Early PIE stops

[WeepingElf]

I am currently pondering an idea about Early PIE stops, namely that the *D set weren't really stops but voiced fricatives. The *T set would then just be voiceless stops and the *Dh set voiced stops. This may explain why *b appears to have been missing: if it actually was [β], it may have easily merged with *w (which would also explain the frequency of the initial clusters *wl- and *wr- . The dental *d would then have been [ð], a notoriously unstable sound, which may explain apparent debuccalizations to *h₁ (which IMHO was [h]), as perhaps in the numeral *h₁wih₁ḱmtih₁ '20' <? *dwi-dḱmt-ih₁ 'twice ten'. Also, *g^w appears to have been less stable than the other labiovelars, becoming w in Anatolian and b early in Celtic. A shift *ɣ^w > *w or *β seems very likely.

I don't know how much sense this makes, though. There are some people here who have good knowledge and good ideas about PIE - what do you think?

[Zju]

Sounds intriguing, but honestly, has about as much going for it as the rest of the dozen or so hypotheses about early PIE stops. All the arguments about *b also work for implosives, for instance.

[Creyeditor]

Isn't it less plausible simply because (non-contextual) fortition is less common crosslinguistically?

[Travis B.]

Sounds intriguing, but honestly, has about as much going for it as the rest of the dozen or so hypotheses about early PIE stops. All the arguments about *b also work for implosives, for instance.

Except distribution - bilabial implosives are significantly more common than velar implosives, which is not true of the PIE *D set.

[Rounin Ryuuji]

I wouldn't be terribly surprised if at least some dialects had the realisations WeepingElf describes.

[hwhatting]

A problem with that account is that where we actually find fricatives in the early PIE languages, they thend to continue the aspirated stops, not the unaspirated stops. (My personal thory is that the T and DH series were both fortis with aspiration (although non-phonemical in the case of the T series), and the D series were lenis.)

[Moose-tache]

One problem here is that /ð/ would need to become /d/ without merging with original /d/ in Germanic, Italic, Armenian, Indo-Iranian, and Greek, probably by original /d/ changing in some way, such as gaining voiced aspiration. That's just the existing theory with extra steps.

[WeepingElf]

Points taken. In fact, I have abandoned the idea again in favour of what I call the "aspiration theory": *T=aspirated voiceless, *D=voiced, *Dh=breathy-voiced. It would be nice to know what happened in Anatolian. What we know is that *D and *Dh merged there, but as what? Apparently something that could not be rendered properly in cuneiform, so they had to write it according to Sturtevant's Rule: medially, the reflexes of *T are spelled double, those of *D/*Dh single. IMHO, the most likely solution is a fortis vs. lenis contrast of the kind we find in Upper German. In Lycian and Lydian, *D/*Dh seem to have given voiced fricatives, but these languages are attested late and their phonologies are not well understood, either.

[Zju]

Sounds intriguing, but honestly, has about as much going for it as the rest of the dozen or so hypotheses about early PIE stops. All the arguments about *b also work for implosives, for instance.

Except distribution - bilabial implosives are significantly more common than velar implosives, which is not true of the PIE *D set.

I was implying that **ɓ could have easily shifted to *w in most positions in PIE already, thus explaining the frequency of *wR, etc.

[Travis B.]

Sounds intriguing, but honestly, has about as much going for it as the rest of the dozen or so hypotheses about early PIE stops. All the arguments about *b also work for implosives, for instance.

Except distribution - bilabial implosives are significantly more common than velar implosives, which is not true of the PIE *D set.

I was implying that **ɓ could have easily shifted to *w in most positions in PIE already, thus explaining the frequency of *wR, etc.

I understand that. What I was saying is that for the *D series to be implosive, it would require an awfully high frequency of [ɠ], which is a quite rare phone crosslinguistically.

[Zju]

Except distribution - bilabial implosives are significantly more common than velar implosives, which is not true of the PIE *D set.

I was implying that **ɓ could have easily shifted to *w in most positions in PIE already, thus explaining the frequency of *wR, etc.

I understand that. What I was saying is that for the *D series to be implosive, it would require an awfully high frequency of [ɠ], which is a quite rare phone crosslinguistically.

I'm not sure I follow. We have three statements:

1. ɓ typically occurs more often than ɠ in any language that has implosives.
2. PIE *D series used to be implosive at an earlier stage.
3. PIE **ɓ shifted to PIE *w (almost) unconditionally.

How does any of that require that PIE *g / *ǵ is very common? If anything, there is an upper limit of its restriction.

[Travis B.]

I was implying that **ɓ could have easily shifted to *w in most positions in PIE already, thus explaining the frequency of *wR, etc.

I understand that. What I was saying is that for the *D series to be implosive, it would require an awfully high frequency of [ɠ], which is a quite rare phone crosslinguistically.

I'm not sure I follow. We have three statements:

1. ɓ typically occurs more often than ɠ in any language that has implosives.
2. PIE *D series used to be implosive at an earlier stage.
3. PIE **ɓ shifted to PIE *w (almost) unconditionally.

How does any of that require that PIE *g / *ǵ is very common? If anything, there is an upper limit of its restriction.

The thing is that PIE *g / *ǵ is, to my knowledge, not rare at all, unlike PIE *b.

[Zju]

I understand that. What I was saying is that for the *D series to be implosive, it would require an awfully high frequency of [ɠ], which is a quite rare phone crosslinguistically.

I'm not sure I follow. We have three statements:

1. ɓ typically occurs more often than ɠ in any language that has implosives.
2. PIE *D series used to be implosive at an earlier stage.
3. PIE **ɓ shifted to PIE *w (almost) unconditionally.

How does any of that require that PIE *g / *ǵ is very common? If anything, there is an upper limit of its restriction.

The thing is that PIE *g / *ǵ is, to my knowledge, not rare at all, unlike PIE *b.

Isn't the question rather if it's more common than both *b and *w?

[WeepingElf]

[...]
I'm not sure I follow. We have three statements:

1. ɓ typically occurs more often than ɠ in any language that has implosives.
2. PIE *D series used to be implosive at an earlier stage.
3. PIE **ɓ shifted to PIE *w (almost) unconditionally.

How does any of that require that PIE *g / *ǵ is very common? If anything, there is an upper limit of its restriction.

It is a fact - a fact that speaks against implosives - that PIE *ǵ/g/g^w are common and *b is almost non-existent. That labial implosives usually occur more often than velar ones in languages that have them is a typological fact which probably means that velar implosives are more likely to change into something else than labial ones, and in PIE according to the "implosive theory", we observe the exact opposite. That of course doesn't mean that the theory is impossible, but it invalidates one major argument adduced by Gamkrelidze/Ivanov for their glottalic theory (i.e., that the *D set were ejectives, which show the opposite trend to implosives - there are languages with /ɓ/ and /k'/ but neither /p'/ nor /ɠ/ - and the *D set conforms to this opposite trend). (Also, the Kartvelian languages Gamkrelidze and Ivanov used as a typological reference point have *p', and allow two ejectives in a root. A constraint against two ejectives in a root exists in Akkadian, though, and it also lacks p', although the latter - but not the former - is a Semitic family trait.)

I feel that my "aspiration theory" is the most parsimonious solution; and it seems as if the *D set was somewhat "weaker" than the *T and *Dh sets, which, together with observations about Uralic and other "Mitian" languages (none of which have any sort of non-pulmonic consonants as a family trait - the Itelmen ejectives are quite clearly an innovation of that language, which is the only non-IE "Mitian" language with non-pulmonics I can think of) had inspired me to come up with the "*D set were fricatives" idea in the first place. But then, "weak" is not really a phonetic concept, it is doubtful whether Proto-Uralic had voiced fricatives at all (there are many hypotheses under discussion what *δ really was - ask six Uralicists and you get at least seven different answers) and "Mitian" is not a solid foundation to build theories on!

[Travis B.]

I'm personally most convinced that *D was a fricative series, because that'd perfectly explain the instability/absence of *b, much more so than *b being an implosive, and it would allow for common *ǵ/g/g^w unlike those being implosives (especially if *g were a uvular implosive, which is a simply ridiculous concept), which seems very far-fetched to me typologically. It also perfectly explains the development of *d in some branches of IE as well.

[WeepingElf]

I'm personally most convinced that *D was a fricative series, because that'd perfectly explain the instability/absence of *b, much more so than *b being an implosive, and it would allow for common *ǵ/g/g^w unlike those being implosives (especially if *g were a uvular implosive, which is a simply ridiculous concept), which seems very far-fetched to me typologically. It also perfectly explains the development of *d in some branches of IE as well.

Fair. I haven't abandoned the idea completely, but have grown mildly skeptical of it. Perhaps there was an early stage with a fricative *D set, and a later one for which the "aspiration theory" holds.

[Zju]

I'm personally most convinced that *D was a fricative series, because that'd perfectly explain the instability/absence of *b, much more so than *b being an implosive, and it would allow for common *ǵ/g/g^w unlike those being implosives (especially if *g were a uvular implosive, which is a simply ridiculous concept), which seems very far-fetched to me typologically. It also perfectly explains the development of *d in some branches of IE as well.

The problem with it being a fricative series is that unconditional fortition is typologically unusual, and just having a voiced non-sibilant fricative series without the corresponding voiceless is.
Uvular implosives do occur, and we don't know that PIE even had uvulars. *b having been **ɓ does explain it changing to *w.

[Travis B.]

I'm personally most convinced that *D was a fricative series, because that'd perfectly explain the instability/absence of *b, much more so than *b being an implosive, and it would allow for common *ǵ/g/g^w unlike those being implosives (especially if *g were a uvular implosive, which is a simply ridiculous concept), which seems very far-fetched to me typologically. It also perfectly explains the development of *d in some branches of IE as well.

The problem with it being a fricative series is that unconditional fortition is typologically unusual, and just having a voiced non-sibilant fricative series without the corresponding voiceless is.

Consider the fate of the voiced obstruent series in Upper German; they became plosives unconditionally, unlike in many other Germanic varieties, and it is most parsimonious to posit they were originally voiced fricatives in many positions (i.e. when not initial), due to the nature of Verner's Law (which makes most sense when one views it as voiceless fricatives becoming voiced fricatives).

Uvular implosives do occur, and we don't know that PIE even had uvulars.

Uvulars are the best answer to the frequency of IE "palatal/prevelar" versus "velar" series, because it is improbable that palatal/prevelar consonants would be so much more common than velar consonants, whereas if the "palatal/prevelar" consonants were velar and the "velar" consonants were uvular their frequency would make a whole lot more sense.

*b having been **ɓ does explain it changing to *w.

Except it is hard to explain typologically common consonants all basically disappearing in a practically unconditional sound shift while typologically rare to mythical consonants are both very improbably common and are highly conserved outside of satem languages.

[Travis B.]

The reason to believe the *T series was aspirated to my knowledge mostly rests on Grimm's Law and that the fortis plosive series tends to be aspirated in modern Celtic languages (but this could possibly be explained away as Germanic influence). I am not sure if this is strong enough evidence to trace an aspirated *T series all the way back to PIE, since there is little evidence of aspiration of *T elsewhere in IE.

[WeepingElf]

The reason to believe the *T series was aspirated to my knowledge mostly rests on Grimm's Law and that the fortis plosive series tends to be aspirated in modern Celtic languages (but this could possibly be explained away as Germanic influence). I am not sure if this is strong enough evidence to trace an aspirated *T series all the way back to PIE, since there is little evidence of aspiration of *T elsewhere in IE.

It's not just Germanic, Celtic and Armenian - the idea is that the PIE root structure constraints suggest that the *T and *Dh sets once formed a class at the exclusion of the *D set, for which the aspiration of the *T set is IMHO the easiest solution. At least, it has the advantage over the various flavours of glottalic theory that it doesn't propose anything that is not reflected as such in any IE language. Of course, in Greek and Indo-Iranian, the *T set is most emphatically not aspirated, as these languages have a distinct Th set from other sources (Indo-Iranian: mostly *T + laryngeal; Greek: *Dh, but also *T + laryngeal); also, it is pretty certain that the *T set was not aspirated in Latin, and probably also not in Balto-Slavic - but in the latter, the *T set may have lost its aspiration at the same time as the *Dh set.

What concerns the "*D set as fricatives" idea, the "Uralic evidence" is actually non-evidence, because (1) we don't know whether IE and Uralic are particularly closely related or not and (2) few Uralicists now believe that *δ was actually /ð/ and *δ' a palatalized counterpart thereof, and *γ (now written *x) was /ɣ/. There are many different ideas about what sounds these were - ask six Uralicists and you get at least seven different answers - but there is apparently some reason to assume that the "deltas" were some sort of laterals and *x some kind of velar-to-glottal fricative or approximant, as mysterious as the PIE laryngeals. Given that, there is no valid reason to assume that PIE *d corresponds to Uralic *δ and PIE *g to Uralic *x, and nothing can be concluded from there!