Of armaments and Old Sinitic reconstructions, part 6

« previous post | next post »

From March through July of 2016, we had a long-running series of posts comparing words in Indo-European and in Old Sinitic (OS),  See especially the first item in this series, and don't miss the comments to all of the posts:

Today's post is not about a sword per se, but it is about an armament for parrying sword thrusts.  It was inspired by seeing the following entry in Paul Kroll, ed., A Student's Dictionary of Classical and Medieval Chinese (Leiden: Brill, 2015), p. 104a:  fá 瞂  pelta; small shield — Middle Sinitic bjwot.  I asked Paul where he got that beautiful word "pelta", and he replied:  "One of the benefits of my early classical studies. I got it from Vergil, but it’s originally Greek."

Because I previously was familiar neither with "fá 瞂" nor with "pelta", my eye lingered on that entry, and within a couple of seconds I was struck by the comparable sounds of the two words:  Middle Sinitic (MS) bjwot and "pelta".  Curious about the resemblance in sound and meaning of two short, practical terms, neither of which I had encountered before, I decided to look into them a bit more deeply.

From Wiktionary, I learned that a pelta is "A small shield, especially one of an approximately elliptical form, or crescent-shaped."  English acquired the word from Latin ("a shield", which derived it from Ancient Greek πέλτη (péltēshield).


From Ancient Greek πέλτη (péltē).


(Classical) IPA(key)/ˈpel.ta/[ˈpɛɫ.ta]


pelta f (genitive peltae); first declension

1. a small crescentshaped shield of Thracian design.

A peltast carrying a pelte shield (pelta)

Now that I could see with my own eyes what a nicely decorated pelta looked like more than two thousand years ago and was captivated by the stylish garb of the peltast wielding it, I was hooked and had to find out more about it.

William Smith, LLD. William Wayte. G. E. Marindin, ed., A Dictionary of Greek and Roman Antiquities (Albemarle Street, London. John Murray. 1890):

PELTA (πέλτη), a small shield. Iphicrates, observing that the ancient CLIPEUS was cumbrous and inconvenient, introduced among the Greeks a much smaller and lighter shield, from which those who bore it took the name of peltastae [EXERCITUS Vol. I. p. 776]. It consisted principally of a frame of wood or wickerwork (Xen. Anab. 2.1, § 6), covered with skin or leather, without the metallic rim. [ANTYX]. (Timaeus, Lex. Plat. s. v.) Light and small shields of a great variety of shapes were used by numerous nations before the adoption of them by the Greeks. The round target or cetra was a species of the pelta, and was used especially by the people of Spain and Mauretania. [CETRA] The pelta is also said to have been quadrangular (Schol. in Thuc. 2.29). A light shield of similar construction was part of the national armour of Thrace (Thuc. 2.29; Eurip. Aloes. 498, Rhes. 410; Max. Tyr. Diss. 19.1, 23.2) and of various parts of Asia, and was on this account attributed to the Amazons, in whose hands it appears on the works of ancient art sometimes elliptic, as in the bronzes of Siris (woodcut, p. 79), and at other times variously sinuated on the margin, but most commonly [p. 2.364]with a semicircular indentation on one side ( “lunatis peltis,” Verg. A. 1.49011.663). Varro, L. L. 7.43, compares this to the ancile. [SALII] A vase fragment in the British Museum (No. E 793) shows clearly the form and construction of the lunata pelta; two Persians exhibit the two sides of the shield.

Peltae, from a vase in the British Museum.

The Spanish Wikipedia has a good article on pelta that includes one of my favorite objects from antiquity, this gold comb preserved in the State Hermitage Museum (St. Petersburg), which was found in a kurgan (burial mound) at Solokha, Dneiper River region on the northern Black Sea:

Here we see another crescent pelta in action.  What is most suggestive about this piece is that the warriors are Scythians, who flourished from the 9th c. to the 1st c. BC in the western and central Eurasian steppes, but ranged all the way to East Asia, with their influences extending far into what is now South China.  This puts them in the right time and right place to be involved in the possible cultural transmission of the peltae which we know they employed.  The Scythian languages belonged to the Eastern branch of the Iranian languages, so we shall keep them in mind when below we go looking for possible cognates or borrowings of "pelta".

For those who wish to get a better idea of who the Scythians were, what they looked like, how and where they lived, etc., I highly recommend the wonderful exhibition, "Scythians:  warriors of ancient Siberia", which is being shown at the British Museum from 14 September 2017 – 14 January 2018.  If you cannot make it in person, there is an excellent catalog with the same title as that of the exhibition:  Scythians: warriors of ancient Siberia.

Before adducing late Classical and Medieval languages that may be operative in our search for possible links between the eastern and western manifestations of this particular type of shield, we would do well to pay attention to the origins of the word in Greek.  This is what Frisk has to say in his Griechische Etymologishces Wörterbuch, p. 501:

Seems likely it derives from IE words for covering or blanket (but poss. related to a stretched skin, like Latin pellis [skin], which may have a different IE root in *pel).

We would do well, however, to take into account these off-the-cuff remarks by Don Ringe:

I think Frisk's caution is very much in order.  Of the various words for weapons and pieces of armor, not one clearly goes back to PIE–and that's what you'd expect, because weapons technology was constantly being upgraded even in the bronze age (the only difference being that "constantly" probably meant "every few generations" instead of "every few years"), and new things are usually denoted by new words.  Under pélma Frisk lists a grabbag of words with divergent meanings, all beginning with *pel- (which could be a root, but there's no verb with such a root–'approach' is *pelh2-) but with all sorts of apparent suffixes.  The only really reconstructable word is *pel-n- 'skin', and that's Germanic and Italic only, so far as I can remember.  Not good enough.

In general, it's a bad idea to work with roots unless there's an actual verb attested somewhere.  What gets transmitted from generation to generation is *words*, so those should be the basis of etymology.  In the case of a basic IE verb, the root *is* the verb; otherwise roots are mostly too abstract.  The only exception I can think of is the Caland system.

Jumping from the earliest layers of the history of "pelta", I would like now to examine mostly medieval words for shields (especially smaller ones) that circulated in Central Asia.  This is by no means to say that I believe any of the particular words mentioned here may have served as the basis for a borrowing into Sinitic as fá 瞂 (MS bjwot), though some of them might be reflexes of earlier forms that may have been involved in the transmission.

I asked colleagues who are specialists in premodern Central Asian languages what words they know of for "shield" in the languages they study.  Their replies follow herewith.

Nicholas S-W:

I know Sogd. ptsāδ, MP/Parth. ispar and magind, Man. MP also maginn (Semitic LW).

Hiroshi Kumamoto:

Middle Persian has spar (Mackenzie Pahl. Dict. q.v.), which goes back to OP *spara- (spara-barai in Hesychios; see LSJ).

Peter Goldin:

You might want to look at H.W. Bailey’s Dictionary of Khotan-Saka (CUP, 1979, pb printing: 2010): p. 266: baṭha- “cuirass” – from *varθra-, cf. Osset. (Digor) ŭart “shield”, Zor.P. vartik, gurtik, gurtakh “defensive armor"

1. 305: be̬sa- “shield” from var- “to cover”
2. 228: pāḍaka “covering, envelope, missive document” < *partaka. ZorP pltkˈ *partak “covering”, NPers. pardah “covering , veil”, ZopP. spar “shield”  NPers. sipar

see also p. 65 kaucāha̬ra “covering”  and explanations noted there.

Martin Schwartz:

Offhand, for Central Asiatic Iranian I only know a word for 'shield' whose correct meaning I supplied long ago, Sogdian, Man. <pts'd>, S. <pts'∂>, both = /ptsâ∂/ < OIr. *patisâda- ' a covering'.  Man. MPers.
<mgynd> /magind/  'shield' vel sim. is from an Aramaic. form with *-nn-, of which *-nd- is a dialectal variant to be expected in  a MPers. borrowing.  I wonder if MPers. spar rightly mentioned by Professor Kumamoto is not from the same root as Pers. parda 'curtain' (s- movable? or more likely < *us-in Hesychius' time, 5th or 6th cent. CE?).

As to Sogd. /ptsâ∂/'shield' (also Christian in Estrangela script as /pts'd/).

OIr. pati is a preverb often indicating 'facing, counter to, again, against', cf. the functions of Lat. re-.The root is *sad = Skt. chad 'to cover'  (not in Chung, Dict. Ir. Verb).The *-sâda- part is a usual noun formation from the root. Iranian cognates include Avestan sâ∂antî-, in context 'who wears reeds? and s.s and leather', ergo a (protective??) garment of some sort'; with  preverb, Pashto psôl- 'to wear'.

*pati-sâda- > 'shield' as *'that which covers someone against something'

Btw, as for the Aram. *maginn- cf. Heb. mVginnâ(h) 'shield', Sem. root g-n-n 'to cover'; inner-Aram. geminate nasal dlalectally developing stop, and reflected by MPers. $ambad 'Saturday,

gund 'military troop', $ambalîd 'fenugreek', and zandîk 'manichæoid heretic' < zaddîq-, *zandîq'

self-designation 'righteous' (NOT 'adherent of the Zand', as still commonly found in Iranistic lit. ).

$ = s-hacek (sh), ^ = macron.

It occurred to me that in the obscure Av. passage with sâ∂aiiaNtî-, the word which looks like 'reeds' can probably mean 'canes', which figure in some kinds of armor and shields.

Marcel Erdal:

Kalkan was a light shield, and then there was tura, a shield one could stand behind. Kalkan was also borrowed into contact languages. Iranian terms (starting with Avestan spâra) do not seem to have had a chance with the warlike Turks (note Turkish siper, though). (There also was vrėdra, related to English ‘guard’.)

Mehmet Olmez:

kalkan (from Clauson) could be an answer for your question?

kalkan ‘shield’; s.i.a.m.l.g. except NE(?). L.w. in Mong., Pe., etc. Doerfer III 1518. Uyğ. xıv Chin.—Uyğ. Dict. ‘shield’ kalkan R II 254; Ligeti 161: Xak. xı kalkan ‘shield’ (al-turs) in one of the two dialects (al-luğatayn) Kaş. I 441 (verse); kalkaŋ al-turs dialect form (luğa) of kalkan III 386; o.o. of kalkan II 356, 19; III 82 (yapın-); 221 (tura:): KB 4263 (tayaklık): xııı (?) Tef. kalkan ‘shield’ 196: xıv Muh. al-turs kalka:n Mel. 71, 8; Rif. 173:Çağ. xv ff. kalkan  sipar ‘shield’ San. 275v. 27 (quotn.): Xwar. xııı (?) ditto Oğ. 38, 98: xıv ditto Qutb 129: Kom. ditto CCG; Gr.: Kıp. xııı al-turs kalka:n Hou. 13, 15: xıv kalkan al-micann ‘shield’ İd. 74: xv al-daraqa‘leather shield’ kalka:n, with a sound between -k- and -ğ- but nearer to -k- Kav. 64, 1: daraxa kalkan Tuh. 15b. 7: Osm. xıv ff. kalkan noted in phr. TTS I 406; II 568; III 398. (Clauson 621 a)

I just remembered tura. Here is TURA from Clauson 531a (and may be there is more information at Doerfer II 958):

2 tura: basically ‘something to shelter behind’ used both for permanent fortifications, and for portable ‘breastworks’ which could be moved about and fixed temporarily to the ground. A l.-w. in the first meaning in Mong. (Kow. 1879, Haltod 432) and Pe. and other languaes, see Doerfer II 958; it survives in most NE languages R III 1446, and Khak. where the meaning has attenuated, through ‘stockade, fortified village’ to ‘town’ and even ‘house’. Xak. xı tura: kalkan al-turs wa’l-daraqa wa kull mā tasattara bihi’l-racul mina’l-aduww ‘brastwork, shield, and anything that a man shelters behind from the enemy’ Kş. III 221; o.o. II 356, 19 (kalkan tura: daraqatuhu wa tursuhu); III 106, 14 (?, text perhaps corrupt): KB kara baylıkın kıldı özke tura ‘he made the wealth of the common people a protection for himself’ 256; (some men expose themselves to swords and battle axes in battle) kayusı turada yuluğda karır ‘some grow old behind breast works in security (?, or as hostages)’ 1736; o.o. 5263 (ordu:), 6434: xııı(?) Tef. bustānnuŋ turası  ‘a garden wall’ 312: xıv Muh. al-qal‘a  ‘fortress’ tu:ra: Mel. 75, 15; Rif. 179: Çağ. xv ff. tura  (‘with -u’) ‘a shield (kalkan) the height of a man which soldiers hold in front of them in battle and fight behind’ Vel. 203 (quotns.); tura ‘iron rods and plates of iron which they fasten together with chains and hooks on the day of battle and make into a line of defence (isār-i laşkar) behind which they stand to fight’ San. 173r. 16 (quotns.).

Stefan Georg:

On the last mentioned term, consult Doerfer „Russ. tury „Schanzkörbe“ in: Zeitschrift für Slavische Philologie 29, 1961, 288 – 301 – not exactly a portable shield, of course.

The Tk. kalkan was of course rather successful as a loan, as evidenced by Mongolian  qalq-a.

In Middle Iranian lgg. we have Khotanese be’sa- ‚shield‘, from the same root which was mentioned by M. Erdal (Old Khotanese baṭha- ‚cuirass', Osset. wart ‚shield‘ (noun)

Soghdian has pts’δ (patsāδ), with a different background < *sa:d- ‚cover, protect‘, with a preverb, to this root belongs also Npers. chador ‚veil‘ (but via indirect transmission, involving Indic/Skt., rather than pure Iranian paths).

Middle Persian has also mgyn (/maginn/) and Parthian has mgynd – which is Aramaic, cf. Syriac magn/mgenna:, acc. to Payne-Smith 2006 „clypeus“, which would be a kind of small round bronze shield, if the gloss is to be taken literally.

Khwarezmian seems to have 'βnyk ‚shield‘ (with several compounds), which might be called like that because it had military signs on it, „banner shield“ or the like (but this is far from certain)..

For Tocharian, a term for this doesn’t seem to be recorded – peaceful Tocharian monks.

So much for pelta and its congeners, from Greek πέλτη (péltē) through various Iranian terms.  I don't think that Tocharian kalkan is relevant, though tura might conceivably be for dùn 盾 ("shield") — I'm on the road as I write this paragraph, so I don't have Schuessler's dictionaries to check the OS reconstruction, but Zhengzhang Shangfang has /duənX/.

Now what can we say for fá 瞂?  Not much, I'm afraid, and that would explain why I wasn't familiar with it before beginning the research for this post.

It is noteworthy that the fá 瞂 (MS bjwot) is associated with the Rong 戎, an ancient people of the west who were famous for their military prowess.  Indeed, their name stands for "arms", "armaments", and "military affairs".

The first we hear of the fá 瞂 (MS bjwot) is in the Shījīng 詩經 (Poetry Classic), which might date to the 6th c. BC or even a bit earlier (though undergoing editing several centuries later), where it appears in the third line of the third stanza of the "Qín fēng, 'Xiǎo Róng'" 秦風小戎 ("Airs of Qin, 'Lesser Rong'").  It is interesting that here it is written as fá 伐 ("cut; chop; attack; strike; cut down; fell"), clearly a homophonous character borrowing for whatever word they were trying to write.

The fact that fá 瞂 shows up later as a newly constructed graph is instructive, since it is composed of the character for dùn 盾 ("shield") on the left as its semantophore and bá 犮 ("pull up") on the right as phonophore.  Further evidence that fá 瞂 is merely an attempt to record the sound of a borrowed word or a word that was being passed around in the oral realm without a fixed written form is provided by the fact that this same word is also tellingly written with bá 犮 ("pull up") as the phonophore on the right, but gé 革 ("leather" [N.B.!]) on the left as semantophore.  Equally revealing is that the same word may also be written with dùn 盾 ("shield") on the left as semantophore and fá 伐 ("cut; chop; attack; strike; cut down; fell") on the right as phonophore.

Schuessler's OS reconstruction for fá 伐 is *bat and his OS reconstruction for bá 犮 is *bât.

As many of my historical phonologist friends never tire of telling me, we should not take OS reconstructions too literally.  They are not meant to reflect the actual pronunciation of Sinitic as it would have been spoken circa 600 BC, but rather are to be thought of more as formulaic representations of phonological relationships and principles.  They are only very rough, theoretical approximations of OS pronunciations.  (Some historical phonologists do take their OS reconstructions literally, but it is my impression that most do not.)

Aside from simply working out the history of early trans-Eurasian cultural exchanges, my purpose in pursuing the linguistic aspects of these inquiries is to try to provide some solid phonological pegs upon which to hang future OS reconstructions.  Since I have hundreds more Sino-IE comparanda linked by sound, meaning, and historical, archeological, visual, and / or material evidence, I'm fairly confident that some of them will contribute to the investigations of historical linguists who have an interest in cross-cultural contacts.  If all goes well, I may be able to present two or three more within the next few weeks before classes begin again.  Meanwhile, the conceptual framework for this type of research may be found in these books:

Victor H. Mair, ed., The Bronze Age and Early Iron Age Peoples of Eastern Central Asia (Washington, D.C.: Institute for the Study of Man Inc. in collaboration with the University of Pennsylvania Museum Publications, 1998).  2 vols.

J. P. Mallory and Victor H.Mair,The Tarim Mummies: Ancient China and the Mystery of the Earliest Peoples from the West (London:  Thames & Hudson, 2000).

"Early Indo-Europeans in Xinjiang" (11/19/08).

Victor H. Mair, ed., Contact and Exchange in the Ancient World (Honolulu:  University of Hawai'i Press, 2006).

Victor H. Mair, "Language and Script: Biology, Archaeology, and (Pre)History," International Review of Chinese Linguistics, 1.1 (1996), 31a-41b.


[Thanks to Ralph Rosen, Peter Lorge, David Graff, Judith Lerner, Petya Andreeva, and the other colleagues whom I have mentioned by name in the body of the post.]


  1. Victor Mair said,

    December 23, 2017 @ 10:05 pm

    From Hiroshi Kumamoto:


    Martin Schwartz wrote:

    It occurred to me that in the obscure Av. passage with sâ∂aiiaNtî-, the word which looks like 'reeds' can probably mean 'canes', which figure in some kinds of armor and shields.


    See the Dec. 20 post on téng jiǎbīng 藤甲兵 ("soldiers [wearing] cane / rattan armor") here:


    The story referred to there from Sānguó yǎnyì 三國演義 (Romance of the Three Kingdoms), Chapter 90, is, according to the translation of Charles Henry Brewitt-Taylor, as follows:


    His warriors wear rattan armor. This rattan grows in gullies, climbing over rocks and walls. The inhabitants cut the rattans and steep them in oil for half a year. Then they are dried in the sun. When dry they are steeped again, and so on many times. Then they are plaited into helmets and armor. Clad in this, the men float across rivers, and it does not get wet. No weapon can penetrate it. The soldiers are called the Rattan Army.


  2. Victor Mair said,

    December 24, 2017 @ 7:20 am

    Another publication that would fit well in this series of posts is:

    Robert S. Bauer, “Sino-Tibetan *kolo 'Wheel',” Sino-Platonic Papers 47 (Aug. 1994), 1-11. (free pdf)


  3. Paul R. Goldin said,

    December 24, 2017 @ 8:36 pm

    Two comments.

    "Further evidence that fá 瞂 is merely an attempt to record the sound of a borrowed word or a word that was being passed around in the oral realm without a fixed written form is provided by the fact that this same word is also tellingly written with bá 犮 ("pull up") as the phonophore on the right, but gé 革 ("leather" [N.B.!]) on the left as semantophore.  Equally revealing is that the same word may also be written with dùn 盾 ("shield") on the left as semantophore and fá 伐 ("cut; chop; attack; strike; cut down; fell") on the right as phonophore."

    The same is true of virtually any Chinese word, whether Sino-Tibetan or borrowed. Show me a word that has, throughout history, been written in one and only one way! So the fact that the word is written in different ways does not, in and of itself, tell you that it must have been borrowed from another language.

    Better evidence might be the fact that neither 伐 nor 犮 normally means "shield." (If anything, they mean the opposite: "strike" and "uproot.") But this is also far from sufficient to establish that the word was borrowed. (Another word undoubtedly related to 伐 is 發, but of course that doesn't normally mean "shield" either.)

    So … maybe the word really is borrowed from some other language, but this can't be determined by analyzing the graph(s) used to write it.

    "[OS reconstructions] are only very rough, theoretical approximations of OS pronunciations."

    This statement is misleading because reconstructions are phonemic. Treating them as though they were pronunciations fosters serious confusion. And the phrase "OS pronunciations" reflects a category mistake, doesn't it? As I understand it, the reason why you use the word "Sinitic" instead of "Chinese" in these contexts is to emphasize that Sinitic is a language family, not a single language. So then "Sinitic" cannot have a single set of pronunciations. Maybe you mean Proto-Sinitic, i.e. the single language (if there was one) from which all the Sinitic daughter languages derive?

  4. Chau said,

    December 25, 2017 @ 8:49 pm

    I thoroughly enjoyed reading this magnificent post.

    Regarding 盾 'shield' I might add a bit about an historical aspect of its pronunciation. Currently it is pronounced dùn in MSM and tún in Taiwanese. The differences in tones (去聲 in MSM vs. 上聲 in Tw) and the initial (d- in pinyin vs. t- in POJ) between the two topolects pare greatly when it is compared with what is given in Kangxi 康熙字典. For the meaning of 'shield', historically, the 切韻 system indicates 食尹切 or 豎尹切 and a 上聲 tone. This would lead to sún in Tw. or shŭn in MSM. From the Classical times all the way down to the Early Qing, it was presumably always this way. Only in a special usage – a personal name 晉卿趙盾 – is it pronounced something like dùn in MSM (tún in Tw). Something must have happened between Kangxi's time and now that allowed a minor alternate reading dùn to usurp the long-established reading shŭn for the 'shield' meaning.

    If we look at words with 盾 as the phonophore, we see that its pronunciation vacillates between t- and s- initials (in Taiwanese):

    t-: 楯 tún, 輴 tun, 遁 tūn, 腯 tún/thún;
    s-: 偱 sûn, 循 sûn, 䋸 sûn, 揗 sûn/sūn.

    This vacillation may have facilitated the initial sound shift to the current reading of 盾 for the meaning of 'shield'.

    Taking the historical reading sún for 盾, then it is quite straight forward to connect it with Latin scūtum 'an oblong shield, made of wood covered with leather'. With sk- turning into sh-/s- (Cf. Danish waskmaskine and English wash machine) and -t into -n by homorganic nasalization, we have:

    L. scūtum > scūt- > Tw sún / MSM shŭn for 盾 'shield'.

  5. ~flow said,

    December 26, 2017 @ 8:50 am

    @Paul concerning phonemic vs phonetic reconstructions, single vs multiple language comparisons—I think you're throwing hairballs at the wrong target.

    'Phonemic' can mean a number of things, depending on the author and the subject of study, but generally what people come up with and label as 'phonemes' will be roughly what can be paraphrased as 'group of sounds that are somewhat similar in nature and linked by historical / synchronic processes or morphophonemic alternations'.

    Secondly, I understand those historical reconstructions of earlier stages, at least in the case of Chinese / Sinitic, as being less of 'inventories of sounds that were used in a single language at a given point in time', but more as 'inventories of systematic sound classes that hint at the phonologies as were prevalent among speakers of more or less closely related variants of a language, such that these inventories and their internal relationships suffice and are phonetically plausible to explain the data points (historical and present-day) that we can observe' (e.g. modern Korean /pul/, Japanese /butu/, Mandarin /fo/, Sanskrit /buddha/ point to a reconstruction where that word had a bilabial plosive in initial and an alveolar stop in the final in some earlier stages of Chinese).

    I do not think that such reconstructions should be taken to mean necessarily that any individual speaker ever used the full gamut of distinctions as are present in the system (impossible if you look at e.g. Karlgren's vowel systems); rather, I take these projections (like the classical rime tables) as amalgams of several 'dialects' that try and show what can, was, and should be differentiated in order to explain the data (or write poems such that they work for speakers from far flung quarters).

    This, I imagine, is not unlike a cross-dialectal sound system of modern English where you might come up with as many Lexical Sets (https://en.wikipedia.org/wiki/Lexical_set) as there are different groupings of vowel phonemes in the many chief varieties of (Southern British, Scottish, American, Australian, Indian…) English.

    So then "[OS reconstructions] are only very rough, theoretical approximations of OS pronunciations." becomes a perfectly cromulent statement.

  6. Chris Button said,

    December 26, 2017 @ 9:11 am

    @ Flow

    'Phonemic' can mean a number of things..

    I would agree with that sentiment since it really depends on how deep your phonemic analysis goes. For example, Baxter & Sagart reconstruct a phonemic /e/ vowel but I would say that is too "surfacey" for want of a better word and definitely overly prescriptive. For me the phonemic representation would be /ʲa/ or /ja/ with their "e" sometimes surfacing due to assimilation, but this should not ignore all the other surface phonetic representations that may result too at any synchronic stage (e.g. /ia/ /ie/ /je/ etc etc)

    @ Victor Mair

    Another publication that would fit well in this series of posts is:

    Robert S. Bauer, “Sino-Tibetan *kolo 'Wheel',” Sino-Platonic Papers 47 (Aug. 1994), 1-11. (free pdf)

    It's interesting to see how the word represented by 車 was completely appropriated in Chinese as something native regardless of its Indo-European origin. Sagart (1999) rightly points out a word family with 車, but I don't see this as any reason to doubt the original external origin discussed by Bauer. I would view 牙 as a similar case due to its Austroasiatic origin, although I'm aware it has been proposed that the loan direction was the other way.

  7. Paul R. Goldin said,

    December 26, 2017 @ 10:28 am

    One response to "~flow," then I have to move on to other things.

    "I do not think that such reconstructions should be taken to mean necessarily that any individual speaker ever used the full gamut of distinctions as are present in the system (impossible if you look at e.g. Karlgren's vowel systems); rather, I take these projections (like the classical rime tables) as amalgams of several 'dialects' that try and show what can, was, and should be differentiated in order to explain the data (or write poems such that they work for speakers from far flung quarters)."

    First, sorry, you're wrong about Karlgren. Karlgren made it perfectly clear that he was trying to reconstruct real pronunciations. Just read his introduction to Grammata Serica Recensa, where he leaves little doubt that he was trying to establish phonetic values. In fact, I chalk up the indefatigable Sinological misconceptions about modern reconstructions to scholars who think Karlgren still represents the state of the field.

    Second, suppose you're right: suppose it's true that reconstructions are "amalgams of several 'dialects' that try and show what can, was, and should be differentiated," just like rime tables. It's an interesting idea (and credit for it should go to Jerry Norman and David P. Branner). But look at the consequence of what you're saying: if you're right, then such reconstructions are assuredly not pronunciations. The phrase "Old Sinitic pronunciations" makes about as much sense as "Old Germanic pronunciations."

    It's been fun, but I have to move on. I think we can agree that (a) 瞂 is an interesting word, (b) we don't know where it comes from, and (c) the shape of the graph(s) used to write it won't tell us the answer.

  8. ~flow said,

    December 26, 2017 @ 5:32 pm

    @Paul I think in fact it was Karlgren himself who dismissed those new-fangled phonological / phonemic trends of the field, but that doesn't preclude that whatever he came up with isn't better understood as a 'phonological amalgam' of several dialects, e.g. when he proposes to distinguish, for his Ancient Chinese:

    kən, ki̯ĕn, ki̯ən, ki̯əm, kuən, ki̯uĕn, ki̯uən, ki̯ʷĕn, kien, ki̯ɒn, kʻi̯än, kan, kân, kuân, kʷan, gʻi̯ʷän, ngi̯ʷɒn, ɣiʷen, kâng, ki̯ang, kång, kʷâng, kʻi̯ʷang, kung, ki̯ung, ki̯ʷong, ki̯ak, kåk, ɣʷɒng, kʻi̯ʷäng, χi̯ʷɒng, χi̯ang, ɣiʷeng…

    it's hard to imagine a single individual's speech.

    Sure others have come up with other, hopefully simpler solutions.

    You say, "The phrase "Old Sinitic pronunciations" makes about as much sense as "Old Germanic pronunciations.""—but sure that does make sense. Today we can very well go and compare 'sounds' like p, b across languages and sure is that valid. If we don't do that it'd make no sense to state that e.g. "/ð/ and /θ/ are found in English but not in German". Sure you can abstract phonetics out of phonology and treat it all as a glass-beads game of logic and relationships, but you then foreclose any way to take phonetic plausibility into account, too.

    That is to say that such an ethereal phonological theory which denies that I can compare, say, pre-modern Chinese /ki-/ with, say, early Romance /ki-/ *on the grounds that they sounded very much alike*—also denies us to say that the respective historical developments of these sounds share a common auditory and articulatory base that makes their parallel evolutions plausible.

  9. Jonathan Smith said,

    December 26, 2017 @ 8:49 pm

    "We should not take OS reconstructions too literally […] [they are] formulaic representations of phonological relationships and principles."

    This is not entirely wrong… but mostly it has become code for "don't take my reconstructions too literally wherever they're phonetically preposterous; in those cases the IPA representations (cough cough) are just symbolic and shit."

    Phoneme systems are subconscious — insidious — analyses on the gestalt that hits our ears. The 36 zimu of traditional yinyunxue, to take one example, are obviously more contrived but nonetheless parallel sorts of analyses for which "phonemes" is thus a reasonably adequate (certainly best available) characterization. Ditto for other traditional categorizations. So yes, Old Chinese reconstructions feature wrong stuff and fake stuff, but at their core are trying to be phonemic. If one wants to be formulaic, one might consider purely symbolic devices — in OC, capital "A", the dash "-", Schuessler's A/B diacritics, etc., are examples.

  10. Victor Mair said,

    December 27, 2017 @ 5:58 am

    Bernhard Karlgren (1889-1978) was a pioneer in the reconstruction of Old Sinitic. When I was talking about historical phonologists who do not take such reconstructions literally as representing speech sounds more than two millennia ago, I was thinking of contemporary scholars such as W. South Coblin and David Prager Branner, and I'm sure their views would not differ much from that of the late Jerry Norman (1936-2012).


    "Better evidence might be the fact that neither 伐 nor 犮 normally means 'shield.'"

    [VHM: I said as much in the o.p.]


    "I think we can agree that (a) 瞂 is an interesting word [VHM: that's what this post is all about, and so far everyone who has read the post and communicated with me concerning fá 瞂 {pelta; small shield — Middle Sinitic bjwot} thinks that it is well worth discussing], (b) we don't know where it comes from [VHM: well, goodness, that's what we're working at], and (c) the shape of the graph(s) used to write it won't tell us the answer [VHM: couldn't agree more; that's why I'm focusing on sounds, not the graphs, as explained in the o.p.]."


    "category mistake / error"

    [VHM: anti-comparativists are fond of levelling that charge when confronted with empirical data — e.g., archeological, visual, material, genetic, biological… evidence — that they don't know how to process]


    "One response to ~flow, then I have to move on to other things."

    "It's been fun, but I have to move on."

    [VHM: I'm here for the duration.]

  11. Chris Button said,

    December 27, 2017 @ 10:48 am

    I think this will be a tough one to prove conclusively, but here are some random ruminations that might trigger some thoughts by other people:

    – If "pelta" is indeed from PIE *pel- "skin" with a suffix (as Calvert Watkins treats it) then the presence of 革 rather than 盾 noted by Prof. Mair in a variant form of 瞂 is very interesting.

    – Looking at other Chinese borrowings where one character is being used for two syllables, we can probably ignore the final "a" to leave us with something like "pelt". The final cluster -lt is of course a problem since in PIE all the sonorants could pattern as syllabic bases to give clusters whereas in OC only the glides j and w, and depending on its analysis rhotic r, could do so. Since l and t are both coronal, perhaps OC just went with one coronal to cover both.

    – Final -l in OC was not a "dark l" /ɫ/ since it merged with final -j early on (most OC reconstructions unfortunately ignore its existence completely). If the -l in "pelta" was dark (is there a way of confirming this?) then this could perhaps have contributed to the lowering of the vowel to "a" in OC and its segmental omission in favor of -t

  12. Victor Mair said,

    December 27, 2017 @ 1:31 pm

    @Chris Button

    Thank you very much for your good comment. I was hoping that someone would notice that (about the leather semantophore).

  13. Jake said,

    December 27, 2017 @ 4:08 pm

    I don't know as many as a single thing about Middle Chinese, how should I think of "bjwot" as being pronounced?

  14. David Branner said,

    December 27, 2017 @ 9:04 pm

    It's long past time to retire the Karlgren model of Chinese historical phonology. (Dare I hope that people will stop reprinting his reconstructions, too — even in the LFK or B-S recensions?) I'll be presenting about this at the National AOS in March, in case anyone aboard here wants to come for a wrassle.

  15. ~flow said,

    December 28, 2017 @ 6:18 pm

    @Jake if bjwot was my analysis for *modern* Chinese I'd intend -jw- (or exchangeably i for j, u for w) to stand for a single vowel or a diphthong (depending on the overall syllable in question) that combines the features (gestures) 'high', 'palatal/front', 'labial/rounded'; all of [ü] (as in French "tu"), [iu] (as in E. "you") and [ui] (as in E. "we") fill that bill.

    (For Mandarin the choice to view [ü] as a combination of /i/ and /u/ is tempting because only one of /i/ and /u/ may co-occur in one syllable; thus, /ni/, /nu/, /nau/, /gua/, /guai/ are all legal, but */-iai/, */-uau/ and crucially */üai/, */üau/ are not.)

    If I had to pronounce bjwot, I'd say [büot] or [biuot] (with /i/ as a glide) ; [buiot] (with /u/ as a glide) seems unlikely (to me for reasons of sonority, syllable complexity).

  16. Paul R. Goldin said,

    December 28, 2017 @ 7:41 pm

    Against my better judgment …

    "If I had to pronounce bjwot, I'd say [büot] or [biuot]"

    I'll bet a billion dollars it wasn't büot or biuot, and guess what–I'll get to keep my billion dollars, because nobody can go back in time and record a native speaker in Tang-dynasty Chang'an. That's the whole point. Historical linguistics cannot reconstruct pronunciations, so linguistic reconstructions don't even try. It's a very fundamental mistake (a category mistake) to misinterpret them as though they were pronunciations.

    Jake, don't try to pronounce bjwot. It's a perfect illustration of what I was saying, because (a) it's based on Karlgren's reconstructions, which are hopelessly obsolete (because they were based on invalid assumptions), and (b) see the preceding paragraph. (The -w- is based on a misunderstanding of a phenomenon called hekou 合口.) As David Branner said a couple of posts ago, dictionaries shouldn't mislead their readers by perpetuating monstrosities like bjwot.

    In fact, the very name "Middle Chinese" is misleading. It's not like Middle English or Middle High German because we have texts in those languages that are written in a reasonably transparent phonetic script. Middle Chinese is a construct.

    And this comment reminds me of something else I meant to say:

    "if bjwot was my analysis"

    I think it's pretty tacky to participate in a conversation like this anonymously. If you're going to say something, take responsibility for it.

  17. Chris Button said,

    December 29, 2017 @ 12:26 pm

    @ Paul R. Goldin

    Are "cot" and "caught" or "poor" and "pour" homophonous pairs? For some speakers they are, for others they are not. If we were trying to make a synchonic statement about today's English looking back from a couple of thousand years in the future, should we reconstruct them as the same or not? Therein lies one of the major problems with the reconstruction of Old Chinese.

    Further to my first post above, regardless of whether people like Karlgren or Baxter & Sagart claim to be reconstructing surface phonetic realisations or underlying phonemes, the problem is exactly the same when they insist that something should unequivocally rhyme a certain way.

    The statistical data on variations between British and American English pronunciation in the Longman Pronunciation Dictionary would have been incredible useful to our hypothetical future selves 2000 years from now if no recordings existed for some reason. However, it would be incredibly foolhardy of us to use the weight of a statistical majority, however big or small, to make a definitive statement in no uncertain terms for all speakers about whether "caught" or "cot" were homophones. As I have said previously on Language Log, the rhyme statistics in Baxter's Handbook of Old Chinese Phonology are an incredibly valuable contribution to scholarship (as is the entire work), but his use of the data to reconstruct immutable Old Chinese rhymes (i.e. there can be but one surface phonetic realization for all speakers) is fundamentally misled in my opinion.

  18. Wolfgang Behr said,

    December 29, 2017 @ 3:52 pm

    If you'd like to push these Central Asian 'shield' comparisons even further, you might be tempted to think of a relationship between Tk. kalkan 'shield' and Chinese 干 OC (BS) *k!ar 'shield', of course, which is — speaking of orthographic variation — sometimes written as , i.e. with a determinative ("semantophore") ge1 戈 (OC *kw!aj) 'dagger-axe' and a derived phonopohore 旱. Since the xiesheng series of 干 clearly had final *-r, rather than *-n in OC, the contact with Turkic would have to postdate that change for which there are competing datings in the literature. dun4 盾, on the other hand, which is used as a gloss for in the Shuowen, is likely to have come from *-n, rather than *-r in OC. The reason why it would not have been a good match for Tk. tura is that it clearly came from a lateral series (OC *l!unʔ > MC *dwonX). Even if it was borrowed after the hardening of laterals in Han Chinese by a form of Turkic it remains unclear to me why an unvoiced early MC *dw- would have been rendered by Turkic *t-.

  19. ~flow said,

    December 29, 2017 @ 9:05 pm

    @David Branner I'm looking forward to that. I believe I made it clear that I see some issues with Karlgren's analyses.

    @Paul R. Goldin ""if bjwot was my analysis" … tacky ". 'bjwot' as *such* is no monstrosity. It's the packaging of so many similar, yet allegedly distinctive features into the space of single syllables of an allegedly single dialect that is problematic (apart from other details). That Karlgren's reconstructions can be regarded as outdated now doesn't mean we cannot or should not discuss them (without getting personal).

    You say "The -w- is based on a misunderstanding of a phenomenon called hekou 合口"; well, hekou(hu) is all about labiality / lip rounding, isn't it?

  20. Chris Button said,

    December 29, 2017 @ 10:22 pm

    It's long past time to retire the Karlgren model of Chinese historical phonology

    True, although I can think of at least one aspect that while requiring modification seems to have fallen by the wayside in many modern reconstructions. As Pulleyblank quite correctly pointed out, there must have been at least some kind of velar component such as an approximant or fricative (albeit not Karlgren's g but still in someways preferable to a completely open syllable) in words like 來. Otherwise, how does one account for the appearance of a j coda in its Middle Chinese reflex or for the k coda in a word like 麥?

  21. ~flow said,

    December 30, 2017 @ 6:28 am

    @Chris Button

    If I understand you correctly, then one should rather assume 'degrees' and / or 'modes' of lets-call-it 'harmony in rhyme', correct? Such that E. "hat", "bad" is a worse rhyme than "hat", "cat", and also different in nature from "hat", "hot", which is another 'non-optimal, but still possible' rhyme. We'd have to factor in differences between dialects that may or may not merge some of the words in speech that appear in rhyming positions.

    If we were to build a table of Lexical Sets (a la Wells) from such observations (maybe with the objective of giving sets of words for use in lyrics that do not grind on any listener from any of the considered dialect backgrounds), that table would sure look somewhat puzzling for any speaker of just a single modern English dialect. If we'd then go and condense / reconstruct 'one single true' form of speech from that table—we'd do what you call "fundamentally misled" in the above.

  22. Jonathan Smith said,

    December 30, 2017 @ 2:17 pm

    See List, Using Network Models to Analyze Old Chinese Rhyme Data

    For some reason, Old Chinese "reconstruction," or whatever you want to call it in this special case, is a puzzle which proves hard to resist… witness Pulleyblank or indeed Norman, a some-time advocate for focus on comparative work on modern varieties whose various controversial "reconstructive" proposals at the same time now account for most of what is novel about Baxter & Sagart (2014)…

  23. DD'eDeN said,

    December 30, 2017 @ 2:27 pm

    I'm very interested in the most ancient round-shields ever used by early humans. Why? I think that they were the very first shelters, portable, sun-shading, rain-shedding, predator-shielding, prey-hidehunting concave domiciles (dome shields) lifted for entry/exit, only much later becoming permanent structures with doorways. Since I've already explained, I'll give what I think was the name: xyambuangolu. I noticed the term Clipeus Argŏlĭcus. Argolicus/Argos reminds me of arigolu (India: father's/sun bowlboat=coracle = parical/parisal), which I consider the inversion of mongolu (Mbuti: mother's/moon dome hut).

  24. ~flow said,

    December 30, 2017 @ 7:36 pm

    @Jonathan thanks a lot for the interesting link! The page also links to https://shh-mpg.academia.edu/JohannMattisList which in turn contains a lot of material on Chinese, historical linguistics and so on. Thanks!

  25. Chris Button said,

    December 31, 2017 @ 5:28 am

    @ ~flow

    Sort of.

    If we have for example data for a rhyme -wat showing that it rhymes in the majority of cases with other words with -wat but in a minority of cases with words with -at, then a reconstruction of -ot might reflect a common surface assimilation of -wat > -ot but shows no flexibility in being able to capture the rhymes in -at.

    The irony is that people who reconstruct things like -ot in Proto-Tibeto-Burman / Proto-Sino-Tibetan languages often then call on a supposed "vowel breaking" at later stages of the languages when they need to get back to a -wat rhyme as a result of new phonological/phonotactic considerations. Needless to say, such a miraculously clean split of features (also without positing any conditioning environment) is not how such dipthongization processes (aka vowel breaking) generally works!

    Incidentally, Old Burmese is a prime example of this where the script unequivocally demonstrates a whole series of rhymes with -wa- and -ja- (such j and w distinctions with the ablaut variant -ə- were lost due to its lower salience/sonority) such that people have tried to posit unsupportable wholescale "breaking" to account for it in their desire to posit "regular" triangular vowel systems. I'm not sure what Paul R. Goldin's point above regarding "hekou" in Old Chinese is, but perhaps it is connected in some way to this fact?

  26. Paul R. Goldin said,

    December 31, 2017 @ 12:38 pm

    The -w- in bjwot is a phantom because hekou is not contrastive after labial initials, as Yuen Ren Chao demonstrated over seventy years ago:

    "Distinctions within Ancient Chinese," Harvard Journal of Asiatic Studies 5.3-4 (1941), esp. 217-27

    "I'm not sure what Paul R. Goldin's point above regarding "hekou" in Old Chinese is"

  27. Chris Button said,

    December 31, 2017 @ 1:44 pm

    @ Paul R. Goldin

    Just for a brief moment there I thought I might have found a ə/a supporter, but alas maybe not :(

    Thank you for clarifying your point. Since the Middle Chinese form proposed does indeed have "o" rather than "a" for the vowel, the "w" is technically redundant (at least phonemically) and I suppose something like bjot would suffice which is how Baxter&Sagart transcribe it in their "notation". However, unless I'm mistaken, don't Baxter& Sagart use "o" to represent an unrounded /ʌ/ anyway? I would of course prefer Pulleyblank's reconstruction of Early Middle Chinese buat with the infamous yod "j" being replaced by "u".

  28. Chris Button said,

    December 31, 2017 @ 1:59 pm

    If you want the rounded feature to disappear entirely, you will need to derive the Early Middle Chinese form from an Old Chinese "Type-A" syllable such as 茇 (rather than "Type B" in 瞂) where you would simply have bat in Pulleyblank's Early Middle Chinese and also I believe in Baxter &Sagart's Middle Chinese notation.

  29. ~flow said,

    December 31, 2017 @ 2:24 pm

    @Paul R. Goldin that a given phonetic factor is not distinctive according to some phonological analysis doesn't mean it's not there, does it? As in Mandarin 'bo', where more or less of an intermittent -w- is audible across speakers, one can analyze that as /bo/ or else as /bw€/, with '€' being a sign for a mid-vowel that sometimes surfaces as [e] and sometimes as [o] (no narrow transcription used here).

RSS feed for comments on this post