07.9.22.23:32: TSAK LAK RAK?

(i.e., does the *tsak phonetic series lack *rak graphs?)

When I proposed *skak as a possible reconstruction for the Old Chinese reading of 作 'make' in "Ten Suns", I was hoping to find a velar-initial cognate:

OC *skak : OC ?*(N-)kak(s) or ?*(N-)kak(s)

That pair would be similar to this pair of readings (Sagart 1999: 63):

挾 OC *s-kep ~ *N-kep 'grasp'

But 作 'make' has no velar-initial cognate (though the phonetic 故 OC *ka's of its graphic variant 做 implies one - see "Personal Reasons"). No graph containing 乍 OC *rdzaks 'suddenly' as a phonetic has a velar initial. All 乍-graphs have alveolar or retroflex affricate initials in Middle Chinese, and it would be safest to derive the retroflexes from rts-type clusters and project the alveolars back into Old Chinese (the numbers are from Karlgren 1957):

806a-e 乍 OC *rdzaks (< ?*N-rtsaks > MC *Jæh 'suddenly' (*J = voiced retroflex affricate)

806g 詐 OC *rtsaks > MC *Chæh 'treacherous' (*Ch = voiceless retroflex affricate)

806h 祚 OC *dzaks (< ?*N-tshaks) > MC *dzah 'blessing'

9.23.11:13: Schuessler (2007: 201) regarded 806h, i, t-u (see below) as cognate to 錯 OC *tshak 'crossing', though the semantic similarity is opaque to me. If Schuessler is correct, then the *dz- is probably from *tsh- plus a prefix that caused voicing.

9.23.11:16: If 錯 OC *tshak 'crossing' is from *s-tsak, then perhaps it and 806h, i, t-u share a root *tsak with an unaspirated initial.

806i 胙 OC *dzaks (< ?*N-tshaks) > MC *dzah 'sacrificial meat and wine'

806j 阼 OC *dzaks > MC *dzah 'east-side staircase leading up to hall'

806k 笮 OC *rtsak > MC *Chæk 'treacherous' (*Ch = voiceless retroflex affricate)

806l 作 OC *tsak > MC *tsak 'get up; start; make'

9.23.11:21: Schuessler (2007: 638) linked this to 措 OC *tshaks 'establish; place; lay aside' presumably via 'start'. Perhaps 措 was once *s-tsak-s.

806m-o 迮 OC *tsak > MC *tsak 'start' (same word as 806l?)

806p-q 柞 OC *tsak ~ *N-tsak > MC *tsak ~ *dzak 'oak'

806r 怍 OC *dzak > MC *dzak 'ashamed'

806s 昨 OC *dzak > MC *dzak 'previous day'

806t-u 酢 OC *dzak (< ?*N-tshaks) > MC *dzak 'present and drink a cup in response to a pledge cup'

9.23.11:13: I can vaguely see a semantic connection between this and 錯 OC *tshak 'crossing'.

806v 筰 OC *dzak > MC *dzak 'bamboo rope'

乍-graphs represented OC *tsak-like syllables.

Although MC *ts-, *tsh-, and *dz- can also reflect OC *s-stop clusters (Sagart 1999: 63-65, 69) -

OC*s-k-, *s-t-, *s-p- > MC *ts-

OC*s-kh-, *s-th-, *s-ph- > MC *tsh-

OC*s-g-, *s-d-, *s-b- > MC *dz-

- I don't know of any stop-initial cognates* for any of the words written with 乍.

Sagart (1999: 67-69) proposed lateral and rhotic clusters as additional sources of MC alveolar affricates:

OC *s-l-, *s-r- > MC *dz-

He reconstructed *l and *r in two of the above words written with 乍:

昨 OC *s-lak > MC *dzak 'previous day'

柞 OC *s-rawk > MC *dzak 'oak'

(I have converted Sagart's superscript a notation to the underlining that I use to indicate 'emphasis'.)

If these words really did have initial *s-l- and *s-r- in OC, why would they be written with a phonetic for OC *tsak-like syllables? Why weren't they written with phonetics for *lak and *rak?

Next: An explanation and an experiment.

*9.23.11:59: Matisoff (2003: 614) reconstructed Proto-Tibeto-Burman *l-tak 'ascend, lift, raise, top' that could be connected with 作 OC *tsak 'get up' (if it is from *s-tak) and with 上 OC *dang' 'rise' (if it is from *N-tak-N-' or *N-tag [Sagart (1999: 134) proposed *-g as a source of OC *-ng']). There are two reasons why I am hesitant to link PTB *l-tak with 作 OC *tsak 'get up':

- There is no Chinese-internal evidence for reconstructing root-initial *t- for 作. I cannot find any Chinese word for 'get up' which has the shape *tak (my derivation of 上 OC *dang' 'rise' from a root *tak is highly speculative). If 作 had an initial *st-, that might imply that all 乍-graphs had *st-type initials instead of *ts-type initials. It is dangerous to change the reconstruction of a set of Chinese words on the basis of a single external etymology. I prioritize internal over external evidence, and facts (the use of phonetics) over hypotheses.

- The semantics don't match. Schuessler (2007: 638) glossed 作 as 'To get up (in the morning) ... to start ... to sprout'. These all seem to be based around the notion of starting rather than rising. Getting up to start the day is not the same as getting up after, say, a fall at any time. However, Karlgren (1957: 213) did gloss 作 as 'rise; stand up'. But the earliest attestations for these meanings are in Shijing and the Analects, long after the oracle bones and bronze inscriptions. It is possible that 'rise' etc. had been there all along but wasn't preserved in writing until later, but it is also possible that 'rise' is a newer word that may have nothing to do with PTB. (Maybe 'make' was *tsak, 'rise' was an unrelated *s-tak, and the two merged as *tsak in middle OC?)

Schuessler has an alternate etymology for 上 OC *dang' 'rise' that I will discuss elsewhere.

9.23.12:21: Karlgren (1957: 213) did find 迮 OC *tsak 'start' in bronze inscriptions, but not 'rise' or 'stand up'.

07.9.21.23:59: PERSONAL REASONS

I still don't understand why Md zuo 'make, do, be' is written as 做, a combination of 亻 Md ren 'person' and 故 Md gu 'reason'. In early Mandarin, 作 'make' had several pronunciations (*tsaw', *tsO, *tsu), and 故 'reason' was *ku (tone aside, identical to modern gu [ku]). Why create a second graph 做 for 'make' with 故 'reason', whose initial and meaning do not match 'make'?

The *tsu pronunciation of 'make' and 故 *ku 'reason' both belonged to the 'departing tone' category. Was a shared rhyme -u (departing) sufficient to justify the choice of 故 *ku 'reason' as a phonetic in 做 *tsu 'make'?

Maybe the creator felt he had no choice. 做 *tsu 'make' had no homophones which could have been used as phonetics. A seemingly obvious alternative 伹 with a common phonetic for *tsu was already in use for*tsü 'clumsy' (with a nonmatching vowel and level tone) and *dzu 'to advance' (with a nonmatching initial and level tone; usually written 徂).

I have not been able to find any evidence for a velar-initial word like *ku < Old Chinese *kaks meaning 'make'. 作 'make' has a phonetic 乍 OC *rdzaks 'suddenly' which only represents *ts- and *dz-syllables. 作 'make' originally meant 'get up'**, and there is no similar velar-initial word other than 起 *khə' 'rise'.

9.22.9:06: None of the variants of 作 'make' contain velar-initial phonetics.

*The phonetic 且 *tshye 'moreover' is found in sinographs for *tsu with other tones:

*tsu (level tone): 租蒩

*tsu (rising tone): 組祖

All the OC ancestors of these early Mandarin *ts(h)- and *dz- syllables shared a vowel *a:

作 EM *tsaw, *tsO < OC *tsak

作 做 EM*tsu < OC *tsaks

伹 EM *tsü < expected to be from OC ?*tsa (not attested in OC)

徂 EM *dzu < OC *N-tsa

且 EM *tshye < OC *tshya'

租 蒩 EM *tsu < OC *tsa

組 祖 EM *tsu < OC *tsa'

9.22.8:33: The different vowels of the early Mandarin readings for 且-graphs reflect the preceding consonants in OC:

OC *a preceded by emphatic *ts-: > EM *-u (徂租蒩組祖)

OC *a preceded by nonemphatic *ts- > EM *-ü (伹)

OC *a preceded by nonemphatic *tshy- > EM *-ye (且)

I don't know what OC *a would become if it were preceded by emphatic *tshy- (if *Cy- clusters even existed - maybe OC *y had no emphatic counterpart).

**9.22.8:44: Schuessler (2007: 638) proposed that 'make' is derived from an earlier meaning 'get up':

'To get up' (in the morning) [Lunyu] > 'to start, start work' [Shu], 'to sprout' [Shi] > 'to do, perform, work, set up, build' [OB, BI, Shi] > 'act as, be' [Shu]; intr. 'to be active' [Shi].

This progression is a bit strange, as Lunyu postdates the oracle bone (OB) and bronze inscriptions (BI). I wonder what the earliest attestation of 'get up' is. If 'make' is a secondary meaning, the meaning 'get up' should already be present in OB.

07.9.20.23:45: WRONG CALL

(告 Md gao [kaw] and Cantonese kow are not too far from my pronunciaition of call as [kɔɰ].)

I posted "Ten Suns" too quickly. Right after I put it up, I discovered that Schuessler's (2007: 601) entry for 造 continued on the other side of the page (p. 602):

The element gào 告 'report' with initial k- is not phonetic, it was part of the original word 造 *tshûh 'to go and offer' (a sacrifice), 'go and appear in court' which usually would involve some announcement or report. Therefore there is no need to postulate an *sk-like initial [for 造].

Then a reader sent me an article by 陳劍 Chen Jian pointing out that 造 in fact did not originally contain 告 Old Chinese *kuk(s) at all, despite the modern form of the sinograph. 告 in 造 is actually a distortion of an early variant of 草 OC*tshu' 'grass'. Therefore there is no pressing need to reconstruct a velar in the OC reading of 造 or to consider 告 'report' a semantic element in 造. I will reconstruct 造 with alveolar initials:

*tsu' 'make; begin; achieve'

*tshus 'proceed/reach'

(I left the aspirated reading out of "Ten Suns"; if I still believed 'proceed' had a velar, I would have reconstructed *skhus.)

Schuessler (2007: 600-601) lists other OC words with similar semantics and *ts-type initials which may be related to the words written with 造 'make; begin; achieve; proceed/reach to' (the reconstructions are mine):

早 OC *tsu' 'early'

(if < *tsung', cognate to 宗 OC *tsung 'ancestor' [< 'early one'?]?)

(if root was *ts-', cognate to 祖 OC *tsa' 'ancestor' with vowel change?; no, Schuessler [2007: 636] derives from root *tsa 'go' - i.e., 'the one who is gone'.)

遭 OC *tsu 'meet, encounter'

曹 OC *N-tsu 'come together; crowd' (< 'make meet'; 'those having met'?)

遒 OC *N-tsu 'collect, bring together' (Should this be *N-s-t-lu? Its phonetic is 酉 *lu'.)

綜 OC *tsu-ng(')-s 'bring together' (cf. my proposal in "Ten Suns" to derive *-u' in 造 from *-ung')

卒 OC *tsu-t 'group' (with nominal *-t suffix)

萃 OC *N-tsu-t-s 'collect; crowd' (Baxter and Sagart reconstructed this with *-p, but the phonetic 卒 OC *tsu-t implies *-t.)

If 造 is cognate to those words, I would reconstruct it as *N-tsu' 'make',*tshu' (< ?*s-tsu') 'proceed'.

07.9.21.8:01: Schuessler (2007: 601) noted that Bodman reconstructed an even larger word family 'to collect, accumulate' including

屯 OC *dun 'accumulate; hill' (< *N-tun if cognate to 頓 OC *tun-s 'hill' < 'accumulation [of earth]'?)

最 OC *tsots 'collect, accumulate; highest degree'

which Schuessler linked to 聚 OC *dzos (< ?*N-tso-s) 'collect, store'.

The last two might be relevant if the root were *ts- or *ts-w. (Perhaps *ts-w with zero vowel would phonetically be [tsu] and *ts-ə-w would be [tso].) However, 屯*dun (< ?*N-tun) does not have an affricate initial like the others. If it were cognate, I would have to derive the *ts- of all the other words from *s-t-, and I'm not willing to alter the reconstructions of many words to make them cognate to only two words (屯 ?*N-tun and 頓 *tun-s). Nor am I willing to reconstruct an *-l- in many words to make them cognate to a single word (遒 *?N-s-t-lu).

07.9.19.23:59: THE MAKING OF TEN SUNS

Ten Suns sounds like a movie but actually refers to the apparent components of 早 'early': 日 'sun' and 十 'ten'. Karlgren (1957: 272) does not explain this sinograph.

Schuessler (2007: 601) proposed that 早 "is prob. derived from 造 'do, make, begin'." He reconstructed those two words in Old Chinese as

早 *tsû' (my *tsu' or *sku')

造 *dzû' (my *sgu' or *Nsku')

(Schuessler's circumflex corresponds to my underlining for emphasis.) Although there is no semantic problem preventing 'begin' from being related to 'early', 造 'begin' (usu. 'make' in modern Mandarin) has a phonetic element 告 OC *kuk(s) (Schuessler's OC *kûk[h]) with an initial velar. Shuowen states that 告 is phonetic and it is hard for me to think that 告 'tell' is semantic in 造 'do, make, begin'. Hence I follow Sagart (1999: 63, 241) in reconstructing 造 as OC*sgu' with *-g- on the basis of its phonetic. OC *Nsku' is also possible. 早 has no embedded phonetic, so I cannot be sure if it was *tsu' or *sku'. Obviously, *tsu' and *sgu' could not be related - a shared rhyme -u' is not enough. However, *sku' and *sgu' (< ?*s-N-ku') or *Nsku' might share a root *(s)ku', though the function of the prefixes would require an explanation.

9.20.00:27: Middle Chinese has no *-awng' or *-ung' from an OC *-ung' after emphatics and nonemphatics. This is suspicious. Is it just pure chance that OC glottal stop could follow *-ng except if it were preceded by *u? I wonder if a pre-OC *-ung' was reduced to *-u', just as OC *-əng' became later *-ə' in most cases (see Sagart 1999: 61).

I suspect that 造 was pre-OC *sgung' (< ?*s-N-kuk-N-') and that it might be cognate to 作 'make' which could have been OC *tsak or *skak (with a root *[s]k-k). 作 can also be written as 做 with the phonetic element 故 OC *ka's implying a velar. But 做 couldn't reflect OC pronunication because it seems to date from the Ming Dynasty, when 作 had initial *ts- and 故 had initial *k-. More on this next time.

07.9.18.23:59: -A MIN

In "See-Saw", I proposed that Min -a came from LOC *-æw via an intermediate stage *-O. This is an awkward idea for two reasons:

First, it involves a front vowel æ backing to O and then bouncing forward to a. Possible, but unlikely.

Second, I wonder if early Min might have had an *-O from LOC *-aw: e.g.,

(But I can't find any other examples of these correspondence patterns in this database. Chaozhou usually has -au instead of -o in this rhyme class. Could these -au be from later strata of borrowings?)

I've refined my proposal slightly:

LOC *-æw > PM ?*-Ö (a front vowel like *æ) > ?*-ə > modern -a

LOC *-aw, *-ow > PM ?*-O > modern -O, -o

(The nasalization in Xiamen hÕ reflects the nasality that was once in its initial: ?*hmOh > *hmÕ > hÕ. Xiamen has no rhyme -õ. *-O raised to -o in Xiamen unless it was nasalized.)

9.19.1:13: Schuessler's (2001) modification of Norman's Proto-Min (SNPM) has the following reconstructions corresponding to my guesses:

SNPM *-au : PM ?*-Ö

SNPM *-ɑu : PM ?*-O

Presumably there are Min languages which still have -Vw for these rhymes. If so, then did SNPM *-au become -a only in a subgroup of Min? I would imagine that subgroup would have to include both Eastern Min (Fuzhou) and Southern Min (Longdu, Xiamen, and Chaozhou), which all have -a in their colloquial (native) strata.

Pulleyblank (1984: 189) proposed that early Min had a contrast between *-aaw and *-aw, and that both lost their final glides (the intermediate forms are my guesses):

*-aaw > -a

*-aw > (?*-Ow) > -O, -o

What complicates any of these scenarios is 早 Chaozhou tsa, implying LOC *Chæw'. But the mainstream LOC form was *tsow', which would imply a nonexistent Chaozhou tso.

Expected correspondence: LOC *-æw : CZ -a (in 敎 炒 包)

Actual correspondence: LOC *-ow : CZ -a (in 早)

Chaozhou is real, and LOC is only a theoretical generic construct, so it's obvious which one should be prioritized. Nonetheless, other evidence points to LOC *tsow', which cannot be the ancestor of Chaozhou tsa.

There are two SNPM reconstructions for 早, *tsau' and *tsɑu'. (Underlining represents PM 'softened' initials rather than emphatics, and the apostrophes represent PM tone B.) Chaozhou tsa seems to descend from the former, whereas Xiamen tso seems to descend from the latter. (Fuzhou tsau looks like a loan. The expected Fuzhou forms would be tsa < *tsau' or tsO < *tsɑu'.)

SNPM *tsau' may ultimately be from a prefixed Old Chinese *r-N-tsu' (*N- = unknown nasal prefix; in OC, underlining does indicate emphasis) whereas SNPM *tsɑu' is from OC *N-tsu' without an *r- prefix. Mien (not Min, and not even Chinese!) dzyou 'early' < *ntz- (should this be *nts-?) is a loanword reflecting a variety of OC with a nasal prefix. See Schuessler (2007: 601). (Mandarin zao [tsaw] and its cognates in most Chinese languages do not have a trace of that prefix.)

07.9.17.23:59: THE SEE-SAW SOLUTION: HERE AND NAW

In 1996, I was puzzled by the complex correspondences between Tangut, Chinese, and Tibetan rhyme categories in transcription. (Sanskrit might have made the picture even more complex if only there were more Sanskrit transcription data.) "Cerebral Vegetables" dealt with a simple case of one Tangut rhyme pair (1.22/2.19) corresponding to three Tangut period northwestern Chinese rhyme categories. Here's an even more extreme case:

This particular instance seems to have a simple explanation. The consensus is that 1.27/2.25 was some sort of non-low central vowel:

Nishida (1964): -uH

Shi et al. (1983: 1.27 only): -ər (real -r coda or retroflex vowel?), -ə̃, -ə

Li Fanwen (1986): -u

Sofronov (1968) and Gong (1997): -ə

Tibetans had no letter for schwa, so they used all their vowel letters in an attempt to write this alien sound.

On the other hand, TPNWC speakers probably did have a schwa, but there was no guarantee that each and every Tangut onset-ə sequence had an exact counterpart in TPNWC: e.g.,

TT3119 FIVE ngwə 1.27

was transcribed in the Pearl in the Palm as 魚骨, a fanqie spelling ?*nggü + ku for TPNWC ?*nggu because TPNWC had no syllable like *ngwə.

What about rhyme category 1.22/2.19? It has been reconstructed in three basic ways:

Type 1: -u-like final element:

Nishida (1964): -aw

Sofronov (1968): -aɯ

Li Fanwen (1986): -âu

Type 2:

Shi et al. (1983: 1.22 only): -ar, -ær (real -r codas or retroflex vowels?)

Type 3:

Gong (1997): -aa

If type 1 were correct, this rhyme should correspond only to *-Vw-type rhymes in TPNWC. But it corresponds to TPNWC *-a and *-O as well as *-aw. Conversely, if type 3 were correct, this rhyme should only correspond to TPNWC *-a and perhaps *-O. And if type 2 were correct, tangraphs of this rhyme class might not be used to transcribe TPNWC at all. TPNWC probably did not have retroflex vowels or final *-r.

In 1996, I proposed what I might now call a type 4 solution. What if tangraphs had more than one reading? What if they were like English read, which can be [riid] or [rEd] depending on context? Just as read has a base form [riid], 1.22/2.19 words might have had an -aa base form with variant readings ending in -w.

Suppose that English were written with sinographs, and 見 represented both see and saw. If one found見 transcribing both Sanskrit sii and saa (my pronunciation of saw is [sɒɒ]), should one conclude that the reading of 見 was something like both sii and saa, or that 見 had two readings?

There are two big problems with the see-saw solution.

First, there is no evidence for the expected variant readings in the Tangut dictionaries. Using Gong's reconstructions, -aa 1.22/2.19 tangraphs don't reappear under, say, -ew 1.43/2.38.

Second, related forms of Tangut verbs were written with different tangraphs: e.g.,

TT0195 SEE (form 1) lyiy 2.33

TT1838 SEE (form 2) lyi 2.9

(The two top elements atop a common bottom element suggest a Tangut B stem with different prefixes.)

Guillaume Jacques' 2006 paper provides more examples.

Last night I came up with a defense for Gong's type 3 solution. Modern NWC dialects have -aw, -O, or even -o corresponding to the Tangut period NWC rhyme that I had been reconstructing as ?*-aw. Perhaps it is my reconstruction that is wrong. If the dialect of TPNWC known to the Tangut had already shifted *-aw to *-O, then it would make sense to use -aa 1.22/2.19 to approximate a foreign -O. Maybe Tangut a and TPNWC *O were phonetically closer (e.g., [ɑ] and [ɒ]?).

The -a found in Min languages corresponding to Late Old Chinese *-æw (e.g., 教 Xiamen ka : LOC *kæw) might be from an earlier *-O:

*-æw > *-O > -a

9.18.00:21: Until Sunday, I had assumed that LOC *-æw > *-æ > Min -a. But I just remembered that LOC *-æ became -e in Xiamen: e.g., 家 LOC *kæ > XM ke. How could LOC *-æ become -a in some words but -e in others?

Are the -e words borrowed from a dialect with a different sound change? My *-O solution provides a way out:

(I propose *-æ instead of *-e as an intermediate stage to account for the -a < LOC *-æ in Xiamen's sister languages Fuzhou and Lungtu. In Xiamen, *-æ raised to -e, whereas in Fuzhou and Lungtu, *-æ remained low and backed to -a.)

For more examples of Min reflexes of LOC *-æw and *-æ, see Pulleyblank (1984: 188-189). (Those LOC rhymes correspond to his late Early Middle Chinese *-E^rw and *-E^r.)

*In my dialect of English, the final liquid -l has become the velar glide [ɰ], the consonantal counterpart of the high back unrounded vowel [ɯ]. In modern NWC, nearly the reverse has occurred: the high central unrounded vowel*ï has become a syllabic liquid [l] (which is retroflex [L] after retroflex initials). See Coblin (1994: 108).

07.9.16.23:46: WHY NO CEREBRAL VEGETABLES?

Thanks to Guillaume Jacques for pointing out to me that there are many examples of [Tangut period northwestern] Chinese -Vw transcribed as open vowels in the Tangut translation of Cixiao zhuan. I still haven't given his new book on Cixiao zhuan the thorough examination that it deserves, but I can say that I have seen similar phenomena in the Pearl in the Palm and in the Tangut translation of the Forest of Categories.

Here's an example of the opposite phenomenon: Tangut -VG corresponding to -V in TPNWC transcription:

TT1790 VEGETABLE 2.19 (according to Precious Rhymes of the Tangraphic Sea)

Transcription: 那

Pre-Tangut NWC *nda (cf. its Tib. transcription Hda *?[nda])

TPNWC ?*nda

Modern NWC dialects: na (but nE < *nay in colloquial Lanzhou)

Tangut reconstructions:

Nishida: naw

Sofronov: naɯ

Gong: naa

This transcription occurs repeatedlly in Pearl in the Palm (14.4.8 onward).

Gong's reconstruction naa best fits the transcription 那 TPNWC ?*nda. If VEGETABLE were something like naw, it would have been transcribed with a sinograph like 腦 TPNWC ?*ndaw.

However, if Tangut rhyme 2.19 and its level tone counterpart 1.22 really were -aa,

- why weren't tangraphs for 1.22/2.19 syllables used to transcribe Sanskrit -aa-syllables?

- why were tangraphs for 1.22/2.19 syllables used to transcribe TPNWC ?*-aw syllables:

金+ 敖 ?*nggaw (Pearl in the Palm 23.4)

(all other examples are from The Forest of Categories; 1.22/2.19 tangraphs were not used in transcriptions in Cixiao zhuan)

熬 and 奡 ?*nggaw

as well as

末 ?*mbO (why not use an -o tangraph to transcribe this? There is no lower mid vowel O in Gong's Tangut reconstruction, so a foreign -O would have to be approximated by a Tangut upper mid -o or a low -a.)

納 ?*nda

Next: How did I try to account for apparent transcription mismatches in 1996? And how would I account for them naw?

9.17.1:03: As far as I know, there are no Tibetan transcriptions of 1.22/2.19 syllables or any Tangut transcriptions of Sanskrit using 1.22/2.19 tangraphs. TPNWC is the only type of external evidence for this rhyme.