Prakritic "Kroraina" and Old Sinitic reconstructions of "Loulan"

« previous post | next post »

Inquiry from Doug Adams:

As you know I’m working on a review for JIES [Journal of Indo-European Studies] on KT Schmidt’s Nachlass [VHM:  see here].  I need to say something about the name Loulan itself and, not unusually, I’m sinking uncontrollably into the quicksand of reconstructed Chinese. The question arises concerning the first syllable, represented by Karlgren’s character 123b. The modern pronunciation is lóu. Because it is assumed to be the Chinese transcription of the first syllable of the native word Kroraina, one finds, in discussions of Loulan, reconstructions like *gləu or *γləu, with the (unstated) assumption that the *l stands for a yet earlier *r. But, when the name Loulan is not part of the discussion, i.e., in general reconstruction, the initial is just *l– or, earlier, *r– (Schuessler gives OCM * or roʔ [and Late Han (about the turn of the millennium) *lo or lioB]) The Khotanese word referring to Loulan/Kroraina is raurana– and is obviously the same word as the Chinese and, indeed, very probably a borrowing therefrom.         So where does the *gl-/*γl– come from? Or is the Chinese Loulan not a transcription of Kroraina but merely an accidental (partial) look alike?

Any elucidation you can give would be appreciated.

As a matter of fact, as someone who has worked for decades in Eastern Central Asia, where Loulan is located in the northeastern part of that region, I have often pondered exactly this conundrum:  how do we get from Kharosthi / Kharoshthi / Kharoṣṭhī Prakrit "Kroraina" to Sinitic "Loulan" (that's the Modern Standard Mandarin [MSM] pronunciation of the two characters used to write the name:  Lóulán 樓蘭 / 楼兰)?  The initial consonant cluster (kr-) of the Prakrit carried over into Sogdian (krʾwrʾn; in Ancient Letter no. 6) and Uyghur (Kroran / Кроран‎ / كروران).  In his Xīyù shǐdì shìmíng 西域史地釋名 (Compendium of Historico-Geographical Terms of Turkestan [Western Territories] — original English title in the book) (Kaohsiung:  Zhongshan Daxue Chubanshe, 2002), p. 76, Francis K. H. So (Sū Qíkāng) 蘇其康 states that the Greek version of the name is Khaurana, but that this is pronounced as Phraurana.  (It would be good to have the source [Ptolemy?] and details of this interesting information confirmed.  Ptolemy [c. AD 100-c. 170] places Khaurana at 150° 37° 15' on his scale).

Loulan is a fabled name in Chinese history.  It was an ancient kingdom centered on a key oasis city of the Silk Road and was known already by the 2nd century BC.  The city was deserted in the 4th c. AD during a severe dry period that also had an enormous impact in the East Asian Heartland (EAH), e.g., with the collapse of the Western Jin Dynasty (266-316) and a move south to establish the Eastern Jin Dynasty (317-420).

Situated just to the northeast of the desiccated Lop Nur (coordinates 40°31′39.48″N 89°50′26.32″E), the remains of the city have been discovered and extensively studied by archeologists during the 20th century, starting with the Swede Sven Hedin (1865-1952) and then followed by the Hungarian-British Marc Aurel Stein (1862-1943), both of whom located the remains of the ancient city and conclusively identified it by its Prakrit and Chinese names.

Among the spectacular finds from this area is a female mummy known as the "Beauty of Loulan" and another from the same area that I call the "Marlene Dietrich of the Desert".  During the last half-century, hundreds of well-preserved human remains, together with an abundance of burial goods, have been recovered from Bronze Age and later sites (2nd and 1st millennia BC) in the area around Kroraina / Loulan.

The kingdom was renamed Shanshan (鄯善) after its king was assassinated by an envoy of the Han dynasty in 77 BCE; however, the town at the northwestern corner of the brackish desert lake Lop Nur retained the name of Loulan. The kingdom included at various times settlements such as Niya, Charklik, Miran, and Qiemo [VHM:  all of which lie along the southeastern rim of the Tarim Basin]. It was intermittently under Chinese control from the early Han dynasty onward until its abandonment centuries later. The ruins of Loulan are near the now-desiccated Lop Nur in the Bayingolin Mongol Autonomous Prefecture, Xinjiang, and they are now completely surrounded by desert.


Since Kharosthi Prakrit government documents from the Loulan / Shanshan kingdom contain Indian terms, it is evident that it must at one time been under the influence of administrative practices from South Asia.  The underlying linguistic substrate, however, is now known to have been Tocharian C.

The task for this post is to determine whether current MSM Lóulán 樓蘭 / 楼兰 is legitimately reconstructible with velars at the beginning of both syllables for Sinitic of the Han period or earlier.  In other words, I begin by asking how strong the (internal) evidence is that both syllables should begin with a velar consonant in Old Sinitic and Han period Sinitic.

I'm especially interested in that initial consonant cluster, "kr-", since it seems to have had an effect on the reconstruction of Han and earlier forms of Sinitic (at least for some historical phonologists), e.g., Zhengzhang /*ɡ·roː//*ɡ·raːn/ for Old Sinitic, about the velar of which I'm dubious, strictly on internal grounds.

BaxterSagart system 1.1 (2014)

Reading # 1/1
‹ lan ›
English orchid

Zhengzhang system (2003)

Reading # 1/1 1/1
No. 8453 6119
0 1
MC rime
/*ɡ·roː/ /*ɡ·raːn/

Axel Schuessler (2009)

樓 Late Han lo    Old Sinitic

蘭 Late Han lan   Old Sinitic rân    Old Sinitic (Baxter 1992)  *g-ran

South Coblin

Both words have velar initial forms in their xiesheng* series, so people who do that sort of work would reconstruct velars. That’s the standard view.

[*VHM:  A xiesheng (Chinese: 諧聲; pinyin: xiéshēng) or phonological series is a set of Chinese characters sharing the same sound-based element. Characters belonging to these series are generally phono-semantic compounds, where the character is composed of a semantic element (or radical) and a sound-based element, encoding information about the meaning and the pronunciation respectively.]


Jonathan Smith

Loulan 樓蘭 and similar are hugely interesting. For what it's worth, both characters belong to the MC velar-lateral mixed phonetic series which have motivated "cluster"-type proposals for OC from Maspero and Karlgren forward and which I tried to consider anew via Sagart 1999 in my CLAO paper*. But the problem is determining exactly what this is worth, whether for Han or for "OC" as such….

*“A/B Type segregation in mixed-onset phonetic series is the key to Early Chinese onset complexity.” Cahiers de Linguistique Asie Orientale, Vol. 47 No. 2 (2018), pp. 165–197.

Doug Hitch

Have you ever been accused of being psychic? Or maybe this is just a case of Serindian Serendipity. Just a few hours before receiving your [VHM's] message, I read the page in Schmidt dealing with the name and I thought I should scan this page* and send it to you.

[*VHM:  From KT Schmidt’s Nachlass, the page on which he discusses the name of Kroraina / Loulan.]

About the earliest Pkt, it is from the "Niya" docs. I ruffled through Burrow's works. He called it Krorainic (which I like better than Lolanish, though I prefer Krorainian) but, as far as I could tell, does not deal with the name. In the fat Kharoṣṭhī Inscriptions, in the index under kroraiṃnaṃmi they note, "For Kroraiṃna = Lou-lan, v. Stein, Ser., p. 415."  I checked that on Volume One. Actually, pp. 413-416 is most useful. It appears that Rapson first identified Kroraina as the place in which a letter was found, the Lou-lan site, in a personal communication to Stein (if I read that right). Then Stein makes the phonological connection Kroraina=Loulan on p.416.

The Oldest Prakrit Evidence for Kroraïna

The name, in several spellings, occurs in documents from both the Niya and Loulan sites. Aurel Stein tells us, "The date of the Niya tablets is approximately fixed by the dated Chinese record of a.d. 269 … brought to light on my former expedition" (Serindia I:329). He remarks several times that Niya was abandoned in the late third century, which gives us the terminus ante quem for the earliest spelling.

Kroraïna was abandoned maybe a half century later. From among the dated Chinese documents, the latest is a narrow slip of wood found by Stein which was shown by Chavannes to date to 330 AD (Serindia I:371).

Kroraïna = Loulan

Stein presented the evidence from a number of Chinese documents from the L.A. site which demonstrate that the area was then known to the Chinese as Loulan (Serindia I:413). There is a little less Prakrit evidence but it is still conclusive.

Rapson provided Stein with an abstract of the three Kharoṣṭhī tablets mentioning the name Kroraïna (Serindia I:414–415). L.A. iv. ii. 3 (=CKD 678) records a land sale from a man originally from Kroraïna to several people. The plot is described as being at Kroraina ‘in the south of the great town’. L.A. iv. ii. 0234 (=CKD 696) is an almost complete paper document. From Burrow’s later translation we read:

I came here from Krorayina and brought the rete camels. Up till to-day there has been no buying and selling. This I make known at your feet. I wish to return to Krorayina. Whatever news there is of you there, you should send me a letter. I will bring it to my father the guśura in Krorayina, at the time you have to go. (BurTrans:140).

L.B. iv. I 6+7 (=CKD 706) is a double wooden tablet found among ruins seven miles from L.A. It conveys the king’s order about the disposition of a farm at Kroraiṃna. The purport of these documents is that the general area was locally known as Kroraïna.

Stein concludes that Loulan,

must reproduce an indigenous local name, and in view of the identity of the ground designated I am tempted to recognize this original form of the name in the Kroraina or Krorayina of our Kharoṣṭhī documents. If we take into account the difficulties which attend all rendering of foreign names by Chinese sounds, and the total absence of any system of transcription before the attempts made in T'ang times, Lou– for Kro– is as close a phonetic reproduction as could be expected, seeing that the semi-vowel r, which is wanting in Chinese phonetics, is regularly replaced by l. In the same way -lan may be recognized as a sufficiently close approximation to –raina or -rayina. (Serindia I:416).

The Gāndhārī Spellings

I list here the ten forms I was able to find in the available documents. I use the transliterations found on With the lone exception of the form in CKD 15, these are listed in the index to the Kharoṣṭhī Inscriptions (Boyer, Rapson, Senart 1920+). That edition ends at number 764. I scanned the additional CKD entries, 765–896 and found no others. I found CKD 15 from a reference in the dictionary (Baums and Glass, see below). It is not listed in KharInscr because the document involves two pieces of wood and the second piece with the form was likely not available to the original editors.

Niya Site

CKD 15, Und. Obv2 Kroray⟨*in⟩aṃmi

CKD 277 B1 Kroraïṃciyana

CKD 370 Cov.Rev. 3 Kroraïṃci

CKD 383 Rev 9 Kr(*o)r[a]yiṃci

Loulan Site

CKD 678, Und. Obv2 Kroraïṃci,  Kroraïṃnaṃmi

CKD 696 Obv. 3 Krorayinade, Obv. 4 Krorayina, Obv. 5 Krorayinaṃmi

CKD 706 und.Obv. 2 Kroraïṃnaṃmi


In CDK 678, in KharInscr the editors give for the akṣara –iṃ– in both forms forms the alternate reading –hiṃ-. Because it occurs twice, it is less likely to be a scribal error. It is possible that the intended spellings were krorahiṃnaṃi and krorahiṃci, and that this then may provide more information about the phonology of the name or about orthographic practice. It looks like the name had a sequence of two vowels within it, presumably /a/ plus /i/. The script does not provide an easy or obvious way to write a sequence of vowels. When writing in Gāndhārī a vocalic akṣara normally implies the beginning of a word. It appears that some writers avoided this implication by using a consonantal akṣara without phonological import to write the second vowel. That is, besides the -aï- spelling with the vocalic akṣara i in the middle of the word, some scribes wrote -ayi- and at least one wrote -ahi-. From a phonetic point of view it is easy to see how Kroraīna could be written Krorayina. It is less easy to understand Krorahina. Possibly, in the local language there was no intervocalic /h/ as is the case in many languages that have initial /h/ including English (which has intervocalic /h/ only in loans). Perhaps maharaya ‘great king’ was pronounced [maaraya] locally.

A similar orthographic device occurs in Khotanese Brāhmī. The loss of intervocalic /ẓ/ led to the sequence of two vowels in many words. This sequence is commonly written with a consonantal akṣara together with the apostrophe, a subscript hook not found in Sanskrit writing. For historical *haṃjsäṣe ‘I intend’ we have haṃjsäte’ and haṃjsä’te with apostrophe and the “peg akṣara” t on which the vowel e is written (also spelled haṃjse’ without peg, and haṃjsate without apostrophe). Historical *huṣe ‘man’ is spelled huve’ with peg v and apostrophe (Ars Metrica 2014:23–27). *päṣa ‘powers’ is written päga’ with peg g and apostrophe, pä’ga with the apostrophe under the preceding akṣara, and pä’ga’ with apostrophe under both akṣaras (Hitch diss: 95—96). It also appears with peg t, päta’ (92). Akṣara y is not used as a peg perhaps because the non-Indic diacritic ei for the diphthong /aᵉ/ may be used with apostrophe where y+apostrophe might be expected. For instance, *bäṣa- ‘poison’ and *näṣa- ‘nectar’ show /ĕĕ/ written ei’ as in bei’ (*bäṣä) or nei’na (*näṣäna) (thesis: 97 fn. 105).

Because the various spellings of the local name in Prakrit indicate a vocalic sequence, I view Kroraïna as the preferred Roman transcription.

The dictionary entry

A Dictionary of Gāndhārī (, Stefan Baums and Andrew Glass, has the following entry.


(unclear; cf. Sogdian Kr’wr’n, Kwr’ynk, Chinese 樓蘭 Ló­u­lá­n, ONWC *Loulɑn; Reichelt 1931: 4, 49) m. Krorayina (place name).

    1. loc.Kroraïṃnaṃmi, Kroray⟨*in⟩aṃmi.

CKD 15, Und. Obv2 Kroray⟨*in⟩aṃmiCKD 678, Und. Obv2 Kroraïṃnaṃmi.

Harold Bailey

VHM:  I suspected that if anyone had attempted to determine the etymology of Kroraina, it would have been Bailey in Khotanese Texts VII.  Checking that volume online, I found six mentions of Kroraina, but they don't attempt to explain the name itself.

Hiroshi Kumamoto

In the famous itinerary (or name-list of the cities around the Tarim Basin) in the Stael-Holstein scroll*, the name occurs as raurata.

See "The Staël-Holstein Miscellany," Asia Major New Series vol. 2, part 1 (1951) p. 11 (of pp. 1-45):

12.1 raurata, a local name, possibly a development of the older Krorayina, to which the legend of the West Indian name Roruka, Roruva, capital of Sauvira [long "i"] (see H. Luders, Weitere Beiträge zur Geschichte und Geographie von Ostturkestan, 29 ff.) was attached.


*According to the preface of Frederick William Thomas and Sten Konow, Two Medieval Documents from Tun-Huang (Oslo:  A.W. Brøggers boktrykkeri, 1929), the one side of the scroll is a Chinese Buddhist text, and the other side has a text in Tibetan and Khotanese, which are somewhat related but not translations of each other. The editors (Thomas and Konow) had only photos of the second side (sent to them by the Baron) to work with, so the exact nature of the Chinese text is unknown. The translation of the Tibetan text is, as far as I know, only published by F. W. Thomas in this 1929 booklet.

Oktor Skjærvø once mentioned that there is a record that the Baron donated the MS to Harvard, and he was looking for its actual whereabouts in the libraries, but I haven't heard the result of his search.

See also G. L. M. Clauson, "The Geographical Names in the Staël-Holstein Scroll", The Journal of the Royal Asiatic Society of Great Britain and Ireland, No. 2 (April, 1931), 297-309 (preview).



VHM:  The following section is for those who are advanced in Chinese philology and phonology, although others may wish to consult these materials without being able to read all of them (e.g., for crucial information, such as the date of the earliest known occurrence of the Sinitic transcriptional name Lóulán 樓蘭).  As such, I will not provide the customary Language Log transcriptions and translations of all the Chinese characters.

Xinchang Li

The earliest occurrence of Loulan 樓蘭 I can find is in Shiji 史記·匈奴列傳 (Records of the Grand Historian, "Account of the Xiongnu / Huns"). In a letter presented to Emperor Wen of Han 漢文帝, the Chanyu 單于 ("nomadic supreme ruler") mentions that he has conquered Loulan, Wusun, Hujie and 26 other kingdoms in the vicinity, and made them all under the control of the Xiongnu. "定樓蘭、烏孫、呼揭及其旁二十六國,皆以為匈奴。" Loulan 樓蘭 also appears in the 大宛列傳 (see below under the next section) in the Shiji 史記.


Diana Zhang

For the first question, "when the earliest occurrence of Loulan 樓蘭 is," my answer would be "according to extant textual records (only; because archaeological findings, etc. are not considered here in the scope of this email reply), perhaps the earliest mention was in 175 BC." The earliest textual occurrences of Loulan as a kingdom's name are in Shiji 史記. Two chapters mention it: 匈奴列傳 (110th juan, 50th liezhuan) and 大宛列傳 (123rd juan, 63rd liezhuan). In the former there is one occurrence and in the latter there are five, adding up to six in total.

All the page numbers cited below are according to the Zhonghua shuju 中華書局 1975 edition.

A. "Xiongnu* liezhuan":

[*VHM:  Xiongnu 匈奴 (Old Sinitic /qʰoŋ.nˤa/, Wade–Giles Hsiung-nu, cf. Sogdian xwn]), a large confederation of Eurasian nomads who dominated the Asian Steppe from the late 3rd century BC to the late 1st century AD.  Their name is probably related to that of the later Huns (1st-7th c.) and perhaps also that of the Huna (5th-6th c.), but the composition of the people who made up the latter two groups — judging from historical, archeological, and physical anthropological sources — had changed by the time they ranged respectively as far west as Europe and the northwestern part of South Asia.  Source]

1) Shiji 110.2896 — 其明年 (the fourth year of Emperor Wen's reign, 175BCE),單于(Modu 冒頓 Chanyu)遺漢書曰: "天所立匈奴大單于敬問皇帝無恙 …… 定樓蘭、烏孫、呼揭及其旁二十六國,皆以為匈奴。諸引弓之民,並為一家。….."

In this context, Loulan is mentioned in a letter from Modu Chanyu 冒頓單于 to the Emperor Wen of Han.

    1. "Dayuan* liezhuan":

[*VHM:  From (, “great”) + the transcription of a foreign word, believed to be Pali yona, yonā or Sanskrit यवन (yavana), both of which are appellations for the Greek, derived from Ancient Greek Ἴωνες (Íōnes), plural of Ἴων (Íōn, “Ionian”).  Source]

2) Shiji 123.3160 — 騫 (Zhang Qian) 身所至者大宛、大月氏、大夏、康居,而傳聞其旁大國五六,具為天子言之。曰 …… 而樓蘭、姑師邑有城郭,臨鹽澤。……

Context (the whole paragraph in Zhang Qian's report about Dayuan [in the Ferghana Valley]): “大宛在匈奴西南,在漢正西,去漢可萬里。其俗土著,耕田,田稻麥。有蒲陶酒 (grape wine!)。多善馬,馬汗血,其先天馬子也。有城郭屋室。其屬邑大小七十餘城,眾可數十萬。其兵弓矛騎射。其北則康居,西則大月氏,西南則大夏,東北則烏孫,東則扜罙、于窴。于窴之西,則水皆西流,注西海;其東水東流,注鹽澤。鹽澤潛行地下,其南則河源出焉。多玉石,河注中國。而樓蘭、姑師邑有城郭,臨鹽澤。鹽澤去長安可五千里。匈奴右方居鹽澤以東,至隴西長城,南接羌,鬲漢道焉。”

Here, Loulan is included in Zhang Qian's report to the Emperor Wu of Han after he returned from his first thirteen-year trip to the Western Realms. His report thus refers to what he experienced between 139 BC to 126 BC. — this report by Zhang Qian is the same one that first mentions Dunhuang 敦煌!

3, 4, 5) (three occurrences) Shiji 123.3171-2  — 自博望侯開外國道以尊貴,其後從吏卒皆爭上書言外國奇怪利害,求使。…… 而樓蘭、姑師小國耳,當空道,攻劫漢使王恢等尤甚。而匈奴奇兵時時遮擊使西國者。使者爭遍言外國災害,皆有城邑,兵弱易擊。於是天子以故遣從驃侯破奴將屬國騎及郡兵數萬,至匈河水,欲以擊胡,胡皆去。其明年,擊姑師,破奴與輕騎七百餘先至,虜樓蘭王,遂破姑師。因舉兵威以困烏孫、大宛之屬。還,封破奴為浞野侯。王恢數使,為樓蘭所苦,言天子,天子發兵令恢佐破奴擊破之,封恢為浩侯。於是酒泉列亭鄣至玉門矣。

This attack on Loulan happened before the first Han-Dayuan war (104-103 BCE) in 110 BCE.

6) Shiji 123.3174 — 天子已嘗使浞野侯攻樓蘭,以七百騎先至,虜其王。

This attack on Loulan was given an exact date: 太初元年 (104 BCE). This time, the attack and capture should already be during the first Han-Dayuan war.

To sum up these materials, I believe that the first mention of Loulan as a kingdom's name was in 175 BCE, through a letter from Modu Chanyu to the Emperor Wen of Han, in the text 史記·匈奴列傳.


For the second question, "how strong is the (internal) evidence that both syllables should begin with a velar consonant in Old Sinitic or Han period Sinitic," I would tentatively incline to answer, "not too much for 樓, but much for 蘭."

[N.B.  The acute mark, "á," is a notation for type A syllables that contrasts with the naked "a" that signifies type B syllables. These are my colleague's way of notation, and there might not be an established scholarly work that also does so (to my limited knowledge so far; more works need to be checked to confirm). Thus this small issue is pointed out in order to give credit to my linguist colleague's system, which I use quite often as upāya.]

樓:B-S ('14) OC *ro; Schuessler ('09) OCM *rô, LHan lo; Coblin (BTD [Buddhist Transcriptional Dialect] '93) Han *lou > ONWC lou. Not much possibility for a velar beginning before this nucleus. It should just have *r-/*l-.

蘭:more complicated. I believe that 蘭 (MC lan) should have a velar in its OC onset because it has a doublet: 蕳 (MC kean). (Ode 95: 士与女,方秉蕳兮; Ode 145: 彼澤之陂,有蒲與蕳).

Coblin has a nice short treatise on the reconstruction of Han *gl- for some cases of the MC l- in his A Handbook of Eastern Han Sound Glosses (1983), 48-50. He gave a sample list of words that should have velar onsets in their older Chinese/Sinitic forms. 蘭/蕳 is not on the list but I believe that this case must belong to it. One could easily see how similar the 蘭/蕳 case is to his examples.

Baxter-Sagart's 2014 list (not their monograph, but a separate addendum on their website, last updated in 09/20/2014) also acknowledges this pair of doublets. They thus give 蘭 *k.rˤan (p. 64) ; 蕳 *k.rˤan, dialect *k.rˤ > *kˤr- (p.51).

For myself (discussed a little bit with my colleague, too), I believe that the distinction between 蘭 and 蕳's historical sound change processes from their near-homophonous OC forms, may be due to the glottalization of 蘭's pre-initial *k-: 蘭 *k-rán > *ʔrán > MC r-/l- VERSUS 蕳 *krán > *krán > MC k-. This analysis of mine is still tentative. However, we can still almost be sure that 蘭 did have an OC *Kr-/Kl- structured onset. (I use r-/l- all the time, because the issues of how much OC distinguishes *r- and *l- and to what extent it does as it comes to a specific word's sound, are still not quite sure among the scholarship that I know so far.)

Tsu-Li Mei

楼蘭 (i)I looked up GRS (or rather its Chinese translation 汉文典)。 In No. 123, the phonetic series which contains 楼,also contains 屦 MC kju and 窭 MC gju.  This means we can reconstruct *gr for 楼。 In No. 185, the phonetic series which contains 蘭 also contains 谏 MC kan and 柬 MC kAn.  This means we can reconstruct *gr for 蘭。

(ii) August Conrady (1920) Die Chinesischen Handschriften und Sonstigen Kleinfunde Sven Hedins in Lou-lan.  In the Preface, Conrady thanks Cai Yuanpei / Tsai Yuan-pei 蔡元培, formerly Minister of Education under President Sun Yat-sen, for helping him to decipher Chinese handwriting on wooden tablets.  Tsai Yuan-pei returned to Leipzig in 1913, after Yuan Shih-kai mounted a counter-revolution.

Yesterday I searched the web and found Mallory and Mair (2000), The Tarim Mummies, and guessed that the book will say the people of Loulan are partly Tocharians.  I also looked up Tocharian mummies on the web and learned of Burrow's account of Kroraina C.  Good luck on your ongoing work on Loulan 楼蘭。

Loulan 楼蘭 should be in Han Chinese *gru-gran, instead of your *glu-glan.  It was always  believed that Middle Chinese 来母 l- goes back to OC *l-.  Axel Schuessler showed that MC l < OC r and 喻四 MC ji- < OC l.


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 Nicholas Sims-Williams and Ben Zimmer]


  1. Victor Mair said,

    May 14, 2019 @ 8:06 am

    From Dan Boucher:

    The discussions by Loewe and Brough on Loulan would also be relevant to the discussion, since they both deal explicitly with Chinese interventions and eventual abandonment of the site.

  2. Chris Button said,

    May 14, 2019 @ 8:38 am

    I follow Schuessler's "minimal" reconstructions in seeing no a priori internal reason to reconstruct a velar k- prefix in either syllable. On purely typological/phonotactic grounds, it should not be surprising that we find evidence for it before liquid onsets as opposed to elsewhere (I'm not generally convinced by some other reconstructions of OC that claim to show evidence for prefixes all over the place with other onsets). Having said that, if external evidence suggests an earlier kr- cluster, then the prefix may simply have been dropped, which is not entirely untoward.

  3. Victor Mair said,

    May 14, 2019 @ 11:25 am

    From Wolfgang Behr:

    The identification Kroraina = Loulan was first made by Karl Himly (1836-1904), whose notes August Conrady (1864-1925) had used in his famous Die chinesischen Handschriften- und sonstigen Kleinfunde Sven Hedins in Lou-lan. Herausgegeben, übersetzt und unter Benutzung von Karl Himlys hinterlassenen vorarbeiten behandelt von August Conrady, Stockholm: Generalstabens Litografiska Anstalt, 1920 [freely available here:

    Conrady discusses the sources in App. I "Zur Identifizierung von Lou-lan", pp. 143-149.

    Already in Early Buddhist transcriptional data Lokakṣema (fl. 168-188)
    consistently uses lou2 樓 for Sanskrit/Prakrit -ru- and -ro-, so there is no trace of an earlier cluster initial implied in these materials, cf.

    jia1lou2luo2 迦樓羅 : garuḍa
    lou2tuo1luo2 樓陀羅 : rudra

    fan4fu4lou2 梵富樓 : brahmapuro[hita]
    lou2shi4huan2 樓耆洹 *ruvivarṇa
    hui1lou2yan2 堕樓延 : vairocana

    (Data from Coblin, Handbook of Eastern Han Sound Glosses, s.vv.)

    However, apart from the xiesheng series argument, already mentioned by several of your correspondents, some scattered _internal_ evidence for an earlier cluster initial in 樓 during the late Warring States period can be found in loan graph relationships. Thus, the disyllabic surname Dong1lou2 東樓 (=OC *tˤoŋ+(Cə.)rˤo[k] in the BS system), well known from the edited literature, is written as dong1gu3 東谷 *tˤoŋ+*C.qˤok (or -yu4 < MC -yowk < -*ɢ(r)ok ?), i.e. with an uvular stem initial in some 3rd c. B.C. seals (e.g. 《璽彙》1033). In the same xiesheng series, the Shangbo 《上博》 vol. III Zhouyi《周易》 Ms. Hexagram 《井》 writes gou1 句 (*[k]ˤ(r)o) instead of lü3 縷 (OC *[r]oʔ) 'thread' in the silk Ms. version (but lou4 漏 'leak' < OC *[Nә-r]ˤok-s in the transmitted text!), pointing to a velar cluster initial. (cont. in next comment)

  4. Victor Mair said,

    May 14, 2019 @ 11:31 am

    From Wolfgang Behr, pt. 2:

    In the Xuanquan 懸泉 slips, internally datable to 111 B.C.-107 A.D., on the other hand, Loulan appears as one of the 24 countries from which the desert relay station (置 zhi4) some 60 km East from Dunhuang receives a delegation, but it is already uanimously written in its orthography 樓欄 familiar from the slightly earlier Shiji entry for 175 B.C. Loulan also occurs with these characters in the Juyan 居延 slips (range: 102 BCE –169 A.D.), in some cases possibly after the change of the name to Shanshan 鄯善 in 77 B.C.

    Using such evidence it is thus exceedingly difficult to determine the time of the loss of the cluster initial in Early Old Northwest Chinese. One would have to look systematically into the corpus of Xuanquan
    transcriptional data to get a better grip on this, but that hasn't been
    done, so far, as far as I am aware of.

    Moreover, the whole problem of identifying Chinese transcriptions of/for Kharoṣṭhī Krora'ina, Sogdian Krʾwrʾn, Greek Chauranaioi (Χαυραναῖοι Σκύθαι, indeed in Ptolemy 15.3) etc. is complicated by the fact that, as Yu Taishan has been arguing a while ago, _geographically_

    Rong1lu4 戎盧 (OC *nuŋ+*C.rˤa)
    Lie4lou2 埒婁 (OC *[r]ot+*[r]o)
    Lun2tai2 or -yi2 輪台 (OC *[r]u[n]+*l(ˤ)ә)
    Lun2tou2 侖頭 lun2tou2 (*[r]u[n]+*[m-t]ʕo)

    in the Chapter on the Western Territories of the Hou Hanshu may _all_ be renderings of the very same place name as Loulan, cf. his:

    Yu Taishan 余太山, "Loulan, Shanshan, Jingjue deng de mingyi — jian shuo Xuanzang zi Yutian donggui luxian" 樓蘭、鄯善、精絶等的名義——兼說玄奘自于闐東歸的路線 [The meaning of Loulan, Shanshan, Jingjue and other names — including a discussion of the route Xuanzang took on his way back East from Khotan], Xiyu Yanjiu 西域研究 (2), 2002, 32-37.

    Other important articles on this problem include the attempts by Lin Meicun and Huang Shengzhang, in one of his last articles, to identify the capital(s) of Loulan:

    Lin Meicun 林梅村, "Loulanguo shidu kao" 樓蘭國始都考 [On the initial capital of the Kingdom of Loulan], Wenwu 文物 (6), 1995, 79-85.

    Huang Shengzhang 黄盛璋, "Chu lun Loulanguo shidu Loulancheng yu LE cheng wenti" 初論樓蘭國始都樓蘭城與 LE 城問題 [A temptative discussion of the first capital "Loulan Town" of the Kingdom of Loulan and the problem of the Town LE], Wenwu (8), 1996, 62-72.

    Whether Gk. Chaurana may be identified with Kroraina (rather than Khotan) in Ptolemy has been discussed by Étienne de la Vaissière, "The Triple System of Orography in Ptolemy’s Xinjiang", in: W. Sundermann, A. Hintze, F. de Blois (éd.), Exegisti Monumenta. Festschrift in Honour of Nicholas Sims-Williams, (Iranica, 17), Wiesbaden : Harrassowitz, 2009, p. 527-535. [available online here

  5. Victor Mair said,

    May 14, 2019 @ 7:00 pm

    From Tsu-Lin Mei:

    First, a brief history of the existence of consonant clusters of *gr- type in Old Chinese. Joseph Edkins 1874 The State of the Chinese Language at the Time of the Invention of Writing first noticed 蓝l- 监k-,洛l-各k- and came to the conclusion that Old Chinese must have consonant clusters of the kl-, gl- (gr-) type. Conrady 1896 Eine indochinesische Causativ-Denominativ-Bildung und ihr Zusammenhang mit den Tonaccenten , citing Edkins, came to the conclusion that Old Chinese had consonant clusters gr-, gl- etc. and WT also had consonant clusters gr-, gl-, sm- and this fact shows that WT and OC can be genetically related. I am not surprised that Conrady 1920 Die Chinesische Handschriften und Sontigen Kleinfunde Sven Hedins in Lou-lan, in discussing the phonology of 楼蘭would come to the conclusion that both characters had initial consonant clusters gr- or kr-. Edkins theory was inherited by Karlgren “Theory of Phonetic Compound” which is the third chapter of Karlgren 1923 Analytic Dictionary of Chinese and Sino-Japanese.

    Second, when did consonant clusters disappear ? 何大安。1999 <古汉语声母演变的年代學>,《林炯阳先生六秩寿庆论文集》台北 gives the best answer: OC consonant clusters begin to disappear during the Western Han (-206-25) and totally disappeared during the Eastern Han (25-220). Since the earliest Buddhist sutras translated into Chinese date from the Eastern Han onward, we would expect that 蘭transliterate Skt. ran, 蓝 transliterates Skt. ram etc.

    Third, Yakhontov 1960 “Consonant combinations in Archaic Chinese” ushered in the new era of OC reconstruction (of consonant clusters). The new element in Yakhontov’s reconstruction is that WT cognates play a major role in supporting Y’s reconstruction of *sm-, *gl- (*gr-) etc. But then, Conrady 1896 is the granddaddy of Old Chinese reconstruction from the perspective of Sino-Tibetan comparative linguistics. Conrady tries to show that causative s- prefix is a morphological process shared by all Sino-Tibetan languages, e.g. OC, WT, WB, thus duplicating what Franz Bopp did for Indo-European languages. And such endeavor was only possible in Leipzig, the headquarter of the Junggrammatiker.

  6. Chris Button said,

    May 14, 2019 @ 10:52 pm

    @ Wolfgang Behr

    Thus, the disyllabic surname Dong1lou2 東樓 (=OC *tˤoŋ+(Cə.)rˤo[k] in the BS system), well known from the edited literature, is written as dong1gu3 東谷 *tˤoŋ+*C.qˤok (or -yu4 < MC -yowk < -*ɢ(r)ok ?), i.e. with an uvular stem initial in some 3rd c. B.C. seals (e.g. 《璽彙》1033).

    In a Pulleyblankian reconstruction, I would go with 樓 as *ráw and 谷 as *ɣákʷ / *ɣàkʷ (*ɣ- shifting to k- or j- respectively depending on the syllable type). For the association of -w with -kʷ (which ultimately comes via -ʷ-ɣ with -ʷ-k) compare a case like 流 with 毓. The association of *r- (probably something like [ɹ] due to its affinity with laterals) with *ɣ- presumably comes from the more approximant-like articulation of *ɣ- as [ɰ] in a compound when not occurring as an isolated onset, which in this case is notably preceded by the velar coda of 東.

  7. Victor Mair said,

    May 15, 2019 @ 7:18 am

    From Hiroshi Kumamoto:

    I was a bit puzzled to see "khot.-sak. raurana" on the first page of K. T. Schmidt's Nachlass on Tocharian C. Also puzzling is Bailey's 1973 article (no other works by Bailey listed there) in the references, which has nothing to say about Loulan or Kroraina. If he vaguely remembered the Khotanese name for the city in Bailey's article with the title "… miscellany", everything fits. We can't blame the author because this was an unfinished draft. On the other hand the editor of the Nachlass must have the obligation of checking and correcting the reference before going to print (in addition to misspelled "Tramblay" and "Rhie, Maylin Martin").

    Bailey's 1951 article is so definite that the earlier works by Konow and Clauson are no longer relevant (except of course the facsimile plates and Thomas's part on Tibetan, which definitely should be reworked). Two major improvements on Bailey are:

    E. G. Pulleyblank, "The Date of the Staël-Holstein Roll", AM, NS 4/1, 1954

    James Hamilton, "Autour du manuscrit Staël-Holstein", TP 46, 1958

  8. Victor Mair said,

    May 16, 2019 @ 1:01 pm

    From Hiroshi Kumamoto:

    Dr. Adams appears to take Khotanese *raurana for Lolan as correct, but the word in StH is clearly raurata, and I'm not aware of any other occurrence of the word.

    The legend of the city that sank in the sand that Bailey referred to (as per Lüders) is now available in a good English translation with copious notes in the last chapter of:

  9. Chris Button said,

    May 16, 2019 @ 3:48 pm

    Some random speculation here…

    I wonder if in the attempt to attribute a cluster *kr- to 樓 when there is little to support anything but *r- (and, moreover, no evidence that a putative cluster would even still be retained had it ever even existed), the mistake might be in treating "k" and "r" in "kr" as independent of one another. While OC *r- in terms of how it regularly evolved into Middle Chinese was not a back articulation, that does not preclude certain speakers perhaps having made such an articulation. If we go with a shift of /r/ > /ʀ/ > /ʁ/ > /γ/ (as opposed to the more standard evolution to Middle Chinese of /r/ > /l/), then the use of /ʁ/ or /γ/ to transcribe a foreign /kr/ is perhaps less surprising.

  10. Chris Button said,

    May 16, 2019 @ 9:58 pm

    Some alternative random (perhaps slightly more plausible) speculation…

    I am reminded of how 虎 *ʰráɣʔ "tiger" (Schuessler has *hlâʔ), attested in the earliest inscriptions, is an old loan from Mon-Khmer *klaʔ "tiger".

    I wonder if rather than looking for some unattested and supposedly dropped velar cluster *kl- or the like, we should rather be looking at a vacillation between *ʰr- and *r- whereby an original 樓 *ʰráw ended up developing a later Middle Chinese reflex as if from *ráw. A certain instability of pre-aspiration before sonorants can be expected on phonetic, and hence typological, grounds (I personally came across just such a case in some northern Kuki-Chin languages where h- from earlier *ʰr- was attested as g- from earlier *r- instead).

  11. Chris Button said,

    May 16, 2019 @ 9:59 pm

    Sorry, I meant *kr- for *kl-

  12. Victor Mair said,

    May 17, 2019 @ 12:02 am

    @Chris Button

    I think that is a brilliant suggestion, one that has powerful explanatory power for what may have transpired to get from kr- to r- / l-. This is especially so when we consider that Kroraina was almost certainly a foreign word that was being Sinitically transcribed by what would ultimately shake out as "Loulan", comparable to what happened when the Kuki-Chin word for "tiger" was borrowed into Sinitic.

  13. Chris Button said,

    May 17, 2019 @ 8:56 pm

    Thanks – I think *ʰráw.rán for 樓蘭 could work with the assumption that 樓 later developed as if from *ráw

    I also think it's important to note that *ʰr- (as with other "voiceless" sonorants) would have been partly voiced, so transcriptions like /r̥/ with a subscript circle should be avoided as misleading.

    Incidentally, the typical evolution of *ʰr- to x- actually happens to chime quite nicely with my earlier speculation regarding some kind of velar fricative articulation. That also reminds me of the interesting behavior of 孝 ʰrə́ws (whose phonetic 㐬 as in 流 *rə̀w, mentioned above, has been inverted and abbreviated to graphically merge with 子 as also found in 好 ʰrə́wʔ ) which for some reason retained its rhotic component to give xr- (instead of the usual x-) thereby leaving us with Early Middle Chinese xaɨwʰ (instead of xawʰ)

  14. Chris Button said,

    May 17, 2019 @ 11:17 pm

    Also by assigning *ʰr- (> x-) to 樓 we get a much tighter association with the *ɣ- of 谷 noted in the alternation of 東樓 with 東谷 above.

RSS feed for comments on this post