The Great Proto-Indo-European Thread's Sequel

[WeepingElf]

Indeed, among Beekes' idiosyncratic ideas is the very same glottalic theory. Idiosyncracies aside, Beekes's book is a strange and not very satisfying read; the late Szemerényi's Introduction to Indo-European Linguistics is a harder read but much more rewarding, even if he disagreed with the laryngeal theory (and made a good case against it, too).

Beekes also maintained that (Early) PIE was an ergative language, and the sigmatic and asigmatic nominatives originally were two different cases, namely an ergative and an absolutive. But it is IMHO clear that the lack of final *-s in non-neuter nasal and liquid stem nominative singulars is phonological, and does not reflect a lost distinction between two different cases (the s-less neuter nominative-accusatives are another thing, though; it may be that the neuters indeed inflected ergatively in Early PIE, and this nominative-accusative was actually an absolutive).

And Szemerényi, he was just an old fart who could not wrap his mind around a theory that was not even new (it dates to 1879!) in his student years, but merely not yet accepted by the majority of scholars. Szemerényi's case may seem convincing to the layman, but actually, the laryngeal theory simplifies many things and is not all that hard to get at all, and even Szemerényi had to grudgingly admit that there must have been something in PIE where Hittite shows /h/. There really is no need to waste your time (and money) on his utterly obsolete book (unless you are interested in the history of IE studies, or just obsolete theories amuse you, and it is small and found in many libraries, so it isn't a big investment anyway).

[alice]

And Szemerényi, he was just an old fart who could not wrap his mind around a theory that was not even new (it dates to 1879!) in his student years, but merely not yet accepted by the majority of scholars. (etc)

And another piece of my youth is tainted forever... although his book also says something like, in italics, "the existence of voiceless aspirates in PIE cannot be denied".

More seriously, though, are there any good and up-to-date books on PIE which don't rely on or promote dubious or idiosyncratic theories?

[WeepingElf]

And Szemerényi, he was just an old fart who could not wrap his mind around a theory that was not even new (it dates to 1879!) in his student years, but merely not yet accepted by the majority of scholars. (etc)

And another piece of my youth is tainted forever... although his book also says something like, in italics, "the existence of voiceless aspirates in PIE cannot be denied".

You may be relieved by knowing that Szemerényi's book was the first I read on PIE (and the second was Gamkrelidze & Ivanov), and it is indeed easy to understand - but as I said, it is obsolete and no longer reflects current opinion. Like an old physics textbook that talks about the luminiferous aether all the time.

More seriously, though, are there any good and up-to-date books on PIE which don't rely on or promote dubious or idiosyncratic theories?

Fortson and Mallory & Adams (as I recommended yesterday) are good and generally up to date, and don't peddle dubious ideas; otherwise, I wouldn't have recommended them! But as with any science, people find out new things all the time, and there are many ideas in circulation (e.g., on the origin of ablaut) where nobody can tell yet whether they will prevail or not.

[hwhatting]

Re Szemerényi - IIRC, he doesn't reject the laryngeal theory (= that there were such things as laryngeals in PIE) as such; he is actually the cause for Pokorny, who is otherwise totally pre-laryngeal, including one or two reconstructions with a laryngeal in his dictionary. He specifically rejects the 3-laryngeal reconstruction that is mainstream nowadays and assumes only one laryngeal, reconstructing it where Anatolian attests it.
One issue where Mallory & Adams deviate from the mainstream is that they assume 4 laryngeals, which is decidedly a minority position, even if they are not the only ones.
Re gsandi: His views on PIE are here. But I don't know how much this reflects his current views; when I last met him a year ago, his interest had mostly switched to reconstructing areal stages in the development of the Western Romance languages.

[WeepingElf]

Re Szemerényi - IIRC, he doesn't reject the laryngeal theory (= that there were such things as laryngeals in PIE) as such; he is actually the cause for Pokorny, who is otherwise totally pre-laryngeal, including one or two reconstructions with a laryngeal in his dictionary. He specifically rejects the 3-laryngeal reconstruction that is mainstream nowadays and assumes only one laryngeal, reconstructing it where Anatolian attests it.

Right. Pokorny mentions the laryngeal theory in his preface to the IEW, and gives arguments why he didn't use it - it was not accepted yet by the majority of scholars. That was perfectly OK in 1959; it is annoying today. Writers of handbooks, dictionaries and the like should keep to the mainstream opinions of their day; Pokorny did nothing else than that - but Szemerényi, writing later, was too conservative for his time, and his book is now out of date. But of course, when just building an IE conlang, you are facing less rigid demands of correctness than when doing academic research on PIE, and you can always wing controversial points, e.g. by dropping all laryngeals (which most IE languages did anyway).

One issue where Mallory & Adams deviate from the mainsztream is that they assume 4 laryngeals, which is decidedly a minority position, even if there are not the only ones.

Yes, "*h₄" is a minority position. Most of them are probably just *h₂, though where Mallory & Adams reconstruct "*h_1/4", you should read that as "*h₁". Also, there is some disagreement among scholars as to where a laryngeal is to be reconstructed and where not, and Mallory & Adams tend to be cautious here and not to reconstruct a laryngeal where that seems doubtful.

But Gamkrelidze & Ivanov are far worse; while the glottalic theory can be rewritten out quite easily but may still be confusing, they posit seven extra phonemes and have a lot of entirely spurious items (such as "*leb^hont^h- 'elephant'") in their vocabulary. This, however, doesn't forbid exploiting their theories in a conlang! (I was into the glottalic theory for quite some time, though I assumed that the system shifted to the conventional system well before PIE broke up, I have meanwhile found an IMHO more elegant solution to the questions it tried to answer - the voiceless stops were aspirated, otherwise everything as in the conventional reconstruction - and none of my drafts of Old Albic ever actually included glottalized stops, and I am still into the notion that an early stage of PIE was an active-stative language, though my argumentation and my reconstruction differ from Gamkrelidze & Ivanov's.)

[bradrn]

Re gsandi: His views on PIE are here.

I’m curious about his reconstruction of a five-vowel system /a e i o u/.

[Lērisama]

Re gsandi: His views on PIE are here.

I’m curious about his reconstruction of a five-vowel system /a e i o u/.

I assume they're just treating syllabic *y *w as phonemic, and part of the minority that recognises a distinct *a

[Travis B.]

(I was into the glottalic theory for quite some time, though I assumed that the system shifted to the conventional system well before PIE broke up, I have meanwhile found an IMHO more elegant solution to the questions it tried to answer - the voiceless stops were aspirated, otherwise everything as in the conventional reconstruction - and none of my drafts of Old Albic ever actually included glottalized stops, and I am still into the notion that an early stage of PIE was an active-stative language, though my argumentation and my reconstruction differ from Gamkrelidze & Ivanov's.)

I for a bit was interested in glottalic theory, until I learned that aspirated voiced consonants are directly attested in Armenian, and that Germanic strongly implies them (i.e. out of T D Dh both T and Dh were likely aspirated in PrePGmc considering their parallel behavior therein). Now I'm of the view that the conventional reconstruction is correct, and if the glottalic system ever did existed it was deep in the mists of time before conventional reconstructions of PIE.

[WeepingElf]

Re gsandi: His views on PIE are here.

I’m curious about his reconstruction of a five-vowel system /a e i o u/.

Gabor's reconstruction (thank you for posting the link, Hans-Werner!) differs from the mainstream in some other points, too, and he doesn't give us the reasons to do so.

As for PIE *a, I think it originally existed only as an allophone of *e next to *h₂ (I think the Early PIE values of *e, *a, *o and *h₂ were [æ], [ɑ], [ɒ] and [χ], respectively), but there are some items with *a that can't be explained by laryngeals, such as *g^hans- 'goose', which are probably loanwords from other languages, or onomatopoetic formations. PIE *b is a similar story: it was originally absent as its antecedent had merged with *w in an early stage when the voiced unaspirated stops were spirants (as in Uralic), but loanwords brought in this new phoneme which conveniently filled a gap (compare how Slavic acquired /f/ from loanwords).

(I was into the glottalic theory for quite some time, though I assumed that the system shifted to the conventional system well before PIE broke up, I have meanwhile found an IMHO more elegant solution to the questions it tried to answer - the voiceless stops were aspirated, otherwise everything as in the conventional reconstruction - and none of my drafts of Old Albic ever actually included glottalized stops, and I am still into the notion that an early stage of PIE was an active-stative language, though my argumentation and my reconstruction differ from Gamkrelidze & Ivanov's.)

I for a bit was interested in glottalic theory, until I learned that aspirated voiced consonants are directly attested in Armenian, and that Germanic strongly implies them (i.e. out of T D Dh both T and Dh were likely aspirated in PrePGmc considering their parallel behavior therein). Now I'm of the view that the conventional reconstruction is correct, and if the glottalic system ever did existed it was deep in the mists of time before conventional reconstructions of PIE.

Yes. Germanic and Armenian essentially just devoiced *D (though Germanic turned aspirated stops into spirants later); these new voiceless stops did not merge with *T because the latter was aspirated (whether that aspiration was old or innovated; if it is old, Indo-Aryan and Greek, which innovated new voiceless aspirated stops, must have deaspirated *T before those innovations, and it seems likely that Italic deaspirated them early as well - all other branches are actually equivocal on this question, but in Celtic, *T appears to have been aspirated, too). That the Old Armenian voiced stops were aspirated is not only shown by some Eastern Armenian dialects where they still are, but also by the Western Armenian shift which only makes sense if the voiced stops were aspirated. That the conventional reconstruction features a rare system makes perfect sense, consider that no branch preserved it unchanged, showing that it was unstable!

As for the glottalic theory, I wouldn't say that it can't be right, but I see no reason to posit it, especially considering that in the probably nearest living kin of IE, Uralic, there are no glottalic consonants. As I said above, I entertain the notion that the *D set once were voiced spirants, with the labial member merging with *w, explaining the (near-)absence of *b as well as the clusters *wl- and *wr-. However, PIE may have been influenced by a substratum related to NW Caucasian, which may have introduced glottalic stops - but why then are they preserved nowhere? Systems of the type T-T'-D are quite common and stable, while Th-D-Dh and T-D-Dh are rare and unstable.

[Travis B.]

As for PIE *a, I think it originally existed only as an allophone of *e next to *h₂ (I think the Early PIE values of *e, *a, *o and *h₂ were [æ], [ɑ], [ɒ] and [χ], respectively), but there are some items with *a that can't be explained by laryngeals, such as *g^hans- 'goose', which are probably loanwords from other languages, or onomatopoetic formations. PIE *b is a similar story: it was originally absent as its antecedent had merged with *w in an early stage when the voiced unaspirated stops were spirants (as in Uralic), but loanwords brought in this new phoneme which conveniently filled a gap (compare how Slavic acquired /f/ from loanwords).

This is similar to my view. While I haven't thought much about the prehistory of the *D series, to me the vowel system of Early PIE makes most sense if it's considered an "S4" vowel system, excluding a marginal/allophonic *a, where the primary vowels were [i æ ɒ u] (I find the idea of "syllabic semivowel phonemes" to be a silly idea, as much as it may make the existence of syllabic nasals neater to some).

[Travis B.]

As for the glottalic theory, I wouldn't say that it can't be right, but I see no reason to posit it, especially considering that in the probably nearest living kin of IE, Uralic, there are no glottalic consonants. As I said above, I entertain the notion that the *D set once were voiced spirants, with the labial member merging with *w, explaining the (near-)absence of *b as well as the clusters *wl- and *wr-. However, PIE may have been influenced by a substratum related to NW Caucasian, which may have introduced glottalic stops - but why then are they preserved nowhere? Systems of the type T-T'-D are quite common and stable, while Th-D-Dh and T-D-Dh are rare and unstable.

Part of why I don't believe in the glottalic theory anymore is indeed because T T' D systems are quite stable -- if such a system existed in the history of PIE, why do we see it nowhere at all, not even a trace (aside from occasional ponderings that English glottalized coda fortis plosives are somehow a relic of a glottalized past, which seems unlikely because this is not seen anywhere outside of Anglic).

[WeepingElf]

As for PIE *a, I think it originally existed only as an allophone of *e next to *h₂ (I think the Early PIE values of *e, *a, *o and *h₂ were [æ], [ɑ], [ɒ] and [χ], respectively), but there are some items with *a that can't be explained by laryngeals, such as *g^hans- 'goose', which are probably loanwords from other languages, or onomatopoetic formations. PIE *b is a similar story: it was originally absent as its antecedent had merged with *w in an early stage when the voiced unaspirated stops were spirants (as in Uralic), but loanwords brought in this new phoneme which conveniently filled a gap (compare how Slavic acquired /f/ from loanwords).

This is similar to my view. While I haven't thought much about the prehistory of the *D series, to me the vowel system of Early PIE makes most sense if it's considered an "S4" vowel system, excluding a marginal/allophonic *a, where the primary vowels were [i æ ɒ u] (I find the idea of "syllabic semivowel phonemes" to be a silly idea, as much as it may make the existence of syllabic nasals neater to some).

Yep! An "S4" system, and "syllabic semivowels" are vowels - semivowels are essentially non-syllabic vowels after all. My personal theory on the origin of ablaut is that the immediate pre-ablaut stage had three "normal" vowels, **a i u and a "strong" (probably long) vowel **â, perhaps also **î and **û. Strong vowels attracted the accent, otherwise the accent was penultimate. Word stems had at least two syllables and ended in a vowel. If the final vowel was **â, the resulting word was thematic (barytone if the first vowel was also strong, otherwise oxytone); if it was **i, it was a proterokinetic i-stem, accordingly for **u. Other stems were athematic, acrostatic if the first vowel was strong, and hysterokinetic/amphikinetic otherwise.

This doesn't explain everything, there certainly was quite much more going on, but it seems to account fairly well of the overall pattern. This system in turn resulted from a richer inventory (as found, for instance, in Proto-Uralic) by the merger of many vowels into **a. In the process of the merger of those vowels, palatalization and labialization of velars by neighbouring front and rounded vowels, respectively (apparently, there were no front rounded vowels in the language at that stage) became phonemic, resulting in the three velar series (the same happened to **χ, except that that consonant never was adjacent to a front vowel because it backed those vowels, just try to say [χe] without either fronting the consonant or backing the vowel, and you get the picture; thus, no palatalized laryngeal - *h₁ was just /h/). All of this may be ascribed to a substratum related to NW Caucasian, where similar things seem to have happened but on a grander scale.

As for the glottalic theory, I wouldn't say that it can't be right, but I see no reason to posit it, especially considering that in the probably nearest living kin of IE, Uralic, there are no glottalic consonants. As I said above, I entertain the notion that the *D set once were voiced spirants, with the labial member merging with *w, explaining the (near-)absence of *b as well as the clusters *wl- and *wr-. However, PIE may have been influenced by a substratum related to NW Caucasian, which may have introduced glottalic stops - but why then are they preserved nowhere? Systems of the type T-T'-D are quite common and stable, while Th-D-Dh and T-D-Dh are rare and unstable.

Part of why I don't believe in the glottalic theory anymore is indeed because T T' D systems are quite stable -- if such a system existed in the history of PIE, why do we see it nowhere at all, not even a trace (aside from occasional ponderings that English glottalized coda fortis plosives are somehow a relic of a glottalized past, which seems unlikely because this is not seen anywhere outside of Anglic).

Just that. T T' D systems are stable, and one would expect that in many branches of IE - but actually no branch has it, except one Armenian dialect spoken in Georgia, which certainly innovated it under the influence of Georgian. (Ossetic also has such sounds, but they are clearly innovated there, too - again influenced by neighbouring Caucasian languages.) There really is no good reason to assume such a system in PIE.

[Travis B.]

I also would say that a posited "S4" vowel system for PIE also makes sense because of the significant tendency of *o in Late PIE to be merged with Late PIE/post-PIE *a in multiple branches, as seen in both Germanic and Indo-Iranian. If *o were really [ɒ] this becomes very understandable, much more so than if *o were a higher vowel.

[Travis B.]

What is your guys opinion on the view that *ḱ *ǵ *ǵʰ were really [k g gʰ] and *k *g *gʰ were really [q ɢ ɢʰ] (with the labialized series being either velar or uvular) in Early PIE? I myself am pretty convinced of this, because the idea that prevelar consonants were significantly more common than velar consonants seems pretty nonsensical typologically, whereas a system where the more common "prevelars" were really velar and likewise the less common "velars" were really uvular in Early PIE makes a lot more sense.

(Of course, you could precisely argue, similarly with the glottalic theory, that the fact that we don't see three series in IE daughters is specifically because such an arrangement is unstable -- but one could argue that the uvular series shifted forward, which is plausible because uvulars tend to be more unstable (look at what happened to them in different Arabic varieties), resulting in a chainshift and/or a series merger in different IE daughters.)

[WeepingElf]

What is your guys opinion on the view that *ḱ *ǵ *ǵʰ were really [k g gʰ] and *k *g *gʰ were really [q ɢ ɢʰ] (with the labialized series being either velar or uvular) in Early PIE? I myself am pretty convinced of this, because the idea that prevelar consonants were significantly more common than velar consonants seems pretty nonsensical typologically, whereas a system where the more common "prevelars" were really velar and likewise the less common "velars" were really uvular in Early PIE makes a lot more sense.

(Of course, you could precisely argue, similarly with the glottalic theory, that the fact that we don't see three series in IE daughters is specifically because such an arrangement is unstable -- but one could argue that the uvular series shifted forward, which is plausible because uvulars tend to be more unstable (look at what happened to them in different Arabic varieties), resulting in a chainshift and/or a series merger in different IE daughters.)

With my model of the development of the Pre-PIE vowels as I posted it earlier today, the traditional model makes more sense than the "back-shifted" model you propose (in which BTW the labiovelars would be labialized uvulars, too), though it could shift either way easily, hence I prefer the rather agnostic terms "front velars" and "back velars". But the front velars palatalize in the satem languages, while no IE language preserves the back velars as uvulars; and as you add in parentheses, the fact that no IE language keeps the system intact without change suggests a typologically marked, unstable system. Also, Uralic and other Mitian languages have palatal stops or affricates, as does NW Caucasian, and such a palatal series could have merged with the palatalized velars, explaining their otherwise unexpected high frequency.

[Travis B.]

What is your guys opinion on the view that *ḱ *ǵ *ǵʰ were really [k g gʰ] and *k *g *gʰ were really [q ɢ ɢʰ] (with the labialized series being either velar or uvular) in Early PIE? I myself am pretty convinced of this, because the idea that prevelar consonants were significantly more common than velar consonants seems pretty nonsensical typologically, whereas a system where the more common "prevelars" were really velar and likewise the less common "velars" were really uvular in Early PIE makes a lot more sense.

(Of course, you could precisely argue, similarly with the glottalic theory, that the fact that we don't see three series in IE daughters is specifically because such an arrangement is unstable -- but one could argue that the uvular series shifted forward, which is plausible because uvulars tend to be more unstable (look at what happened to them in different Arabic varieties), resulting in a chainshift and/or a series merger in different IE daughters.)

With my model of the development of the Pre-PIE vowels as I posted it earlier today, the traditional model makes more sense than the "back-shifted" model you propose (in which BTW the labiovelars would be labialized uvulars, too), though it could shift either way easily, hence I prefer the rather agnostic terms "front velars" and "back velars". But the front velars palatalize in the satem languages, while no IE language preserves the back velars as uvulars; and as you add in parentheses, the fact that no IE language keeps the system intact without change suggests a typologically marked, unstable system. Also, Uralic and other Mitian languages have palatal stops or affricates, as does NW Caucasian, and such a palatal series could have merged with the palatalized velars, explaining their otherwise unexpected high frequency.

Of course, your hypothesis implies that, assuming the Mitian hypothesis is true, 'front velars' in PIE which are not in what would be palatalizing environments should map to palatals in other Mitian languages. This is something that can be tested.

[WeepingElf]

Of course, your hypothesis implies that, assuming the Mitian hypothesis is true, 'front velars' in PIE which are not in what would be palatalizing environments should map to palatals in other Mitian languages. This is something that can be tested.

In theory, yes. In practice, Proto-Mitian may be out of reach of the comparative method, similar to how bones (or other organic remains) more than about 50,000 years old are out of reach of C14 dating. The deeper you dig, the smaller and more numerous the correspondence sets become because more and more conditioned sound changes accumulate, and st some point, the correspondence sets become so fragmented that each contains only a single item, at which point reconstruction is no longer possible because there is no way telling real correspondences from chance resemblances. All we have are some morphological elements, chiefly pronouns, that resemble each other in several language families, but hardly any lexical cognates, which, while best explained by descent from a common ancestor (pronouns are unlikely Wanderwörter), are insufficient to build a reconstruction of that common ancestor on. Even most of the Proto-Uralic words resembling PIE words are more likely loanwords from PIE into Proto-Uralic, as can be told by the near-trivial sound correspondences: each PIE phoneme is mapped to the nearest PU morpheme, with ablaut grades and even vowel-colouring effects of laryngeals, both of which are certainly not of Proto-Indo-Uralic age, faithfully reflected in the PU words. So we get into a quagmire of speculation where many things seem plausible but nothing is certain. (But at least, we can build conlangs on the things seeming plausible .)

[Travis B.]

Hmm, I'm reading stuff like this that implies that most of the similarities between PIE and Proto-Uralic are likely either borrowing or chance, and that the Urheimat of PU is likely as far east as Yakutia, contrary to past views that set it in the Urals. As for Mitian, I'm seeing it suggested that possibly Mitian is a real thing but does not include IE at all.

[Skookum]

Not an expert by any means but I‘ve taken an interest in Uralic linguistics recently and it seems like people are increasingly skeptical of even PIE > PU loans, at least on the Uralicist side. I think the consensus is that PU is a considerably younger language than PIE and therefore the timelines don‘t match up, leading to the reinterpretation of most early loans as being from (Proto-)Indo-Iranian.

As Travis mentioned, if an eastern origin of Uralic is accepted, this makes the possibility of PIE > PU loans even less likely, although I’m not sure what the state of the research on that question is.

[hwhatting]

Re gsandi: His views on PIE are here.

I’m curious about his reconstruction of a five-vowel system /a e i o u/.

I assume they're just treating syllabic *y *w as phonemic, and part of the minority that recognises a distinct *a

Treating [i,u] and [y,w] as positional variants is usual; AFAIK, the somewhat more frequent position is to treat the glide variant as the archiphoneme and only have the ablauting vowels /e/, /o/ (and perhaps /a/) as "real" vowels.
BTW, gsandi is He/Him.