Kiy is the Yale romanization of the Old Japanese word for 'fortress'. It is not intended to represent [kiy]. The syllable represented by kiy has been reconstructed several different ways. I prefer to reconstruct it as *[kï] or *[kïy], and I will use my reconstructions hereafter.

OJ *kï is thought to be a contraction of one or more syllables ending in diphthongs: *kəi, *koi, *kui, (and possibly even *kïi?). Many of these diphthongs are believed to be combinations of root-final *-kV plus a suffix *-i which may have been preceded by a lost consonant: ?*-Ci. The bare roots appear in compounds whereas the suffix is generally in free forms: e.g.,

OJ *kə- ~ OJ *kï < (*kə-i) 'tree'

OJ *ku-, post-OJ *ko- (< *ko-) ~ post-OJ *ki (< OJ*kï [unattested] <*ko-i) 'yellow'

OJ *tuku- (< ?*tuko-) ~ OJ *tukï < (*tuko-i) 'moon'

OJ *kuku- ~ OJ *kukï < (*kuku-i) 'stalk'

There are, however, cases of nonalternating root-medial*ï: e.g.,

OJ*kïsi (< **kV[C]isi) 'shore'

OJ *kïr- (< **kV[C]ir-) 'to mist'

OJ *kï 'fortress' also does not alternate, but it is not a *CVC(V) root. There are two possibilities:

1. *kï 'fortress' once had a *kV- root counterpart in compounds, but only the complex, suffixed form has survived.

2. *kï 'fortress' is a bare root without an *-i suffix:

*kV(C)i (not *kV(C)-i) > *kï

I prefer the latter, though I will mention some very weak evidence for a suffix in part 2.

Next: From 己 selves to 城 cities. KIY TO THE FORTRESS (PART 0.5)

I came home late last night from a business trip to Niuyue, so I couldn't write this until tonight.

When I looked up 城 'wall, city wall, walled city, city' in Schuessler (2006: 185), I found that he reconstructed it as Old Chinese *geng ? with a question mark and traced it back into a bigger word family following Bodman's (1980: 160) proposal:


Old Chinese (reconstructions are Schuessler's)

Written Tibetan


*geng ? 'complete, achieve, build; put; load, pack'

'fill, fulfill'









*geng ? 'a vesselful', *gengh ?

'abundant, highest degree'











*geng ?  'wall', etc. (< 'filled-in earth, stamped earth')

gyang, gyeng 'pisé, stamped earth, wall'; r-gyang 'wall'.

Since I am wary of proposing a Proto-Sino-Tibetan word for 'wall' (PST speakers may not have had such things), I assume that the development of 'fill' > 'filled-in earth' > 'wall' occurred independently in Chinese and Tibetan.

I also assume that the Chinese 成-words and the Tibetan g/k-ng words are not related, because I prefer to reconstruct 成城盛 with initial *d: *deng(s). (I am not even sure if the Tibetan words for 'fill', etc. and 'wall' are related.)

Although Karlgren (1957: 216) did not regard 丁 OC *teng 'fourth Heavenly Stem' as phonetic in 成 (my *deng), Xu Shen did at the end of the first century AD. This means that 丁 and 成 must have sounded similar to Xu Shen, even if the two graphs originally had no relationship. Moreover, Gao You (early 3rd century AD) gave

成 OC *deng > Late Old Chinese *dieng 'complete', etc.

as a paranomastic (punning) gloss (Coblin 1983: 236, #239) for

定 OC *dengs > Late Old Chinese *dengh 'determine'

(The phonetic of 定 OC *dengs 'determine' is 丁 OC *teng 'fourth Heavenly Stem'.)

This gloss would make no sense if 成 had a velar initial *g-. I should, however, point out that

[d]etermining the degree of phonetic similarity between glossed and glossing words in the paranomastic definitions is a fundamental and particularly vexing problem ... there are many anomalous contacts among the initials. (Coblin 1983: 15-16)

One could therefore defend Schuessler's reconstruction of *g- for 成 by proposing that the phonetic similarity of 成 and 定 was limited to their rhymes (*-ieng and *-engh in Late Old Chinese). However, that would still not explain why Xu Shen thought that 丁 OC *teng 'fourth Heavenly Stem' was phonetic in 成 OC *deng (AMR), *geng (Schuessler) 'complete'.

Xu Shen wrote a paranomastic gloss with irreconcilable initials (Coblin 1983: 193, #1246):

名 OC *meng 'name' : 成 OC *deng 'complete'.

One could posit cluster initials like *dg-, *gd-, *dm-, *md-, or even *dgm-, etc. to explain these glosses. However, I know of no other evidence for such clusters in these words, and I agree with Coblin (1983: 17):

In my view it is very likely that many of these [paranomastic gloss] examples represent riming pairs and that this sort of sound similarity was an acceptable if less common criterion for the construction of paranomastic glosses. This is an alarming conclusion, for, as foreseen by Serruys, it makes the use of paranomastic glosses a risky and subjective business. How are we to decide when a pairing involves both initials and finals? Ultimately we must conclude that paranomastic glosses are most useful when a significant number of them can be brought to bear on a particular problem or when they can be used in conjunction with other types of evidence. A single unsupported paranomastic equation is a very slim reed upon which to base the reconstruction of a syllable initial.

Hence I would rather interpret Xu Shen's gloss as evidence for a common rhyme *-eng in 名 OC *meng 'name' and 成 OC *deng 'complete'.

I have found potential word family evidence for a dental initial in 成 OC *deng. 成 is phonetic in

誠 OC *deng ''sincere, truly, really'

This *deng may have been a prenasalized *N-teng with a root *t-ng also found in

正 OC *teng-s 'straight, correct'

整 OC *teng-' 'orderly; arrange'

and possibly

亭 OC *deng (< ?*N-teng?) 'settle, regulate'

定 OC *deng-s (< ?*N-teng-s) 'settle, establish, determine'

眞 OC *tin (< ?*ting) 'true, real'

貞 OC *r-teng 'test, try out, correct, verify'

偵 OC *r-t(h)eng 'test, verify'

Did *t-ng have a stop-final variant in

直 OC *rdək (< ?*r-N-tək) 'straight, right'? KIY TO THE FORTRESS (PART 0)

Before I actually start talking about the title topic in part 1, I'd like to provide the keys to the mystery of my four-day absence from blogging.

On Friday and Monday, I fell asleep after coming home from work.

On Saturday and Sunday, I wrote an entry that I discarded. Read it and see if you can guess why:

<start of my weak-end entry>

When I started looking at tangraphs with [フ+キ] on the left, I came across what appeared to be an intriguing analysis from Tangraphic Sea:


TT4798 WASH zwər 1.84 =

(left of) TT4793 WALLED-TOWN kã 1.24

+ (right of) TT3690 RECEIVE lhyiy (rhyme unknown)

+ left of WASH lwu 2.1

Notice that none of the components from the three source graphs precisely match the components in TT4798 WASH:

[フ+キ] is equated with [コ+ㄇ+ニ]. Since both those elements can occur on the left and right, they cannot be explained as positional variants of each other (unlike the WATER elements below).

One might guess that there were two synonymous or homophonous syllables that could have been written as

[コ+ㄇ +ニ] + 几 + [フ+フ+フ]

and that one was written with [コ+ㄇ+ニ] and the other was written with [フ+キ] to differentiate them in writing. However, there is no graph

[コ+ㄇ+ニ] + 几 + [フ+フ+フ]

contrasting with TT7498:

[フ+キ] + 几 + [フ+フ+フ]

so the substitution of [フ+キ] for [コ+ㄇ+ニ] seems to be pointless.

几 is equated with [几+ノ]. Perhaps [几+ノ] has been abbreviated to 几 to accomodate the next element. As far as I know, [几+ノ] does not occur with anything adjacent to its ノ.

[フ +フ+フ] WATER is equated with [ㄈ+フ] WATER. This is expected since [フ+フ+フ] WATER is the right-hand variant of [ㄈ+フ] WATER, which only occurs on the left.

The most interesting inexact equation is the first one, which is not unique to that graph. It also occurs in these analyses (and possibly others):


TT0369 EXAMINE gyi 1.16 =

left of TT0355 (phonetic transcription character) kyĩ 1.16 +

left of TT4793 WALLED-TOWN kã 1.24


TT4427 己 SELF dzywï 1.69 =

( Is the resemblance to Middle Chinese 自 *dzih 'self' coincidental? *-w- corresponds to nothing in Chinese.)

left of TT4412 自 SELF 'yiy 1.36 (Tib. tr. [g-ye] [Nevsky 1960 I: 323]) +

left of TT4793 WALLED-TOWN kã 1.24


TT4793 WALLED-TOWN kã 1.24 =

right of TT4427 己 SELF dzywï 1.69 +

right of TT2199 SMALL-TOWN phe 1.8 (which doesn't look like [ー+ヒ]!)

Thus to understand [フ+キ], I should also understand [コ+ㄇ+ニ].

<end of my weak-end entry>

There is one huge problem with all of that ... namely, the fact that TT4793 is actually

Mojikyo #405 WALLED-TOWN dzywï 1.69

with the expected [フ+キ] on the left and [ニ+几+ニ] on the right


TT4503 / Mojikyo #404 SWEET kã 1.24

which is a loan from Chinese 甘 'sweet' (probably pronounced *kã in the dialect known to the Tangut).

The Tangraphic Sea analysis of TT4503 is


right of TT4572 DRY rorw 1.91 +

right of TT5716 EARTH tser 1.77

The right side of DRY ([コ+ㄇ+ニ]) is apparently being used as a Chinese-based phonetic, since its Chinese translation equivalent would have been 乾 *kã 'dry' - a homophone of 甘 *kã 'sweet'.

The analysis of DRY derives its right side [コ+ㄇ+ニ] from the left side of TT4503 SWEET kã 1.24.

One might hope that all graphs with [コ+ㄇ+ニ] were pronounced kã 1.24, but TT4503 is an isolated instance. All other tangraphs with [コ+ㄇ+ニ] on the left have very different readings:

TT4504 khwə 1.27

TT4505 shywiy 2.32

TT4506 BE-SILENT dwerw 2.78

TT4507 dywu 1.3 (cf. TT5396 below)

TT4508 REALIZE dwerw 2.78

TT4509 zyi 1.11

TT4510 u 2.1

TT4511 lyọ 2.64

I don't have time to look up the 34 tangraphs other than TT4572 DRY rorw 1.91 with [コ+ㄇ+ニ] on the right (Grinstead 1972: 89-90) but I wouldn't expect any phonetic consistency among their Tangut (A) readings. Here is a sample of every fifth [コ+ㄇ+ニ]-right tangraph with a gloss:

TT3529 TOGETHER gu 2.1

(TT5171 MARKETPLACE lhyor 1.94 doesn't count since it has [亠+ㄇ+ニ]; Grinstead seems to have miswritten it with [コ+ㄇ+ニ])

TT5396 KNOW dywu 2.3 (cf. TT4507 above)

TT5593 BANANA mya 1.20

How did I confuse

TT4793 / Mojikyo #405 WALLED-TOWN dzywï 1.69 and

TT4503 / Mojikyo #404 SWEET kã 1.24?

I was relying on the Internet Tangut Studies Association's online database which has some three-digit Mojikyo numbers mapped onto the wrong TT numbers.

It turns out that even Grinstead's gloss for TT4793 may be wrong, as Nevsky (1960 I: 587) translated it as

垣 'wall', 牆 'wall', 壁 'wall' (but not 城 'walled city')

ограда 'fence', стена 'wall'

It is true that in Chinese, 城 can mean 'wall', 'city wall', and even 'city'. (The 'Great Wall' is the 萬里長城 'Ten Thousand Mile Long Wall' and the 'Forbidden City' is the 紫禁城 'Purple Prohibited City'.) But I can't assume that Tangut TT4793 WALL could also refer to city walls or cities.

Next: A Sino-Tibetan word for 'wall'?

Tangut fonts by Mojikyo.org
All other content copyright © 2002-2007 Amritavision