This Chinese character

的

represents a Mandarin suffix -de that is partly equivalent to the Tangut genitive/dative suffix written with TT5283. -de can indicate a possessor (but not an indirect object):

Xde Y 'Y of X'; 'X's Y'

cf. Tangut X TT5283 Y 'Y of X'; 'X's Y'

(-de also has other functions unlike those of TT5283.)

But one might not guess that -de was similar in meaning to English -'s on the basis of its written form which consists of

白 + 勺

bai 'white' + shao 'ladle'

This seems almost as bizarre as the analysis of TT5283:

< +

TT5283 (genitive-dative suffix) (Gong 'yiy 1.36; Tibetan transcriptions g-ye, ye, yi [Nevsky 1926: 21 #80]) <

left of TT5273 WISH (Gong 'yiy 1.36; homophonous with TT5283) +

right of TT5545 RELATIVES (Gong thyu 1.3)

White ladles have even less to do with possession than wishing for relatives. (Everyone possesses relatives, even if those relatives are dead or unknown. Nobody just materializes out of nowhere.)

Without recourse to Tangut B, I have no way to explain the structure of TT5283. On the other hand, I could explain the structure of 的 de '-'s', but I won't. Can you guess

- the functions of 白 bai 'white' and 勺 shao 'ladle'

- how 白 bai 'white' + 勺 shao 'ladle' came to represent -de '-'s'

Next: The answer.

HINT: 的 de '-'s' has a variant

旳

with 日 ri 'sun' instead of 白 bai 'white'. Some other variants are more puzzling.

06.9.28.23:44: RELATIVE TO GENITIVE

In "Genesis of the Genitive", I forgot to mention tangraphs which contain the components of

TT5283 (genitive-dative suffix) (Gong 'yiy 1.36; Tibetan transcriptions g-ye, ye, yi [Nevsky 1926: 21 #80])

in the opposite positions. Let's see if (フ+刂) and (乙|||) have any single semantic or phonetic function when placed respectively on the right and on the left. I am not optimistic:

(フ+刂)-right tangraphs (Grinstead 1972: 79)

(乙 |||)-left tangraphs

Optimally, one might expect four sets of tangraphs:

1. (フ+刂) as semantic element: semantic unity, with or without phonetic unity

(Semantically related words can be phonetically similar if they share a common root.)

e.g., 3909 TONGUE, 4307 INSIPID, 4775 (a surname that sounds like Chn 'sweet')

2. (フ+刂) as phonetic element: phonetic unity, with or without semantic unity

(Phonetically similar words can be semantically related if they share a common root.)

e.g., 0612 and 3261 (both pron. Gor 2.80)

3. (乙|||) as semantic element: semantic unity, with or without phonetic unity

e.g., 5081 RELATIVE? (those whom one is tied to?), 5083 TIE-ROPE, 5545 RELATIVES

4. (乙|||) as phonetic element: phonetic unity, with or without semantic unity

e.g., 5081 nyï 1.30 and 5082 nyi 2.12

However, there are more than four sets, and many are singletons with unique meanings and readings. What if there are six sets consisting of the four above (with a couple of modifications) plus two more?

1. (フ+刂) as semantic element: semantic unity, with or without phonetic unity

2. (フ+刂) as Tangut A phonetic element: phonetic unity, with or without semantic unity

3. (乙|||) as semantic element: semantic unity, with or without phonetic unity

4. (乙|||) as Tangut A phonetic element: phonetic unity, with or without semantic unity

5 (new). (フ+刂) as Tangut B phonetic element: phonetic unity, with or without semantic unity

(フ+刂) could have the same phonetic value in both Tangut A and B, or different values

6 (new). (乙|||) as Tangut B phonetic element: phonetic unity, with or without semantic unity

(乙|||) could have the same phonetic value in both Tangut A and B, or different values

The singletons might belong to sets 5 and 6; i.e., their structure only makes sense in terms of their Tangut B readings.

Next: White ladles.

06.9.27.23:30: GENESIS OF THE GENITIVE

In "Whose Who?" (Part 1 / Part 2) I discussed the tangraph

TT4593 WHO (Gong sywï 2.28)

which was used before the genitive/dative* suffix

TT5283 (Gong 'yiy 1.36; Tibetan transcriptions g-ye, ye, yi [Nevsky 1926: 21 #80])

A genitive/dative suffix is hard to draw. Let's put dative aside for now. What does possession look like? It's an abstract concept. The first thing that comes to my mind is a hand holding a stick. That's not what TT5283 is. It was analyzed as

left of TT5273 WISH (Gong 'yiy 1.36; homophonous with TT5283) +

right of TT5545 RELATIVES (Gong thyu 1.3)

TT5283 (gen/dat) might be semantophonetic: WISH supplies the sound and RELATIVES suggests possession (since everyone has relatives). One might conclude that the left side of WISH (フ+刂) represented the syllable 'yiy 1.36 and the right side of RELATIVES represented possession, but other tangraphs indicate that this is not the case:

Tangraphs with the left side of TT5283 (gen/dat) and WISH (both 'yiy 1.36); asterisks indicate that these appear with a different left side in the Mojikyo font; phonetic subsets are indicated with letters A-J.

There are ten types of readings associated with tangraphs sharing (フ+刂), the left-hand component of WISH and TT5283 (gen/dat). Four only occur once (F, G, H, J). Half (C, D, F, H, I) have rhymes like -yiy with different types of initials; the only thing all the rhymes have in common is a medial -i- or -y- in Gong's reconstruction - is that all (フ+刂) represented, or did it have a single consistent sound value in Tangut B?

Tangraphs with the right side (乙|||) of TT5545 RELATIVES (Gong thyu 1.3)

(Only tangraphs glossed in Grinstead [1972: 80] are listed.)

(Although Gong reconstructs both rhymes 1.7 and 1.3 as -yu, presumably they were similar but non-homophonous in some way. Sofronov [1968 I.136] reconstructed 1.3 as -Ü and 1.7 as -yũ.)

Only SEVENTH-SON has anything to do with RELATIVES semantically, and SECOND-SON, THIRD-SON, FOURTH-SON, FIFTH-SON, SIXTH-SON, and EIGHTH-SON do not contain the right side of RELATIVES (乙|||) at all. (The tangraph for FIRST-SON, if any, is not glossed as such in Grinstead 1972.) There is no consistent phonetic or semantic characteristic of (乙|||)-right tangraphs (though DEER and RELATIVES do have similar rhymes and initials that have similar points of articulation).

In short, nothing about the structure of TT5283 (gen/dat) unambiguously hints at what the tangraph means or how it was pronounced. It could have represented, say, a word for a kind of deer (cf. TT4537) pronounced kiew 1.44 (cf. TT5271).

(I say "unambiguously" because even though it does share a left-hand component with the tangraphs in phonetic subset C, there is no way for the reader to guess that it is not nearly homophonous with the tangraphs in the nine other subsets.)

Next: How do the Chinese write their equivalents of TT5283 (gen/dat)?

*Using a single suffix to indicate possession (genitive) and indirect objects (dative) is not surprising from an English perspective: cf. how English to appears in both belong to (a possessor) and in give to (an indirect object).

06.9.26.23:59: WHOSE WHO? (PART 2)

In Part 1, I forgot to mention the Tangraphic Sea's analysis of

TT4638 WHO (Gong sywï 1.30)

as

right of TT5251 LANGUAGE (Gong dạ 2.56) +

all of TT3415 ASK (Gong yï^r 1.86)

How would deleting the left side of ASK (the element PERSON) tell a reader that

TT4593 WHO (Gong sywï 2.28)

was to be used before the suffix

TT5283 (genitive suffix) (Gong 'yiy 1.36)?

Unfortunately, the second volume of Tangraphic Sea with analyses of tone 2 tangraphs is missing, so no analysis of TT4593 WHO (Gong sywï 2.28) is available. But presumably its left side is also derived from TT5251 LANGUAGE, which interestingly resembles a mirror-image version of Chinese

誰

'who' (< 言 'speech' [semantic] + 隹 'short-tailed bird' [phonetic])

though the tangraphic element resembling 隹 'short-tailed bird' is not phonetic (at least from a Tangut A perspective*).

I don't think the presence of a 'language' element (コ over 乙; cf. the cursive form 讠 of Chn 言 'speech') on the left side of TT4638 WHO, TT4593 WHO, and 誰 'who' is coincidental. I suspect that the two WHO tangraphs are partial visual calques of Chn 誰 'who'. But what about the non-calque parts of the WHO tangraphs?

TT3415 ASK (Gong yï^r 1.86)

cannot be a Tangut A phonetic in

TT4638 WHO (Gong sywï 1.30)

since the two do not sound alike. If ASK is a Tangut B phonetic, then were ASK and TT4638 WHO homophones in Tangut B distinguished in writing by the presence or absence of the LANGUAGE element on the left**? Or did TT4638 WHO have a Tangut B reading that sounded like the Tangut B reading of ASK plus the initial syllable of the Tangut B reading of LANGUAGE?

ASK was analyzed as

right of TT4478 HOW (Gong lyọ 2.64) +

right of TT4638 WHO (Gong sywï 1.30)

Thus ASK and TT4638 WHO form a circle: each was analyzed as comprising part of the other. Their common denominator is a right-hand element that appears in tangraphs with various meanings and readings (Grinstead 1972: 126):

(I have not listed tangraphs with the right side of WHO and ASK which Grinstead has not glossed.)

Did some or all of the above have Tangut B readings with a common final syllable?

*The Tangut A readings of tangraphs containing the left-hand component of

TT5251 LANGUAGE (Gong dạ 2.56)

have no single common semantic or phonetic thread running through all of them (though similar subsets exist within the group as a whole):

The element also appears on the bottom (TT0223) and on the right side (the others below):

**If the LANGUAGE element on the left of

TT4638 WHO (Gong sywï 1.30) and TT4593 WHO (Gong sywï 2.28)

had a zero phonetic value in Tangut B, then the Tangut B readings of the two tangraphs were

XY vs. Y

(X = phonetic value of PERSON, Y = phonetic value of the right element)

Let's assign arbitrary sound values to X and Y for ease of illustration. If X = a and Y = i, then there was an alternation between

ai 'who' (strong form not preceding genitive suffix)

i 'who' (weaker form preceding genitive suffix)

Such an alternation of strong and weak forms reminds me of this Sanskrit set:

vaach 'voice' (cf. voice, vocal, vox, etc.)

vachmi 'I say' < vach 'say' + -mi (1st person singular)

uchyate 'was said' < uch- 'say' + -ya- (passive) + -te (3rd person singular)

vaach, vach, and uch contain the same root (vac) in its 'strong' (long vowel: v-aa-ch), 'mild' (short vowel: v-a-ch), and 'weak' (no vowel; v-ch > uch) flavors.

06.9.25.23:59: WHOSE WHO? (PART 1)

What did my recent posts on Sanskrit have to duu with Tangut?

Sanskrit has a rich and well-understood morphology and phonology. Words and affixes (e.g., duu-) have many forms (e.g., dur-, dus-, duH-) which can be predicted on the basis of a complex rule system.

Tangut, on the other hand, has a poorly understood morphology and phonology. Although we are fortunate to have extensive documentation of the rhymes and tonal classes of Tangut, the data at hand is insufficient to give us precise phonetic values. We know that Tangut words X and Y belonged to rhymes A and B and had a 'level' tone ('tone 1') and a 'rising' tone ('tone 2'), but we cannot say what rhymes A and B were with absolute certainty. Nor can we be sure that the 'level' and 'rising' tones were in fact level and rising.

Moreover, the Tangut script (tangraphy) seems to have very little correlation with the sound system presented in the dictionaries. That is unexpected given the high phonetic content of all other writing systems, including some even less successful than Tangut (e.g., the Khitan small script, thought to be entirely phonetic). For that reason, I hypothesized that there were in fact two Tangut languages, one recorded in dictionaries (Tangut A, a 'high' language?) and another (Tangut B, a 'low' language?) reflected more directly in the writing system which did double duty for both languages.

But how can the Tangut B hypothesis explain this pair of tangraphs?

TT4638 WHO (Gong sywï 1.30)

TT4593 WHO (Gong sywï 2.28)

Rhymes 1.30 and 2.38 were identical except for their tones (1 and 2). Thus the two tangraphs were homophonous if tones were ignored. (And they were listed as homophones in the Homophones dictionary which did ignore tones [Sofronov 1968 II.195].)

Both sywï 1.30 and sywï 2.28 may be related to Written Tibetan su 'who', whose resemblance to Sino-Korean 誰 su 'who' (< Late Middle Chinese shHuy < Old Chinese duy) is coincidental.

TT4593 sywï 2.28 seems to be a variant of TT4638 sywï 1.30 used before the genitive suffix (Sofronov 1968 I.240-241):

WHOSE = TT4593 WHO (Gong sywï 2.28) + TT5283 (genitive suffix) (Gong 'yiy 1.36)

Questions:

1. Why would sywï 1.30 become sywï 2.28 (i.e., change its tone from 1 to 2) before 'yiy 1.36? Is this an instance of tonal dissimilation (tone 1 + tone 1 > tone 2 + tone 1)? Or does the tone shift have some other significance? Could tone 2 in Tangut A be a remnant of some earlier, now-lost consonant or vowel (still preserved in Tangut B? see below)?

2. What does the presence or absence of the element

TT3344 PERSON (Gong dzywo 2.44)

signify in

TT4638 WHO (Gong sywï 1.30) and TT4593 WHO (Gong sywï 2.28)

Was PERSON arbitrarily subtracted from TT4638 to create a derivative tangraph TT4593? I would prefer this not to be the case. That leaves two possibilities:

A semantic explanation: Doubtful. How could TT4593 WHO (without PERSON) be less 'person-like' than TT4638 WHO (with PERSON)?

A phonetic explanation: Problematic. PERSON is not a known indicator of tone 1. It appears in many tone 1 and tone 2 tangraphs with or without any obvious semantic relationship to people (e.g., certain numerals and DOG [calendrical term]). Such widespread usage implies that it had a phonetic function in Tangut B. (If so, its Tangut B reading must have been simpler than its Tangut A reading dzywo, as it is unlikely that many Tangut B words contained the syllable dzywo.) Could PERSON represent a vowel present in the Tangut B reading of TT4638 but absent in the Tangut B reading of TT4593?

If arbitrary Tangut B sound values are assigned to the components:

left element = ka, PERSON = i, right element = pu:

TT4638: kaipu

TT4593: kapu

If so, is the vowel alternation (ai ~ a using the arbitrary sound values) a trace of an earlier ablaut system?

NEXT: How was WHO analzyed in Tangraphic Sea?

06.9.26.00:49: Did Tangut B speakers reduce a stressed diphthong in 'who' to a single unstressed vowel because the following genitive suffix was stressed? E.g., káipu- + -té = kaputé (once again using arbitrary phonetic values and assuming only one stressed syllable per word)?

06.9.24.23:36: DUU-RIVATION (PART 3)

(Preceded by "En-dur-ance Test", "Dus-tortion", Part 1, and Part 2.)

If Sanskrit speakers didn't convert a single 'underlying' prefix (dus-) into the many 'surface' forms of the negative prefix (duu-, dur-, dus-, duSh-, dush-, duH-), how might they have invented new du...-words at will?

I don't really know. I wasn't there, and even if I were, I couldn't have looked into speakers' brains, not even with 21st-century technology. So all I can do is guess within a few common-sense constraints:

1. the solution must not involve abstract forms no one has ever heard

2. conversely, the solution must be based on actual forms people have heard

3. the solution must generate the correct forms

Constraints 1 and 2 rule out an 'underlying' prefix like duC- with an abstract consonant C whose exact value would be determined by the following consonant.

The "actual forms people have heard" would consist of pairs like

aatman 'self' : duraatman 'evil-natured'

tarka 'reasoning' : dustarka 'bad reasoning'

karman 'act' : duShkarman 'wickedness' (< 'bad act')

chakShas 'eye' : dushchakShas 'evil eye'

shruta 'heard' : duHshruta 'badly heard'

rakShya 'guarded' : duurakShya 'difficult to be guarded'

New forms would be created by analogy: e.g.,

given tarka 'reasoning' > dustarka 'bad reasoning'

tara 'conquering' (another t-initial word like tarka) > dustara 'invincible' (< 'difficult to conquer')

In practice, these analogies were not applied consistently, and hence there was some degree of variation:

t-initial words originally had duShT- negative forms (e.g, duShTara 'invincible');

later, duShT- gave way to dust- (e.g., dustara 'invincible')

duH- as well as duSh- before k(h)-, p(h)-

dus-, duSh-, dush- as well as duH- respectively before s-, Sh-, sh-

The variation in the prefix was originally not due to analogies (botched or otherwise) but to sound changes affecting the final -s of the original negative prefix dus-. After those sound changes were applied, future generations heard the words resulting from those changes and created new words by analogy: e.g.,

generations of speakers before sound changes

stage 1: dus-r... (r... = some r-initial word)

first generations of speakers put the word through gradual sound changes

stage 2: dus-r... > duz-r... (-s- > -z- before r-)

stage 3: duz-r... > duZh-r... (-z- > -Zh- before r-)

stage 4: duZh-r... > dur-r... (-Zh- > -r- before r-)

by stage 4, it is no longer obvious that dur- < dus-

a later generation of speakers

heard dur-r words and created more dur-initial forms of r-words by analogy

but did not recreate the sequence of sound changes that turned -s- into -r-;

instead they came up with a sound change of their own:

stage 5: dur-r- > duu-r... (-rr- > vowel lengthenening +r-)

even later generations of speakers

heard duu-r-words and created more duu-initial forms of r-words by analogy

but did not recreate the sound change (-rr- > vowel lengthenening + r-)

Once a sound change is gone, it's gone unless it can be easily recoverable on the basis of audible information (and the sequence of changes that turned -s- into vowel lenghthening before -r- wasn't). But the effects of sound changes are not gone, and speakers can use affected words as a basis for creating new words that only appear to have gone through the sound changes: e.g.,

if one wanted to say 'bad robot',

one could create duurobot by analogy with duurakShya, etc.:

rakShya : duurakShya

robot : duurobot

and duurobot looks like it could have come from some earlier dus-robot (via the sound changes described above), even though no dus-prefixed form of robot ever existed.

Next: What does this have to duu with Tangut?