10.12.18.21:09: S-B-ECULATIONS ON PROTO-INDO-EUROPEAN STOPS
Compare the following two tables:
Selected Greek reflexes of Proto-Indo-European (PIE) stops
Selected stop correspondences between Indic and Indo-Thai borrowings
Sanskrit and Pali
|Thai||b or p||d or t||ph||th|
Can Thai somehow shed light on the development of stops in Indo-European?
There are several key differences between these tables.First, Greek inherited PIE stops whereas Thai borrowed Indic stops.
Second, PIE is a reconstruction whereas Sanskrit and Pali are attested.
Third, Sanskrit and Pali had simple p and t, not ejective *p' and *t'.
Fourth, Thai has two correspondences for *p and *t. The voiceless stops (p, t) are from a later layer of borrowing of Indic words. Disregarding that later layer results in a table that looks even more like the PIE-Greek table:
Sanskrit and Pali
Since it would make no sense to borrow Indic p, t, b, d as b, d, ph, th, the earlier Thai stops must have been closer to Indic p, t, b, d. The reconstruction of earlier Thai *b and *d is certain, but the sources of modern Thai b and d are more problematic. J. Marvin Brown (1962) reconstructed those sources as *ʔm and *ʔn, which bear only the vaguest resemblance to Indic p and t. Li Fang-Kuei's (1977) *ʔb and *ʔd are closer, but are voiced (and could be reinterpreted as implosive *ɓ and *ɗ). Since I doubt anyone would hear p and t and reproduce them as *ʔb and *ʔd or *ɓ and *ɗ, I have wondered if those stops were actually *ʔp and *ʔt. However, I know of no language with ʔp and ʔt (or Korean-style tense pp and tt) but no ʔk or kk. (Indic k corresponds to Thai k. Modern Thai has no g.)
Indic borrowings in modern Khmer have a pattern of correspondences not unlike those of Thai:
Sanskrit and Pali
|Khmer||ɓ or p||ɗ or t||p||t|
But note that Indic voiced unaspirated stops do not correspond to voiceless aspirates as in Thai.Ferlus (1992: 83) did not reconstruct implosives (or aspirates) in Proto-Khmer:
A une époque difficile à préciser les occlusives p et t se sont glottalisées en ɓ et ɗ devant voyelle tout en restant inchangées hors de ce contexte, comme premier élément d'un groupe.
This would explain why the Indic letters for p and t were used to write the Old Khmer sources of the stops now pronounced as ɓ and ɗ - those letters stood for Old Khmer *p and *t at the time of borrowing. (In modern Khmer, the Indic letters for p and retroflex [!] ʈ are used to write modern ɓ and ɗ.) Ferlus' Proto-Khmer only had two stops series: e.g. (not a complete list),
voiceless: *p *t
voiced: *b *d
One might be tempted to derive Thai b and d from Proto-Tai *p and *t, but Proto-Tai had three stop series (like PIE!): e.g. (not a complete list),
voiceless: *p *tglottalized: *ʔp *ʔt (= Li Fang-Kuei's *ʔb *ʔd; could also be written as *ɓ *ɗ)
voiced: *b *d
Note, however, that Proto-Tai *ʔp is common (as expected for an implosive), whereas PIE *p' is rare (as expected for an ejective).
I wonder if the Proto-Tai glottalized series originated from earlier clusters, just as the tense consonants of Korean (pp, tt, etc.) originated from earlier clusters. Pre-Proto-Tai might have had only two series like Proto-Khmer.
Aspirates were traditionally reconstructed in both Proto-Tai (see slide 8 of Pittayaporn 2009) and PIE (*p *b *bh or even *p *ph *b *bh) but they may be unnecessary in both languages. Pittayaporn derived Tai aspirates from *Cr-clusters, uvulars, and loanwords.
In the Khmer-based Thai alphabet, the letters for b and d (< *ʔp *ʔt) are basic and the letters for p and t are derived:
|Original sound value of Khmer source letter||*p||(nonexistent in Khmer)||*t||(nonexistent in Khmer)|
|Sound value of Khmer source letter when the Thai alphabet was created||*ɓ||*ɗ|
|Earlier Thai||*ɓ (*ʔb? *ʔp?)||*p||*ɗ (*ʔd? *ʔt?)||*t|
(The sound values in the table above are only for initials directly before vowels.)
The letter shapes indicate that the Thai alphabet was created after Khmer had shifted *p and *t to implosive *ɓ and *ɗ, forcing the Thai to create new letters for *p and *t.
Summing up the possible parallels between PIE, Khmer, and Thai:
|Indo-European||SE Asian parallel|
|PIE three-way distinction for stops: voiceless, glottalized, voiced||Same three-way distinction in Proto-Tai, though the glottalized series might be secondary (whereas the glottalized series of PIE seems to be primary)|
|In most IE branches, the voiceless glottalized series
became voiced: *p' > b
Exceptions: Germanic, Armenian - did their substratum languages also have glottalized consonants, or did they have a Chinese-like *ph *p *b three-way system that was mapped onto the slightly different PIE three-way system: *p(h) *p' *b?
|Proto-Khmer voiceless > Khmer voiced implosive: *p
Tai voiced implosive > Thai voiced nonimplosive: *ɓ > b
Did most IE branches have an implosive transitional phase? *p' > *ɓ > b?
Did IE substratum languages have implosives?
|In Greek and Indo-Iranian, the voiced series became
PIE *b > Greek ph
PIE *b > PII *bh
|In Thai, the voiced series became aspirated: *b > ph|
17th century Middle Vietnamese had a series of aspirates which have largely lenited in modern mainstream Vietnamese dialects:
|Modern Vietnamese||[f]||[th] (still a stop!)||[x]||[h]|
For convenience, I will regard h as a member of the aspirate series even though it is a fricative. I will call this group of consonants 'aspirates' even though only one member is still an aspirate in modern mainstream Vietnamese.
Vietnamese has six tones. Three (huyền, nặng, ngã) are associated with proto-voiced initials. Thompson (1976) would reconstruct voiced aspirates for native words such as
hàm 'jaw' < *ɦ-
hành 'onion' < *ɦ-
hèn 'vile' < *ɦ-
hùm 'tiger' < *ɦ-
phồng 'swell' < *bh-
phù 'swell' < *bh-
thịt 'meat' < *dh-
Thompson does not reconstruct *gh-, though he reconstructs a *gɦ- that becomes Vietnamese s- [ʂ]. I would rather reconstruct *gr- instead of *gɦ-. *gh- could then be reconstructed for native words like
khà 'drink with gusto'
khà khà 'laugh loud'
khập khà khập khiễng 'limp'
khè khè 'snore, snoring'
khề khà 'drawling and hoarse'
khò khè 'wheeze'
khò khò 'snoring'
khù khờ 'silly'
khù khụ 'cough loudly'
khừ khừ 'groan'
lủ khủ lù khù 'obtuse'
However, I'd rather not reconstruct voiced aspirates at all.
Most of the kh-words listed above are sound-symbolic: e.g., khạc [xaak] 'spit' even sounds like English hack (in the sense of 'cough'). The exception is khè 'yellow'. Onomatopoeia may contain anomalous sound combinations absent elsewhere in a language. I am hesitant to reconstruct a phoneme to account for a small number of exceptions.
On the other hand, the h-, ph-, th-words cannot be set aside as sound-symbolic. But such words are uncommon, so perhaps their initials go back to some voiced preinitial/presyllable plus voiceless aspirate series: e.g.,
*N(V)-ph- > *mbh- > *bh- > ph-
*N(V)-kh- > *ŋgh- > *gh- > h- (instead of kh-)
thịt 'meat' cannot be from *dh- because its cognates elsewhere in Vietic have s-: e.g., siit in Ruc, Sach, Liha, Tho, and Thanh Hoa Muong (but thiit in Bi and Hoa Binh Muong). Moreover, the vietographs for thịt contain the phonetic 舌 Sino-Vietnamese thiệt < Middle Chinese *ʑiet 'tongue':contain the phonetic 舌 Sino-Vietnamese
舌 (phonetic in isolation)
月 (meat) + 舌
肉 (meat) + 舌
舌 + 肉 (meat)
Since Sino-Vietnamese th- can be from Middle Chinese *ɕ- or *ʑ-, I could reconstruct 'meat' as *ʑit, but I would rather reconstruct *C(V)-ɕit with a lost voiced preinitial/presyllable conditioning the voicing of *ɕ. Ferlus (ms. cited at SEAlang) reconstructed 'meat' as Proto-Vietic *-siit.
My last two posts have been attempts to find parallels for the development of Proto-Indo-European ejectives. Did some learners of PIE hear voiceless *p' and imitate it as voiced b?
Are there examples of ejectives in North American languages corresponding to voiced obstruents in Anglicized names?
Georgian ejectives (k' p' t' q' ts' ch') are written with voiceless Cyrllic letters:
Are there any nonstandard Cyrillicizations with voiced letters for ejectives?
к п т (regular voiceless letters; aspirated kh ph th are қ ҧ ҭ; Georgian has no regular k p t)
ҟ ҵ ҷ (modifications of the voiceless letters к ц ч)
UPSID lists a single language (!Xu) with both voiceless and voiced ejectives. Did PIE have voiced ejectives? Ladefoged and Maddieson (1996: 79-80) "do not know of any linguistic use of voiced ejectives" and regard their articulation as "most unlikely". They interpret the 'voiced ejectives' of !Xu as "prevoiced" clusters of nonejectives followed by ejectives: e.g., d' is really d + t'. Since there is no evidence for reconstructing two series of ejectives in PIE, it would be bizarre if PIE had such clusters without corresponding simple ejectives: e.g., dt' but not t'.
Googling "ejectives borrowing voiced", I found this strong statement by Fallon (2002: 257; emphasis mine):
I would feel more comfortable about the GT if a non-IE example of that exact type could be found.
[...] there is no language in which ejectives unconditionally, and across, the board, all turned to voiced stops, unlike deglottalization [...]
Of course, if the Ejective Model is correct, then many branches of Indo-European would also show voicing [...] But as Job (1989) cautions, this is simply based on reconstruction, rather than well documented evidence. And if we are testing the plausibility of the Glottalic Theory, it would be circular reasoning to include the Indo-European languages. Nevertheless, we have seen numerous cases of diachronic ejective voicing*, though not of the exact type required by the Glottalic Theory.
*The first one that comes to mind is Proto-Afroasiatic *k' > q > g in Libyan Arabic: hence the Q ~ G variation in spellings of Qaddafi's name.
In Korean, foreign voiced obstruents have been borrowed as tense obstruents (Lee and Ramsey 2000: 125-126):
Eng bus > modern Korean 뻐스 ppŏsŭ (also 버스 pŏsŭ)
Eng back > modern Korean 빽 ppaek 'connections'Eng dam > modern Korean 땜 ttaem (also 댐 taem)
Eng jam > modern Korean 쨈 tchaem (also 잼 chaem)
Eng gum > modern Korean 껌 kkŏm
(But note that Korean tense obstruents can also correspond to foreign voiceless obstruents: e.g., North Korean kkonggo 'Congo', kkahira 'Cairo' < Arabic قاهرة Qaahira [Sohn 1997: 204]).
Chinese voiced obstruents were also borrowed as tense obstruents in the idealized Sino-Korean pronunciation of the 15th century*: e.g.,
佛陀 ppurʔ-tta < Middle Chinese *bur-da < *but-da 'Buddha'
cf. actual modern Sino-Korean 불타 pultha 'id.' without tense obstruents
Could the opposite process have occurred in Indo-European?
David Boxenhorn suggested that the development of Indo-European stops could have been influenced by substratum languages. If
- Proto-Indo-European had tense obstruents (instead of ejectives)
- learners of PIE spoke languages without tense obstruents
those learners could have substituted voiced obstruents for tense obstruents: e.g.,
|Obstruent series||PIE (3 series)||Substratum language (2 series)||Substratum-influenced IE dialect (2 series)|
Lee and Ramsey (2000: 126) wrote,
To the ear of an English speaker, these [Korean] reinforced [i.e., tense] consonants sound closer to the [English voiced] sounds being imitated, because they have no aspiration, and the voice onset time is the closest of all the [Korean] consonant types to zero.
To the ear of this English speaker, Korean tense consonants sound more like Japanese voiceless geminates than English voiced consonants.
IE branches which distinguish between the three obstruent series may reflect substratum languages with three (or more?) series and developments unique to specific PIE dialects: e.g.,
3 series from 3 series: Italic?
|Obstruent series||PIE pre-Italic dialect (3 series)||Italic substratum language (3 series)||Proto-Italic (3 series)||Latin (3 series)|
|III||*b > *ph (as in Chinese and Thai)||*ɸ||*ɸ||f-, -b(-)|
The inventory of the PIE pre-Italic dialect (*p, *pp, *ph) is similar to the phonemic inventory of modern Korean (p, pp, ph)**.
Korean speakers approximate foreign f with native ph. Conversely, Italic substratum language speakers could have approximated PIEPID *ph as native *ɸ. (But is there any known case of aspirates borrowed as fricatives?***)
4 series from 3 series: Indo-Iranian?
|Obstruent series||PIE pre-Indo-Iranian dialect (3 series)||Indo-Iranian substratum language (4 series)||Proto-Indo-Iranian (4 series)|
|(no PIE aspirate series)||*ph||*ph|
|III||*b > *β||*bh||*bh|
The inventory of the PIE pre-Indo-Iranian dialect (*p, *pp, *β) is similar to the inventory of Middle Korean (p, ph, pp, β). Middle Korean β is from *b, just as my proposed PIEPII *β is from *b.
Indo-Iranian substratum language speakers could have approximated PIEPII *β as native *bh.
PII voiceless aspirates developed from *sC- and *CH-clusters under the influence of substratum languages which already had voiceless aspirates.
Here are a few problems with the above scenario:
1. Tense consonants are very rare. pp occurs in only five languages in UPSID, whereas the rare ejective p' occurs in 44 languages.
2. Tense pp is not uncommon relative to other tense consonants, whereas ejective p' is the rarest ejective. Perhaps PIE *p' was 'nativized' as tense *pp which later shifted to *b.
3. Modern Korean tense consonants originated from clusters which in turn generally originated from unstressed syllables. There is no reason to believe that PIE series II obstruents had such complex origins.4. Although ph and bh as well as p and b occur in Indo-Aryan words from non-IE substrate languages, that does not necessarily mean that the substrate languages had four series of obstruents. b ~ bh variation in such words imply that one or more substrate languages had a sound unlike either b or bh - perhaps *β.
14:45: Worst of all, only one series can be reconstructed in Proto-Dravidian and voiced aspirates in Munda are probably due to Indo-Aryan influence (Neukom 1999: 131). It's possible that there were non-Dravidian, non-Austroasiatic substrate languages with four or more series, but I'd rather not posit them to make my hypothesis work. (Kharia is a Munda language with five series [p ph b bh ʔb] according to Biligiri 1965.)
*There are only three modern Sino-Korean readings with tense obstruents:
雙 쌍 ssang < Middle Chinese *ʂɔŋ
氏 씨 sshi < Middle Chinese *ʑi < *dʑie
喫 끽 kkik (but kik in North Korea) < Middle Chinese *khek
雙 and 喫 never had voiced initials in Chinese. Their 15th century idealized Sino-Korean readings are swang and khyərʔ. The origins of their anomalous modern tense initials are unknown. The reading ssang dates from 1677 (Martin 1992: 112). I don't know when kkik was first atttested.
The ssh- of 氏 is not a retention of the ss- of its idealized reading ssi. Its earlier real reading was si and its "fairly new" tense consonant incorporates a genitive -s-: -s-si 'clan of' (Martin 1992: 112).
**Modern Korean also has [b] as an allophone of /p/.
***Sino-Vietnamese ph [f] corresponds to Middle Chinese *ph as well as *f, but the [f] pronunciation of ph is recent and hence cannot be used as an example of borrowing aspirates as fricatives. Vietnamese ph was still [ph] in some dialects as late as the 20th century (Gregerson 1969: 149).