Galactic glimmers: of milk and Old Sinitic reconstructions

« previous post | next post »

Often have I pondered on the origin and precise meaning of the Sinitic word lào, luò (reading pronunciation) 酪 ("fermented milk; yoghurt; sour milk; kumiss"); Old Sinitic (OS) /*ɡ·raːɡ/ (Zhengzhang).  My initial impression was that it may have been related to IE "galactic" words.

Possibly from a Central Asian language; compare Mongolian айраг (ajrag, fermented milk of mares) and Turkish ayran (yoghurt mixed with water). The phonetic similarity between Sinitic (OS *ɡ·raːɡ, “milk”), Ancient Greek γάλα (gála, milk) and Latin lac (milk), from Proto-Indo-European *ǵlákts (milk) is worth noting (Schuessler, 2007).


Paul Kroll, ed., A Student's Dictionary of Classical and Medieval Chinese, p. 256a:

1. kumiss, fermented mare's milk (also cow's or sheep's) < Khotan-Saka ragai (with metathesis)

    a. yogurt, milk curdled by bacteria

As Schuessler (2007), p. 345 notes, the fermented drink "arrack" may be a different etymon, a loan from Arabic 'araq ("fermented juice").  (Pulleyblank 1962:  250 contra Karlgren 1926) [VHM:  full references below]

The above are merely random, tantalizing observations about words for "milk" in various European, Near Eastern, Central Asian, and East Asian languages.  As presented, by no means do they constitute a brief for a connection between Sintiic and IE or, for that matter, any other language or group of languages.  We will have to dig much deeper into the semantics and phonology of words from many different sources before coming to any firm conclusion concerning the origin of Sinitic lào, luò 酪 ("fermented milk; yoghurt; sour milk; kumiss").

Before explaining succinctly why I think lào, luò 酪 ("fermented milk; yoghurt; sour milk; kumiss") had a non-Sinitic source, I shall first cite the views of the great Swedish Sinologist, Bernhard Karlgren, who wrote about lào, luò 酪 towards the end of his 1926 book Philology and Ancient China (Oslo:  H. Aschehoug [W. Nygaard]), pp. 138-139.  Karlgren narrates how the Chinese general Li Ling (d. 74 BC) deserted to the Huns (Xiongnu).  In a letter written to his friend Su Wu (140-60 BC) in 91 BC, Li Ling tells how they had a drink lào, luò 酪 Ancient Chinese (i.e., Middle Sinitic) lâk, derived from Archaic Chinese (i.e., Old Sinitic [OS]) glâk, where the l- stood for r-, a sound that was lacking in MS and OS.  Karlgren goes on to say that this lâk, through "an interesting case of sound-substitution", entered Japanese as "sake".  He further states that the word exists in Ainu in the form "arakke", which shows that the Ainus did not get it from Japanese, but from some other source

We're still dancing around the edges of the question of the derivation of lào, luò 酪 ("fermented milk; yoghurt; sour milk; kumiss"), though we're becoming more familiar with its dimensions and nature.  Let us go back to the basics, the beginning, of why I think lào, luò 酪 has a non-Sinitic origin.

Since animal milk production and consumption were part of the Secondary Products Revolution that occurred in the Near East and Europe during the 4th-3rd millennia BCE (c. Middle Chalcolithic), and since lactose intolerance is as high as 95% of the population in parts of Asia, including China, we may safely say that animal milk drinking spread from west to east, and with it words for animal milk.

Naturally, Sinitic has a native word for human milk, and that is nǎi 奶, old form .  Actually, Sinitic has more than one word for human milk, with another being rǔ 乳 .  It is interesting, though, that these Sinitic words for human milk signify both the human breast and human breast milk, whereas lào, luò 酪 refers specifically to animal milk products.

nǎi 奶 / 嬭 OS /*rneːlʔ/

Similar words in the area include Proto-Hmong-Mien *niaʔ²ᴰ (mother), Tibetan ཨ་ནེ (a ne), ནེ་ནེ (ne ne, paternal aunt), Khmer ញី (ñii, female). It may be comparable with Proto-Sino-Tibetan *s-nja-n (breast; milk; suck), whence Sichuan Yi (nyip, liquid milk), ꀉꆂ (ax nie, breast; milk).

It is unknown how the Min forms relate to forms in other dialects. See this article for a discussion of the Min Nan etymon. For similar nasalisations in Min, compare (ěr).

Colloquial words in different Min dialects show considerable variation – most have an n– initial, but finals and tones differ greatly (although the tone is never a rising tone). Some propose that this is substrate influence, passed on from the maternal Baiyue ("Hundred Viet") lineages since the intermarriages between southward-migrating Han Chinese and native non-Han women. Other southern dialects also show remnants of this native word: Hakka [Meixian] nɛn5, Cantonese [Guangzhou] nin1 ( or ). Compare Thai นม (nom, breast; breastmilk), Zhuang noemz (breast; breastmilk).


rǔ 乳 OS /*njoʔ/ (Zhengzhang), cf. Proto-Sino-Tibetan *s-nəw.

Human mother's milk (nǎi 奶; rǔ 乳) does not go bad, because it is drunk directly from the breast.

In contrast, the animal milk products collectively known as lào, luò 酪 ("fermented milk; yoghurt; sour milk; buttermilk; kumiss") go through a process of fermentation or souring.  This would have been necessary in premodern times before the invention of refrigeration.

Koumiss is alcoholic.  It is produced through the fermentation of mare's milk. Buttermilk and yoghurt are not alcoholic, but they undergo a process that makes them sour.  That process is also called fermentation.  If bacteria are involved, then you can distinguish it from alcoholic yeast fermentation as non-alcoholic bacterial fermentation.  The bacterial fermentation of lactic acids is often referred to as making "cultures".

Thomas Allsen, who is writing a book about about alcohol among the steppe peoples, says:

Certainly luo/lao means any fermented dairy product. If, however the term is associated with horses it means kumiss, for nomads never "waste" horse milk or any other product. In general, nomads consume very little fresh milk except for tea and for children and ferment everything. It is revealing as well that in western texts the stock phrase describing the nomads' diet is typically "meat and kumiss." In a real sense, then, the Chinese "meat and lao" is the more accurate description.

An interesting paper by Isabelle Bianquis, "Les alcools de lait en Mongolie", distinguishes two types of alcoholic drinks based on milk:

l’aïrag    lait de jument fermenté (plus connu sous le nom
koumiss, d’origine turque)

l’arkhi    yaourt distillé

We shall now proceed to march through a mass of relevant linguistic data, beginning with Khotan-Saka ragai, which is linked to lào, luò 酪 in Kroll's dictionary cited above.  Before we accept this equation, Hiroshi Kumamoto advises us to consult Emmerick's note in Akiner and Sims-Williams, eds, Languages and Scripts of Central Asia, Ldn: SOAS, 1997, 28.

Martin Schwartz:

As for arrack, the Arabic etymon is definitely 'araq, a DISTILLED alcoholic drink, and its underlying meaning is 'sweat'—  = droplets forming on the still.


I would imagine the Khot. is a LW from Sinitic. I would not follow an etym by Bailey, if he wanted to take the Khot. from PIE **glag-. As for a PIE 'milk'–Sinitic conn, apart from Paleoglot's remarks [VHM: see below], it's far less than compelling formally or semantically, and no clear geographical scenario for a borrowing of the restrictedly Greco-Latin word into Sinitic.


I don't have access at present to Akiner and S-W, but I looked at Bailey's ragai in his Etym Dic and came away with the thought that it is probably unreliable in many ways, and that there is no basis to use the entry for Indo-European, Indo-Iranian, or Sinitic etymological connections. Scholars outside of Iranian Studies may be unaware that there is a consensus among us that Bailey was a very learned man who made many important contributions, but was weak as to etymological methodology and correct definitions, and was responsible for a number of ghost words, as his student, the late Emmerick, has frequent occasion to point out.

Juha Janhunen:

I think Mo. araki / ariki/n 'spirits' (ultimately from Arabic 'araq) is a different etymon from airag 'fermented milk'. The latter must be connected with Tü. airan (ayran), but the time and direction of borrowing is unknown (?). The Mo. diphthong ai, written <aii> and also Romanized as <ayi>, could go back to *axi < *agi, in which case the earlier form of the Mo. item would be *agirag – somewhat more similar to OChi. grag.

Marcel Erdal:

Seeing that the Italic-Greek word for 'milk' hasn't turned up anywhere else in IE, it may as well be of Mediterranean origin and not PIE.

There is no way to connect general Mongolic ayirag 'kumiss' with Arabic ‘araq, a doubly distilled alcoholic drink (which reached Mongolic as araki, ariki).


The work of Sagart and Garnier [VHM:  see below] does not support the view of an Italic-Greek term for 'milk' which might go back to PIE; the 'milk' word and its cognates are much more likely to be PIE.

Nor would I necessarily connect Mo. ayirag with Turkic ayran 'churned milk'.  ayran still has not turned up in Old Turkic proper, and Kâshgharî (1005-1102) who mentions it, also has words with the d > y change. The proposal  of Larry Clark and others for adIr- 'to separate' as source of ayran is therefore not yet completely off the table. So Janhunen's Mongolic – Old Chinese connection would actually (in spite of the millennia in between) look quite viable.

András Róna-Tas:

On pages  464-470 of West Old Turkic I have dealt with Hungarian író 'buttermilk' and its Turkic background, including Mongolian ayirag "kumiss etc." and words like aguz, agurag, "beestings" and also the idea of Pulleyblank 1962, 253 to see a Hunnic word in Chinese which I found highly problematic. (file available upon request)

Alexander Vovin

I think that the context in Li Shizhen's (1518-1593) work that Pulleyblank 1962 also quotes rather clearly shows that OC *Grak (酪) is a foreign word. It is quite likely that like many other names of milk products discussed by Pulleyblank in the same article we deal here with a Xiongnu word. As Pulleyblank pointed out, Mo. ayiraq (possibly from *agirak, as mentioned by Juha), is likely to be from the same source. Much more speculative and problematic (but not completely impossible) is Pulleyblank's comparison with Yeniseic, cf. PY *uɁk (2nd tone) (Werner 2001.2: 376) 'flour/milk soup', with a hypothetical development *Grak > *Gak > uɁk (attested in Ket, Yug, and Kott).

There is no missing sign in *uɁk: u — glottal stop — k.

Pulleyblank, Edwin G. 1962. The Hsiung-nu Language. Asia Major, New Series. Vol. IX, pt. 2, pp. 239-265.

Laurent Sagart and Romain Garnier presented a paper 2 years ago at the SLE conference in Naples on IE words for milk and the genetic backgrounds. I am reproducing their abstract here:

Milk and the Indo-Europeans
Laurent Sagart and Romain Garnier
(CNRS, Paris;
Université de Limoges et Institut Universitaire de France)


Even more than the existence of a diversified Indo-European vocabulary of animal husbandry, the lack of a significant agricultural vocabulary (already known to C.C. Uhlenbeck, as reported in Kortlandt 2009), and especially of terms for specific domesticated cereals, is a strong indication that the Proto-Indo-Europeans were not farmers, although they were probably contemporary, and in contact with, the expanding populations who introduced farming into Europe from Anatolia beginning in the 9th millennium BP. Recent work by Haak et al. (2015) argues that a massive migration of Yamnaya culture steppe hunters into the Corded Ware culture area c. 4500 years ago established a new population component, distinct from both palaeolithic hunter-gatherers and from farmers out of Anatolia, in northern Europe. They linked this migration with the spread of Indo-European languages in Europe. This fits well with the Steppe hypothesis of Indo-European origins. The Corded Ware culture could be equated with a node postdating the separation of Anatolian in the Indo-European tree. This question remains: what allowed the Indo-European languages to prevail over earlier languages in a large continuous area in Europe and central/south Asia? Here we integrate the findings just described with new insights in Indo-European etymology and with different strands of recent work on dairying in neolithic Europe: we note the significant geographical overlap between the presumably Indo-European-speaking Corded Ware culture and the area of maximum lactase persistence in modern Europeans according to the map by Leonardi et al. (2012), and suggest that a major factor behind the Indo-European expansion was the capacity to consume fresh milk in adulthood.

While the importance of milk and milk products in early Indo-European culture was already described by Anthony (2007), we underline here the fact as recently as 7000 years ago all human populations were lactose-intolerant: adults could not digest the sugar lactose contained in milk. The invention of cheese, a milk derivate poor in lactose, by early farmers in NW Anatolia c. 8500 BP (Evershed et al.) for the first time allowed humans to turn animal milk into a stable source of food, contributing to the positive demography of farming populations. However, the early farmers themselves were largely lactose-intolerant (Burger et al., 2007): the capacity, originating in an ancient genetic mutation, to directly drink animal milk only started to become prevalent a few millennia later. We follow Burger (as quoted in Owen 2010) in supposing that contact with cheese-making farmers revealed the lactase persistence gene in certain hunter-gatherer individuals from the Pontic steppes. We hypothesize that the incidence of this gene rapidly increased among the relevant communities, which we associate with Proto- Indo-European. We assume that this unusual ability provided Indo-European speakers with a demographic edge over contemporary farmer and hunter-gatherer populations, fueling their geographical expansion and facilitating the replacement of earlier languages by Indo-European varieties. We support these hypotheses with an examination of the Indo-European dairying vocabulary, showing the unusual importance of milk-related terms, and describing historical changes in this vocabulary which testify to the growing importance of animal milk in early Indo-European society. We also point out that the oldest textual evidence describing milk drinking in adults concerns speakers of Iranian languages.


Kortlandt, Frederik. 2009. C.C. Uhlenbeck On Indo-European, Uralic And Caucasian. Historische Sprachforschung 122: 39-47.

Haak, Wolfgang and Lazaridis, Iosif and Patterson, Nick and Rohland, Nadin and Mallick, Swapan and Llamas, Bastien and Brandt, Guido and Nordenfelt, Susanne and Harney, Eadaoin and Stewardson, Kristin and Fu, Qiaomei and Mittnik, Alissa and Banffy, Eszter and Economou, Christos and Francken, Michael and Friederich, Susanne and Pena, Rafael Garrido and Hallgren, Fredrik and Khartanovich, Valery and Khokhlov, Aleksandr and Kunst, Michael and Kuznetsov, Pavel and Meller, Harald and Mochalov, Oleg and Moiseyev, Vayacheslav and Nicklisch, Nicole and Pichler, Sandra L. and Risch, Roberto and Rojo Guerra, Manuel A. and Roth, Christina and Szecsenyi-Nagy, Anna and Wahl, Joachim and Meyer, Matthias and Krause, Johannes and Brown, Dorcas and Anthony, David and Cooper, Alan and Alt, Kurt Werner and Reich, David. 2015. Massive migration from the steppe was a source for Indo-European languages in Europe. Nature 522: 207-211.

Leonardi Michela, Pascale Gerbault, Mark G. Thomas and Joachim Burger. 2012. The evolution of lactase persistence in Europe. A synthesis of archaeological and genetic evidence. International Dairy Journal 22 (2012): 88-97.

Anthony, David. 2007. The Horse, the Wheel and Language. Princeton: Princeton University Press.

Evershed, Richard P., Sebastian Payne, Andrew G. Sherratt, Mark S. Copley, Jennifer Coolidge, Duska Urem-Kotsu, Kostas Kotsakis, Mehmet Özdoǧan, Aslý E. Özdoǧan, Olivier Nieuwenhuyse, Peter M. M. G. Akkermans, Douglass Bailey, Radian-Romus Andeescu, Stuart Campbell, Shahina Farid, Ian Hodder, Nurcan Yalman, Mihriban Özbaşaran, Erhan Bıçakcı, Yossef Garfinkel, Thomas Levy & Margie M. Burton. 2008. Earliest date for milk use in the Near East and southeastern Europe linked to cattle herding. Nature 455, 528-531.

Burger, J., M. Kirchner, B. Bramanti, W. Haak, and M. G. Thomas. 2007. Absence of the lactase-persistence-associated allele in early Neolithic Europeans. Proceedings of the National Academy of Sciences 104 (10) 3736-3741.

Owen, James. 2010. Stone Age Milk Use Began 2,000 Years Earlier. National Geographic News.


Chau Wu:

Lào, luò 酪 is a classic example of Old Chinese words with an initial consonant cluster kl- (or gl-), which was later split into k- and l- initials, for examples, compare 各, 格, 閣, 鉻, 硌, vs. 洛, 落, 駱, 絡, 珞, 咯 (陳新雄, 古音研究, pp. 657-662). Therefore, the reconstructed sound for 酪 is *ɡ·raːɡ (鄭張尚芳 – I'm sad to note that he passed away just a little more than two weeks ago on May 19). The original meaning of 酪 is 'milk'. Examples, the literary Taiwanese lo̍k 酪 is defined as 'lin-chiap 奶汁, lin-chúi 奶水' (Campbell, p. 475), whereas Sino-Japanese raku 酪 means 'dairy', for example, rakuseihin 酪製品 'dairy products'.

Peter B. Golden:

I have usually seen it (Mong. ayraq) explained as coming from Arabic ‘araq.


Michiel de Vaan’s Etymological Dictionary of Latin and the other Italic Languages (Leiden: Brill, 2008): 320 under Lat. lac, cognate with Grk. γάλα et al. and Arm. kaxc (dial.) < *glg-t-sm kat’n < acc. sing. *glg-t-m, concludes that "Lat. lact- goes back to *glgt- > *glagt” PIE *glg-t “milk.” Robert Beekes, Etymological Dictionary of Greek (Leiden: Brill, 2010), I: 256, has much the same in slightly greater detail – and also dismisses Hittite galaktar “soothing.”

The new Russian edition/translation of Kāšġarī, ed. I. Kormušin, trans. A.R. Rustemov (Moscow, 2010), I: 141, n. 2, prefers to render the  Arabic translation of ayran (al-mākhiḍ) as “skimmed milk” (снятое молоко), rather than the “churned milk” (сбитое молоко) of Dankoff and Kelly- but makhaḍa usually means “churn.”

Stefan Georg:

Curent IEist wisdom adds to Latin and Greek now (more and more confidently) arm. /katʽn/ (no, it doesn’t look obvious, decent Armenian etymologies never do) – if correct, this would add one more witness to Latin and Greek (note in passing that no less a figure than Szemerényi viewed L. as a loan from Gk., but this is generally rejected, and probably should be), which would *slightly* increase the possibility of its PIE status – but of course still doesn’t rule out a „mediterranean“ (cover term for „loan from an unknown ultimate source from the geographical region roughly defined by this geographical term) origin. If you are familiar with the discussion around Indo-Aryan Bangani from the 1990s, you will know that this language is, to say the least, problematic for IE comparisons (but one should know the whole discussion, not only the voices of the detractors, but also those of the supporters of Zoller’s work) – so, whatever this may mean (and it may mean much or nothing), one should note that /lɔktɔ/ has been recorded in this language, which, if true and significant, would change the picture some more in the direction of PIE. The literature (among which some handbook-ish publications, Gamkrelidze / Ivanov, Mallory / Adams etc.) also point out Hittite /galaktar/ „soothing substance, balm, nutriment“ – which then *would* score the last-second goal (or, for our American friends, the „Catch“, Dwight Clark, RIP!) for PIE, but this is, again, generally rejected by connoisseurs of Anatolian like Kloekhorst and probably others (this might rather denote the white liquid of the opium poppy – admittedly something milky (but also certainly something soothing), and is probably to be derived internally within Hittite, so very probably no PIE – and thus no „Catch“ – here).

Chris Button:

H. W. Bailey, in his "Khotanese Texts" (vol VII, 1985), agrees with Pulleyblank that 'araq should be kept separate. The Burmese evidence is interesting in this regard since Modern Burmese /əjeʔ/ "liquor" goes back to Written Burmese /ərak/ and is clearly identified as a loan, first attested in the 16th century, from Mon via Malay and ultimately from Arabic by Hla Pe (1967). What's interesting is that there is another Written Burmese word /rak/ "liquid extract" which occurs in compounds and appears to go back to /rjak/ since Inscriptional Burmese /rj-/ merged with Written Burmese /r-/. It occurs in Inscriptional Burmese in the compound /rjak tak/ "buttermilk", the second syllable of which Hla Pe notes to come from Pali /takka/ "buttermilk" again via Mon and first attested in the 15th Century. Since prefixal /ə/ is often a nominalizer in Burmese, superficially /ərak/ as a noun seems to be related to /rak/ in a compound noun as if the disyllabic loanword has been reanalyzed internally as a /ə/ prefix on /rak/. However, this doesn't support the fact that /rjak/ seems to be attested a century earlier and it has a medial /j/ unlike /ərak/ in which one would not be expected based on its Arabic origin. Although the date of attestation does not rule out it having occurred earlier without any written record, and the presence of a /j/ could theoretically be erroneous since /rj-/ and /r-/ show vacillation in the inscriptions (yet we do have Kuki-Chin /hriak/ "grease" to support it), it does all tend towards the notion that we have two separate morphemes in Burmese: /r(j)ak/ and /(ə)rak/. As if that's not confusing enough, Gordon Luce in his "Comparative Wordlist" (1985) compares /rjak/ with 液 *làk which is suspiciously similar to 酪 *rák.

"Indo-European (*)*ǵalak- 'milk'", Paleoglot (3/26/10):

This is another rant about hideous Proto-Indo-European roots still reconstructed in the 21st century that should have been dumped in the 1960s along with Woodstock. I love Proto-Indo-European (PIE) and am fascinated by it but I also hate unjustified reconstructions and "junk linguistics". On that subject, let's now talk about (*)*ǵalak- or similar forms designated as the PIE word for 'milk' in addition to the more substantiated root *melǵ-. I don't think a person truly understands PIE until they recognize the myriad of shoddy reconstructions out there in its name that need to be dismissed.

We have Greek γάλα (gen. γάλακτος) & γλάγος 'milk' along with Latin lac 'milk' (gen. lactis). Based primarily on this, Douglas [VHM:  Mallory?] and Adams have reconstructed *ǵl̥lákt- , have then attempted to add dubious cognates from Indo-Iranian loaded with assumptions, and have concluded (or merely asserted without firm basis rather) that "[…] both the archaic morphological shape and the geographical distribution would seem to guarantee this item as at least a regional word in PIE."[1]

In my view, a more reasonable, alternative view is that the Greek and subsequent Latin forms are from Hittite kalaktar meaning more generally 'nutriment'[2] and have nothing to do with PIE at all. This would be one of those Greco-Anatolian Wanderworts which spread during the 2nd millennium BCE along μέλι 'honey' which I've just talked about before. Whether directly or through an intermediary, this must be where internal -kt- comes from while word-final -r has been deleted in the Greek loan. The Latin form must then be from Greek. The word even finds its way into Egyptian as ỉrṯ.t 'milk' (*yarāṯat /jəˈɾɑ:cəʔ/).[3] We know that the word must be from Hittite or similar Anatolian dialect because it can be further derived from the native verb root kala(n)k- 'to soothe, satiate, satisfy'. Reconstructing a protolanguage root that's unanalysable despite an etymology already available with a clear historical source is the kind of sloppy, unacademic nonsense I loathe with a passion.

[1] Douglas/Adams, Encyclopedia of Indo-European culture (1997), p.381 (see link).
[2] Puhvel, Hittite Etymological Dictionary: Words beginning with K (1997), p.19: "kal(l)aktar, galaktar (n.) 'soothing substance, balm, nutriment'" (see link).
[3] Woodard, The Ancient Languages of Mesopotamia, Egypt and Aksum (2008), p.181 (see link).

J. P. Mallory, D. Q. Adams, Encyclopedia of Indo-European Culture (London and Chicago:  Fitzroy Dearborn, 1997), pp. 381-382:

*ǵ(l̥)lákt (gen. *ǵlaktós) ‘milk’. [IEW 400–401 (*glag- ~ *glak-); Wat 41 (*g(a)lag- ~ (g(a)lakt-); GI 85; Buck 5.87].  Lat lac (gen. lactis) (< *lakt < *dlakt with regular reduction of dental stop + -l- cluster < *glakt with regular dissimilation) ‘milk’, Grk gála (gen. gálaktos) ‘milk’ (with generalization of the Lindeman variant *gl̥lakt), glaktophágos ‘living on milk’ (without the *-t we have glakôntes [pl.] ‘full of milk’, glágos [with voicing assimilation] ‘milk’), Hit galaktar (= /glaktar/) ‘sap, milky fluid from trees and plants’.  Since Latin, Greek, and Hittite are all centum languages the reconstructed initial is ambiguous; it could be *g- or *ǵ-.  If the latter, it is very tempting to add the various Nūristāni words for ‘milk’: Ashkun zō, Kati zu, Tregami dzor, Waigali zōr.  These words reflect a Proto-Nūristāni *dzara-, Proto-Indo-Iranian *ź(h)ara- or *ź(h)r̥ra-. A *źr̥ra- would match Grk gála exactly.  There is also an Ancient Chinese *lak ‘dairy product, cottage cheese, or similar commodity, imported from northern barbarians’ that would appear to reflect an even older Chinese *g/krak or the like and it has been suggested that this word reflects a borrowing on the part of Chinese from some IE group in eastern Central Asia.  With or without the evidence from Chinese, both the archaic morphological shape and the geographical distribution would seem to guarantee this item as at least a regional word in PIE.  Possibly the original noun ‘milk’ since it has no known root connections within PIE.

In their whole section on "Milk" (pp. 381a-383b), Mallory and Adams "reconstruct a rich vocabulary for PIE concerning milk and milk products, a testimony to the importance of these things to a people who were heavily dependent on animal husbandry for sustenance." (p. 381a). Terms covered include those for "to milk", "milk", "coagulated (sour) milk", "(skim) milk, whey", "rich in milk", "milk (butter)", "milk (ghee)", "buttermilk", "cream", "butter", and "curds, curdled milk".  In the latter part of their article, Mallory and Adams discuss the archeological, economic, and cultural aspects of milk and milk products among IE peoples.

At this point we still have no hard and fast conclusions, but a goodly amount of data and some interesting leads toward understanding the connections between lào, luò 酪 ("fermented milk; yoghurt; sour milk; kumiss") and various non-Sinitic sources, including possible IE terms for milk and milk products.

Bottom line for this post:

The various milk products grouped together as lào, luò (reading pronunciation) 酪 ("fermented milk; yoghurt; sour milk; kumiss"); Old Sinitic (OS) /*ɡ·raːɡ/ (Zhengzhang) have, at best, a low level of alcohol content, so they are unrelated to Arabic 'araq and its cognates in other languages such as Mongolian araki, ariki, which refer to distilled liquors.


[Thanks to J. P. Mallory, Douglas Adams, Patrick McGovern, Georges-Jean Pinault, Wolfgang Behr, Françoise Bottero, Petya Andreeva, Gene Hill, and Annie Chan]


  1. Nick Kaldis said,

    January 8, 2019 @ 10:28 am

    There is an online short lecture on the etymology of the Greek word for milk, I cannot vouch for its accuracy, but it does include some claims that will be of interest to your investigation:

    "Etymology οf the word MILK (ΓΑΛΑ)"

  2. Nicky said,

    January 8, 2019 @ 5:19 pm

    There was recently study that combined ancient DNA and protein analysis about earliest signs of dairying in Mongolia:

  3. Victor Mair said,

    January 8, 2019 @ 8:21 pm

    From Brian Spooner:

    In Central Asia there's also kefir (which I now buy every week in Whole Foods!), and I've been wondering how I might learn more about the history of that. When I was living with nomads and people in small isolated villages in eastern Iran back in the 60s and 70s I was fascinated by their relationship with milk, and especially how they processed it so that it never went bad. I produced a special issue of Expedition on it in 1980. You might find these pages interesting:

    "Pastoral Production: Milk and Firewood in the Ecology of Turan" (pdf), by Mary Martin

    VHM: This invaluable article, though short, introduces 15 different types of milk products and is accompanied by 9 clear photographs that illustrate how they are made, as well as the gathering and use of scarce firewood, which is essential for making milk products under the circumstances in which the nomads of eastern Iran lived.

    The article was written by Mary Martin during my second year at Penn. She was working in the Near East Center just down the hallway from me at that time. I knew that she had expertise in certain aspects of nomadic culture, but I had no idea that it was in such an esoteric subject as the making of dairy products in remote villages of eastern Iran.

  4. David Marjanović said,

    January 8, 2019 @ 9:01 pm

    The book chapter by Garnier, Sagart & Sagot was published in 2017 and is freely accessible here. Unfortunately I haven't finished reading it myself, but it looks very good so far.

    I'm pretty sure what Vovin means by G is [ɢ], the voiced uvular plosive – an unstable sound, highly prone to becoming a fricative or approximant that can then disappear, less prone to becoming [g].

    Armenian and Greek are probably pretty closely related, so the Armenian /katʰn/ does not substantially increase the probability that the Latin/Greek word goes down to PIE. I would not at all be surprised to learn that Gordon had this one right and it's a Wanderwort originating in Hittite.

    That can't explain the Nūristāni forms, but those lack any trace of *-kt-, so there's no good reason to think they're related. The Bangāṇī word, on the other hand, is tantalizing; a lot more work needs to be done on that language.

    Anyway, I agree that the meaning of 酪 makes it likely that it's a loan, but IE is not a good place to look for its source.

    "Etymology οf the word MILK (ΓΑΛΑ)"

    No, -λκ- does not randomly become -λγ- in some Greek words but simultaneously not in others; no, mulgere is not a loanword from Greek or anywhere; no, milk isn't a loanword from Latin or anywhere. The whole video is just pseudoscience. Words aren't related just because they're vaguely similar.

    There was recently study that combined ancient DNA and protein analysis about earliest signs of dairying in Mongolia:

    Thank you! In short, dairy products were staple food in Mongolia in 1300 BC, for people who were as lactose-intolerant as today's Mongolians and who were not "Indo-European" (Yamnaya/Afanasievo) in the rest of their genetics either. What kind of language they did speak is anybody's guess.

  5. Chris Button said,

    January 8, 2019 @ 10:21 pm

    the Arabic etymon is definitely 'araq, a DISTILLED alcoholic drink, and its underlying meaning is 'sweat'— = droplets forming on the still.

    This makes me wonder if Old Chinese 液 *làk is related to Arabic 'araq.

  6. Chris Button said,

    January 8, 2019 @ 11:42 pm

    I'm pretty sure what Vovin means by G is [ɢ]…

    Yes I wonder where he is getting a uvular from? Baxter & Sagart don't reconstruct 酪, but I don't think they would use one there.

    … an unstable sound

    It's certainly a rare sound. It's also why it is highly unlikely that Baxter & Sagart's reconstruction of an onset *ɢ- is correct in Old Chinese. At least *ʁ- would have been more plausible which I would reconstruct (although in different environments from Baxter & Sagart) and which, following Pulleyblank, also occurs in coda position.

  7. Jonathan Smith said,

    January 9, 2019 @ 12:31 am

    According to this new article, early words in phonetic series were in general disyllables of form Kərak, etc., from internal , in which light the idea of a source like Mo. *agirag or some such looks highly cowpoo 靠谱.

  8. Jonathan M Smith said,

    January 9, 2019 @ 12:31 am

    *phonetic series 各

  9. Jonathan Smith said,

    January 9, 2019 @ 12:32 am

    *from internal principles
    dodgy author though

  10. Philip Taylor said,

    January 9, 2019 @ 5:00 am

    David M ("In short, dairy products were staple food in Mongolia in 1300 BC, for people who were as lactose-intolerant as today's Mongolians and who were not "Indo-European" (Yamnaya/Afanasievo) in the rest of their genetics either"). I am missing something here, but I cannot see what. How can dairy products be a staple food for people who are lactose-intolerant ?

  11. Anne Mendelson said,

    January 9, 2019 @ 6:17 am

    How can dairy products be a staple food for the lactose-intolerant? By fermentation, of course. The nomadic peoples of Central Asia always consumed the milk of their animals in fermented form. This was especially crucial with mares' milk, which has much more lactose than milk from goats or cows and creates more alcoholic fermentations, partly through the action of yeasts.. With other dairy animals, most of the fermentation is through lactic acid bacteria.

  12. Philip Taylor said,

    January 9, 2019 @ 7:39 am

    Ah, right, understood. Thank you, Anne.

  13. Victor Mair said,

    January 9, 2019 @ 8:45 am

    From Anne Mendelson:

    I’m a culinary historian with no training in Sinitic etymology, but at the moment happen to be writing a book about the history of milk in Western societies where lactose tolerance or adult lactase persistence is at least the perceived norm. My general thesis, or bird in the head, is that the consumption of cows’ milk in the form of fresh unfermented drinking-milk — i.e., full-lactose milk — is one of the great nutritional aberrations in the history of food, an eighteenth-century English dietary fad that made an extraordinary transition to mainstream medical doctrine.

    One of the consequences has been to obscure the fact that in the original lands of dairying in the Middle East and pasts of Central Asia, milking took place only during the warm seasons when local daytime high temperatures triggered very rapid spontaneous fermentation by lactic acid bacteria (L.A.B). The peoples most dependent on milk regularly consumed it in fermented (or naturally lactose-reduced) forms similar to yogurt and koumiss, which seriously inhibited most pathogens and allowed people who lacked adult lactase persistence — meaning everybody — to consume milk without digestive problems.

    Cheesemaking (mentioned by a few people in the thread) doesn’t enter the picture to any great extent — certainly not the classic European cheeses, which require cool conditions for complex sustained fermentations. Rapid L.A.B. fermentations, or at most the combined action of bacteria and some yeasts to produce some alcohol, would not have been rocket science for Asian nomadic peoples, though they may look like it to people who never encounter milk in anything but supermarket cartons. It makes sense that among some peoples the words for milk and fermented milk would have been the same, because milk became fermented milk almost as soon as drawn from a goat’s or mare’s udder.

  14. Chris Button said,

    January 9, 2019 @ 2:15 pm

    I wonder if the -j- in Inscriptional Burmese rjak (lost in Written Burmese to merge with unrelated rak in ərak "liquor" from Arabic 'araq) of rjak tak "buttermilk" is connected to some kind of metathesis of the /y/ (i.e. /j/) component in Mongolian ayraq? If there is a connection to Old Chinese 酪 *rák then the -j- cannot have been original there since we would then expect Old Chinese *rác (from **rákʲ < **rják) in accordance with the evidence (at least as I see it) that any pre-OC palatal (and separately labial) elements were thrown over onto any velar codas shifting -k, -ŋ, -ɣ to -c, -ɲ, -j respectively.

    @ Jonathan Smith

    Thanks for letting us know about your article. I'd love to read a copy. Just looking at the abstract, I'm encouraged that you are bringing Ferlus' work into consideration. His is one of the very few proposals for the Type A/B distinction that I find somewhat convincing. It actually chimes well with Pulleyblank's ruminations on the topic, although I'm not sure whether either of them realized it.

    In brief, Pulleyblank reconstructs a prosodic accent that he compares with Stern’s moraic analysis of Sizang (Northern Kuki-Chin language) accordingly: prominence on the first mora, where Sizang has secondarily lengthened tense vowels, led to the insertion of a high vowel in Old Chinese either replacing the nucleus or forming a diphthong with it; prominence on the second mora, where Sizang has secondarily lengthened (i.e. tense) sonorant codas, resulted in no alternation. As typological support, Pulleyblank adduces the vowel alternations associated with register distinctions in Mon-Khmer languages on the basis that the Old Chinese prosodic accent and the pitch changes associated with onsets in Mon-Khmer could have had a similar effect on the nucleus. Although Ferlus’ related attempt to attribute the split to an actual fortis-lenis contrast in Old Chinese onsets is somewhat problematic since his proposal for such widespread prefixation is largely unsubstantiated, the common phonological ground with Pulleyblank’s proposal hints at a possible indirect association via analogical extension.

  15. RfP said,

    January 9, 2019 @ 3:19 pm

    @ Anne Mendelson:
    When discussing lactose intolerance, it's important to keep in mind that unpasteurized cow's milk contains lactase, "an enzyme that catalyzes the hydrolysis of lactose to glucose and galactose."

    Because pasteurization destroys lactase, lactose intolerance is a much greater issue in today's developed countries, even beyond the lack of fermented milk products in our normal diets.

    The nutrient dense food movement (including people like Jessica Prentice, who coined the term "locavore", which was Word of the Year ten or twelve years ago) has been doing everything they can to help people understand these issues, with a lot of the work focused in the Weston A. Price Foundation, who are involved in promoting the use of raw milk for many reasons, including this one.

  16. Jonathan Smith said,

    January 9, 2019 @ 5:05 pm

    @Chris Button Pulleyblank is definitely always worth approaching seriously. Ferlus's phonetic interpretation is speculative and yes the parallel to register in MK is dubious; my goal is to point to some MC onset distributions within phonetic series that are arguably interpretable in terms of his 1998 idea (which was based on the tendency of graphic loaning and apparent derivations by prefixation to proceed B > A.) If you like email me ( and I'll send you the article, or alternatively you can find a draft on

  17. Anthony said,

    January 9, 2019 @ 8:43 pm

    A recent discussion in Science:

  18. Chris Button said,

    January 9, 2019 @ 10:42 pm

    Yes I wonder where he is getting a uvular from? Baxter & Sagart don't reconstruct 酪, but I don't think they would use one there.

    Turns out Baxter & Sagart reconstruct 液 *làk as *ɢrAk so perhaps there is some confusion there? As an aside, ignoring the inherent issues mentioned above regarding the likelihood of an actual *ɢ- onset in OC, I wonder how Baxter & Sagart can then account for their reconstruction of 狄 as *lˤek whose 火 component is shown by bronze forms to be a deformation of a phonetic 亦 (腋).

    @ Jonathan Smith

    Thanks – will do!

  19. David Marjanović said,

    January 10, 2019 @ 5:12 am

    On Type A/B syllables and initial uvulars, this paper promises interesting things.

  20. Chris Button said,

    January 10, 2019 @ 1:03 pm

    Pulleyblank, Edwin G. 1962. The Hsiung-nu Language. Asia Major, New Series. Vol. IX, pt. 2, pp. 239-265.

    It's probably worth noting that this is actually the "appendix" to his "The Consonantal System of Old Chinese (Part II)" which starts on p.206.

    Regarding /ɢ/, on p.253 Pulleyblank says the following: "For the hypothetical Hsiung-nu form, we should prefer to reconstruct a monosyllable, something like *ɣrak or *ɢrak". However, his OC system at the time does not include uvulars which he explicitly rejects. Later, in 1977-8 he introduces uvulars as codas and then raises the possibility of them as onsets in 1982 (further discussed in his "ablaut" paper of 1989). However, by the time of his 1991 "Ganzhi" paper he has abandoned uvulars altogether. Personally I like (with certain limitations) the approach in his 1989 paper where he extends the association of uvulars with labialization in coda position to onset position.

  21. Jonathan Smith said,

    January 10, 2019 @ 3:40 pm

    @ David Marjanović Do you mean "and pharyngealized onsets [in A]"? The way Baxter & Sagart (2014) use uvular onsets is orthogonal to A/B, thus contrasting q qˤ qh qhˤ ɢ ɢˤ (in addition of course to k kˤ kh khˤ g gˤ).

  22. David Marjanović said,

    January 13, 2019 @ 3:18 pm

    I know; Gong's paper touches on both of these (and their interaction in Tangut, not in Sinitic).

    Incidentally, I read your recent paper today! It looks promising. I hope you can eventually find more morphological prefixes that could explain the seemingly chaotic pattern where clusters sometimes simplify and sometimes don't.

RSS feed for comments on this post