Correspondences between Ancient Greek doȗle (voc.) 'slave' and 奴隷 Jpn dorei / Tw lô·-lē
« previous post | next post »
[This is a guest post by Chau Wu]
The word 奴隷 Jpn dorei (ドレイ) / Tw lô·-lē ‘slave’ is of great interest to me. My study of West-to-East lexical loans suggests that the origin of this word is Ancient Greek δοȗλos (doȗlos, m.) and δοȗλα (doȗla, f.), which mean ‘slave’. The figure below is a funerary stele of Mnesarete, daughter of Socrates (not the philosopher), showing a female servant facing her deceased mistress. There are some other terms for slave in Ancient Greek, depending on the context, but doȗlos and doȗla are historically the most commonly used, from Mycenean, Homer, Classical, Koine, down to Modern Greek.
Figure. Funerary stele of Mnesarete, daughter of Socrates (not the philosopher*); a young servant (left) is facing her dead mistress. Attica, c. 380 BC.
(From Glyptothek, Munich, CC BY-SA 3.0 via Wikimedia Commons)
*"Grieve for Mnesarete", Ancient World Magazine (6/20/18).
The forms doȗlos (m.) and doȗla (f.) are the nominative case; the vocative case for both genders is δοȗλε (doȗle). I believe it is this form that slave owners would use to call their slaves if not by name. And this is most likely the form heard and taken to be the word for ‘slave’ by foreign borrowers. Hence, we have: Anc.Gk. doȗle > [Intermediate(s)] > Jpn dorei. Taiwanese lacks the voiced plosive /d/; therefore, foreign d is usually substituted with an l (or sometimes with a t).
Thus, an equation may be proposed for the loan transfer:
Anc. Gk. doȗle (voc.) > [Intermediate(s)] > Jpn dorei /Tw lô·-lē 奴隷
However, things may not be as simple and naïve as the above equation suggests. The current situation shows that an n- initial for 奴隷 distributes widely in the vast majority of Sinitic topolects (e.g., MSM núlì) as well as in Sino-Korean (noye 노예) and Sino-Vietnamese (nô lệ). This is also reflected chronologically as early as in the Buddhist transcription data of Eastern Han dynasty, showing 奴 being used to transcribe for the Indic [no] sound, e.g., Skt. anomiya / Pali anomā is transcribed as 阿奴摩 (*ˀa-no-ma >) ˀȃ-nwo-mwȃ (Coblin, W.S., 1983, A Handbook of Eastern Han Sound Glosses, p. 254). [VHM: Wisdom Library] So, the n- initial is widely and deeply entrenched in Sinitic.
To account for the non-nasal initial in the Japanese and Taiwanese data, there are two possibilities. One is that they represent descendants of an early stage of borrowing before nasalization took place. It is known that the Min topolects are the most archaic group of Sinitic languages, having split off from the mainstream during the Qín 秦 and Hàn 漢 dynasties, around second or third centuries BC (Schuessler, A., 2007, ABC Etymological Dictionary of Old Chinese, p.1 & p.125). They are thereby unaffected by the later nasalization. I did not know how to explain the Japanese data, but now with John Whitman’s excellent explanation of d ~ n alternation (in the previous post on this subject), which is related to the second possibility (vide infra), my problems are resolved.
The other possibility is that both Japanese and Taiwanese are descendants of Middle Chinese which already has an n- initial. Pulleyblank reconstructs nɔ-lεjh for 奴隷 in Early Middle Chinese (Pulleyblank, E.G. 1991, Lexicon of Reconstructed Pronunciation in Early Middle Chinese, Late Middle Chinese, and Early Mandarin, p. 227 & p.189). In Japanese and Taiwanese, 奴 has undergone a secondary de-nasalization, resulting in the present non-nasal forms. If the Anc. Gk. doȗle is taken as the source for 奴隷, this is in essence a round-about route from d- to n- and then back to d- again (and Tw l-).
A similar situation arises for the proposed derivation of the Sinitic word for ‘cow’ 牛 niú from PIE *gwou- (nom. sg. *gwōus) as discussed in a previous post in Language Log.
Understanding the interplay between nasalization and de-nasalization in loan processes may hold the key to unlock a treasure trove of secret relationships between European source words and their Asian descendants.
Selected readings
"Sino-Japanese n- / d- initial interchange" (101/25)
"Stay hyDRAEted" (9/29/25)
———————————
Afterword on Mnesarete
VHM
The Mnesarete (meaning 'remembering virtue'), mentioned above as the daughter of a certain Socrates, is not the same as the other famous Mnesarete, better known as Phryne, mentioned in Athenaeus' The Deipnosophists.
Phryne: The Ancient Greek Courtesan Who Disrobed For Her Freedom – GreekReporter.com
Theodoros Karasavvas (April 16, 2025)
|
|||||
| Artist | Jean-Léon Gérôme | ||||
| Year | 1861 | ||||
| Medium | Oil on canvas | ||||
| Dimensions | 80.5 cm × 128 cm (31.7 in × 50 in) | ||||
| Location | Kunsthalle Hamburg, Hamburg | ||||
_-_01.jpg)
Wikipedia — and see this article for other artistic renditions of this celebrated scene
Click here for a large, high-resolution reproduction of this painting. Note how stunned the judges are upon beholding her naked body.
_-_01.jpg){kind=link}
A depiction of Phryne, a famous hetaera (courtesan) of Ancient Greece, being disrobed before the Areopagus. Phryne was on trial for profaning the Eleusinian Mysteries, and is said to have been disrobed by Hypereides, who was defending her, when it appeared the verdict would be unfavourable. The sight of her nude body apparently so moved the judges that they acquitted her. Some authorities claim that this story is a later invention.
Because of the intrinsic interest of this gripping story, and to clarify certain confusing aspects it bears with the illustration above, we quote the account as given in the Greek Reporter:
Phryne the Thespian was a notable ancient Greek hetaira, or courtesan, of Athens, who is remembered throughout the millennia for her dramatic trial which she won by baring her naked body.
Her real name name was Mnesarete, but people referred to her as Phryne (“toad”) because of the yellow undertone of her skin.
Her story has survived for thousands of years with the famous model and courtesan becoming a symbol of freedom against sexism, as well as repression disguised as piety.
Phryne was born around 371 BC in Thespiae (Boeotia) but spent most of her life in Athens. Because of her stunning looks, she became a model, posing for various painters and sculptors, including Praxiteles, who was also one of her most frequent clients.
Unlike most Athenian women, who rarely left their homes and had very little voice in society, courtesans like Phryne were granted much more freedom.
They could leave the home and were seen as educated and intelligent so that they could have engaging discussions with their clients.
One of the statues Praxiteles modeled after Phryne, the Aphrodite of Cnidus, was purchased by the city of Cnidus in Kos after the city that had originally commissioned it objected to its being a nude. The statue became such a notable tourist magnet that the city managed to pay off its entire debt.
Phryne’s beauty also became the subject of many ancient Greek writers, who praised her looks, with Athenaeus openly worshiping her in his work titled The Deipnosophists. From this work we also know that Phryne was the wealthiest self-made woman in all Athens at the time.
She became so rich and powerful during her lifetime that she even proposed paying for the reconstruction of the walls of Thebes, which had been destroyed by Alexander the Great in 336 BC.
Intimidated by the idea that a female model and courtesan could restore what a great king like Alexander the Great had destroyed, Phryne’s offer was rejected by the local authorities of Thebes, and the walls remained in their ruined condition.
Regardless of her incredible wealth and beauty, and prominent clients, what keeps the memory of Phryne alive to this day is her famous trial.
According to Athenaeus, Phryne was prosecuted on a capital offense and was defended by the orator Hypereides, one of her lovers. Athenaeus does not specify the nature of the charge, though some other historical sources state that she was accused of profaning the Eleusinian Mysteries.
Although there is great debate among scholars about what really happened that day in court, Athenaeus wrote that Hypereides tore off Phryne’s dress in the middle of the courtroom to show the judges her beautiful body.
His reasoning was that only the gods could sculpt such a perfect body; thus, killing or imprisoning her would be seen as blasphemy and disrespect to the gods.
What appeared to be an unfavorable verdict for Phryne turned into a glorious victory for her after the inspired action of Hypereides.
Phryne walked out the court triumphant, and her story went on to inspire many works of art, including the iconic painting Phryne before the Areopagus by Jean-Léon Gérôme (1861) and the sculpture Phryne Before the Judges, by Albert Weine, from 1948.
Additionally, Baudelaire wrote two poems about her, the composer Saint-Saëns wrote an opera about her (Phryne, 1893), and several modern writers have penned novels about her controversial trial.
There's an embedded video near the end of the above account which has a lot to say about community forum deliberations in Athenian life and the role of slaves in an Athenian household.
wgj said,
October 9, 2025 @ 8:56 am
The earliest plausible path for a Hellenic word to enter the Sinosphere must be via Alexander's campaign into India, and then the spread of Buddhism into China. In other words, the same way those Greek elements showed up in Chinese visual art (most visibly the wavy outlines of clothings on statues). If that's how the word 奴隸 came to be, then what else were slaves called prior to the Han dynasty?
Chau said,
October 9, 2025 @ 10:00 am
wgj: "If that's how the word 奴隸 came to be, then what else were slaves called prior to the Han dynasty?"
I know of 嬯 and 儓 (variant forms of the same word). Please see, for examples:
Wiktionary on 嬯 (OC *tə̂; Zengzhang: /*dɯː/), 'servant, slave woman' (Bronze Inscription)
https://en.wiktionary.org/wiki/%E5%AC%AF#Chinese
Wiktionary on 儓 (OC *tə̂; Zhengzhang: /tʰɯːs, *dɯː/), 'the lowest kind of servant' (according to Zuo Zhuan)
https://en.wiktionary.org/wiki/%E5%84%93#Chinese
Note that the reconstructed OC pronunciations for both forms, especially the d- initial, correspond to the first syllable of Anc. Gk. doȗle (voc.) 'slave'.
J.W. Brewer said,
October 9, 2025 @ 11:00 am
A quick glance suggested that one theory is that it came into the Greek lexicon as a loanword of Semitic origin and thus has no cognates in other IE branches, although apparently someone argues for a Sanskrit cognate bearing (as can of course happen with cognates) somewhat different meaning.
But in any event one other obvious question is whether there are any relevant words in any of the other numerous languages spoken along the Silk Road in between Greek-speaking areas and Sinitic-speaking areas that could have come from a borrowing of δοȗλos. If there's no cognate in e.g. the Iranian languages that should make identifying loans easier.
TR said,
October 9, 2025 @ 11:20 am
Small correction, δοῦλε is the masculine vocative only; the feminine is δούλη, same as the nominative.
wgj said,
October 9, 2025 @ 11:29 am
奴 has already existed in oracle script (嬯 and 儓 obviously not) and is almost certainly directly related to 女. Are you suggesting 女 has a Hellenic origin?! If so, the entire Chinese civilization might as well be Greek.
Jonathan Smith said,
October 9, 2025 @ 11:51 am
Taiwanese lú 女 'woman' is from Indo-European (?) *dhugh(ə)ter via [Intermediate(s)] such as *du- (Wu 2016 [?]: 151).
Chau said,
October 9, 2025 @ 11:54 am
Many thanks to TR for the correction.
katarina said,
October 9, 2025 @ 12:10 pm
Wiktionary gives the Old Chinese (1250 BCE to 250 AD) pronunciation of 奴 NU "slave" as */na:/. It gives an image of the character 奴 "slave" in Shang dynasty Oracle Bone Script (1250-1046 BCE), which cannot be copied here.
Baidu baike 百度百科 (online) shows the development of the character 奴"slave" from the Western Zhou dynasty (1046-771 BCE) to the present:
https://baike.baidu.com/pic/奴/4993379/0/10dfa9ec8a13632760f3ed6b9e8fa0ec08fac74f?fr=lemma&fromModule=lemma_content-image
Victor Mair said,
October 9, 2025 @ 12:37 pm
"Daughter of Holy Cow" (11/18/20) — with many scholarly references
"More on Persian kinship terms; 'daughter' and the laryngeals" (3/18/20)
"Turandot and the deep Indo-European roots of 'daughter'" (3/16/20)
Chris Button said,
October 9, 2025 @ 2:01 pm
Meanwhile I still favor the proposal that 女 originally represented a slave.
David Marjanović said,
October 9, 2025 @ 2:14 pm
The development of initial [m n ŋ] into [b d~l~ɾ g] unless a nasal vowel followed is a well-known innovation of Southern Min. Here it is in Wikipedia; one of the cited works is this chapter in an edited book from 1985 which you can read for free if you create an account. The references in that paper go back at least to 1959.
It is a regular sound change, applying to every inherited word. That's why "Min" itself is Bân in Taiwanese (but, for example, Mìng in the Eastern Min dialect of Fuzhou), why 馬, which has been prominently featured on LLog a few times, is bé in Taiwanese (but mā in Fuzhou), and why 語, a word with *ŋ-, is gí or gú in Taiwanese (but ngṳ̄ in Fuzhou). For more examples see the Wikipedia article and the papers.
Chris Button said,
October 9, 2025 @ 3:08 pm
@ David Marjanović
Not necessarily an innovation. It could be inherited in the same way it was inherited in Sino-Japanese. Check out the first link in the selected readings above. Regardless, it does not alter your overall point.
Jonathan Smith said,
October 9, 2025 @ 7:06 pm
It is always possible to make stuff up, but normal principles have that this is an innovation, in particular "denasalization at a post-Proto Coastal Min date" which is "one criterion for separating [Northeastern Min] and [Southern Min-Hingua] into separate languages". This from Bodman (1985: 9), or see literally any pertinent study.
David Marjanović said,
October 9, 2025 @ 7:57 pm
The denasalized Kan-on words were borrowed into Japanese in denasalized form from a Sinitic lect that had undergone this denasalization (independently of Southern Min).
Southern Min did it with inherited words (as Seoul Korean is doing right now, independently again). So I don't understand what you mean.
Bodman '85 is the book chapter I linked to.
Chris Button said,
October 9, 2025 @ 8:50 pm
@ David Marjanović
In his "Late Middle Chinese" (1970) article, Pulleyblank says the following:
"On the other hand, the pronunciation of nasal initials as prenasalized stops in T'ang, which gave, for example, Kan'on ba in contrast to Go-on me for the word 'horse', was very likely a peculiarity of the Ch'ang-an dialect, no doubt pre-existing the T'ang period, but one that did not spread to other regions. A possible exception would be the Southern Min area, if Forrest is right in attributing the allophonic alternation between voiced stops and nasals found in that region exclusively to what he calls T'ang-Min, that is, the literary pronunciation of characters imported into the area during T'ang. On the other hand it may be that this feature of Southern Min and the prenasalized voiced stops of T'ang (which still survive in some Shansi dialects) both reflect a type of pronunciation which had once been more widespread and may have relevance, therefore to Old Chinese, or proto-Chinese."
Jonathan Smith said,
October 9, 2025 @ 9:40 pm
Re: that, one may consult e.g. same Bodman paper (p. 15) — "[contra Forrest 1965] [t]he Amoy forms with denasalized initials are not at all limited to the literary or character readings but are typical also of the popular or colloquial stratum which reflects a much earlier origin" — or anyone who knows how to say e.g. 'month' 'person' 'six' 'not have' etc. in Taiwanese. Re: Old Chinese, yes features of modern languages may indeed be relevant to it.
Martin Schwartz said,
October 9, 2025 @ 10:38 pm
For Greek doûlos, see the comments in the excellent 2014
languagehat doulos. @Jonathan Smith: Why take a
Taiwanese word for 'woman' from an unspecified form of the PIE
'daughter' word?
Nelson Goering said,
October 10, 2025 @ 1:12 am
Out of curiosity, does anyone know when the disyllabic 奴隸 is first attested in Chinese?
Chris Button said,
October 10, 2025 @ 5:19 am
@ Jonathan Smith
And with allophonic variation in one place, you are even more liable to analogical extension in another place …
No-one is saying it absolutely wasn't innovative. The point is that it might not have been innovative.
Chris Button said,
October 10, 2025 @ 5:26 am
And as Pulleyblank says: "… no doubt pre-existing the T'ang period … a type of pronunciation which may have once been more widespread… "
Lucas Christopoulos said,
October 10, 2025 @ 6:02 am
It seems that during the Western Han period (206 BCE – 9 CE), 奴隸 had stabilized as the regular term for slave, distinguishing such people from servants (僮僕), convicts (徒), or indentured dependents (客 or 庶人). Was it a borrowing from the Greco-Bactrian political system? Possible, as the new term marks the moment when slavery became a recognized and legal institution in imperial China during the Western Han. The word 奴隸 did not yet appear as a compound in Qin inscriptions or bamboo texts, but both morphemes were in full administrative use. 奴 (nú) appears frequently in Qin legal and administrative texts, meaning male servant, captive, or slave-like dependent.
Example (Shuihudi Qin lü zaji 秦律雜抄, ca. 217 BCE):
「人有奴者，毋得妄出其縣。」
“Those who have male servants (nú) may not send them out of the county without authorization.”The innovation of combining them—to designate a human as both subordinate and owned—was likely a Han-era lexical creation, built on Qin legal foundations that already defined those two servile categories.
Victor Mair said,
October 10, 2025 @ 7:03 am
https://languagehat.com/doulos/ is mostly about slang words for hats > police
David Marjanović said,
October 10, 2025 @ 8:31 am
That's very interesting (especially about current Shanxi), but it's from 1970; surely Bodman was familiar with it in 1985 and Norman in 1988 and 1991?
Yes; maybe the reference was specifically to this comment, which in turn links to this open-access paper about the etymology of δοῦλος (m., attested from Thucydides onward) and δούλη (f., already in the Iliad).
languagehat said,
October 10, 2025 @ 8:47 am
Martin Schwartz is presumably referring to the comments by David Curwin and Xerîb.
Chris Button said,
October 10, 2025 @ 11:46 am
I can certainly recommend "Le Doulos". Not as good as Jean-Pierre Melville's later classics, but it's still a worthy entry in his oeuvre.
Jonathan Smith said,
October 10, 2025 @ 6:33 pm
"It might not have been innovative" is influencer-think (cf. "Tylenol might cause autism"), that is, neither categorically ruled out nor positively supported by any good-faith scientific argument. Pretending reason might sway such a crowd, note further e.g. Hokkien b/l/g for EMin and other Min m/n/ng in characteristically Min (and often non-Sinitic) colloquial items that are trivially not "Tang" loans: (POJ) bô-êng : (BUC) mò̤-èng ('busy'), gû-leng: ngù-nèng ('[cow's] milk'), lâng-kheh : nè̤ng-káh ('guest'), bó͘ : muō ('wife'), etc., etc…
Yves Rehbein said,
October 10, 2025 @ 7:38 pm
隸 alone is attested in Warring States period, before Han, @ Nelson Goering, Lucas Christopoulos. So, did the Buddha come into China this early, @ WGJ?
To read 女 as a slave seems incorrect to me, @ Chris Button. The descriptions are ambiguous if it has breasts, which is the interpretation following 母, or if the lines are arms, hands folded in front, which, in addition to the kneeling pose, would be similar to the Egyptian hieroglyph (B1), and different from 奴, which has the hands behind the back, which would be similar to the respective Egyptian hieroglyph (A13), but the hand radical was added later (for clarity?).
Chris Button said,
October 10, 2025 @ 10:28 pm
@ Yves Rehbein
Qiu Xigui (1988) notes some scholars to have remarked on the graphic form being "a person kneeling with joined hands" and the "close pronunciations" of 女 and 奴.
Ferlus (1996) then tries to establish an etymological link, and Takashima (2020) has an article "Paleography, Historical Phonology, and Historical Lexicology: 'Kneeling Women with Their Wrists Crossed' and 'Slaves with Their Eyes Gouged'."
@ Jonathan Smith
I think you missed the point above about allophonic variation and analogical extension.
How about we move from the LMC pre-nasalized stops to the LMC voiced aspirated stops then? From Pulleyblank (1984):
"The voiced aspirates of LMC may have developed fully out of voiced stops of the kind that are reconstructed for EMC, but they could also represent a separate dialect tradition … In this case LMC pʱ- might not derive from earlier *b- … Unfortunately it is difficult to test this hypothesis directly in the absence of information about the pronunciation of such initials in pre-Tang dialects."
@ David Marjanović
No idea. But Karlgren was writing about it in Shanxi as well in other dialects in his "Études…" way back in 1915!
So it is most definitely nothing new. Having said that, note how Marjorie Chan begins her “Post-stopped nasals in Chinese” (1987) article:
"What has been observed in a number of modern Chinese dialects, however, is syllable-initial nasals which are followed by homorganic voiced stops… Although post-stopped nasals are described in many Chinese dialects, few linguists are aware of the post-stopping phenomenon (or 'denasalization'), and even fewer realize that the phenomenon is not restricted to two or three dialect groups, nor to a small geographic location."
Chris Button said,
October 11, 2025 @ 8:06 am
@ Lucas Christopolous
Qiu Xigui has an article entitled 奴隸 in volume 6 pages 259-267 of his collected works about the historical background of slaves. The note at the end says that it is the full version of an abridged piece that appeared in an encyclopedia entry first published in 1986.
David Marjanović said,
October 11, 2025 @ 12:20 pm
No, you didn't make that point. You just waved in its general direction. What is it? What do you think happened there, and how could it have worked?
Well, it is rather important whether they were unaware of potentially important facts or whether they thought Pulleyblank's (1970) description was wrong or otherwise irrelevant.
Where does it occur, then?
Chris Button said,
October 11, 2025 @ 1:30 pm
Pulleyblank talks about "allophonic alternation" in the citation I provide above. I simply noted how such alternation is conducive to analogical extension, which is hardly a controversial point (you can probably find some good information on how analogical change works on the internet). Having said that, it might also be an irrelevant point depending on how widespread prenasalized stops were. But as with the voiced aspirates, who knows?
Unfortunately people seem blissfully unaware of it since the detail is often omitted for simplicity in LMC reconstructions (unless concerned with Sino-Japanese). It also requires a level of interest in grassroots phonetics that has largely been missing since — dare I say it — Karlgren's time!
Even Pulleyblank (1984), who was acutely aware of it — and paid close attention to phonetic reality along with Sino-Japanese and Min evidence — takes the following approach: "One may therefore write these phonemes as simple nasals for LMC, leaving the more precise phonetic realization to be specified as required when one is dealing specifically with Tang pronunciation."
I can't possibly answer this without being overly presumptuous.
If you search for "UCLA Working Papers in Phonetics, No. #68, July 1987", you can read all about it yourself!
David Marjanović said,
October 12, 2025 @ 7:48 am
Thank you! But LMC is irrelevant to Southern Min given that the denasalization there is not limited to loanwords, as Pulleyblank apparently thought, but is regular throughout the inherited vocabulary.
Chris Button said,
October 12, 2025 @ 9:35 am
@ David Marjanović
Sorry, I'm clearly not explaining myself well enough.
To be absolutely clear, Middle Chinese (EMC/LMC) is not a director ancestor to Min. There is a myth that Min languages are therefore more conservative than other Chinese languages. In fact, they just seem to have broken off before Middle Chinese.
So, with that out of the way, two points:
1. Prenasalized stops could go back to a time before Middle Chinese (as noted by Pulleyblank in the citation above), which would render the "inherited vocabulary" point irrelevant
2. Even if the prenasalized stops emerged in the Tang koine of LMC and then spread to Min, the allophonic variation could have influenced the rest of the lexicon by analogical change,, which again would render the "inherited vocabulary" point irrelevant.
By the way, I try to write here about things that people cannot find by simply googling on the internet and then pasting in the comments here. While I can share a lot of relatively unknown or hard-to-access information on Old Chinese, Middle Chinese, and Sino-Japanese, I cannot do that for details on Min dialects.
Meanwhile, I am very much looking forward to Jerry Norman's posthumous dictionary though!
Yves Rehbein said,
October 12, 2025 @ 4:28 pm
@ Chris Button, thanks for the links. Ferlus cites “esclave, serf, slave” (par le sens
intermédiaire de “wives and dependent children”), reconstructs separate forms 奴 < k-N-(r)a, 女 < *k-N-raʔ, mechanically, and notes that the problem is that those terms all derive somehow from the graphics of a base 女 (1999). The Proto-Sino-Tibetan route seems to be uncertain with only a irregular, palatalized, Tibetan cognate and a separate word family in other branches and a separate root s.v. STEDT: "servants". Takashima's discussion is aware of 視, but I have not read it yet. I was going by "象雙手在身後交疊", MFCCDB, through translate.
Now that you mentioned it, Egyptian ḥmt B1 ("woman") and ḥm ("slave") may be related, but ḫfty A13 ("enemy") is a different word [1, 2]. Recall that .t is a feminine suffix which developed like Proto-Indo-European and Semitic gender from neutral. Notice that Linear B ⟨ra⟩ (in do-e-ra δούλη) looks similar to B1 in hieratic, provided that B1 is a determiner after the word and that it can be read rmṯ.t "women", earlier "humankind; people" [3, 4] (cf. rmṯw [1]).
1: ḥmt (B1) https://en.wiktionary.org/wiki/%F0%93%81%90
2: ḫfty (A13) https://en.wiktionary.org/wiki/%F0%93%80%8F
3: rmṯ.t (B1) https://thesaurus-linguae-aegyptiae.de/lemma/853446
4: rmṯ.t (A1-B1) https://thesaurus-linguae-aegyptiae.de/lemma/94550
Chris Button said,
October 12, 2025 @ 5:47 pm
These two comments were playing on my mind since I couldn't believe Forrest, and Pulleyblank, would be ignorant of such matters:
It turns out they were not ignorant of such matters after all. To put Pulleyblank's comment in context, here's what Forrest actually says:
p.243: "Sometimes a popular form was too deeply implanted to be uprooted by the T'ang Min, and has forced its way into the book pronunciation of the locality; rarely, the two forms seem to have changed places, as where (221), P. /ma, 'horse', has -ma as its literary form in Amoy as against the popular -be; normally the denasalized form belongs to the literary language."
p.248: "Further differences between Swatow and Amoy on the one hand and the northerly dialects on the other are seen in the treatment of initial nasals; characteristically ŋ- > g- in Amoy, Swatow, and Hinghwa, but remains elsewhere; Amoy and Swatow have popular forms in b- < ACh. m- unknown in other varieties. Analogically one might have expected in the same dialects *d- < ACh. n-; but n- was apparently saved from denasalization by changing into l-. The denasalization must, therefore, be here a relatively later phenomenon."
p.251: "The sound change m > b, etc., is an unusual one, and is seldom found to occur 'spontaneously', i.e., unconditionally. It cannot be due to local Min speech habits, otherwise we should have found it with even more regularity among the native forms than among the literary, which is plainly not so. Everything, therefore, points to the introduction of this pronunciation from north China, where also the Japanese learnt it, and, in view of the rarity of this change, it is hardly likely, though we must admit it to be possible, that it arose independently in two parts of the Chinese area. Now, we have seen in the last chapter that something like it exists in the Chin dialects of Shansi; we are not bound to believe that T'ang Chinese was imitated with perfect exactitude when taken over by Min speakers, and a substitution, sudden or gradual, of b- for mb-, d- for nd-, etc. is very easy for people unaccustomed to the more complex sound. It is not therefore to strain our actual knowledge too far to hold provisionally that a dialect derived from China was standard in China in T'ang times, and as such was taught to the Japanese inquirers after Buddhist truth, and later taken into Fukien by northern immigrants."
Chris Button said,
October 12, 2025 @ 7:04 pm
Typo in last sentence from Forrest above. It should say "… a dialect derived from Chin was standard in China …"
@ Yves Rehbein
I would reconstruct 女 as nʳàɰɁ and 奴 as náɰ
Nelson Goering said,
October 13, 2025 @ 1:32 am
"Pulleyblank talks about "allophonic alternation" in the citation I provide above."
Yes, Pulleyblank uses that word in a very brief passage — but it's not at all clear what he means by it, and I don't understand how this is actually allophonic variation in any normal sense of the term. A few examples of just what how these would be "allophonic" (in strict use, contextually determined variants of a single phoneme; in looser use, any variation in the precise phonetic realization of a particular phoneme).
It is difficult to see how borrowing could produce allophonic variation extensive enough to influence the larger phonological system, unless (maybe) there was really very widespread borrowing of doublets which differed from their inherited counterparts only with respect to this one sound, leading to a perception of these phones as in free variation — a development whose plausibility could only be assessed by comparison with similar examples (if there are any; I am not aware of any comparable case happening in any other language). Otherwise you'd expect a) assimilation of loans to the native phonology, or b) the introduction of a new contrast.
Did such extensive minimal doublet-borrowing happen? If it didn't, and Min actually shows real denasalization in non-borrowed vocabulary, then this must be an inherited change. If Forrest's characterization is accurate, and denasalization is only typical of borrowed forms, then the sound change must be only characteristic of the source-variety.
Nelson Goering said,
October 13, 2025 @ 3:40 am
My first paragraph should end by saying that a few examples *would be very useful*.
David Marjanović said,
October 14, 2025 @ 6:02 am
This looks like overkill of poor Ockham. Why propose such contortions in order to save such a munificent hypothesis instead of abandoning it for a much more parsimonious one?
And "allophonic variation" – uh, yes, the denasalization did not happen before nasal vowels (which Southern Min has and MC did not), and in part of Southern Min (evidently not in Taiwanese) it didn't happen in syllables with a nasal coda either. At some point, therefore, voiced plosives and nasals must have been allophones in the strict sense: entirely predictable realizations of "the same thing". I suppose MC loanwords disrupted this pattern and created a phonemic distinction between the two. (Loanwords turning allophones into separate phonemes has precedents: the /f/-/v/ distinction in English in word-initial position happened that way.)
[n] > [l] is, of course, a denasalization.
In large parts of Africa, [d] on one hand and [l] or [r] on the other hand are wholly predictable allophones of a single phoneme, while [b] and [g] don't participate in any such phenomenon. Also, the Wikipedia article I linked to, which seems to be entirely a summary of Norman (1988, 1991), says the denasalized outcome of *n isn't [l] all across Southern Min (as it is, apparently, in Taiwanese), but "[d~l~ɾ]", as I already said.
Jonathan Smith said,
October 14, 2025 @ 2:26 pm
Re: allophonic variation, AFAIK this remains the typical way of describing the situation for some lects inclusive of (diverse/emergent) Taiwanese, thus e.g. "/b l g/ > [m n ng]" before specific rimes/codas and phonetic studies noting that ≈[b][g]-onset > post-stopped nasals e.g. after preceding nasals. FWIW ≈[b][g] themselves are observed to feature voiceless prenasalization in some speakers/studies. Previously I noted native descriptions of "Hokkien" which use "b" etc. for a phoneme /[b~m]/.
One can inquire re: allophony in the "strict sense": (1) as I noted before, historical sources show some cases of e.g. [ba] ≠ [ma] ≠ [mã] where the weird middle guy is IMO explicable via means like recent loaning or contraction. E.g. (obviously loaned) Amoy má 'horse, surname' is shown with oral vowel in Douglas (1873); this would have been a recent (not e.g. Tang-era) loan in words like hū-má 'emperor's son-in-law' ("literary" vs. "colloquial" is far from binary — Forrest's comment quoted by Chris Button above is confused as bé is indeed the regular colloquial form/word 'horse'.) (2) Even in e.g. modern Taiwanese IMO mā 'also' etc. usually have oral vowels; I think this one is a new contraction. And (3) in some lects (Teochew?) /b/ ≠ /m/ etc. seems now to be thoroughly structural as a function esp. of loaning.
So reference to "allophonic variation" in (some) modern languages is normal and not (very) wrong. Point is that the "Hokkien" denasalization is definitively *not* a product of loaning + analogical spread yadayada; it is a regular internal development. Note Proto-Southern Min (e.g. Kwok 2018) features contrasting e.g. /b/ and /m/ in many syllable types with the latter joining /b/ by varying regular rules in particular ("Hokkien") daughter lects. The above cited statement by Pulleyblank ("…if Forrest is right in attributing the allophonic alternation between voiced stops and nasals found in that region exclusively to what he calls T'ang-Min") is just misleading/incorrect as Pulleyblank did not ever work on Min languages or modern languages period, basically… he is a favorite of Chris Button, however…
Chris Button said,
October 14, 2025 @ 2:32 pm
@ David Marjanović
The lateral stuff is a side note that I included to buttress the point that, contrary to what was said above, Forrest was actually fully aware of prenasalized stops / denasalization at the literary and the vernacular levels.
However, I must disagree with your first comment. I discuss the phonetic reason for the nasal vowels here:
https://languagelog.ldc.upenn.edu/nll/?p=71316#comment-1634176
As for parsimony, surely Forrest's proposal is easily the most parsimonious? The problem is that there is no definitive evidence–and, when it comes to phonetic vs phonemic issues, limited interest in it too.
Chris Button said,
October 14, 2025 @ 2:54 pm
@ Jonathan Smith
Pulleyblank paid close attention to Min and Sino-Xenic evidence in much of his work. His Middle Chinese book is peppered with such evidence throughout.
Despite originally being a historian and not as far as I'm aware ever having conducting lingustic fieldwork, he was not just a Sinologist playing at linguistics. He knew lingusitics better than anyone. And that included phonetics.
Other than Pulleyblank, I can't think of anyone after Karlgren (who predated the invention/discovery of the phoneme in any case) who actually paid much attention to phonetic reality. Everyone just mostly worked on stripping Karlgren's reconstruction down to its phonemes.
Pulleyblank even proposed a vertical vowel / voweless reconstruction of Old Chinese. Yet, perhaps because of the intense reductionism involved, he also stressed the importance of surface phonetics.
I still recall a discussion with an academic who said to me that they had listened to a recording of Kabardian and that it was full of vowels! I politely asked why they would have expected to hear anything else in a living language.
Nelson Goering said,
October 15, 2025 @ 2:29 am
""/b l g/ > [m n ng]" before specific rimes/codas and phonetic studies noting that ≈[b][g]-onset > post-stopped nasals e.g. after preceding nasals"
@Jonathan Smith, that's just normal allophony, of the routine kind found the world over. (Or rather, conditioned sound change. The idea that phonemicization can only occur once the conditioning environment becomes opaque is flatly contradicted by the actual historical data of some languages. The introduction of loans rendering the conditioning environment obscure would be of course confirmation of this.) The question at issue is how the introduction of loans could introduce a kind of "allophony" that could in turn be "conducive to analogical extension" in this case. An example or two of how Chris Button thinks this could actually have worked in practice would go a long way towards rendering his suggestion assessable.
Chris Button said,
October 15, 2025 @ 5:17 pm
@ Nelson Goering
As simply as possible in the most general terms:
1. External: [ŋ] ~ [ᵑg] vs [g]
2. Internal: [ŋ] vs [g]
Analogy extends the allophony from #1 into #2.
As for a more pervasive Tang koine, I noted here that [ᵑg] may conversely reflect [g] trying to maintain voicing. So [ŋ] > [ᵑg] > [g] and circular [ŋ] > [ᵑg] > [ŋ] are both possible evolutions.
Nelson Goering said,
October 16, 2025 @ 1:37 am
I still think actual examples would help clarify the discussion considerably. I mean actual words and forms meant to have undergone this development, in case it's not clear what I mean by "examples".
But basically, you're suggesting that either a single source language for loans had free variation (let's be precise in our phrasing: this is not classic conditioned allophony), or (probably more likely, given the discussion so far), that loans were coming in, some from varieties with one phone, some with the other (this is, of course, not allophony of any kind at all, though sufficient borrowing from the two sources could result in some free variation in the borrowing language).
OK, so far not totally implausible. One might well imagine (given some evidence) that some loan words would be adopted in doublets, sometimes with a more "foreignizing" pronunciation (the way some Brits say *en route* with French, or French-ish, pronunciation), sometimes with a more nativizing pronunciation. In this case, presumably [ŋg] and [ŋ], respectively (I wonder what some concrete examples of a potential doublet-borrowing might be). Whether this would have resulted in really widespread variation in any given register, even in such loans, I'm a little doubtful of. After all, "free" variation is rarely really free, and tends to actually have social import of some kind. Insofar as the [ŋg] pronunciations were widely adopted by some speakers, this could easily result in a new phoneme.
In any case, what seems more remarkable is the idea that you could go from having a situation where the end result of this was anything other than either the loss of the "foreignizing" pronunciation (assimilation of loans to [ŋ] everywhere), or the establishment of [ŋg] as a separate phoneme. The idea of extending the "free" variation to inherited [ŋ] seems a rather unlikely suggestion, and one that needs support from typological parallels (do you have any in mind?), or else some kind of really good internal evidence (it seems safe to say that this is absent, based on the discussion so far). This is certainly not the kind of routine analogy you see again and again in languages the world over.
(Or in other words, I agree wholeheartedly with DM about the plausibility of this suggestion.)
Chris Button said,
October 16, 2025 @ 5:51 am
At a phonemic level, sure. But not a surface phonetic level, which is what is being discussed here.
I personally I like Pulleyblank's suggestion that the prenasalized stops could go back to a time before Middle Chinese. As I noted above, I also think a circular EMC > LMC > EM (Early Mandarin) shift like [ŋ] > [ᵑg] > [ŋ] is not out of the question.
Chris Button said,
October 16, 2025 @ 5:57 am
(Ok, so technically ŋ- is on its way out in Early Mandarin, but the point remains)
Chris Button said,
October 16, 2025 @ 7:41 am
The line can be blurry too. For example, [k] ~ [q] reflect Middle Chinese variations of phonemic /k/ based on the conditioning environment. The distinction would not normally be noted in a phonemic reconstruction. But (as I think Pulleyblank once pointed out), the distinction would have been so overt to be almost phonemic.
Nelson Goering said,
October 16, 2025 @ 8:41 am
"At a phonemic level, sure. But not a surface phonetic level, which is what is being discussed here."
No: sociolinguistically. If we're talking about phonemes (however defined), we're not already talking about the variants anyway.
"based on the conditioning environment"
If there's a conditioning environment, then that's classic conditioned allomorphy (allomorphy in the strict sense), and completely different from what you've been suggesting for these initial stops. Probably this was, as Pulleyblank seemed to feel, actually phonologized earlier than hardcore structuralists would hold (I've already noted this general principle above).
"As I noted above, I also think a circular EMC > LMC > EM (Early Mandarin) shift like [ŋ] > [ᵑg] > [ŋ] is not out of the question."
Duke of York changes are common, so there's of course nothing remarkable about that scenario, if it accounts best for the evidence.
Chris Button said,
October 16, 2025 @ 8:55 am
That [k] vs [q] stuff was a distracting aside. Sorry, was trying to be helpful but achieving the opposite!
I mean simple free variation along the lines of whether you have a plosive element at the end of your nasal or not. Some speakers may, some may not, some may vary depending on sociolinguistic circumstances or other.
Jonathan Smith said,
October 16, 2025 @ 9:17 am
@Nelson Goering
/b l g/ > [m n ng] (etc.) is "just normal allophony, of the routine kind found the world over"
My point: reference to "allophony" per se is not the issue with Pulleyblank's and by extension Chris Button's assertions.
"Or rather, conditioned sound change."
No, the historical change is rather the reverse; the above formulation is a typical (not my!) description of the contemporary e.g. Taiwanese situation in terms of allophony.
"The idea that phonemicization can only occur once the conditioning environment becomes opaque" is dubious
Sure; point is only that there is no shortage of evidence/analysis/native intuition for allophony down to the present day if one happens to prefer that approach; it's not super relevant to the core question here and I don't personally a have a dog.
"The introduction of loans rendering the conditioning environment obscure…"
If anyone cares, in Taiwanese in particular or say the (closely related) Amoy of Douglas (1873), "early modern" loans from standard languages are in e.g. b- not m- and an allophonic analysis remains totes cromulent again if one likes. (Indeed Douglas says explicitly that e.g. his b- is "often pronounced like m- or interchanged with it." So.) But again (see above), newer loans (from e.g. Mandarin, Japanese…) behave differently and could support a different analysis… if one likes.
Re: that core question — origins / nature of "denasalization" in e.g. Taiwanese — as you will see by now there is no serious alternative historical proposal afoot, just another in a long line of hand-wavy pseudo-Pulleyblankian would-be topplings of "Sinological" "dogma". Laissez les bons temps rouler on…
Chris Button said,
October 16, 2025 @ 4:33 pm
@ Nelson Goering
Think of how some English speakers may sometimes glottalize a final stop coda. That "unreleased" stop, which exists in free variation with a normally "released" stop, may eventually just become a glottal stop without any further distinction.
Yves Rehbein said,
October 16, 2025 @ 6:53 pm
There are many different kinds of allophony. Although I agree it would help to be more precise, I do not agree that "classic conditioned allophony" is helpful, @ Nelson Goering. ".
On the example of J. S. Bach – randomly – I count four different perspectives. 1. English Back is native, Bach is an instance of code-switching. 2. ch /x/ is realised differently across the map, which is predictable only if you know dialect geography. 3. ch /x ç/ is conditioned by the preceding vowel in German. 4. Historically, ch continues *k in a conditioned environment, but local dialect maintains /k/, ick (bin ein Berminer) vs. ich /ç/. This is definitely not obvious from synchronic description; i.e. English is wrong? As for Bach, compare Dutch beek, Icelandic bekkur, also French Bacquerel. Paradigmatic denken ~ dachte on par with think ~ thought shows that this goes back at least to Proto-West-Germanic. So English Back is somehow correct, accidentally, but cognate beach also is. PIE *eǵh₂- is palatal to begin with, so ich is "more correct" (paraphrasing George Orwell); PIE *ten-K– is quite a stretch, though, i.e. thoroughly uncertain. Deck (Ger. Dach "roof", but Deck in the nautical sense, compare thatching) proves a history of Dutch or earlier Frankish–Frenchish influence either way, which is a subject of current research. NB: English I vs. ich and accusative thou, you vs. dich, euch are, I hope, cromulent examples of whatever happened to nu, OC *nʳàɰɁ (vel sim.), PST *naq (Wiktionary q.v.).
My point is, synchronic description is the bulk of linguistic description, but it is only a small part of historical linguistics and perhaps that is why the definitions make no sense.
Nelson Goering said,
October 17, 2025 @ 1:41 am
I seem to have expressed myself poorly. My point isn't that "classical allophony" is the only kind of variation (there would be no need for more precision if this were the case), or that synchronic variation is the only interesting type (my view is actually precisely the opposite).
I had exactly two requests: one (now satisfied; thank you Christ Button) to be precise about what type of variation is envisioned in early Min, and another (not satisfied) for a satisfactory argument as to how this variation could in loanwords could lead to a drastic shift in the phonology of inherited words. Well, and a third request for some specific lexical examples, since in my view historical linguistics is always best discussed concretely.
Yves Rehbein said,
October 18, 2025 @ 5:29 am
I wonder that, too. For example, èr 二 MC nyijH is considered one of the most stable descendents of PST, but Japanese ni has no dental, and Korean dul is not considered related.
Following the Sanskrit comparandum (Wiktionary; Neumann in FS Risch for more, Beekes sceptical), Hindi dās has a synonym borrowed however indirectly from Mongolian nökör ("friend", Wiktionary). Manuscripts of Buddhist–Tibetan–Chinese–Mongolian terms (the Mongolian Mahāvyutpatti, edited by Alice Sárközi) has nökör in some collocations, perhaps also ‘qatun-u nököd buyu’ / ‘qatun-u daγuli [fn.: daγaγuli] ba daγalta’ (“attendants of the wife”), cf. ‘daγuli’ (“accompany”), ‘daγalat’ (“companion”). If γ may reflect h, and ǰarudasun (“slave”) is a different word, as is dayisun (“enemy”); perhaps it’s nothing. As Wiktionary is suggesting *des- (s.v. Sanskrit दास), recall PIE *dus– (“bad”): Persian došman (“enemy”). There is the suggestion that it derived from *dwóh₁ (“two”).
That would be a neat coincidence with èr 二 though I would be rather careful around numeralia, because so much has been written already. One can only wonder. See earlier Tocharian, Turkic, and Old Sinitic "ten thousand" (April 23, 2019).
KIRINPUTRA said,
October 25, 2025 @ 4:00 am
Taioanese LÔ͘-LĒ is a 20th-century bookish loan from Japanese or (less likely) Mandarin. Native word LÔ͘-CHÂI 奴才 is also more or less bookish. (KÁN, "slave girl", is an older word and may have had a broader meaning in antiquity.)
LÔ͘-LĒ & LÔ͘-CHÂI were derived using "Han" readings, which are partly "artificial", or arbitrary, so, no good reading too much into them; but, safe to say (as others have above) that the L- here is a direct or indirect result of wide-ranging denasalisation in Hoklo, or some immediate ancestor of Hoklo.
BTW, 奴 reads NÔ͘ (= NÔ͘ᴺ; the vowel is nasalised) in some "dialects" (not exactly) of Taioanese. But I've never heard *NÔ͘-CHÂI or even *NÔ͘-LĒ.
KIRINPUTRA said,
October 25, 2025 @ 4:16 am
> Analogically one might have expected in the same dialects *d- The sound change m > b, etc., is an unusual one, and is seldom found to occur 'spontaneously', i.e., unconditionally. It cannot be due to local Min speech habits, otherwise we should have found it with even more regularity among the native forms than among the literary, which is plainly not so.
His info was incomplete, and he hadn't seen the arbitrary, "artificial" side of the Han readings.
> Even in e.g. modern Taiwanese IMO mā 'also' etc. usually have oral vowels;
Not AFAIK. My guess is this is based on hearing Taioanese spoken by people that normally don't speak Taioanese (although it may tech. have been their L1).
> At some point, therefore, voiced plosives and nasals must have been allophones in the strict sense: entirely predictable realizations of "the same thing".
Still the case in modern Taioanese. Words like MĀ, and readings like MÛI (梅), have nasalised vowels.
> I suppose MC loanwords disrupted this pattern and created a phonemic distinction between the two.
Not true, as I think J. Smith has pointed out.
KIRINPUTRA said,
October 25, 2025 @ 4:17 am
> Analogically one might have expected in the same dialects *d- < ACh. n-; but n- was apparently saved from denasalization by changing into l-.
As David M. (I think) has pointed out, [d] is a realisation of Taioanese & Hokkien L-. I think it's the dominant realisation before high vowels in stressed syls., but is NOT limited to that environment.
Jonathan Smith said,
October 25, 2025 @ 9:45 am
@KIRINPUTRA
because it's interesting, re: possibility of oral vowel specifically in items like mā 'also' (NOT e.g. in mā 'scold') and mài 'don't', while this is really a job for SuperEar — whom many of us claim to be but aren't — or for serious phonetic study, my thought is that these ultimately represent contractions m̄ + ā / m̄ + ài in which to the contrary it is certain olds who would retain the oral vowel (along possibly with a sense of the words as contractions). Later would be normalization to the regular nasal onset + nasal vowel pattern among many folks. And still more recently, of course, widespread loss of nasalized vowels among the younger/more casual users you refer to. Just a thought…
KIRINPUTRA said,
October 25, 2025 @ 8:29 pm
@ Jonathan Smith
I see. Makes sense that an oral vowel might be used in, say, M̄ ÀI even if it began to contract. (MÀI & MĀ wouldn't illustrate this, though, for diff. reasons; confusingly, modern M̄ ÀI is not directly related.)
Come to think, a speaker of "emergency" Taioanese may pronounce KÁᴺ perfectly but have an oral vowel on MĀ.
Jonathan Smith said,
October 25, 2025 @ 9:23 pm
@KIRINPUTRA
Not sure re: mā, mài; these are historically "new" items and their origin deserves consideration via close study of the POJ textual record… my sense re: the former is it is very rare in early texts and where it appears often in the mouths of girls / young women…
Ys re: youngun Taiwanese which involves a merged phonological system of sorts… and speaking of káⁿ, similar to the above I suspect káⁿ + m̄ > kám…