Given that

The combination of s and i is rare in WT [Written Tibetan] (excluding loanwords and onomatopoeia which can deviate from a language's 'normal' sound pattern). The combination of sh and i, on the other hand, is common in WT.

it appears that pre-WT si became WT shi, with a few exceptions*.

The shift of alveolar s to a palatal s before the palatal vowel i has occurred independently in standard Mandarin, Korean, and Japanese. It has also occurred among WT's Tibeto-Burman relatives: e.g., Miri in India (Matisoff 2003: 33):

Proto-Tibeto-Burman m-sin 'liver' > Miri shin

cf. WT mchhin-pa < pre-WT mshin < msin

Proto-Tibeto-Burman sing 'tree / wood' > Miri e-shing

cf. WT shing < pre-WT sing

PTB msin 'liver' and sing 'tree / wood' look like Old Chinese 辛 sin 'bitter' and 薪 sin(g) 'firewood' with s-. If these words were inherited from Proto-Sino-Tibetan, they might have been PST sin (there is no evidence for an m- in OC) and sing.

If one wanted an s-like OC in spite of all that, one could claim that

PST s- > OC sh- > later Chn s-

but it is simpler to skip the shibilant step:

PST s- > OC s-

Moreover, in that version of the shibilant hypothesis, PSTs- would shift to OC sh- before all vowels. Such a change is rare. I only know of two languages in which s- became sh- before almost all vowels: Bengali and Lahu, another Tibeto-Burman language (and even Lahu has s- before the non-palatal vowel ï; Matisoff 2003: 27). It's more likely that s- > sh- > s- never happened. Thus the equation of

OC (shibilant version) 死 shi' : WT shi 'die'

should be abandoned in favor of

OC 死 si' : WT shi < pre-WT si < PTB səy (Matisoff 2003: 27) 'die'

Using Pulleyblank's two-vowel OC reconstruction (2VOC) results in an even neater equation:

OC 死 səy' : PTB səy 'die'

However, one should not embrace 2VOC** on the basis of one pair of possible cognates.

Next: 2V (or not) TB? ADDENDUM: The OC and PTB (and PST?) words for 'die' resemble Jpn shinu 'die'. Could shinu be a loanword from Chinese? No.

First, the stem of shinu is shin-, not shi-. The -n corresponds to nothing in Chinese.

Second, shin- and in- 'depart' are the only members of the na-hen conjugational class. It's unlikely that a foreign verb would end up in such an exceptional category. There is an verb shi-su 'die' indisputably borrowed from Chinese which belongs to the large category of foreign noun + -su 'do' verbs.

Third, there is an indigenous etymology for shin- 'die': shi 'breath, wind' + in- 'depart'. (shi only appears in compounds: e.g., arashi 'storm' < ara- 'wild' + shi 'wind'.)

* Looking through Goldstein's The New Tibetan-English Dictionary of Modern Tibetan (a somewhat redundant title) which is obviously not a classical Tibetan dictionary but the only Tibetan dictionary I have on hand, si- appears in the following types of words:

Words dating after the si > shi change: e.g., foreign borrowings: si-we-den 'Sweden', si-ri-ya 'Syria', si-an-an 'CNN'; si-ling 'military commander' (< Md 司令 siling 'commander'); sig-ra 'bamboo basket', sil-ra 'basket' (both pronounced [siirə]; borrowed from a foreign [siirə]?)

Words of probable onomatopoetic origin: e.g., si 'whistle', si-li-li 'tinkling sound of bells', sig-sig 'shaking from fear or cold', sing-sing 'ringing/tinkling sounds of bells and metlas striking', sib-sib / sim-sim 'lightly/softly falling rain/snow', sir-sir 'a buzzing sound', sil 'cymbal', sil-sil 'sound of bells rining/tinkling'.

The unexplained remainder: e.g.,

sing (past: sings) 'to choose'

sim 'to seep through'

sim-po 'quiet; happy'

sil- 'fruit' (in sil-tog 'fruits', sil-sdong 'fruit trees' [< + sdong 'tree'])

sil- 'fragment' (in sil-bu, sil-ma 'in fragments', sil-shing 'firewood cut into pieces' [< + shing 'wood')

sil- '?' (in sil-ño 'buying retail' [< + ño 'buy'] sil-Htshong 'retail sales' [< + Htshong 'sell'])

**Pronounced like Tuvok. SINITIC SANS SIBILANTS? (PART 5)

An example of sleepless shallow thinking: For a moment a week ago, I thought right before I went to bed, my shibilant theory makes the Old Chinese (OC) and Written Tibetan (WT) words for 'die' look alike, so I must be on the right track! Yes, it did make them look alike, and no, I wasn't on the right track.

In my OC reconstruction, 'die' is 死 si'. (Initial consonant clusters are also possible, but I have no evidence for them. Hence I'll stick with Occam's Razor and go with a simple s-.) If the shibilant hypothesis were correct (it's not), I would reconstruct OC 'die' as shi'. This does resemble WT shi 'die'.

Matisoff (2003: 475) links OC 辛 'bitter'* (sin or shin, depending whether you are anti- or pro-shibilant) to the Proto-Tibeto-Burman word for 'liver' (not the best semantics, I know - and I won't give away his PTB reconstruction yet). The WT word for 'liver' is mchhin-pa (-pa is a suffix). Since msh- is impossible in WT (see Guillaume Jacques' chart), perhaps mchh- < msh-**, and pre-WT ?mshin is not far from shibilant OC shin.

If OC 薪 sin(g) 'firewood' (containing an altered form of OC 辛 sin/shin 'bitter' as phonetic) were re-reconstructed as shing, it would be a dead ringer for WT shing 'tree, wood'.

So does all that tell us that Chinese shifted sh- to s- whereas WT retained Proto-Sino-Tibetan sh-?

No (and not just because I am agnostic about Sino-Tibetan). The combination of s and i is rare in WT (excluding loanwords and onomatopoeia which can deviate from a language's 'normal' sound pattern). The combination of sh and i, on the other hand, is common in WT.

Next: How does that skewed distribution contribute to the death of the shibilant hypothesis?

* The graph 辛 for OC sin/shin 'bitter' can also represent the eighth heavenly stem which Pulleyblank has regarded as a letter for OC sy- (1979) or s- (1991). The modern version of the graph appears to be a compound of 立 OC rəp 'stand' and 十 OC gip 'ten', but the early version of the graph is quite different. It looks like a triangle atop a capital T. It does not clearly picture anything bitter (unless it's a bitter plant?), so I assume it was used to write a homophonous word.

** Cf. English prince [prins] which can sound like prints [prints] due to an 'intrusive' stop:

[ns] > [nts]

One can view the [t] as [n] with its nasality and voicing turned off as a transition between [n] and [s]:


Similarly, one could break down WT mchh- into m-c-sh-. -c- is a palatal stop serving the same transitional role that [t] played in Eng [nts] above. The +/- chart for WT mchh- would be identical to the chart for Eng [nts].

The alternation of initials in the paradigm of WT 'die' could be similarly explained:


(The pronunciation of the WT letter Ha-chhung 'little a' that I transliterate as H is uncertain. I think it had some sort of nasal quality, so I use the symbol N for its pre-WT predecessor. Matisoff [2003: 116] thought it "represented a syllabic pre-glottalized nasal onset, something like [glottal stop + syllabic m]". A nasal interpretation of H might explain why sh- becomes ch- after both m- and H-.) SINITIC SANS SIBILANTS? (PART 4)

I don't have time to go into Tibeto-Burman tonight, so I'll only answer the last question of my last post:

And if the shibilant hypothesis is wrong, why did the Chinese borrow foreign palatals as alveolar affricates?

The non-shibilant version of my reconstruction of early Old Chinese has the following s-like sounds:

voiceless aspiratetsh

The shibilant version of early OC has a similar inventory at a different point of articulation:

voiceless aspiratechh

There were far more s-sounds in Middle Chinese (MC):

voiceless aspiratetshChhchh

(z- < sl-. The retroflexes mostly came from earlier r-clusters. The palatals are from earlier dentals, except for zh- < the labial + dental cluster ml-.)

So what do the following sound correspondences in borrowings tell us?

Foreign Sh-, sh-: LOC ? > MC Sh-

Foreign ch-, -ch-, -sh- : LOC ? > MC -ts-

Foreign -j- : LOC ? > MC -dz-

The first correspondence might indicate that LOC had retroflex Sh- (and other retroflexes?) in addition to alveolars or palatals:

Non-shibilant LOC:

voiceless aspiratetsh?Chh

Shibilant LOC (cf. Middle Vietnamese which had [Sh] and [sh] but no [s]):

voiceless aspirate?Chhchh

If we assume that foreign Sh- was borrowed as Chinese Sh-, then foreign palatals were either

- borrowed as alveolars:

Foreign ch-, -ch-, -sh- : LOC -ts- > MC -ts-

Foreign -j- : LOC -dz- > MC -dz-

- or as palatals:

Foreign ch-, -ch-, -sh- : LOC -ch- > MC -ts-

Foreign -j- : LOC -j- > MC -dz-

Which scenario is correct? According to Pulleyblank (1962: 109):

Indirect evidence of the absence of the later [Middle Chinese] palatals in the [late Old] Chinese of the Han period is provided by the use of the dental [stricly speaking, alveolar] affricates for foreign palatals. Unfortunately Chinese dental affricates where spellings in Indian alphabets have palatals in names from the Northern and Eastern Tarim basin are equivocal ... Pelliot regarded the Chinese spellings as evidence of the presence of dental affricates in the native names and this gains support from the continued use of dental affricates even after Chinese palatals were being used for Indian palatals [in MC] (Pelliot 1923, p. 126) ...

kucha: 丘慈 MC khutsï, 屈茨 MC khutdzi

borrowed into LOC as 龜 茲 kwïtsïə (kwïchïə?)

chalmadana: 左末 MC tsa'mat

borrowed into LOC as 且 末 tshïa'mat (chhïa'mat?)

[MC reconstructions are mine. -AMR]

A similar doubt applies to a word like 師子 'lion', which must be derived from Tokh[arian].

Perhaps the above should then be rewritten as:

Foreign ts-, -ts-, -sh- : LOC -ts- > MC -ts-

Foreign -dz- : LOC -dz- > MC -dz-

assuming Pelliot was right. But even if Pelliot were wrong, and the foreign names really did contain palatal affricates ch and j, one could argue that LOC speakers borrowed palatal affricates as alveolar affricates (ts and dz) because they had no exact equivalents (i.e., palatals). In either case, there is no need to reinterpret LOC alveolars as palatals.

One could still argue that the LOC alveolars were fronted early OC palatals, but languages with sh as their sole sibilant fricative are rare. Middle Vietnamese did lack s, but it had both Sh and sh. Bengali and Lahu have sh in most positions but even they have s in a few positions (before dentals [t th n r l] in Bengali and before the non-palatal vowel ï in Lahu; other Lahu palatals also become alveolars before ï). Perhaps early OC was like Bengali and Lahu with palatals in most positions with nonpalatal, alveolar variants in other positions. If so, what would be the consequences for comparative Sino-Tibetan studies?

Next: sh-ino-Tibetan shibilants? SINITIC SANS SIBILANTS? (PART 3)

So why did I even consider the shibilant hypothesis in the first place? Here's a demonstration of how a little knowledge can be dangerous:

Pulleyblank (1962: 226) regarded Md 獅子 shizi [ShRtsï] < Middle Chinese (MC) Shitsï' 'lion' (originally written 師子 without the 犭 'dog' semantic element) as a Tocharian borrowing into late Old Chinese (LOC). The Tocharian forms* were Shechake (with initial retroflex [or, more likely, postalveolar] Sh-) and shishäk [shishïk]. Although many Mandarin nouns end in the 'extender' suffix 子 -zi 'son' that forms disyllabic nouns out of monosyllabic roots, according to Pulleyblank,

There is no reason to regard here as the noun forming suffix of Modern Mandarin. In the earliest passages it is always treated as an inseparable part of the word and it is only much later that shih [獅] alone comes to be used for 'lion'.

'Lion' was Shitsïə' in my reconstruction of late LOC:

Tocharian Sh-, sh- (retroflex/postalveolar, palatal) > borrowed as LOC Sh- (retroflex)

Tocharian -ch-, -sh- (palatal) > borrowed as LOC -ts- (alveolar)

Tocharian -k(e) (velar) > borrowed as LOC -' (final glottal stop)

The word-medial consonant correspondence would be even more exact if the shibilant hypothesis were correct and if early OC shibilants were still intact in LOC:

Tocharian -ch-, -sh- (palatal) > borrowed as LOC -ch- (palatal)

If Shechake were the source of 'lion', two out of three LOC consonants would match:

Tocharian Sh-ch-k-

LOC Sh-ch-'

One could also cite other transcriptions of foreign words implying palatal shibilants instead of alveolar sibilants in LOC: e.g. (Pulleyblank 1962: 109):


LOC transcripition

LOC (non-shibilant version)

LOC (shibilant version)


龜 茲




且 末



kujula kadphises


khwïdzuhkïak (why kïak?**)



子 合



I am not alone in proposing a palatal(ized) origin of later Chinese ts. I first encountered the idea in Pulleyblank (1991: 52):

In the case of 子 [the first branch and one of the letters in Pulleyblank's OC alphabet], one has to suppose that a palatalized *ky- first became a palatal affricate *ch- and then fronted to a dental affricate ts-. Such a process has many parallels in other languages. As I have pointed elsewhere, Written Burmese has palatal c, ch, now pronounced as dental [s] and [s-h [aspirated s]], corresponding to Chinese [ts], [tsh] and Tibetan ts, tsh, dz, which supports the assumption that this process has taken place in all three languages.

[Pulleyblank seems to be assuming that WB retained Proto-Sino-Tibetan palatals long after Chinese and Tibetan shifted them to alveolars.]

In his 1979 article on an earlier version of his OC alphabet, Pulleyblank proposed that the OC origin of later Chinese ts was py (p. 36):

The supposition that *py would have been replaced by a palatal stop c, becoming an affricate ch and later a dental affricate ts can easily be supported by examples from other languages. In many modern Tibetan dialects written Tibetan phy-, by-, my- have been replaced by ch-, j-, ny-, except where -y- has dropped out, especially in front of i or e. A very similar palatalization of labials took place in Old French, e.g., sachant < sapiente, cage < cavea. *py is probably not the only source of Middle Chinese ts-.

Later, I saw that Starostin (1989) derived some (but not all) Chinese alveolar affricates from earlier palatal affricates: e.g. (from his online database),

即 (Karlgren 399a-c) 'thereupon': Starostin OC chit > Starostin MC tsyit > Md ji

節 (Karlgren 399e) 'joint'***: Starostin OC chiit > Starostin MC tsiet > Md jie

Note, however, that neither Pulleyblank nor Starostin went so far as to derive later Chinese s- from a palatal sh-. Written Tibetan (WT) shi 'die' corresponding to Md si 'die' < MC si' 'die' < OC si' (or shi' according to the shibilant hypothesis) seems to preserve an earlier sh- lost in later Chinese. Or does it?

Next: How can the correspondence of WT sh- : Chinese s- be explained without resorting to a Proto-Sino-Tibetan (PST) shibilant? What would be the implications for PST - and for Proto-Tibeto-Burman - if the shibilant hypothesis were correct? And if the shibilant hypothesis is wrong, why did the Chinese borrow foreign palatals as alveolar affricates?

*It's not clear which forms belonged to Tocharian A and B. Pulleyblank (1962: 109) identifies Shechake as TB and shishäk as TA, but later (1962: 226) he states that Shechake is TA and shishäk is TB.

** 却 LOC kïak appears to be a scribal error for 劫 LOC kïap 'rob', used to transcribe Skt kalpa 'a long period of time'. It is tempting to reconstruct an -l- in 劫 kïap (klïap?) so that it sounds more like Skt kalpa and the -la of kujula. However,

- there is no other evidence for an -l- in 劫 kïap (or in the rest of the 去 phonetic series [Karlgren 642])

- medial -l- may have been lost in LOC, so even if 劫 kïap once had an -l-, it was gone by the time this transcription was devised

- 劫 kïap could have represented the kadph- of kadphises

***節 LOC/MC tset (my reconstruction) was borrowed into early Vietnamese as (t)set, which became modern Tết (best known to Americans because of this offensive). How did 'joint' come to mean 'Vietnamese New Year'?

'joint' > 'juncture' (a metaphorical joint) > 'festival' > 'New Year' (the festival) SINITIC SANS SIBILANTS? (PART 2)

One good thing about 'publishing' hypotheses on a blog is that they're not as permanent as in a publication. I can delete entries, whereas I can't delete journal articles or (parts of) books. (And even if entries are archived somewhere without my permission, I might be able to request their removal.) Nonetheless, I'm still embarrassed when I post ideas late at night that are obviously wrong in the morning.

I've been wondering why Old Chinese (OC) alveolar sibilant initials (s-, ts-, tsh-, dz-) didn't palatalize in Middle Chinese (MC) like t-, th-, d-, n-, hn-:

OC t-, th-, d-, n-, hn- > MC ch-, chh-, j-, ñ-, sh- (all palatals)


OC s-, ts-, tsh-, dz- > MC s-, ts-, tsh-, dz- (no change; still alveolar)

I came up with a complicated (and unlikely) scenario in which OC had palatal shibilants instead of alveolar sibilants which 'moved out of the way' for palatalized dentals to take their place.

The simple answer is to assume that t- th- d- n- hn- were dentals rather than alveolars and that palatalization applied only to dentals. (But are alveolars really less susceptible to palatalization than dentals?)

The difference in treatment between the t- and s- groups might have a parallel in Russian. To my ear, Russian palatalized t and d sound like palatal ch and j (cf. OC t-, th-, d- > MC ch-, chh-, j-) because they can be pronounced with "slight frication" (also see Lyovin [1997: 64, a native speaker]) whereas Russian palatalized s- does not sound like a palatal sh-. In future Russian, palatalized t and d might become palatal ch and j, but palatalized s might still be s, just as OC s remained s in MC.

MC s and its brethren did not, however, remain intact in modern standard Mandarin. dz devoiced and the group palatalized before Old Mandarin i, ü, and y*:



Old Mandarin (OM)

Modern standard Mandarin (Md) spelling





x (phonetically [sh], not [ks] or [x])

須 'need': OC sno > MC suə > OM > Md xu





j (phonetically [ch], not [j])

津 'ford': OC tsin > MC tsin > OM tsin > Md jin [chin]; this is the -jin in 天津 Tianjin 'Heaven Ford', where Tangutologist Ksenia Kepping grew up




q (phonetically [chh], not [kw])

親 'close; relative, parent': OC tshin > MC tshin > OM tshin > Md qin [chhin]



tsh (aspirated in 'level' tone syllables)

q (see above)

秦 'Qin dynasty': OC dzin (or ntsin?) > MC dzin > OM tshin > Md qin [chhin] (level tone)

ts (unaspirated in non-'level' tone syllables)

j (see above)

捷 'quick': OC dzap (or ntsap?) > MC dziəp > OM tsye > Md jie [chye] (non-level tone)

Some of these palatalizations occurred before vowels that were not palatal until (late) MC: e.g.,

in 須 'need': OC o > late OC uo > MC > late MC ü

in 捷 'quick': OC a > late OC ïa > MC > late MC ye

Conversely, the archetypal palatal vowel i lost its palatality ('i/y-ness') before the s-class in late MC open syllables: e.g.,

死 'die': OC si' > MC si' > late MC > Md si [sï]


津 'ford': OC tsin > MC tsin > late MC tsin (not tsïn) > Md jin [chin] (not [chïn], which is impossible in Md)

親 'close; relative, parent': OC tshin > MC tshin > late MC tshin (not tshïn) > Md qin [chhin] (not [chhïn], which is impossible in Md)

秦 'Qin dynasty': OC dzin (or ntsin?) > MC dzin > late MC tsHin (not tsHïn; H = voiced h) > Md qin [chhin] (not [chhïn], which is impossible in Md)

Next: The sh-uperficial advantages of the shibilant hypothesis.

* The s-class did not, however, palatalize in front of other Old Mandarin vowels: e.g., [ï] (see 死 'die' above, which did not become Md [shï]; its initial s- has apparently remained intact since OC).