The Great Proto-Indo-European Thread

[Tropylium]

Germanic and Italic are unlikely to be more closely related than a common European subgroup,

Interestingly enough, I've seen a recent talk handout that claims Italo-Germanic to be about as well-supportable as Italo-Celtic.

This does not seem to be online anymore, but here is a list of some exclusive I-G innovations mentioned:
– *tstr > *str
– a zero-grade formation in *pisk- 'fish'
– *tak- > *tak-ē- 'to be silent'
– (s)poimn- > *spoim- 'foam'
– *dwis- > *dwis-no- 'double'
– Prepositions of origin ending in *-nē

Doesn't look super fancy to me, but then again, the list of exclusive I-C innovations I've seen is not super huge either.

[WeepingElf]

Kortlandt has also suggested in his paper Indo-Uralic consonant gradation more or less the same idea as here, but with stress as the trigger (and with widespread later levelling of the resulting *T ~ *Dʰ alternations).

Hmm, interesting. So the difference between *T and *Dh may be reduced to a consonant gradation similar to the consonant gradation systems found in various Uralic languages, and cognate to them. (AFAIK, the standard reconstruction of Proto-Uralic does not feature consonant gradation, but it may simply have been allophonic in Proto-Uralic - and already in place in Proto-Indo-Uralic. Also keep in mind that AFAIK there is a similar thing in Eskimo-Aleut.)

As for Uralic correspondences, at least the PU gemination contrast (*p *t *č *ć *k : *pp *tt *čč *ćć *kk) could be cognate to one of the IE phonation contrasts (and most variations on Indo-Uralic / Eurasiatic / Nostratic propose that it is). But interestingly enough there's hardly any good evidence for a *tt in Proto-Uralic, in which light I suspect that the geminate series could instead result from some sort of earlier cluster simplification (e.g. *jp *jt *jk > *pp *ćć *kk, or *kp *pk > *pp *kk while *pt *kt remain).

Another interesting observation. The PIE *D grade may be cognate to PU geminates, but the near-absence of *tt in Uralic and the near-absence of *b in IE need to be accounted for. Cases of PIE *d may be cognates of PU *čč or *ćć, and the near-absence of PIE *b may be due to a *b > *w merger in PIE, as has been discussed here.

Still, such speculations must be founded on actual cognate sets, of course, but they may guide us to where we should look.

But then does that not just beg the question of why the Hittites did not simply follow the Akkadian practice of not indicating their phonation contrast, rather than introducing a practice (iirc) not present in Akkadian?

Apparently, the phonation contrast in Hittite was different from that in Akkadian. Hence the idea that it was a contrast of aspiration rather than voice.

I think a lot more research needs to go into the actual phonetic realisation of the three series. After all, voiced aspirates are only attested in a single branch (Indic), and aspiration is attested in what could well be an actually genetic subgroup (Graeco-Aryan). In the north-west, both Germanic and Italic point to a realisation of this series as voiced fricatives in their histories, which should also be taken into account. Moreover, Germanic and Italic are unlikely to be more closely related than a common European subgroup, probably also including Celtic and Balto-Slavic

As I said a few days ago, I fancy that there was a dialectal differentiation in the phonetic realization of the *Dh series: breathy-voiced stops in Greco-Aryan and spirants in Northwestern (Italic-Celtic-Germanic-Baltic-Slavic).

(and perhaps also Armenian? Iirc the affinities of Armenian are uncertain)

AFAIK, Armenian is often considered to be affiliated in some way to Greek, and counted among the Greco-Aryan languages (also, some of its dialects at least vote for the *Dh series being breathy-voiced stops). Were you thinking of Albanian?

[KathTheDragon]

Germanic and Italic are unlikely to be more closely related than a common European subgroup,

Interestingly enough, I've seen a recent talk handout that claims Italo-Germanic to be about as well-supportable as Italo-Celtic.

This does not seem to be online anymore, but here is a list of some exclusive I-G innovations mentioned:
– *tstr > *str
– a zero-grade formation in *pisk- 'fish'
– *tak- > *tak-ē- 'to be silent'
– (s)poimn- > *spoim- 'foam'
– *dwis- > *dwis-no- 'double'
– Prepositions of origin ending in *-nē

Doesn't look super fancy to me, but then again, the list of exclusive I-C innovations I've seen is not super huge either.

Personally, I don't think there are subgroupings beyond Common European. However, the exact make-up of that group isn't obvious. Italic, Celtic, and Germanic are almost certain, but you might make a case for Balto-Slavic.

Kortlandt has also suggested in his paper Indo-Uralic consonant gradation more or less the same idea as here, but with stress as the trigger (and with widespread later levelling of the resulting *T ~ *Dʰ alternations).

Hmm, interesting. So the difference between *T and *Dh may be reduced to a consonant gradation similar to the consonant gradation systems found in various Uralic languages, and cognate to them. (AFAIK, the standard reconstruction of Proto-Uralic does not feature consonant gradation, but it may simply have been allophonic in Proto-Uralic - and already in place in Proto-Indo-Uralic. Also keep in mind that AFAIK there is a similar thing in Eskimo-Aleut.)

But then does that not just beg the question of why the Hittites did not simply follow the Akkadian practice of not indicating their phonation contrast, rather than introducing a practice (iirc) not present in Akkadian?

Apparently, the phonation contrast in Hittite was different from that in Akkadian. Hence the idea that it was a contrast of aspiration rather than voice.

Quite. Of course, my money is on gemination, though I can't derive gemination without an intermediate step of aspiration, so it's just a detail at this point.

Kortlandt has also suggested in his paper Indo-Uralic consonant gradation more or less the same idea as here, but with stress as the trigger (and with widespread later levelling of the resulting *T ~ *Dʰ alternations).

Hmm, interesting. So the difference between *T and *Dh may be reduced to a consonant gradation similar to the consonant gradation systems found in various Uralic languages, and cognate to them. (AFAIK, the standard reconstruction of Proto-Uralic does not feature consonant gradation, but it may simply have been allophonic in Proto-Uralic - and already in place in Proto-Indo-Uralic. Also keep in mind that AFAIK there is a similar thing in Eskimo-Aleut.)

(and perhaps also Armenian? Iirc the affinities of Armenian are uncertain)

AFAIK, Armenian is often considered to be affiliated in some way to Greek, and counted among the Greco-Aryan languages (also, some of its dialects at least vote for the *Dh series being breathy-voiced stops). Were you thinking of Albanian?

Maybe? I've never considered Albanian to be anything other than closely related to Graeco-Aryan, but I swear there was one that was sometimes considered clsoe to Graeco-Aryan, sometimes to European, which I thought was Armenian.

[Pabappa]

If Italo-Germanic is real, Venetic might be in it. Its a languager long considered to be a very basal split from Italic, but could actually be a separate family that was merely influenced by Italic.

https://en.wikipedia.org/wiki/Venetic_l ... sification

Aslo, could Illyrian be the answer to your questripon?

[Terra]

Hmm, interesting. So the difference between *T and *Dh may be reduced to a consonant gradation similar to the consonant gradation systems found in various Uralic languages, and cognate to them. (AFAIK, the standard reconstruction of Proto-Uralic does not feature consonant gradation, but it may simply have been allophonic in Proto-Uralic - and already in place in Proto-Indo-Uralic. Also keep in mind that AFAIK there is a similar thing in Eskimo-Aleut.)

Kortlandt says:

The expected distribution of fortes and aspirates can now be specified as
follows. The two types of Uralic weak stops (before and after the vocalic nucleus
of a weak syllable) apparently merged into the Indo-European aspirates while
the Uralic strong stops (before the vocalic nucleus of an open syllable) became
the Indo-European fortes. Initial stops adopted the same pattern, which resulted
in a consonant alternation in roots with fortes before zero grade suffixes and
aspirates before full grade suffixes, e.g. *tekm, *d‘g‘em- ‘earth’. After the analogi-cal introduction of *o in unstressed syllables at stage C, we obtain paradigms
like the following (cf. Beekes 1995: 178):
nom. *g‘eiom ‘winter’ *nepot ‘grandson’
acc. *g‘iem(m) *nepot(m)
loc. *g‘iem(i) *nepot(i)
abl. *g‘imes *neptos
inst. *g‘imet *neptot
In the nominative *g‘eiom, which replaced *keim, the full grade suffix was apparently
introduced from the oblique form at this stage, while nom. *nepot and
obl. *nepot- may represent an earlier paradigm *nept, *nb‘ed‘-, with generalization
of fixed stress on the initial syllable. These examples show how fortes and
aspirates could become associated with fixed and mobile stress patterns, respectively.

I already knew about *dhghem 's strange vowel alteration; Hittite preserves a vowel in-between, pointing to *dheghm, or if Kortlandt is right, actually *tekm. Is there any direct evidence for *tekm or *keiom though?

Also, *delgh and *dlogh come to mind. In Leiden's Italic Ety Dictionary, De Vaan says of "longus":

The adj. longinquus was probably formed on the basis of longe (thus EM) as an
antonym to propinquus. A major question is whether PIE *d(o)Iugh
- (Hit.), *doligh
-,
*delgh
- (Gr.), *dlongh
- (Lat., Gm.) and *dlh;gh
- (Ilr., BSL, Alb.) are different variants
of what was originally a compound petrified in the meaning 'long' (Kloekhorst 2008:
819f. compares cases such as MoE high and dry, safe and sound). The first member
could be *de/ol(hLcompare Cz. del (poet.) [f. i] 'length' < PS1. *dblb, OCS
prodbljg 'I prolong', Ru. dlit'sja, Cz. dliti 'to last', Ru. dlina 'length'; the second
member containing *-gh
-.
Bibl.: WH I: 820, EM 366, IEW 197.

Kortlandt's theory would say yes, they are related; The alternate in the same way that *dhghem and *tekm does. However, shouldn't it actually be *delk and *dlogh ? Again, what direct evidence do we have for *delk ?

Does anybody know any other roots that have alterations like this?

[KathTheDragon]

I already knew about *dhghem 's strange vowel alteration; Hittite preserves a vowel in-between, pointing to *dheghm, or if Kortlandt is right, actually *tekm.

Hittite tēkan cannot point to *tekm, due to the consistent single spelling of the velar, which therefore indicates *ǵ or *ǵʰ. Interestingly, Kloekhorst has found evidence for the plain voiced stop, instead of the aspirated stop, in the Sanskrit oblique cases jmaḥ, jmā́, jmán, and further connects the root to the s-mobile verb *steǵ- "to cover".

Gelb, in 'Old Akkadian Writing and Grammar', states that the writing system for Old Akkadian did not distinguish voiceless, voiced and emphatic. By 'Old Akkadian' he means up to the end of the Ur III dynasty.

Kloekhorst, in his paper Initial Stops in Hittite, claims precisely the opposite for Alalaḫ Akkadian, and supports himself with a fair amount of evidence. The implication is that Hittite adopted cuneiform from them, which would therefore be very strong evidence against any kind of voice contrast in Hittite.

[2+3 clusivity]

Where did initial <a-> in the Balto-Slavic 1st p. sg. come from -- e.g. OCS, Lith.? It seems like the result should be in <e->

[Zju]

e was lengthened to ē due to Winter law, and ē regularly gave ja in word initial positions. This was reversed in roots which had prefixed variants due to analogy, but not in the pronoun. The dropping of initial j in azъ is irregular. I don't know about Lithuanian though.

[Zju]

So apparently PIE *kswéḱs 'six' is borrowed from Proto-Kartvelian *eks₁w- 'id.', but how did exactly the borrowing occur? My wild guess is that reduplication was involved: *eks₁w > **ekswéksw > *kswéḱs, but this is rather problematic, as one would expect **kswéksu in such case and not *kswéḱs.

PIE septḿ̥ 'seven' and PK šwid- 'id.' also seem to be cognates, but it's unclear in exactly what way. Has any researcher looked into this?

[KathTheDragon]

I'm not convinced that reconstructing an initial *k for "6" is at all necessary (only Avestan seems to need it, so a purely Avestan explanation is in order), and it may well be that the *s is due to analogical interference from "7" (observe that it is not present in all branches, or simply replaces the *w), leaving an underlying *weḱs for the numeral.

[Zju]

I'm not convinced that reconstructing an initial *k for "6" is at all necessary (only Avestan seems to need it, so a purely Avestan explanation is in order), and it may well be that the *s is due to analogical interference from "7" (observe that it is not present in all branches, or simply replaces the *w), leaving an underlying *weḱs for the numeral.

How do you derive Proto-Slavic *šes- if not from *xes- < *kses- < *kswéḱs ? Doesn't Sanskrit ṣáṣ also need initial k- to trigger retroflexion?

[Sleinad Flar]

This suggests assimilation for the Sanskrit form (*sáṣ > ṣáṣ), and argues that the Proto-Indic form lost the /w/ due to analogy with *sapta "seven". I don't know about the Slavic forms though. This author doesn't count "six" among the *ksw-words.

[KathTheDragon]

Derksen takes it from *sweḱs via the ordinal *suḱs-t(h₂)o-, with an assimilation like Sanskrit.

[vokzhen]

The numbers six and seven have come up in a thread before.
PIE: swéḱs séptm
Proto-Semitic: šidṯ šabʻ(atum, in feminine) (š is probably /s/)
Egyptian: yVssá(w) sáfḫa(w) (-w is masculine)
Etruscan: śa semph (possibly, 6 may have been huth instead)
Iberian: śei sisbi (and modern Basque sei zazpi)
Proto-Kartvelian: ekšw šwid
Hurrian: šeše šinti (š is probably /s/)
While there may be some chance resemblance, there was probably a lot of borrowing that happened. PIE and Proto-Semitic seem particularly close for 7 and probably borrowed one to the other or from the same donor language, and Proto-Kartvelian and PIE seem particularly close for 6. But I really don't think I've ever of anything more substantial than that in terms of how they might be related.

[2+3 clusivity]

Within Indic, Sanskrit and several of the dardic languages are unusual in the early drop of initial k in six leaving only a hint of ruki. Most show regular developments from /*ks-/. Looking to the east, see pali and hindi in ch-. Oddly, Sanskrit - or at least Vedic - was probably very unrepresentarive of the majority of the later Indic branch. Nuristani also shows ruki triggering.

Edit Khotanese six <kʂäta’, kʂei’> is also tempting but was probably in the first case /ʈʂʰəta-/ (gotta love Iranian languages with retroflex consonants and aspiration). Tumshuqese six is transcribed as <xʃi> for the latter language when they added new letters, but the initial sound is unclear. I do not think it is transcrbed earlier.

[hwhatting]

Derksen takes it from *sweḱs via the ordinal *suḱs-t(h₂)o-, with an assimilation like Sanskrit.

If your source is his etymological dictionary, I think you read the part about *suḱs-t(h₂)o- in a wrong way - he says that the Baltoslavic ordinal was originally *(s)ušto- (still reflected in Old Prussian), but that it was later changed to *šešto- under influence of the cardinal. The (supposed by him) assimilation of *seš- to *šeš- in the cardinal happened Independent of the ordinal, not "via" the ordinal.
Taking into account 2+3's post, we have the following picture:
1) Clear intitial *ks- in Avestan and many Indian languages
2) Something that goes back to either the regular outcome of intitial *ks- or to assimilation in Balto-Slavic
3) Something that goes back to either an irregular outcome of intitial *ks- (namely, with loss of the /k/) or to assimilation in Vedic
4) Western IE languages that show only initial /s/
I think the most parsimonious solution is assuming PIE *kswek's, with Dissimilation to *swek's in Western IE (a trivial simplification of an initial cluster that wasn't frequent in PIE), regular development for 1) and 2), and only one language being irregular (Vedic), while assuming original *swek's would demand irregular devlopments not only in 3), but also in 2), and still would require the initial cluster *ks to have arisen in Indo-Iranian.

[Zju]

Derksen takes it from *sweḱs via the ordinal *suḱs-t(h₂)o-, with an assimilation like Sanskrit.

If your source is his etymological dictionary, I think you read the part about *suḱs-t(h₂)o- in a wrong way - he says that the Baltoslavic ordinal was originally *(s)ušto- (still reflected in Old Prussian), but that it was later changed to *šešto- under influence of the cardinal. The (supposed by him) assimilation of *seš- to *šeš- in the cardinal happened Independent of the ordinal, not "via" the ordinal.
Taking into account 2+3's post, we have the following picture:
1) Clear intitial *ks- in Avestan and many Indian languages
2) Something that goes back to either the regular outcome of intitial *ks- or to assimilation in Balto-Slavic
3) Something that goes back to either an irregular outcome of intitial *ks- (namely, with loss of the /k/) or to assimilation in Vedic
4) Western IE languages that show only initial /s/
I think the most parsimonious solution is assuming PIE *kswek's, with Dissimilation to *swek's in Western IE (a trivial simplification of an initial cluster that wasn't frequent in PIE), regular development for 1) and 2), and only one language being irregular (Vedic), while assuming original *swek's would demand irregular devlopments not only in 3), but also in 2), and still would require the initial cluster *ks to have arisen in Indo-Iranian.

It should be *šeś- instead of *šeš-, though, as Slavic languages only have s as second fricative. It was perhaps something like this: *ksweḱs > *kseḱs > *kseśs > *kseś > *kšeś > *šeś The second fricative being š in Old Prussian seems irregular development and may be due to assimilation of the second sibilant. In Latvian, on the other hand, most likely the development was *šeś- > *šes-, then assimilation *šes- > *ses-, with the second š in seši being secondary development. In Slavic languages the outcome was regular: *šeś- > *šes- > *xes- > *šes-.

[hwhatting]

It should be *šeś- instead of *šeš-, though, as Slavic languages only have s as second fricative. It was perhaps something like this: *ksweḱs > *kseḱs > *kseśs > *kseś > *kšeś > *šeś The second fricative being š in Old Prussian seems irregular development and may be due to assimilation of the second sibilant. In Latvian, on the other hand, most likely the development was *šeś- > *šes-, then assimilation *šes- > *ses-, with the second š in seši being secondary development. In Slavic languages the outcome was regular: *šeś- > *šes- > *xes- > *šes-.

You're right about Slavic, and that makes Derksen's assimilation theory even more unlikely. In Latvian, /s/ for Balto-Slavic /š/ is regular, and secondary palatalization of /s/ to /š/ as well, so there's no need to assume any special developments. Another possible accounting is to assume a dialectal split between Baltic and Slavic, with Baltic continuing *k'swek's- and Slavic a simplified *k'swek'- or *k'swes-, in which case there would be no need to assume assimilation in Baltic.

[2+3 clusivity]

Following up on the above and building off of Hwatting's analysis in particular. Pulling from Indo-European Numerals by Jadranka Gvozdanovic' and my own notes, it seems like the following forms show up in Indic and Iranian:

Iranian

/*xšwaš/ -> Y. Avestan <xšuuaš> /xʃvaʃ/ (for comparative purposes, Old Persian and Old Av. are not attested),
/*(x)šwaš/ -> Parthian <šwh>, Pašto: <špaẓ̌> /ʃpaʐ/ (the final varies by Dial.), Buddhist Sogd. <wγwšw>, etc., many East Iranian minority langauges show /xV[+round]-/
/xšaš/ -> ? Khotanese <kʂäta’, etc.> /ʈʂʰəta-/ (however, <*xšwV-> on occasion gave <kʂä> and <*ʂvi->; regardless <*sw-> gives /hv-/), Tumshuqese <xʃi>, Ossetian (Digor and Iron) <æxsæz>,
/(x)šaš/ -> Zoroastrian Pahl. <šaš>, Modern Persian <šeš>, etc.

Indic (with the major qualification that Old Indic inherited words do NOT show (k)ʂv- – I take this to mean Vedic/Sanskrit, since later langauges of the NW in particular show great diversity)

/*xšwaš/ -> khowar /ʈʂʰoy/
/*(x)šwaš/ -> ? Niya (Tarim Basin, probably related to Gandhari) <ʂo> (based on Skt. <*ʂaʈ> would give <ʂa> in this dialect), Tirahi / x(w)ō/, various Dardic languages show /ʃ~ʂ~xV[+rounded]/, Shina Kohistani /sva/, Romani dialects show up with /ʃov/ (though some have /-eʃ/)
/xšaš/ -> Pali/Prakrit/Sindhi <cha> /tʃʰə/, Dial Pahari /tʃʰe/, Dial Pashai /ʈʂʰa/, most modern Indic languages /tʃʰ-/ or /S-/ developed from /tʃʰ-/ in recent times.
/(x)šaš/ -> Vedic/Sansrkit <ṣaṣ&ʂaʈ> /ʂəʂ&ʂəʈ/, Kashmiri /ʃe/, Dial Pahari /ʃah/, Gilgiti Shina /ʂa/

Nuristani – not much data here due to the languages and difficulty getting the sources actually published. I plan to update it but here is what I have. One big disclaimer is that modern /u/ and /o/ may have been push chained up from /*a/ or another low vowel.
/*(x)šwaš/ Most? W. Kata: /ʂú/, kɒm: /ʂú/, Ashkun: / ʂú, ʂó/, Waigali: /ʂú/
/*(x)šwaš/ too ? but hard to say, because Initial plosives dropped in this language: Vɒʃi of Saitsi: /wūʂ(u)/; Vɒʃi of Uʂüt: /wūʂ(u)/; Vɒʃi by Grierson’s informant from ? <uʃū>.

Given that all the possibilities show up on the two main branches suggests a common /**ksw-/ to me for Indo-Iranian. The greatest diversity of forms is clearly in the Hindu Kush/Pamirs/Tarim Basin.

Seems like loss of /k-/ is common by middle Iranian and, at least in the northern extremes of India, by recent periods certainly (and in parts even in the ancient period as shown by Vedic and the Niya Prakrit). /k-/ descendant forms remain strong in some Dardic areas still and strong though-out the sub-continent. Given the east to west gradient, Slavic practically agrees by having /k-/ dropping forms.

Edit: Also, big lesson from this, old does not equal representative -- Looking at you Sanskrit.

[kanejam]

As for Salmoneus's idea: it may be that an early stage had two types of stops: pulmonic *T and ejective *T'. At that stage, there was a constraint against two ejectives in one root. Permissible root types with two stops were three: TVT, TVT', T'VT, while T'VT' was forbidden. Then some (unknown) vowel feature caused some Ts to attain voicing, leading to the possible root types: TVT, TVT', T'VT, DVD, DVT', T'VD - the antecedents of the "Classical PIE" types TVT, TVD, DVT, DhVDh, DhVD, DVDh.

That would reduce the Pre-PIE stop grades from three to two, and thereby move the whole thing closer to Uralic, which has just one stop grade. Perhaps study of Indo-Uralic sound correspondences will reveal which factors conditioned the splits that happened in pre-PIE - or there were no splits in pre-PIE, just mergers in pre-Uralic.

As Tropylium said, Kortlandt points out that roots with mobile stress tend to have voiced aspirates, whereas static stress tend to have voiceless. Unfortunately they give no examples or figures and I'm sure there are plenty of exceptions, but it is tantalising to connect the stop system with the accent system. This is possibly backed up by a handful of correspondences such as *-bhi (a case/adverbial suffix) with *pi (a preposition? this may not actually exist, I'm drawing from memory) but has a bit of trouble explaining the large number of voiceless consonants in unstressed suffixes.

Along with connecting the IE stops with the single and geminate stops in Uralic, I have also seen an attempt to connect them with the Uralic nasals, of which there are more than in IE and could conceivably come from voiced stops.

I think one of the strengths of the compromise glottalic theory T 'D D is the ease of explaining the lack of *DeD roots, but the bigger strength by far is that it requires far fewer somersaults to get to the attested systems. T D Dh is really only reconstructible for the Greco-Aryan branch/dialect area (I'll argue that the Italic reflex of *Dh never went through a voiceless aspirate stage, instead going straight to voiced fricative as happened in Germanic). Anyone suggesting T T' D > T D Dh before core PIE is just taking the worst parts of both theories.

(and perhaps also Armenian? Iirc the affinities of Armenian are uncertain)

AFAIK, Armenian is often considered to be affiliated in some way to Greek, and counted among the Greco-Aryan languages (also, some of its dialects at least vote for the *Dh series being breathy-voiced stops). Were you thinking of Albanian?

Maybe? I've never considered Albanian to be anything other than closely related to Graeco-Aryan, but I swear there was one that was sometimes considered clsoe to Graeco-Aryan, sometimes to European, which I thought was Armenian.

Vladimir E. Orel, author of the Albanian Etymological dictionary, strongly supports a connection between Albanian and Balto-Slavic. While it also has numerous cognates with Greek, he puts these down mostly to borrowings. He could of course be biased. AFAIK Armenian is usually lumped in with Greek, or at least the Greco-Aryan branch/dialect area. It does seem like you are confusing these but then I could be wrong.

Edit: While I like the idea that phonation and accent are linked, Kortlandt's alternations of *tekm, *d‘g‘em- ‘earth’ seem misguided; if there is a connection, I would expect it to back a bit further than PIE.

[KathTheDragon]

I think one of the strengths of the compromise glottalic theory T 'D D is the ease of explaining the lack of *DeD roots, but the bigger strength by far is that it requires far fewer somersaults to get to the attested systems. T D Dh is really only reconstructible for the Greco-Aryan branch/dialect area (I'll argue that the Italic reflex of *Dh never went through a voiceless aspirate stage, instead going straight to voiced fricative as happened in Germanic). Anyone suggesting T T' D > T D Dh before core PIE is just taking the worst parts of both theories.

The distribution of the realisations of the *Dʰ series is actually very interesting. In the north-west, you have Germanic and Italic pointing to *Ð, and in the south and east, you have Greek, Indic, apparently Armenian, and evidently Tocharian*, all pointing to *Dʰ (or something similar). Albanian, Anatolian, Balto-Slavic and Celtic all merged it with original *D, though of course Balto-Slavic had Winter's Law, and Celtic had *gʷ > *b, both developments before the merger.

*Tocharian underwent Grassman's Law in its development, at the very least for *dʰ, prior to the changes of *d > PToch *tˢ and *dR > PToch *R. Given that Grassman's Law only operated in branches that had aspirates, it's probable that Tocharian also had aspirates.

[hwhatting]

In the north-west, you have Germanic and Italic pointing to *Ð, and in the south and east, you have Greek, Indic, apparently Armenian, and evidently Tocharian*, all pointing to *Dʰ (or something similar).

And nothing in Germanic and Italic speaks against assuming *Dʰ as a stage before *Ð, so we're back at the traditional model. We could chalk up *Dʰ > *Ð
as a further Italo-Germanic isogloss. But that would disturb my favourite reconstruction of the chain of events in Germanic, which would be PIE T D Dʰ > Tʰ D Dʰ (the tenues becoming aspirated by default like in many contemporary Germanic languages) > Þ D Ð (all aspirates become fricatives) > Þ T Ð (the only stop in the system becomes devoiced, as voicing is not distinctive on stops any more).

[Sumelic]

In the north-west, you have Germanic and Italic pointing to *Ð, and in the south and east, you have Greek, Indic, apparently Armenian, and evidently Tocharian*, all pointing to *Dʰ (or something similar).

And nothing in Germanic and Italic speaks against assuming *Dʰ as a stage before *Ð, so we're back at the traditional model. We could chalk up *Dʰ > *Ð
as a further Italo-Germanic isogloss. But that would disturb my favourite reconstruction of the chain of events in Germanic, which would be PIE T D Dʰ > Tʰ D Dʰ (the tenues becoming aspirated by default like in many contemporary Germanic languages) > Þ D Ð (all aspirates become fricatives) > Þ T Ð (the only stop in the system becomes devoiced, as voicing is not distinctive on stops any more).

Seems to me that T D Dʰ > Tʰ T Dʰ > Þ T Ð is more likely. In modern Germanic languages, it seems like the aspiration of syllable-initial fortis stops generally goes hand in hand with near-zero voicing onset time for the lenis stops, which are not actually "voiced" in all languages or in all environments. I guess this could be seen as a kind of chain shift, so I'm not sure if it's possible to say which came first.

Aside from this, though, are there any attested examples of breathy-voiced stops being more prone to fricatization than simple-voiced stops?

[vokzhen]

The very strong caveat that I don't know much about PIE linguistics is in effect, I'm mostly just throwing out ideas to see if they help those who actually know what they're talking about.

As Tropylium said, Kortlandt points out that roots with mobile stress tend to have voiced aspirates, whereas static stress tend to have voiceless. Unfortunately they give no examples or figures and I'm sure there are plenty of exceptions, but it is tantalising to connect the stop system with the accent system. This is possibly backed up by a handful of correspondences such as *-bhi (a case/adverbial suffix) with *pi (a preposition? this may not actually exist, I'm drawing from memory) but has a bit of trouble explaining the large number of voiceless consonants in unstressed suffixes.

Original /t 'd/ system, older layer of words with mobile stress and productive voicing of t>d (traditional t>dh), newer layer of words with fixed stress and no voicing rules? The voiceless consonants in unstressed syllables would then be a younger generation of grammaticalization. Don't know if that can be matched up with other theories about the relative chronology of things, though.

Along with connecting the IE stops with the single and geminate stops in Uralic, I have also seen an attempt to connect them with the Uralic nasals, of which there are more than in IE and could conceivably come from voiced stops.

Which, again, is compatible with the theory that the original PIE system was /T 'D D/ rather than /T D Dh/, with the tradiational voiced/possible glottalized set > nasal in Uralic, maintained in PIE as some kind of glottalized sound:
Late PIE: T 'D D
Germanic: D series fricitivizes, 'D series devoices to (')T. I don't see a lot of precedence for this, but from the other direction Korean and Javanese both have a voiceless, unaspirated series that gained elements of glottalization. Obviously not an exact match, but it seems especially compatible with hwhatting's chronology, adjusted to T 'D D > Tʰ 'D D > Þ (')D Ð > Þ (')T Ð.
Italic: D series fricitivizes, leaving room for the 'D series to lose its (pre)glottalization
Balto-Slavic: 'D denatures into ' + D, with ' becoming part of the vowel nucleus as a (glottalizing) rising tone (compare Chinese, Northern Athabascan where -ʔ > high tone, Tzeltal ɓ>ʔb,ʔC > creaky voice on vowel)
In the other branches, either the glottalization becomes pure phonation (creaky voice) and, in Greco-Armenian-Aryan, then pushes the voiced towards breathy, or preglottalization is lost directly (Khmer, some Mayan ɓ>ʔb>b), after the voiced series becomes breathy in Greco-Aryan. I wouldn't say one is more likely that the other; despite Sindhi having full-blown implosives, it seems just as likely that it was creaky voice that Sindhi reinforced back to full implosion, given that implosives aren't preserved anywhere else (and, with a lot of stretching, might be able to claim areal influence from pre-Indo-European languages, if the Harrapan language was indeed related to Austroasiatic). However, I at least think there's decent reason to believe it was still preglottalized/implosive at least when Germanic and Balto-Slavic split off, as glottalization precedes the consonant rather than following it as is common with stiff/creaky voice (where glottalization is usually realized on the following vowel).

The problem with this is I'm not sure this matches up well with Tocharian and Anatolian, like, at all.

Aside from this, though, are there any attested examples of breathy-voiced stops being more prone to fricatization than simple-voiced stops?

I'm not sure. Gujarati has fricative allophones of breathy stops intervocally, and less widespread of aspirated stops as well. It's only one language but then, not many languages have breathy stops in the first place so one language that fricativizes them is a lot more than one language that fricativizes voiced stops. The only other changes to breathy stops I know of involve vowels, either with tone (Chinese, Punjabi, Zulu) or vowel quality (Khmer).

[kanejam]

*Tocharian underwent Grassman's Law in its development, at the very least for *dʰ, prior to the changes of *d > PToch *tˢ and *dR > PToch *R. Given that Grassman's Law only operated in branches that had aspirates, it's probable that Tocharian also had aspirates.

Actually this is fairly controversial, IIRC only Ringe supports this and it's based on only two words for which there are several alternative possible explanations. So again the only real evidence for voiced aspirated is for Greco-Aryan, and then we can posit the single 'Cao Bang' shift. This is made even easier if we assume that the non-implosives actually had allophonic or free variation between tenuis and aspirated, as supposedly in some dialects of Georgian and Armenian (albeit with ejectives instead of implosives).

As vokzhen points out, Winter's law has a clear mechanism if we assume the implosive theory, and one could argue that the bizarre Celtic shift *gʷ > *b makes more sense if it's an attempt to rebalance the back-heavy implosive series. We can speculate on the ablative singular ending in *-d where voiced consonants usually aren't found in endings by assuming that word final *t in some pre-stage was unreleased and came to be associated with the implosive (I don't know if there's an attested change like that though). It also gives an option for the possible alternation *d ~ *h₁ as in *deḱm̥t ~ *h₁ḱm̥tom if we assume h₁ = ʔ (which I know is really controversial, I only bring it up as a possibility).

Most of this is just regurgitating stuff I've read, and the last two points are pretty weak, but at least there is some evidence for *T 'D D when there is hardly any for *T D Dh (outside of Greco-Aryan).