14.10.11.18:21: WHEN 1825 IS REALLY 1829

What's worse than having to publicly correct a mistake on a blog? Having to publicly correct that correction!

Andrew West pointed out that the correct fanqie for Tangut character 3371 (and its homophones 0596 and 1283) is

=+

3371, 0596, 1283 1dzia = 3031 + 1829 (not 1825!)

I got the idée fixe that

1825

was the final speller and didn't notice that 1829 with the same left-hand radical 'fire' and a similar right-hand radical in the fanqie of the handwritten copy of the Tangraphic Sea in Wenhai yanjiu (1983) or Arakawa's Seikago tsūin jiten (Tangut rhyme dictionary, 1997).

Notice that I have not supplied readings for 3031 or 1829.

I have already explained why 3031 is ambiguous, and I will add one more complication here:

- 3031 is the initial speller for 3371, 0596, and 1283 which are in the MIxed Categories of the Tangraphic Sea. For some reason, all dz-, dʐ-, and ɬ-syllables were placed in Mixed Categories along with a seemingly random smattering of other syllables. That suggests 3371, 0596, 1283 had dz-.

- On the other hand, 3031 is in the 'rising' tone volume of Precious Rhymes of the Tangraphic Sea instead of the Mixed Categories volume. That implies 3031 did not have dz-.

The fanqie for 1829 indicates -w- ... or does it? There is no transcription evidence for the -w- of 1829, its final speller 1289 1lwia or 0259 1lwia, the only homophone of 1289. -w- is an attempt to account for why 1289 1lwia is not in the same homophone group as

3456 1lia 'to come'

whose Chinese transcription 辢 *la has no *-w-. Then again, that transcription is not ironclad proof 3456 didn't have -w-, because the Chinese known to the Tangut had no syllable *lwa. Nonetheless a Tangut lwia could have been transcribed in Chinese as 辢合 *laCLOSED with a small 合 'closed (mouth)' diacritic to indicate -w-.) 1289 and 3456 had the same initial (l-) and rhyme (1-ia), so they presumably had different medials (-w- and zero).

If 1289 was 1lwia, then 1829 was 1tshwia, and 3371, 0596, and 1283 were 1dzwia ... which conflicts with the use of 0596 as a transcription character for Sanskrit ja without -v- (there is no -w- in Sanskrit).

Let's suppose that 3371, 0596, and 1283 were 1dzia without -w- and that their fanqie final speller 1829 was 1tshia without -w-. 1829 and 1825 were in different homophone groups even though they had the same initial (tsh-) and rhyme (1-ia), so they presumably had different medials (zero and -w-). But if 1825 was 1tshwia, why was it transcribed in Tibetan as tsha instead of tshwa? Was the subscript -wa character accidentally omitted?

This is so frustrating. I want to end on a more positive note. Andrew West recently created an online Homophones lookup tool. You can input the Li Fanwen 2008 numbers I use for tangraphs to see that

- 3371, 0596, 1283 1dz?(w)ia are in the same homophone group (31A46-48; all Homophones numbers here are from the A edition; different editions have different numbers)

- 1829 1tsh(w)ia (the final speller for those three syllables) and 1825 1tsh(w)ia (which I confused with 1829) are in different homophone groups (31B36 [which has no homophones] and 33A13-14 [a set of two homophones: 5041 and 1825])

- 1289 1lwia (the final speller for 1829) and 3456 1lia are in different homophone groups (53B78-54A11 [a set of two homophones: 1289 and 0259] and 55A54 [which has no homophones])

Alas, Homophones does not give any concrete information about the homophone groups beyond their initial classes: e.g., 3371, 0596, 1283, 1829, and 1825 belong to the sixth class (alveolars) and 1289 and 3456 belong to the ninth class (liquids). The Tangraphic Sea lists homophone groups organized by rhyme with fanqie, but fanqie for most 'rising' tone syllables are lost, and readings for fanqie spellers are dependent on a mixture of transcription evidence and educated guesswork (e.g., the reasoning for reconstructing -w- above).

14.10.10.20:00: WHEN RHYME 21 IS REALLY RHYME 20

(10.11.18:25: Formerly titled "Tangut Grade III -a('): Rhymes 19 and 21 (Part 2)", but I changed the title since this entry has nothing to do with either rhyme apart from my confusion of rhymes 20 and 21.)

If you don't want to constantly make a fool of yourself in public, don't blog about Tangut.

For the past couple of days, I've been reconstructing 3371 as 1dzɨa' with Grade III rhyme 21 which would be unusual after dz-, but its fanqie in the Mixed Categories of the Tangraphic Sea clearly indicates that it has Grade IV rhyme 20 which is normal after dz-:

=+

3371 1dzia = 3031 2dzi + 1825 1tshwia (sic; should be 1829!)

Even this corrected (?) reading remains problematic for several reasons.

First, the initial might be ts-. The evidence is ambiguous:

1. There is no fanqie for 3031, the initial speller of 3371.

2. 3031 was used to transcribe

Chinese characters with *ts-readings

Sanskrit ci (pronounced [tsi] in the variety of Sanskrit known to the Tangut, probably via Tibetan which had [ts] for Sanskrit c).

3. 3031 was transcribed in Tibetan as both Hdza and Htsa. The phonetic value of H- is uncertain: it could have represented prenasalization or a voiced back fricative.

4. Another character

1290 2?-ew 'ordinal suffix, class, limitation'

with 3031 as a fanqie initial speller was transcribed in Tibetan as tsa, tsi(H), gtsiH, and gdzi(H).

5. 3371 was homophonous with

0596 'to grow'

a transcription character for Sanskrit ja (pronounced [dza] in the variety of Sanskrit known to the Tangut, probably via Tibetan which had [dz] for Sanskrit j).

Second, it would be odd for a -wia graph (1825; sic - should be 1829!) to be a fanqie final speller for -ia without -w-. But it would also be odd for Sanskrit ja [dza] to be transcribed as dzwia instead of dzia.

The Tibetan transcription of 1825 is tsha, not tshwa. So maybe 1825 lacked -w- after all. And maybe it lacked -i- as well. A Sofronov-style reconstruction of 1825 as 1tsha may be best. But then how can one explain the different fanqie for the other 1tsha (or 1tshia) in the Tangraphic Sea?

=+

1829 'to heat up, burn' 1tshia = 0311 1tshwiə + 1289 1lwia

(10.14.20:00: This is actually the fanqie for 1825!)

Maybe 1829 had -w- and 1825 and its homophone

5041 'stove, furnace'

did not. Their fanqie has no -w- in either speller:

=+

1825 and 5041 1tshia = 3278 1tshi + 1693 1sia (used to transcribe Sanskrit sa)

(10.14.20:00: This is actually the fanqie for 1829!)

I will revise my reconstructions accordingly:

(10.14.20:04: No, judging from the corrected fanqie, Sofronov and Gong were right to reconstruct -w- in 1825 and 5041! Which means that the equation below is still 'broken' or 'unbalanced', depending on your preference in metaphors.)

Plugging that revised reconstruction of 1825 back into the fanqie at the beginning of this post results in a balanced equation:

=+

3371 1dzia = 3031 2dzi + 1825 1tshia

The two homophones of 1825 listed in Mixed Categories of the Tangraphic Sea share that fanqie and should also be read as 1dzia:

0596 'to grow' and 1283 'stomach' (attested only in dictionaries)

This entry demonstrates how errors and their corrections can cause chain reactions in Tangut reconstructions.

I have eliminated one type of apparent anomaly in rhyme 21: the combination of an alveolar initial dz- with the Grade III medial -ɨ-. But other anomalies remain, and I will examine them in future entries.

14.10.9.22:55: TANGUT GRADE III -A('): RHYMES 19 AND 21 (PART 1)

Last night I mentioned the words (phrases?)

3371 0378 1dzɨa' 2ʔʊ 'curled hair' and 3371 1144 1dzɨa' 2dị 'bun (of hair)'

and noted that their first syllables had an anomalous initial-rhyme combination. (No, actually they don't!)

3371 has the Grade III rhyme 21 (= 1.21/level tone rhyme 21 and 2.18/rising tone rhyme 18). (10.10.20:01: The true rhyme of 3371 is 20.) Here are the latest reconstructions of that rhyme and its immediate neighbors in the first rhyme cycle:

In Gong's reconstruction, there is no Grade III/IV distinction, and many rhymes are redundant: e.g., rhymes 21 and 24. Hence Gong regarded

3371 1dzjaa (rhyme 21; = my 1dzɨa') 'hair worn in a bun; peak' and 4075 1dzjaa (rhyme 24; my 1dzia') 'thrifty'

(10.10.20:02: 3371 should be 1dzia with rhyme 20.)

as homophones in spite of their placement into different rhymes and homophone groups in the Tangraphic Sea. They are not homophonous in the other three reconstructions.

In Arakawa's reconstruction, rhyme 21 is the only Grade IV rhyme, and it has a combination of the -y- of his Grade II and the vowel length of his Grade III.

Sofronov's reconstruction is very different from all others: e.g., it has Grade II and Grade IV variants of rhyme 21. Sofronov reconstructs five subtypes of a-rhymes corresponding to three subtypes in the other reconstructions.

In my reconstruction, Grade III rhymes are characterized by medial -ɨ- and are distinct from Grade IV rhymes with -i-. Grade III and IV rhymes typically have different initials:

III: v- (= w- in most other reconstructions), shibilants (tʂ-, tʂh-, dʐ-, ʂ-, ʐ-), l- (cf. Grade II which occurs with shibilants but not sibilants or r-)

All of these initials are associated with Grade III in the Late Middle Chinese (LMC) of the rhyme table tradition. (So are many other LMC initials other than sibilants and *ɣ-.) In LMC, Grade III was nonpalatal and Grade IV was palatal. Assuming that the Tangut carried over that distinction into their analysis of their own language, Tangut Grade III initials must have been nonpalatal. Tangut l may have been velarized [ɫ].

IV: all other initials (cf. Grade I which occurs with all non-shibilants)

However, this correlation between grade and initial is not absolute: e.g., 1dzɨa' has a dz- that normally should precede a Grade IV rhyme. Hence the distinction between medial /ɨ/ and /i/ is phonemic as well as phonetic, and the Tangut created separate rhyme categories whenever the medial could not be predicted on the basis of the initial. Minimal pairs like 3371 and 4075 above necessitated the separation of rhymes 21 and 24. (10.10.20:05: 3371 1dzia [not 1dzɨa'!] and 4075 1dzia' actually differ in terms of the presence or absence of the mysterious feature that I write as -', not in terms of medials.)

On the other hand, I presume all medials in rhyme 27 were nondistinctive (and predictable?*) as suggested by the mixture of Grade III and IV in this rhyme 27 fanqie:

=+

1ʂɨã (Grade III) = 2ʂɨu (Grade III) + 1kiã (Grade IV!)

I'll start looking at the unpredictable medials of rhymes 21 and its -'-less counterpart 19 this weekend.

*It is possible that -ɨ- and -i- were completely interchangeable in rhymes like 27: e.g.,

1ʂɨã ~ 1ʂiã (cf. Grade III rhyme 36 1ʂɨe; there is no Grade IV rhyme 37 *1ʂie)

1kɨã ~ 1kiã (cf. Grade IV rhyme 37 1kie; there is no Grade III rhyme 36 *1kɨe)

It is also possible that rhyme 27 had only one medial (-ɨ- or -i-) after all initials, so all rhyme 27 syllables were Grade III or IV.

It is not possible to choose between these alternatives at this point. It might be more accurate to write the medial of rhyme 27 with an algebraic symbol like -I-. However, I have already used that symbol to represent a lost unstressed presyllabic vowel conditioning the raising and fronting of pre-Tangut *a to i. I assign medials to rhyme 27 syllables following the general pattern: -ɨ- after shibilants (there are no v- or l-rhyme 27 syllables) and -i- after other initials.

14.10.8.22:32: WHIP = TSU + SHARP + ?

If 0219 2tseʳw 'whip' has three sources, the first two might be one of three tangraphs with a TSU-type reading and 3767 1reʳw 'sharp, pointed end':

=//?++?

What might be the third? There are nine tangraphs sharing a right side with 0219 that I didn't cover last Saturday:

LFW2008	Tangraph	Reading	LFW2008 gloss	Class(es)
0054		1tswa	hair worn in a bun or coil	HAIR
0375		1ka	second syllable of 2phʊ 1ka 'boots worn in rain or snow'	HAIR (fur boots?)
0378		2ʔʊ	second syllable of 1dzɨa' 2ʔʊ 'curled hair'	HAIR
1144		2dị	second syllable of 1dzɨa' 2dị 'bun (of hair)'	HAIR
2279		1swa	second syllable of 2siọ 1swa 'a kind of grass'	SWA
4021		1swa	second syllable of 1niu 1swa 'ear ornament'	SWA
4045		1swa	hair	HAIR, SWA
4371		1dạ	second syllable of 2me 1dạ 'hair'	HAIR
5133		2rieʳ	wool, feather, fine hair	HAIR

All of the above characters either represent (parts of) words for hair or syllables homophonous with 1swa 'hair'. So 2tseʳw 'whip' is either 'TSU + hair' or 'TSU + sharp + hair'.

=+

3371 1dzɨa' 'hair worn in a bun or coil; peak (< like a bun of hair on the top of the head?)' = 2750 1ɣɤu 'head' + 1lwʊ̣ 'to mix, blend'

They may be adjectives modifying 1dzɨa'.

Both the structure and pronunciation of 3371 are odd to me (10.10.20:15: because I reconstructed 3371 incorrectly! It should be 1dzia with a Grade IV rhyme, not 1dzɨa' with a Grade III rhyme.) I wouldn't describe a bun or coil as mixed and blended hair. And Grade III rhymes with -ɨ- normally don't follow alveolars. I will take a closer took at -ɨa' tomorrow.

14.10.7.23:57: WERE TANGUT WHIPS SHARP?

On Sunday I concluded that the left side of 0219 2tseʳw 'whip' might be an abbreviation of some tangraph with a TSU-type reading, though I admit the phonetic match is poor:

<//?

2tseʳw 'whip' < left of 1tshwiu, bottom left of 2dziu', or right of 2dʐwɨiw?

I also identified the rest of 0219 as being from

2061 2pɤẹ̃ 'hair'

as a whole. And on Saturday I used Google to demonstrate that whips are associated with hair in English, though of course there is no guarantee the Tangut also had such an association.

2061 of course consists of two components. Maybe each of those components in 0219 2tseʳw 'whip' is from a different source. Let's look at eleven possible sources of

the center of 0219:

LFW2008	Tangraph	Reading	LFW2008 gloss	Class(es)
1146		2kạ	tattered	TATTER
1964		?ɬə	smooth	LHY, SMOOTH
2434		1bie	to mend, patch	BE, TATTER (i.e., to fix tatters)
2600		1miaʳ	hair	HAIR
3088		1bie	second syllable of 2bə 1bie 'dung beetle'	BE
3089		1tʂɨọ	ugly	UGLY
3090		2ɬọ	first syllable of 2ɬọ 2ɬwi 'ugly and old'; can it stand alone?	UGLY
3558		2pɤẹ̃	first syllable of 2pɤẹ̃ 2ba 'flattery'	BE
3767		1reʳw	sharp, pointed end	SMOOTH (left and center from 1963 'smooth')
4330		1ʔị	ladle, scoop	I (bottom center and right from 3101 2ʔị 'to repeat')
4817		?ɬə	plane for carpentry	LHY

I have excluded five tangraphs containing 2061.

The classes can be grouped into three families:

TATTER < BE > HAIR

SMOOTH > LHY > (UGLY if 2ɬọ had a ?ɬə tangraph as phonetic)

I

The last is an unusual case, as the shape of the bottom center component of 4330 1ʔị 'ladle' does not match its source 3101 2ʔị 'to repeat' in its Tangraphic Sea analysis:

=+

The source of the top and bottom left of 4330 1ʔị 'ladle' is 4368 2dwʊ 'chopsticks'.

Among these characters, the best candidate for a source of 0219 'whip' is 3767 1reʳw 'sharp, pointed end'. I wish I knew more about Tangut material culture. Did Tangut whips have sharp ends?

14.10.6.23:06: THE APPEARANCE OF ANGER

0924 2niạ 'anger, rage' and 0996 2mə 'appearance, spirit'

were borrowings from Chinese 惱 'angry' and 模 'pattern' according to Li Fanwen (2008: 156, 167).

The first etymology would work only if there was a pre-Tangut prefix *Sɯ- of unknown function (!) added to *nawʔ from Middle Chinese *nawˀ. The *S- of the prefix conditioned vowel tension (indicated by a subscript dot) and the high vowel *ɯ of the prefix conditioned the -i- in the main syllable:

*Sɯ-nawʔ > *Sɯ-nɨawʔ > *Sɯ-nɨaɯʔ > *S-nɨaɯʔ > *nnɨaɯʔ > *ṇɨaɯʔ > *ṇɨ̣ạɯ̣ʔ > 2niạ

The relative chronology of changes is not entirely clear, though *a-breaking must have preceded *ɯ-loss and *S-tension.

I once thought Tangut rhymes ending in the algebraic symbol -' (corresponding to what I used to reconstruct as long vowels) once had final consonants:

-V' (= -VV) < *-VC

If that were the case - and I don't think it was* - then the absence of -' in 0924 2niạ would not rule an earlier final consonant (i.e., *-w) since -' could not occur with tense vowels. This complimentary distribution is a clue to the identity of -' which had to have some phonetic characteristic that was incompatible with tense vowels.

The second etymology is highly improbable because Middle Chinese 模 *mo 'pattern' should correspond to Tangut *2mʊ, not 2mə. (See Gong 2002: 413 for examples of MC *-uo : Tangut -u which is equivalent to MC *-o : Tangut -ʊ in my reconstruction. I regret not include the raising of *-o to *-ʊ in pre-Tangut.)

There are isolated instances of the correspondences

Tangut -ə : Japhug rGyalrong -u < *-o, -ɯ < *-u

in Jacques (2006), but the general pattern is clear:

Tangut -ʊ (= Jacques' -u) : Japhug rGyalrong -u < *-o, -ɯ < *-u

2mə 'spirit' may be an unrelated homophone of 2mə 'pattern' that was written with the same character.

The Precious Rhymes of the Tangraphic Sea analyzed the graph 0996 for 2mə as being from

=+

the top of 1365 and the bottom of 4744 2ʔiõ 'appearance' (a loan from Middle Chinese 樣 *jɨaŋʰ or Tangut period northwestern Chinese *jõ).

Li may have been tempted to have derived 2mə from Middle Chinese 模 *mo 'pattern' since the word appears with the clarifying character 4744 in Homophones:

2ʔiõ 2mə

He translated that collocation as 模樣 'pattern' which would have been read as *mo jɨaŋʰ in Middle Chinese - a near-mirror image of 2ʔiõ 2mə! I think this resemblance is coincidental. In Tangut period northwestern Chinese, 模樣 was something like *mbʊ jõ which would have been borrowed into Tangut as *bʊ 2ʔiõ. (Tangut tones for Chinese loans are unpredictable, so I have not indicated the hypothetical tone of the first syllable.)

The analysis of 0924 2niạ 'anger, rage' is unknown. Perhaps it was from the top and bottom left of 0948 1na 'to steal' (phonetic) plus 'demon' (semantic) extracted from one of forty-nine different possible characters:

=+(+?)/(?+)

('Demon' has left-hand and right-hand forms which are interchangeable in tangraphic analyses.)

None of the other 'demon' characters mean 'anger', so none stand out as more likely sources than others.

*My old -V' < *-VC hypothesis would not predict Tangut-Japhug rGyalrong comparisons such as these from Jacques (2006):

'nose': 5700 2ni' (not *2ni) : J sna

'needle': 4935 1ɣa (not *1ɣa') : J ta-qaβ

'fruit': 2436 1mia' : J sɯ-mat

may be coincidental.

14.10.5.23:59: WHAT PLUS 'HAIR' EQUALS 'WHIP'?

If the center and right components of

0219 2tseʳw 'whip'

are from

2061 2pɤẹ̃ 'hair',

what is the source of the left-hand component

?

None of the 69 other characters with that component are a plausible semantic match for 2tseʳw 'whip' which may belong to the TSU phonetic class:

LFW2008	Tangraph	Reading	LFW2008 gloss	Class(es)	Class codes
0009		1ʂwɨo	to appear; to raise (< 'cause to appear'?)	APPEAR	S1
0020		1tʂɨa	road, way (literal and metaphorical: 'manner'); to lay bricks	CHA, ROAD	P1, S2
0029		2rɪʳ	market	RIR	P2
0033		1tshị	road, way	ROAD	S2
0051		1thaʳ'	obvious	APPEAR	S1
0060		1dõ	street	ROAD	S2
0130		2thʊ	source, resources	TU?	P3?
0486		2paʳ	horse with white trotters	PAR	P4
0503		1tʂɨa	the surname Cha	CHA	P1
0745		2vɨe	the surname Ve	VE	P5
0752		1tʂɨa	ceremony, courtesy	CEREMONY, CHA	S3, P1
0753		2vɨe	face	VE	P5
0760		2dʐɨe	to judge, decide	JE	P6
0924		2niạ	anger, rage	NA	P7
0948		1na	to steal, rob	NA	P7
0996		2mə	appearance, spirit	APPEAR	S1
1003		1lew	full, filled, satisfied	not HOLLOW?, LU? (but analysis has 1630 2dziẽ 'carve'!)	S4
1026		1tʂwɨa	the name Chwa; luck	CHA	P1
1071		2dziu'	first half of 1071 1226 'to hide, conceal'	HIDE, TSU?	S5, P8?
1082		2riʳ	second syllable of surnames ending in Rir	RIR	P2
1094		2ʐɨə	to go without a burden	GO	S6
1226		?T-	second half of 1071 1226 'to hide, conceal'	HIDE, TU?	S5, P3?
1360		1va	to hide, conceal'	HIDE	S5
1364		1ŋa	hollow, void	HOLLOW, NGA	S4, P9
1578		1swiə	ear	EAR	S7
1588		1tʂɨa	sheep guardian god	CHA, SHEEP	P1, S8
1630		2dziẽ	to carve	CARVE, JE	S9, P6
1641		2dʐɨa	lamb	CHA?, SHEEP? (but analysis has 1043 1lew 'full')	P1?, S8
1651		1tshwiu	to salute	CEREMONY, TSU?	S3, P8?
2663		1kwiə̣	to kowtow, worship on bended knees	CEREMONY	S3
2755		2lwəʳ	the surname Lwyr	LWYR	P10
2972		1ŋa	to spread; Grinstead: 'empty'	HOLLOW?, NGA	S4?, P9
3049		1xwaʳ	to melt, thaw; to confess (< 'melt down' and release information?)	XA, SPEAK	P11, S10
3575		2ni	to listen, hear	EAR	S7
3579		2kie	impressive and dignified, eminent	APPEAR (i.e., prominent?), CEREMONY?	S1, S3?
3689		1lʊ'	to dig	LU	P12
3731		1khɤu'	to milk	KHY	P13
3813		2vɨẹ	to see someone off	VE	P5
3821		2lʊ	to enjoin; to tell; to give a present	CEREMONY?, LU, SPEAK	S3?, P12, S10
3828		1tʂɨə	to give a present; to enjoin; to tell; to know	CEREMONY?, CHA, SPEAK? (but no CEREMONY, CHA, or SPEAK graph in analysis which has 3813 2vɨẹ 'to see someone off')	S3?, P1?, S10?
3874		1ʔiə	hunger	HOLLOW (lacking food)	S4
3920		1kiụ	to bow, salute	CEREMONY	S3
4110		2paʳ	awning, shed	PAR	P4
4153		2lɨiw	to gather, assemble; transcription character	LU?	P12?
4170		1dza	to chisel	CARVE	S9
4185		2ʂɨa	musk	SHA	P14
4201		?kha	casket, small box	XA?	P11?
4469		2ʂɨi	to go toward, depart	GO	S6
4475		?xa	to puff, blow; transcription character	XA	P11
4534		2dʐwɨiw	hungry	HOLLOW (lacking food; but analysis has 130 'source')?, TSU?	S4?, P8?
4677		2bə	bull	BA	P15
4681		1niu	ear	EAR, NU	S7, P16
4682		2khiə'	chimney, window, hole, space	HOLLOW, KHY	S4, P13
4696		1bạ	cymbals	BA, CYMBAL	P15, S11
4744		2ʔiõ	appearance, shape; transcription character	APPEAR	S1
4758		1tsiə	big cymbals	CYMBAL	S11
4761		1ʂwɨa	to speak, say	SHA, SPEAK	P14
4762		1tʂhɨe	to go, walk	GO	S6
4766		2bə	a kind of vegetable	BA	P15
4768		1ʂwɨa	ambition, will	SHA	P14
4812		2rioʳ	to brush, wipe, whisk	RIR?	P2
4822		2dzwiə	to go, walk	GO	S6
4849		1niu	the surname Nu	NU	P16
4894		1mio	to listen, hear	EAR	S7
5126		1lɨu'	to carve, engrave	CARVE, LU	S9, P12
5412		2lwəʳ	ceremony, rite; to get a haircut; transcription character	CEREMONY, LWYR	S3, P10
5693		1vɪʳ	to listen, hear	EAR, VE	S7, P5
6010		1kiụ	to bow, salute (= 3920)	CEREMONY	S3

I have numbered phonetic (P) classes by order of first occurrence in the table above. Class names are in my lay romanization for Tangut which ignores the four grades, vowel tension, and the unknown distinction indicated by -'. Y represents central nonlow vowels.

Phonetic classes organized by Homophones chapter

Some of the phonetic classes could be combined (P4. PAR + P15. BA, P1. CHA + P14. SHA, P10. LWYR, P12. LU).

P6 and P8 might be split, as I am not certain that mixing class VI and VII initials was permissible in Tangut phonetic series.

S1. APPEAR

S2. ROAD

S3. CEREMONY

S4. HOLLOW

S5. HIDE

S6. GO

S7. EAR

S8. SHEEP

S9. CARVE

S10. SPEAK

S11. CYMBAL

10.6.0:59: Some of those 27 classes could be combined into even bigger classes using ambiguous graphs as pivots: e.g., 0020 can either be CHA or ROAD, so CHA and ROAD graphs could be grouped together. Here is one particularly large group containing 18 classes:

That diagram is meant to be read from left to right: e.g.,

CEREMONY > LU > CARVE > JE

Two smaller groups are

CYMBAL > BA > PAR

EAR > NU, VE

Three classes cannot be combined with others: GO, NA, RIR.

Thus one could say there are six kinds of :

1. CEREMONY

2. CYMBAL

3. EAR

4. GO

5. NAR

6. RIR

But I doubt literate Tangut actually looked, at, say, 0760 2dʐɨe 'to judge, and thought, 'its left side indicates that it has a JE-like reading like 1630 2dziẽ 'carve', derived from the right side of 5126 1lɨu' 'to carve', in turn derived from the bottom left of 3821 2lʊ 'to give a present', in turn derived from the center of 5412 2lwəʳ 'ceremony':

<<<<

How did the Tangut learn and perceive their own script?