Turandot and the deep Indo-European roots of "daughter"

« previous post | next post »

In recent days, the famous aria from the final act of Giacomo Puccini's opera Turandot, "Nessun dorma" (Italian: [nesˌsun ˈdɔrma]; English: "Let no one sleep"), has surfaced as part of a worldwide movement to encourage the Italian people in their struggle against the novel coronavirus (see here, here, and here).  This article by Claudia Rosett gives the backstory:

"An Uplifting Moment, in the Time of Coronavirus", PJ Media (3/14/20)

This led me to ponder the origins of Turandot's name, especially since the operatic version of the story is set in China and she is alleged to be a Chinese princess.  Right away, I was in for a jolt, since "The name of the opera is based on Turan-Dokht (daughter of Turan), which is a common name used in Persian poetry for Central Asian princesses." (source)

Digging deeper in the same source, we learn:

Turandot is a Persian word and name that means "daughter of Turan", Turan being a region of Central Asia, formerly part of the Persian Empire. The name of the opera is taken from Persian Turandokht (توراندخت‎), with dokht being a contraction of dokhtar (daughter); the kh and t are both pronounced.

The story of Turandot was taken from a Persian collection of stories called The Book of One Thousand and One Days (1722 French translation Les Mille et un jours by François Pétis de la Croix – not to be confused with its sister work The Book of One Thousand and One Nights) – where the character of "Turandokht" as a cold princess was found.

June Teufel Dreyer remarks:

This is fascinating.  Have always assumed that “daughter” came from Nordic languages’ “dottir”;  Töchter in German.  So it came from Persian?  Or they’re accidentally the same?

June is right.  It turns out that English "daughter" does have a Germanic basis, but it also has close cognates in Iranian, Indic, Greek, Slavic, and other branches of Indo-European, all the way back to PIE *dhugheter, but was lost in Celtic and Romance.

Middle English doughter, from Old English dohtor "female child considered with reference to her parents," from Proto-Germanic *dokhter, earlier *dhutēr (source also of Old Saxon dohtar, Old Norse dóttir, Old Frisian and Dutch dochter, German Tochter, Gothic dauhtar), from PIE *dhugheter (source also of Sanskrit duhitar-, Avestan dugeda-, Armenian dustr, Old Church Slavonic dušti, Lithuanian duktė, Greek thygater). The common Indo-European word, lost in Celtic and Latin (Latin filia "daughter" is fem. of filius "son").

(Etymological Dictionary Online)

For the fuller filiation (!) of the PIE word, we have this richly informative Wiktionary entry:


There are two PIE reconstructions that can be obtained using the comparative method:

    • *dʰugh₂ter- on the basis of: PIIr. *dʰugHtar- (Sanskrit duhitár-, PIr. *dʰugʰtar- > *dʰugdʰar- > *dugdar- > Gathic Avestan dugədar-), Ancient Greek thugátēr-, Tocharian A ckācer, Tocharian B tkācer
    • *dʰukter- on the basis of: Iranian *duxθrī (> Old Persian *duhçī) and *duxtar- (> New Persian duxtar) due to the absence of Bartholomae’s Law, possibly Gaulish duxtir, Gothic dáuhtar, Oscan fu-utreí (dative singular), Armenian dustr, Hieroglyphic Luwian t(u)watra/i-, Lycian kbatra-, Slavic *dъkti > *dъťi, Lithuanian duktė̃

The latter form is secondary, occurring due to the deletion of the medial laryngeal in the sequence CHCC in the oblique stem, which was paradigmatically leveled in the daughters. E.g. genitive singular *dʰugh₂tr̥és > *dʰuktr̥és. The CHCC > CCC change was a synchronic PIE phonological rule.

According to Kloekhorst, hieroglyphic Luwian tu(w)atra/i- and Lycian kbatra- reflect Proto-Luwic *duetr-, further reflecting Proto-Anatolian *duegtr- < PIE full-grade stem *dʰwegh₂ter-. The original inflection was thus hysterodynamic bandi-type as described by Beekes (1995: 175): *CéC-R, *CC-éR-m, *CC-R-ós:

    • Nominative singular: *dʰwégh₂-tr̥
    • Accusative singular: *dʰugh₂-tér-m̥
    • Genitive singular: *dʰugh₂-tr-ós

After the split of Anatolian branch from Proto-Indo-European, the other Indo-European languages underwent a common innovation, replacing the nominative stem *dʰwégh₂tr̥ by the accusative stem in the zero-grade *dʰúgh₂tēr which however retained the original accentuation and which further underlies the attested Greek forms *θύγατηρ (Homeric θύγατρα (thúgatra)) > θυγάτηρ (thugátēr) (θυγατέρα, θυγατρός). In the other Indo-European languages the accentuation of the accusative was later on transferred to the nominative form, yielding the oxytonic paradigm listed in the declension table, as retained in Sanskrit (duhitā́, duhitáram, duhitúḥ) and Lithuanian (duktė̃, dùkterį, dukterès).


The original meaning is probably "the (potential) suckler, the one that draws milk"; compare Sanskrit दुहे (duhé) / दुग्धे (dugdhe), and the *-tḗr suffix common to other r-stem kinship terms.


*dʰugh₂tḗr f

    1. daughter


more ▼Athematic, hysterokinetic
nominative *dʰugh₂tḗr
genitive *dʰugtrés



Luciano Pavarotti sings "Nessun dorma" from Turandot (The Three Tenors in Concert 1994)


Nessun dorma! Nessun dorma!
Tu pure, o Principessa,
nella tua fredda stanza,
guardi le stelle
che tremano d'amore, e di speranza!
None shall sleep! None shall sleep!
Not even you, oh Princess,
in your cold bedroom,
watching the stars
that tremble with love, and with hope!
Ma il mio mistero è chiuso in me;
il nome mio nessun saprà!
No, No! Sulla tua bocca,
lo dirò quando la luce splenderà!
But my secret is hidden within me;
no one will know my name!
No, no! On your mouth,
I will say it when the light shines!
Ed il mio bacio scioglierà
il silenzio che ti fa mia!
And my kiss will dissolve
the silence that makes you mine!

Just before the climactic end of the aria, a chorus of women is heard singing in the distance:

Il nome suo nessun saprà,
E noi dovrem, ahimè, morir, morir!
No one will know his name,
and we will have to, alas, die, die!

Calaf, now certain of victory, sings:

Dilegua, o notte!
Tramontate, stelle!
Tramontate, stelle!
All'alba, vincerò!
Vincerò! Vincerò!
Vanish, o night!
Fade, you stars!
Fade, you stars!
At dawn, I will win!
I will win! I will win!

I've been mispronouncing her name my whole life.  The correct pronunciation clearly has a "t" at the end (listen).


"Novel transmission of the novel coronavirus" (3/15/20) — has a long list of "Selected readings" at the end.


The word for daughter in Irish is iníon, in Scots Gaelic is nighean, and in Welsh is merch.

The words for daughter in the major Romance languages are French fille, Italian figlia, Spanish hija, Portuguese filha, and Romanian fiică.

They derive from Latin filia.  Interestingly, filia ("daughter") derives from Latin filius ("son"):

From Old Latin fīlius, fīlios, from Proto-Italic *feiljos (compare Faliscan hileo), from Proto-Indo-European *dʰeh₁ylios (sucker), a derivation from the verbal root *dʰeh₁(y)- (to suck). Related to fēmina, fellō, fētus, Old English delu (nipple, teat), dēon (to suck, suckle), Old Armenian դալ (dal).  Compare with the PIE derivation of "daughter" above.

With descendants in:





  1. Scott P. said,

    March 16, 2020 @ 4:38 am

    So why the "gh" in "daughter"? Were those letters still pronounced, at least as a remnant, when that spelling was fossilized? If it was pronounced more like Nordic, "dottir", I am curious how those extra letters became inserted.

  2. Frans said,

    March 16, 2020 @ 4:53 am

    @Scott P.

    The way I've always interpreted the pronunciation guides in Chaucer and other Middle English texts is that you should more or less pronounce it as if you were a Dutch person who doesn't really know how to pronounce written English, besides a few choice consonants like the voiced dental fricative represented by eth. Which does mean pronouncing the gh/ch like in loch, to paraphrase the example typically given. Or more to the point here, as in Dutch/German dochter/Tochter. Anyway, surely there were regional variations.

  3. Victor Mair said,

    March 16, 2020 @ 8:16 am

    From Brian Spooner:

    dokht, braat, pet, maat

    From the little I remember from the 1960s when I was studying Middle Persian the basic kinship terms in Indo-European languages all go back (as one might expect) to the original some five thousand years ago at least. The terms for mother, father and brother haven't changed very much, except for the addition of the (comparative?) suffix, producing e.g. (Persian) maadar, pedar, baraadar, dokhtar. But in most modern Indo-European languages dokht has changed much more, possibly because of the difference in the way young women have been treated everywhere since the beginning of settlement around 10k BCE. The subject gets more and more interesting as one gets into the broader field of kinship studies as a linguistic anthropologist.

  4. Walt said,

    March 16, 2020 @ 8:36 am

    Incidentally, most people who knew Puccini day that he never pronounced the final "t" in Turandot. Likewise, accounts from the premiere also agree that the final "t" was silent. This isn't surprising since it is difficult for native Italians to pronounce. Nowadays it is common to hear it both ways, although I find it much more likely to be "t"-less in Italy.

  5. Anthony said,

    March 16, 2020 @ 9:27 am

    As they say, in these internet days it's Turandot.com (hence with t).

  6. David Cameron Staples said,

    March 16, 2020 @ 10:53 am

    The Goidelic words iníon and nighean are both from the Classical Irish inghean (Scottish Gaelic via metathesis), which was originally pronounced /ˈinˌɣən/ c.1200, and became something like/ˌɨnˈijən/ by the 17C. (The initial unstressed /ɨ/ being dropped is where the nighean pronunciation came from.) This was spelled in Middle and Old Irish as ingen with the /ˈinˌɣən/ pronunciation. There is another OI word spelled ingen, meaning "nail" (cognate with Lat. unguis), which was pronounced /ˈiŋˌgən/. The difference is that in Primitive Irish, the word for "daughter" was INIGENA, and consonants appearing between vowels were lenited, but if they were protected by another consonant, then they weren't. Which is how we know that ingen = "nail" must have been something like INGENA in primitive Irish, even though it's not attested.

  7. cameron said,

    March 16, 2020 @ 11:52 am

    While those basic kinship terms, as cited in the note from Brian Spooner above, in Modern Persian are all quite conservative with respect to their ancient Indo-European heritage, Persian has borrowed most of the words for uncles and aunts from Arabic. Persian has amu and ame for paternal uncle and aunt, respectively. And it has xâle for maternal aunt. The weird exception is the word for the maternal uncle. The Arabic word is xâl, but Persian didn't borrow that alongside its feminine counterpart. Persian retains a non-Arabic term dâyi for the maternal uncle.

    It's quite perplexing why that one Persian word would have been retained while those closely-related terms would have been replaced by Arabic borrowings.

    (Note that the words above aren't used for the spouses of your aunts and uncles, only for the direct relations. Your paternal uncle's wife would not be an ame, for example, but a zan-amu. )

  8. Dave Cragin said,

    March 16, 2020 @ 1:50 pm

    @Frans, as you noted English spelling often captures old pronunciation. As speaker of Dutch, you likely see the same in "enough." When I explained this to a Belgian friend, he said "Just like in Dutch!" (genoeg) Night shows this as well. In English, the sound of -gh softened overtime, but we retained the spelling.

    John McWhorters' linguistic courses pointed the above out to me. This was interesting to me because despite having learned some German & a little Dutch, I never thought of why the English words had "-gh" until he explained this.

  9. Joe Fineman said,

    March 16, 2020 @ 3:11 pm

    It amuses me that, while something has happened to son & daughter in Latin, and something has happened to brother & sister in Greek, and something has happened to father in Russian, we, by this distant northern shore, have clung to IE for all six nuclear-family words: mother, father, daughter, son, sister, brother.

  10. Gregory Kusnick said,

    March 16, 2020 @ 4:52 pm

    cameron: I'm just speculating wildly here, but men share with their maternal uncles something they don't share with their fathers or paternal uncles, namely their mitochondrial DNA and (on average) about 25% of their X-chromosome. The result is that sex-linked traits such as colorblindness and male-pattern baldness appear to pass from maternal uncle to nephew rather than from father to son. Perhaps the persistence of dâyi owes something to subconscious cultural recognition of this genetic quirk.

  11. Julian said,

    March 16, 2020 @ 6:04 pm

    @Joe Fineman
    So what did happen to father in Russian? That is, what is the origin of the present word for father? Just curious.

  12. martin schwartz said,

    March 16, 2020 @ 6:37 pm

    The Wictionary entry is infinitely better than the Online Etym.
    Dict. The h with subscript 2 is the "second laryngeal', a consonant
    which aspirated the previous *g in Sanskrit and was vocalized as i,
    is relfected as Gr. a,nd disappeared in Iranian and most IE langs.
    In Iranian the stem was *dugdar- (> duxtar-),retained in oblique
    (acc. etc.) forms, but dugdâ (duxtâ) in the nomi., so that we have
    Middle Persian duxt (and duxtar) and similar forms in many modern Iranian languages, and Persian duxt (doxt); similar variation in other kinship terms. The cited form for "Venus' = 8"god's daughter'.
    Eng.gh in daughter, as in night, bright, light, thought, eight, etc. represents an old velar (like ch in German nacht and acht); in Scots dilaects the velar is still maintained, and "8' sounds like the German. I suppose some Scots also have the velar in daughter.
    Martin Schwartz
    Martin Schwartz

  13. Francois Lang said,

    March 17, 2020 @ 4:09 am

    Slightly off topic, but there's a great deal of debate about how to pronounce "Turandot". This fraught topic even has a scholarly article devoted to it.


  14. Scott P. said,

    March 17, 2020 @ 4:09 am

    @Frans, as you noted English spelling often captures old pronunciation. As speaker of Dutch, you likely see the same in "enough." When I explained this to a Belgian friend, he said "Just like in Dutch!" (genoeg) Night shows this as well. In English, the sound of -gh softened overtime, but we retained the spelling.

    John McWhorters' linguistic courses pointed the above out to me. This was interesting to me because despite having learned some German & a little Dutch, I never thought of why the English words had "-gh" until he explained this.

    What gave my pause is that in Geoffrey Sampson's Writing Systems, he says "Many of [English's] non-phonemic properties have no connection with morphophonemic alternations. Chomsky and Halle justify the of right/righteous by arguing from synchronic evidence that the root must be underlying |rixt|; but there is no comparable evidence for an underlying |x| in night or light … The generative phonologist's account of English orthography was very influential for a number of years but I conclude with W.N. Francis that it is 'extravagant and unsupported'. "

  15. Robert Coren said,

    March 17, 2020 @ 4:25 am

    So why did Puccini (or his librettist) move the story to China (while retaining the very un-Sinitic name)? Wouldn't Persia have been sufficiently "exotic"?

  16. Christian Weisgerber said,

    March 17, 2020 @ 6:56 am

    @Joe Fineman
    Modern English sister looks more like Old Norse systir than Old English sweostor though, so it's not all straight inheritance back to PIE.

  17. Coby Lubliner said,

    March 17, 2020 @ 1:58 pm

    It wasn't Puccini but Carlo Gozzi who set the story in China. There have been may versions of Gozzi's play (a commedia dell'arte), including one by Schiller and an opera by Busoni.

  18. Frans said,

    March 17, 2020 @ 11:27 pm

    @Scott P.

    I don't have access to that work, but that sounds like it's talking about present-day English? (Or perhaps I should say '60s-'80s English, but I rather doubt that matters to the point.)

  19. Andreas Johansson said,

    March 18, 2020 @ 8:30 am

    What would be the evidence for underlying |x| in "right"? Genuinely curious?

  20. Leo said,

    March 18, 2020 @ 9:53 am

    Julian: if Wiktionary is right, the Russian for "father", отец, comes from PIE *átta. I wonder if this is related to Turkish "ata", as in Atatürk, "Father of the Turks".


  21. Victor Mair said,

    March 18, 2020 @ 10:33 am

    From John Colarusso:

    “daughter’ is interesting because it gives strong evidence for the nature of the laryngeal-2.

    Greek has an /-a-/ for *h-2, and no change on the preceding /-g-/, while Sanskrit has an /-i-/ and voiced aspiration of the preceding /-g-/.

    The only sound that can split into back low as well as front high vowels is a pharyngeal.

    Pharyngeals have strong low first formants (acoustic resonances) mimicking high vowels (hence pharyngeal or emphatical softening as Trubetzskoy called it; note Phnoecian Ba9al (<9> for voiced pharyngeal) ‘lord’, vs. Hebrew Be?el).

    But pharyngeal articulation involves contraction and lowering of the tongue root, hence it can give low vowels

    So, Greek went with the acoustics of PIE *h.2, while Sanskrit went with the articulation along with some friction reinterpreted as belonging to the preceding voiced sound.

    The kinship sufix may have been *-h.2t-er, akin to the full grade *h.2et-el- (-el- a diminutive of endearment) as in OE athel-, Mod E Ethel, Mod G Edel ’noble’.

  22. Christian Weisgerber said,

    March 18, 2020 @ 12:24 pm

    atta looks like a babble-word (Lallwort). These are constantly reinvented for kinship terms the world over and are useless for demonstrating genetic relationships between languages.

  23. Leo said,

    March 18, 2020 @ 9:01 pm

    Christian Weisgerber: your point is well taken.

  24. Jen in Edinburgh said,

    March 19, 2020 @ 3:38 am

    I have a feeling that I'm missing something here, but:

    Scots has richt/nicht/licht, and German has Recht/Nacht/Licht, with the [x] but assorted vowels.
    (And Norwegian has rett/natt, with the same vowels as German but without the [x].)

    So the [x] is out there, but I don't know which versions came from which.

  25. Frans said,

    March 19, 2020 @ 11:36 am

    @Jen in Edinburgh I'll expand a little on my earlier remark. Since the quote says Chomsky and Halle and synchronic, the question under discussion in it presumably isn't whether there once was an /x/ or a /ç/ or some such (obviously there was, outside of misapplied analogous spellings like delight), but whether current English spelling accurately reflects a hidden inner psychological phonological model that still effectively includes this /x/.

    This inner /x/ would explain why it's righteous instead of raishous and nightingale instead of naishingale. Or something along those lines. Because otherwise that's what you'd get through an application of the rules.

    (If you look up generative phonology it'll probably explain it a lot better; I don't really know much about it other than that generative grammar & phonology were treated as historical stuff in a few linguistics courses.)

RSS feed for comments on this post