The Great Proto-Indo-European Thread's Sequel

[Elizabeth K.]

It is out already, found in Alexander Lubotsky's Festschrift — though the publisher homepage unhelpfully currently has a broken link for the TOC. I'm looking forwards to this arriving online at some point in the future (I think Beech Stave Press allows authors to distribute articles online about a year after publication).

Christ, I don’t have $70 on hand for a Festschrift. I’m a broke college student. It had better come out soon, and un-paywalled.

[Frislander]

Just a quick question on the state of the literature. This year I'm taking a paper on comparative linguistics, and confusingly in Cambridge it's run by the Classics faculty rather than by the Linguistics section, so it's entirely IE-based, and only gives a shit for Greek, Latin and Sanskrit. The main sources/recommended reading seem to be James Clackson's Indo-European Linguistics: An Introduction and Sihler's Comparative Grammar of Greek and Latin, and I was just wondering if people have any particular views on these books, or indeed any recommendations of other good books I should seek out.

[KathTheDragon]

They're decent handbooks, and represent the traditional viewpoints pretty well. As introductions to Indo-European go, I can't fault them. But for serious interest, you want to read literally anything other than handbooks. Academia.edu has an excellent range of papers on just about anything, including a lot of utter BS, but that's to be expected.

[Howl]

It is out already, found in Alexander Lubotsky's Festschrift — though the publisher homepage unhelpfully currently has a broken link for the TOC. I'm looking forwards to this arriving online at some point in the future (I think Beech Stave Press allows authors to distribute articles online about a year after publication).

Christ, I don’t have $70 on hand for a Festschrift. I’m a broke college student. It had better come out soon, and un-paywalled.

If your college has a historical linguistics department, you could try to check if their library has it. But if Kümmel really did find something significant, then there should also be some follow-up papers about it.

Some of the claims of laryngeal reflexes are solid, for example the h in Hittite. Some are more controversial, for example Cowgill's law in Germanic. Tropylium is right to note that when laryngeals have reflexes, they are either back fricatives or back stops. There may be some examples of non-back reflexes in Uralic, though. For example, pre-Finnic seems to have *š as a reflex of h1 in loans. But this later became /h/, so it is possible that it was loaned as *h and that later merged with the sound change *š -> *h. I also have possible examples of Uralic *č as a reflex of h2 (those are probably loans) and *j as a reflex of *h1/*h2 next to high back vowels in Uralic.

My own attitude towards the whole laryngeal theory is that there are some serious issues, but I do not know what the proper way would be to fix those. So I am open to exploring unorthodox theories and I also hope that research into Indo-Uralic can shed more light on this.

[Howl]

*h₂ is, per recent results from Kümmel (though I'm still waiting to see the full paper), retained in SW Iranian as /h/~ /x/. Given also the Anatolian evidence, I think this nails this down as a voiceless back fricative pretty conclusively.

Let us know when the paper comes out and where to find it. That sounds positively fascinating.

A short overview of the results can be found in the following paper by Kümmel:
https://www.academia.edu/31147544/Is_an ... o-Iranian_

[WeepingElf]

I am currently reading Selected Papers on Indo-European Linguistics by the late, lamented Danish scholar Jens Elmegård Rasmussen, who was full of interesting ideas, not all of which, however, strike me as plausible. In some of the papers, for instance, he proposes some extra segmental phonemes whose limited distribution makes the proposals questionable.

One of them is an "infix", which he transcribes with a capital O with an inverted breve below, in order to explain some instances of o-grade which his ablaut theory does not cover properly. This "infix" expressed several disparate inflectional and derivational categories, including the causative-iterative and nouns of the type of Latin toga. His idea is that the phoneme in question was a uvular continuant and the "infix" started as a prefix, but the segment was metathesized into the word (in order to maintain the sonority hierarchy of the onset) such that it ended up just before the root vowel, which it then coloured to *o. I don't think this makes sense. As far as I know, no language has a segmental phoneme that occurs only in a small number of prefixes like this. At most, one can think of some kind of a prosodic feature instead which caused otherwise unexplained o-grades.

The second proposal serves to explain the odd behaviour (from the standpoint of the ablaut theory) of the nominal and verbal thematic vowels. Rasmussen writes the thematic vowel as *e', where the prime represents something like a glottal stop. Again, a phoneme with implausibly limited distribution, and again, a prosodic explanation is probably better. First, what is a "thematic vowel"? It is a succinct way of saying "stem-final non-high vowel". That's all. If a stem has a non-high vowel at the end, it is thematic; if it doesn't, it is athematic. My hypothesis is that there was some sort of prosodic feature causing the special behaviour of thematic vowels at the stem-ending interface. In all stems, even athematic ones, though it clearly manifests only when the stem ended in a non-high vowel.

This leads me to another proposal of his, which many scholars have made, Rasmussen only being one of them. Namely that the *s of the nominative singular ending actually was voiced *z. This would explain why thematic nouns end in *-os rather than the expected *-es in the nom. sg.: the thematic vowel manifests as *o before a voiced segment and as *e otherwise. This again sounds like a segmental phoneme that occurs only in a single morpheme, but nobody can rule out that in an early stage of PIE, *z was more common but merged unconditionally with *s, such that we can recover it only where it followed a thematic vowel.

[Frislander]

As far as I know, no language has a segmental phoneme that occurs only in a small number of prefixes like this. At most, one can think of some kind of a prosodic feature instead which caused otherwise unexplained o-grades.

Arammba has a voiced interdental fricative which appears solely in 3rd person verbal prefixes, so it's not entirely unattested.

[WeepingElf]

Point taken, Frislander. But I still think such cases are so rare that a hypothesis that avoids them is better.

As for multiple phonemes merging in PIE *s, reading further in Rasmussen's book I found an essay where he proposes even more such phonemes. First of all, he proposes *z also to have been the sigma of the sigmatic aorist (which would then have to be called the "zetatic aorist" ...), though I don't understand why.

He also proposes a *c, which occurs in the 2nd person ending, which becomes *s before a word boundary (including the boundary between a word and an enclitic) and *t otherwise. This may be the case, too. This one may have been an affricate *ts; was there also a *dz???

Then, he proposes something he calls *ð (actually, he uses Greek delta, but ð is easier to type on my keyboard ), which becomes *y at the end of the word after non-high vowel and otherwise *s. In my opinion, the allomorphy is different, namely *-s at the end of a word, and *-y- when another morpheme follows (including word-enclitic and proclitic-word boundaries), and the thing appears to be cognate to the Uralic plural marker which shows an allomorphy between *-t and *-j- under the same conditions.

And finally, he proposes something he calls *ř: *r after the thematic vowel, *s after vowel+spirant, *u after consonant+spirant, *i otherwise; this one is the morpheme of the "primary endings". If you ask me, I doubt that all these things are even the same morpheme.

[WeepingElf]

Thinking again about these extra phonemes, I realized that all of them can be easily shot down with standard tools of historical linguistics

First, *z. The thematic nom. sg. *-os may just be the result of paradigmatic levelling, as Rasmussen himself assumes for the instrumental singular *-oh₁ (of course, we can't exclude the possibility of *h₁ having been voiced, but there are no reasons to assume that), considering that such commonly used cases as the accusative, genitive and dative all have the thematic vowel as *o. And as I said yesterday, I can't follow Rasmussen's argumentation that the s-aorist also had *z (BTW, the term "zetatic aorist" is my own coinage, not Rasmussen's).

Second, *c. I think we are dealing with a conditioned sound change here which changed final voiceless stops into fricatives; this would also have affected final voiceless velar stops, turning them into laryngeals. That seems to have happened with the dual marker, for instance. Of course, the final *t-s of the 3sg, and 3pl. verbal endings need to accounted for if this is the case, but they may have been protected by something following them (perhaps a laryngeal?), and I have a hunch that these endings are rather young and may postdate this change (replacing earlier *-s and *-ses with which the 2nd person endings had gotten into conflict due to the very change discussed here).

My solution for *ð I have already posted yesterday. And *ř, as I also posted yesterday, is just a grab bag of disparate suffixes which need no special phoneme or morphophoneme to account of.

[Richard W]

What's the evidence that *h₁ didn't survive into Hittite as /ʔ/ word initially?

[Znex]

What's the evidence that *h₁ didn't survive into Hittite as /ʔ/ word initially?

Mostly that Hittite doesn't have a letter for it. In contrast, Kloekhorst proposes that glottal stops were preserved as plene spellings, such as *h₁és-si > Hit. e-eš-ši /ʔésːi/ {be.PRS-2SG}, *h₁éd-ti > Hit. e-ez-ši /ʔétsːi/ {eat.PRS-3SG}.

[Richard W]

Mostly that Hittite doesn't have a letter for it.

I can't work out what you mean. Hittite mostly doesn't have symbols for consonants; it has syllabograms for CV, VC and CVC. So please explain your statement, rather than leaving me to guess wrong. In the appropriate contexts, Akkadian VC implies an initial glottal stop.

Kloekhorst can be interpreted as saying that rather than h₁ vanishing word internally like h₃, it even survived word internally,

[Znex]

Mostly that Hittite doesn't have a letter for it.

I can't work out what you mean. Hittite mostly doesn't have symbols for consonants; it has syllabograms for CV, VC and CVC.

I mean there are no ʔV, Vʔ, ʔVC syllaboɡrams.

[Richard W]

Mostly that Hittite doesn't have a letter for it.

I can't work out what you mean. Hittite mostly doesn't have symbols for consonants; it has syllabograms for CV, VC and CVC.

I mean there are no ʔV, Vʔ, ʔVC syllaboɡrams.

Which is a pretty weak argument, given that Akkadian VC also serves as ʔVC. Writing systems tend not to make a great effort to write glottal stops with a low contrastive load, e.g German and Filipino.

[Znex]

I can't work out what you mean. Hittite mostly doesn't have symbols for consonants; it has syllabograms for CV, VC and CVC.

I mean there are no ʔV, Vʔ, ʔVC syllaboɡrams.

Which is a pretty weak argument, given that Akkadian VC also serves as ʔVC. Writing systems tend not to make a great effort to write glottal stops with a low contrastive load, e.g German and Filipino.

No need to tell me that! It's an argument from absence, apart from what I already said about Kloekhorst.

[WeepingElf]

Let me add some thoughts about the plural marker, for which I proposed a few days ago that it may have been *-s in word-final position and *-y- before another suffix (such as a case marker), an enclitic or at the end of a proclitic (thus showing similar allomorphy as the Uralic plural marker, which is *-t in word-final position and *-j- before another suffix). I wonder whether the "mess" we see with the plural cases (which to a large part differ between Late PIE and Hittite, and at any rate do not show the forms expected from an agglutinating paradigm, which would be *-y- followed by the singular case endings) has to do with a loss of the *-y- allomorph in the plural case forms in Early PIE (perhaps in a change somehow connected with the rise of ablaut), rendering the plural cases homophonic with the singular cases? I shall address the various plural cases below. What now follows is speculation, read at your own peril, and may be utter nonsense.

The animate nominative plural is the same in Late PIE and Hittite, and what one would expect: *-es, i.e. the plural marker added to the bare stem (though the *-e- raises some thought). It is the animate nominative singular which calls for an explanation. Does the *-s suffix descend from an old topic marker, perhaps related to the pronoun *so?

The animate accusative plural is *-ms in Early PIE. It seems to consist of the acc. sg. *-m and the plural *-s, but in the "wrong" order (one would expect *-y-m from an agglutinating paradigm). One idea I have about the PIE accusative is, that in an active-stative early stage of the language (which in itself is doubtful, of course!), it was an "animacy-neutralizing" morpheme, more derivative than inflectional in character, used to form a "syntactically inanimate" noun out of an animate one. This idea may be utter hogwash, but it would explain the order of the morphemes. Or was it just "re-pluralized" after *-y- was lost?

The genitive plural also throws up questions. In Hittite, the difference between *-os (Late PIE gen. sg.) and *-om (Late PIE gen. pl.) seems to be one of specificity rather than number. So LUGAL-as means 'of the king' (e.g., of the Hittites), while LUGAL-an is 'of kings (in general), kingly' (LUGAL here is the Sumerogram for 'king'). This may mean that *-os was the original genitive suffix, while *-om may have been some sort of denominal adjective-forming suffix.

The dative plural differs totally between Late PIE, which has something like *-bhyos (the consonants before the *o differ enough from language to language to make the reconstruction uncertain), and Hittite, which points at *-os. It has been conjectured that the *-s in *-bhyos is our plural marker, as the dative dual is *-bhyoh₁. But what does the element *-bhyo- mean? Probably, this is another denominal adjective-forming suffix.

Finally, the dual. We do not know whether Early PIE had a dual; Late PIE does, but the Anatolian languages do not. Did Late PIE innovate the dual, or did Anatolian lose it? Perhaps the Late PIE dual *-h₁ is related to the Uralic dual *-k in a similar way as the plural *-s to Uralic *-t (this of course only if Indo-Uralic or some larger unit is real, which is speculative), in which case it would be inherited in Late PIE and lost in Anatolian.

All this probably raises more questions that it would answer, and may turn out to be utterly wrong-headed, but I think it is at least worth discussing.

[Pabappa]

Is there a simple explanation for where the masc plural -oi~ī of gk and latin came from? PIE doesn't seem to have had anything similar. Classical Latin in particular seems to have avoided plurals in -s.

[WeepingElf]

Is there a simple explanation for where the masc plural -oi~ī of gk and latin came from? PIE doesn't seem to have had anything similar. Classical Latin in particular seems to have avoided plurals in -s.

First, Latin -ī is monophthongized *-oi, so Greek and Latin actually have the same ending here. The usual explanation is that this ending was transferred from the pronominal paradigm which has *-oi in PIE already, perhaps when the thematic nom. pl. *-o-es was shortened in some way to make it indistinct from nom. sg. *-os.

[WeepingElf]

I may add to my speculation that adding *-y- before the PIE singular case endings usually leads to forms that look bizarre from the vantage point of PIE phonology as we can reconstruct it. The same holds for the hypothetical dual marker *-h₁-. That one gets phonologically malformed forms this way seems to point at a pre-PIE sound change that did them in.

[Znex]

What's the general consensus on the pre-PIE MOA consonant sets? The basic PIE contrast is voiceless-voiced-voiced-aspirated *T D Dʰ, although from the synchronic and diachronic phonotactics, there seems to be more going on.

While I'm loathe to suppose like some do that the *D series were originally implosives or ejectives, there are signs of apparent glottalisation (eg. the Balto-Slavic Winter's Law, and a similar sound chance in pre-Proto-Anatolian where *HT and *D are treated similarly; *D only leniting to a stop series where chain lenition does occur, eg. in Germanic; the atypical lack of *b; possible old *D...D dissimilation, eg. **bib(h₃)eti > *pib(h₃)eti; apparently sporadic loss of word-initial *d in Tocharian, which may suggest that *d and *ʔ were too similar). While typical proponents of the glottalic theory suppose that this points to a **T Tʼ D contrast, the fact that *D Dʰ are treated as a pair most typically, outside of the IE languages where aspiration appears, leads me to believe alongside the Anatolian evidence that they were, if not both voiced, originally lenis series. Hence perhaps the contrast was originally **TT ʔT T. Maybe the exact nature of the *D series may be further narrowed down by the PIE roots we have; do we have any definite root shapes that forbid the appearance of *...HD or *...DH?

On the other hand, the *Dʰ series remains a bit of a mystery to me. What motivates the appearance of aspiration if it is not original, and its relative frequency in PIE roots? I've read some proposals that Grassman's Law reflects an original allophonic dissimilation in some root shapes, perhaps original **DeD shapes. That there are other prohibited root shapes may suggest a broader solution. My main problem though is that it is either a voiced or lenis series gaining aspiration, where aspiration may be more expected for a voiceless or fortis series. Is there much precedence for this in the world's languages? My first thoughts are admittedly of the Germanic languages: voiceless stops, which have frequent aspiration in Germanic, are well known to have derived from PIE *D; the voiceless stops in the High German languages even undergo further aspiration-motivated lenition.

Or is the appearance of voicing in the *Dʰ series on the other hand because of the needed distinction from the *T series? *T in the older IE languages tend towards frication or affrication rather than voicing, and aspiration does develop from *T in many IE languages; whether as a common allophone (as in the Celtic languages; or as the main allophone, as in Armenian and perhaps pre-PG and pre-PC) or as a new series (as in Indo-Iranian). Other voiceless aspirates occur in cross-linguistic IE onomatopoeia (which interestingly sometimes pattern as if according to Grassman's law: eg. Sans. कखति kákhati vs. Anc.Gr. καχάζω vs. Lat. cachinno vs. PG *kahhatjaną {to laugh,cackle}). The Indo-Iranian series stems from three main positions itself: *TH, *sT (whose distribution also matches Grassman's law except where generalised), and in onomatopoeia; not counting loanwords. So perhaps the development of *Dʰ as an aspirated series is motivated by an old allophonic alternation of *T. Hence glottalic proponents also propose the contrast **T(ʰ) Tʼ D(ʰ), which again to my mind may be simplified to **TT(ʰ) ʔT T(ʰ). Again, are there any cross-linguistic parallels to such a development?