the first place, and the resulting rarity of remaining *bR- may have been subject to hypercorrection to align with the rules KathTheDragon outlined above.

You'd then need some other explanation, however, for why roots of types *d-B, *ǵ-B, *g-B, *gʷ-B are also missing.

Actually, how much evidence for *wr- there is exactly to begin with? Is it just Germanic and Indo-Iranian? I can only find two roots in LIV with reflexes in both (*wreg- and *wreyḱ-). (But for all I know there could be e.g. evidence for ϝρ- in some obscure Ancient Greek dialect.)

Yeah I dunno about that. Why should all PIE words ultimately derive from CVC roots or CVC+CVC compounds?

Not necessarily all of them, but "unstable" clusters like *wr- are unlikely to have been around too long in PIE, and syncope is in general the most common way new consonant clusters arise. Full grade / zero grade alternations additionally already testify for widespread syncope somewhere in pre-PIE.

There is also that most language families that Indo-European has been proposed to be related to (Afrasian, Dravidian, everything in Northern Eurasia) do not seem to allow initial consonant clusters, and the only IE clusters I have seen Nostratic proto-clusters even proposed for are *s + stop (which would allegedly correspond to affricates elsewhere). There is Kartvelian, but has anyone even proposed a binary IE-Kartvelian specifically?

PU reconstruction seems to have the same fixed idea, to derive (nearly) all words from CVCCV words, which then should be CVC- roots with -CV suffixes. (or am I missing something?)

Both (C)VCV and (C)VCCV are considered "canonical" root shapes. But yes, it is often been speculated that the residue of forms pointing to CVCVCV or CVCVC would have been derivatives; and there has even some speculation that the CVCCV roots would come from earlier CVCV-CV in pre-Proto-Uralic. There is essentially no evidence for CVC roots as a separate category from CVCV (to this day most Samic and Finnic languages disallow CVC word stems), and the only CV roots are grammatical elements like pronouns.

(I have a theory cooking however that, rather than suffixation, some CVCVCV forms would have come through vowel epenthesis from CVCCV.)

[KathTheDragon]

Actually, how much evidence for *wr- there is exactly to begin with? Is it just Germanic and Indo-Iranian? I can only find two roots in LIV with reflexes in both (*wreg- and *wreyḱ-). (But for all I know there could be e.g. evidence for ϝρ- in some obscure Ancient Greek dialect.)

There are also scattered reflexes in other languages, such as fl- in Old Irish (which is for *wl-, but still relevant). Sadly, the roots displaying these clusters are few, and not all abundantly attested in languages that retain the clusters. That said, I think it's assured that these roots did occur in PIE, and the evidence for *mlewh₂-, which provides a parallel, is hard to dismiss.

For what it's worth, I'm not married to my idea about the ultimate origin of *wR- and *mR-. It's just idle speculation, really, and I doubt the matter can be resolved at all without good external comparanda, which are sadly lacking.

[Alces]

There's the widely-attested nominal stem *wréh₂d- 'root', with *wr- reflected in Welsh gwraidd and Irish fréamh, and the *w- of the oblique stem *wr̥h₂d- reflected in Goth. waúrts, ON urt, OE wyrt (> NE wort), OHG wurz.

[KathTheDragon]

Yeah, that's not "widely attested", just Celtic and Germanic, two European branches known to have been in contact from an early time.

[Alces]

Well, the root is attested in Indo-Iranian, Greek, Albanian and Italic as well. It's just that Celtic and Germanic are the only ones which attest to the presence of the *w-.

[Alces]

I have a question (or several questions): what's the origin of Sanskrit ahám 'I'? Ringe (2006) reconstructs the PIE nom. 1sg. pronoun as *égh₂, and the only reason I can see why the final laryngeal might be reconstructed is so that áham can be derived from *égh₂-om. In that case, *-om could be the thematic neuter nom. sg. ending. But why would the first-person pronoun take a neuter ending? Also, what grounds are there for supposing that the laryngeal was *h₂ specifically? Alternatively, the ending could be *-Hóm---does anybody have any idea about where that might have come from? (Ringe refers to Katz 1998, but I don't have access to this publication.)

[hwhatting]

What I've seen being speculated is that *-om was a particle that had several uses, inter alia meaning something being in a relationship with something else, whence it's use in the Gen. Pl., and that in the neuters it was firstly a derivational suffix (e.g. *yug-om "related to / used for coupling / yoking" = "yoke"). Its similarity to the ending of the Acc. Sg. of the o-stems then led to this kind of indeclinable word take on the o-declination. Wether the *-om used in the pronoun is the same or a different suffix / particle, I cannot say.
In Sanskrit, -am is not limited to the 1st person - it's also there in the nominative of the 2nd sg. tvam, of the 1. pl. vayam, and the 2nd p.l. yu:vam. Unfortunately, none of these other pronouns allows us to clearly decide wether the Suffix was PIE *-om or PIE *-Hom.
I think at least some scholars think of the laryngeal as part of the pronoun in order to derive the -o: of the 1st sg. in several IE languages via ablaut (*H1eg'H2 in Anatolian, *H1eg'H2-om in Indo-Iranian and Slavic, *H1eg'-oH2 in Latin, Greek ego:), but that's of course not necessary, *-oH could be just another suffix.

[Zju]

I read that the word used to be a verb meaning 'I am here' that took either the thematic or the athematic ending and from there the difference in the forms *ég'oh₂ vs. *ég'om and reflexes.

I've also read that PIE didn't have plural caseforms of cases other than N A D - is that true? If so, has anyone tried to determine if two language subfamilies share a similar origin for some of their plural caseforms and therefore if they could be related?

[Alces]

What I've seen being speculated is that *-om was a particle that had several uses, inter alia meaning something being in a relationship with something else, whence it's use in the Gen. Pl., and that in the neuters it was firstly a derivational suffix (e.g. *yug-om "related to / used for coupling / yoking" = "yoke"). Its similarity to the ending of the Acc. Sg. of the o-stems then led to this kind of indeclinable word take on the o-declination.

That seems like a fairly plausible explanation for the origin of the thematic neuter nom./acc. sg. *-óm ending, but I don't see why such a morpheme might be suffixed to personal pronouns.

I read that the word used to be a verb meaning 'I am here' that took either the thematic or the athematic ending and from there the difference in the forms *ég'oh₂ vs. *ég'om and reflexes.

A personal pronoun coming from a phrase meaning "I am here"? Sounds fishy to me; I'd like to see if there are any attested examples of such a development. Also, if Skt ahám is from an athematic verb ending in *-mi, wouldn't we expect to see some trace of the *-i?

In Sanskrit, -am is not limited to the 1st person - it's also there in the nominative of the 2nd sg. tvam, of the 1. pl. vayam, and the 2nd p.l. yu:vam.

The nom. 1du. vā́m could also reflect the same ending if it's from PIE *we + *-Hóm, with vowel contraction. Likewise the accusative forms mā́m (1sg.) āvā́m (1du.), tvā́m (2sg.), yuvā́m (2du.) could reflect the same ending, added to Ringe's reconstructed forms *m̥mé, *n̥h₃mé, *twé and *uh₃wé (with some stem remodelling too in the 1sg. and 2du.). You might be able to alternatively explain the endings of the acc. sg. forms as coming from a-stem acc. sg. *-am (masculine/neuter), *-ām (feminine), just as the -ān in acc. 1pl. asmā́n and acc. 2pl. yuṣmā́n is probably just the a-stem acc. pl. ending. But, glancing at Wikipedia, it doesn't look like -ām is an acc. du. ending in any nominal stem class, so I don't know how else the endings of these accusative dual forms can be explained.

On a perhaps unrelated note... there's an interesting puzzle with regards to the lengths of nom. 2pl. yūyám. and acc. 2pl. yuṣmā́n. Ringe reconstructs the PIE antedecent of the former as *yū́ but adds that it might be < *yúy (from nom. 2du. *yú + *-y, just as nom. 1pl. *wéy appears to be from nom. 1du. *wé + *-y). If nom. 2du. yuvám is from *yú + *-Hóm, then we'd expect *yū́ + *-Hóm to give *yūvám. The -y- could be accounted for by proposing some sort of retention of the original *-y, but then the length is unexplained; more attractively, we could propose that it's levelled in from nom. 1pl. váyam. But then... why does yuṣmā́n have short u? If Ringe's *uh₃wé reconstruction for PIE is correct, this would become *ūvā́m with long *ū by regular sound change + addition of the *-Hóm ending. So not only the initial y-, but also the -u- after it must be accounted for by levelling in from one of the other nom. forms... and the only nom. form with short -u- is the dual one, yuvám. But isn't a levelling from a dual form into a plural form (as opposed to the other way round) highly unexpected?

(If anybody knows of a good resource that could help me with IE personal pronouns, I'd much appreciate it... the only things I have to refer to here are the very cursory remarks made by Ringe [2006], who says I should look at Katz [1998] if I want more detail, except I don't have that book, and the basic overview in Fortson [2004].)