So, not to sound like a crank, but... I find a lot of details about reconstructed PIE a little hard to believe

[WeepingElf]

Thorn: lax (modal/breathy?), secondarily voiceless
Tyr: tense (creaky/modal?), voice-neutral
Dag: lax (modal/breathy?), secondarily voiced

This is similar to what I fancy:

"Thorn": [+breath][-voice]
"Tyr": [-breath][±voice]
"Dag": [+breath][+voice]

Well, the *D-set ("Tyr") may have been [+voice], and developed from a set of voiced spirants, as they are found in other Mitian languages like Uralic or Eskimo-Aleut (these two families appear to be especially conservative among the eight Mitian member families); of these, the labial one had merged with *w, thus explaining the "labial gap" in PIE and by the same time initial clusters of the type *wr- and *wl-. The *T and *Dh sets would in this scenario have evolved from a split of the single Proto-Mitian stop type. Alas, without sufficient lexical cognates, this remains speculation.

[Nortaneous]

absurd sound changes like /t’/ > /d/

isn't ejective > implosive attested somewhere in Mayan? (but probably pʼ > ɓ and/or qʼ > ʛ without a comprehensive shift)

[Moose-tache]

absurd sound changes like /t’/ > /d/

isn't ejective > implosive attested somewhere in Mayan? (but probably pʼ > ɓ and/or qʼ > ʛ without a comprehensive shift)

Both sides of the debate use the "But it occurs naturally in Upper Plains Borkdork, so it's perfectly reasonable." I'm guilty of it myself above when I used Austro-Asiatic languages as examples of "phonation" distinctions on phonemically voiced plosives. But the goal should never be to find a language somewhere in the cloud forests of New Ireland to come to our aid, but to demostrate the most parsimonious route from point A to point B. What I like about the Clackson model (i.e. that Tyr was a voiced consonant that involved some partial closure of the glottis either on itself or the adjacent vowels) is that it only needs one very simple sound change to become a plain voiced stop, its most common realization in attested languages, but still has the capacity to explain why Tyr behaves like a glottalized consonant.

[WeepingElf]

Well, the *D-set ("Tyr") may have been [+voice], and developed from a set of voiced spirants, as they are found in other Mitian languages like Uralic or Eskimo-Aleut (these two families appear to be especially conservative among the eight Mitian member families); of these, the labial one had merged with *w, thus explaining the "labial gap" in PIE and by the same time initial clusters of the type *wr- and *wl-. The *T and *Dh sets would in this scenario have evolved from a split of the single Proto-Mitian stop type. Alas, without sufficient lexical cognates, this remains speculation.

Let me attempt a crude graphical illustration:

Code: Select all

Stage I       II      III     IV

      *T------*Th--+--*Th-----*T
                   |
      *Ð------*D------*D------*D
                   |
                   +--*Dh-----*Dh

Stage I is the hypothetical Mitian language, perhaps Proto-Indo-Uralic. Stage II is an intermediate step, after which *Dh forks from *Th, Stage III is Early PIE, Stage IV is Late PIE. The resulting system was of course unstable, which is precisely the reason why no IE language has preserved it unchanged.

[Moose-tache]

I usually try to keep conlanging out of the natlang forum, so I won't comment on the Mitian part. But the part about aspiration on the Thorn and Dag series is of course an important question for PIE.

Early versions of the glottalic hypothesis used allophonic aspiration to explain the various realization across languages: Hellenic used unaspirated Thorn and aspirated Dag, while Armenian used aspirated Thorn and unaspirated Dag. But there are two problems. First, while there's nothing impossible about having aspirated and non-aspirated forms in free variation with no conditioning environment, as a theory it's pure handwavium. Second, the German and Armenian shifts look secondary, so only Dag shows compelling evidence of being aspirated at any point during late PIE. Your model above, in which the Thorn series is temporarily aspirated, would certainly "make sense," but there isn't any direct evidence for it.

As I've said before, I like the traditional model, and the fact that it's weird and unstable is (as you say) not really much of a drawback. non-Anatolian PIE may have split up over the course of just a few centuries in the early third millennium, so PIE could be a very brief snapshot indeed.

[WeepingElf]

I usually try to keep conlanging out of the natlang forum, so I won't comment on the Mitian part. But the part about aspiration on the Thorn and Dag series is of course an important question for PIE.

Fair. "Mitian" is a speculative entity that is perhaps better left out of consideration. It would be nice if one could find cognate sets between IE and Uralic, but that would be a major research project which lies well out of the scope of us petty amateurs!

Early versions of the glottalic hypothesis used allophonic aspiration to explain the various realization across languages: Hellenic used unaspirated Thorn and aspirated Dag, while Armenian used aspirated Thorn and unaspirated Dag. But there are two problems. First, while there's nothing impossible about having aspirated and non-aspirated forms in free variation with no conditioning environment, as a theory it's pure handwavium. Second, the German and Armenian shifts look secondary, so only Dag shows compelling evidence of being aspirated at any point during late PIE. Your model above, in which the Thorn series is temporarily aspirated, would certainly "make sense," but there isn't any direct evidence for it.

Fair. The approach was to modify the traditional model minimally such that the *T and *Dh sets form a class at the exclusion of the *D set, as suggested by the root structure constraints, and explain why *b appears to be absent or at least rare. I admit that this is just speculation, and that there is no evidence in favour of aspiration of the *T set outside Germanic and Armenian, which may have (independently) innovated here. But the glottalic theory makes a bigger change, and it is no less speculative, as there is no evidence in support of it!

IAs I've said before, I like the traditional model, and the fact that it's weird and unstable is (as you say) not really much of a drawback. non-Anatolian PIE may have split up over the course of just a few centuries in the early third millennium, so PIE could be a very brief snapshot indeed.

Yes - the traditional model is the best match for the data we have, and that's why it is reconstructed like that. It may be rare among the world's languages and unstable - and indeed, it is preserved in no branch unchanged!