Daughter of Holy Cow

« previous post | next post »

I was just thinking how important cows (and their milk) are for Indian people and was surprised that's reflected in such a fundamental word for a family relationship as "daughter" — at least in the popular imagination.  Even in a scholarly work such as that of D.N. Jha, The Myth of the Holy Cow (New Delhi:  Navayana, 2009), p. 28, we find:

Some kinship terms were also borrowed from the pastoral nomenclature and the daughter was therefore called duhitṛ (= duhitā = one who milks).

That somehow seemed too good to be true, a bit dubious on the surface.  To test the equation, I began by bringing together some basic linguistic information acquired on a preliminary web search.

American Heritage Dictionary of the English Language, 5th ed., Appendix of IE roots:



Oldest form *dhugh2ter-. daughter from Old English dohtor, daughter, from Germanic *dohtēr.

[Pokorny dhug(h)əter 277.]


From Middle English doughter, from Old English dohtor, from Proto-West Germanic *dohter, from Proto-Germanic *duhtēr, from Proto-Indo-European *dʰugh₂tḗr.

Sanskrit Dictionary

duhitṛ f. a daughter (the milker or drawing milk from her mother[ confer, compare Zend dughdar, Greek , Gothic dauhtar, Lithuanian dukte, Slavonic or Slavonian dushti]).

Then I approached the real experts with the following:

For "daughter", Vedic has duhitṛ ("one who milks").


    1. Is that analysis reliable?
    2. Is duhitṛ cognate with "daughter"?

Here are the replies I received:

Whitney Cox:

duhitṛ is indeed cognate with Eng. daughter.  While I'm no kind of Indo-Europeanist, I do recall hearing that connecting it with √duh, dogdhi, etc. is spurious.  But I don't have any references to hand.

George Cardona:

Yes, duhitṛ is cognate with daughter, Tochter and so on: as in Iranian, medial laryngeals are deleted in polysyllables.  I doubt the analysis as 'who milk', on a par with pitṛ being 'he who protects'.  Derivates of his type, with a suffix *ter, should have full-grade vocalism in the root.

Michael Witzel:

This is  an old, more or less accepted etymology:  the daughter milking the animals in the morning, etc.

*dhug-h2ter from *dhug ‘to milk’ —>Skt duh.

The suffix h2 (laryngeal 2) is also seen in p-h2ter ‘father’ and ma-h2ter ‘mother’, etc., whatever the original meaning  of p- and mā- may have been (some think of baby language, almost world-wide spread, note Chinese!)  

[ph2ter —> Skt. pitar … is normal]

Thus duhitṛ  = Engl. daughter along with other IE languages.

Jay Jasanoff:

The root *dheugh- has “dairy” associations only in Indo-Iranian. The “daughter” word, moreover, is reconstructible as *dhughh2-ter-; the *h2 is unexplained under the “milk” etymology.

Joshua Katz:

The Sanskrit and English words are indeed cognate, with both going back to PIE *dhugh2-ter-.  As for whether this means something like "producer (of milk)," I suppose it's possible, though the idea isn't especially compelling to me semantically (still, who knows …) and, more to the point, isn't actually all that easy phonologically.  We know that there's a laryngeal at the end of the "root" (or whatever it is) of the word for "daughter," whereas the root that means "produce" (it's the source of English doughty) really does have a voiced aspirate *gh.  The matter is tricky enough that you'll want a second and even a third opinion on this, I think.

I know that Szemerényi in his Indo-European kinship book from the 70s says something about how the word for woman (Greek gyne, etc.) is underlyingly the same as the word for cow …  No one believes this!  I would expect that book to have one of the fuller discussions of the options for the word for daughter.  Unfortunately (?), I don't own it and don't otherwise have it to hand.

Alexander Lubotsky:

Yes, Vedic has duhitṛ for "daughter", but it does not mean "one who milks".  And, yes, duhitṛ is cognate with "daughter".

The correct reconstruction is *dhugh 2-ter-. First of all, it has nothing to do with *d hugh-. Secondly, the latter root means 'to fit, be fitting' and not 'to milk', which is due to a secondary development of 'the cow is fitting' = 'the cow gives milk'.

Michaël Peyrot:

The connection is suggestive in Vedic, but the reconstructions of the two words now have become nevertheless clearly different: duh- "milk" is from PIE *dʰeugʰ- (same root as German taugen "be fine, useful") and duhitā- "daughter" is from PIE *dʰugh₂tēr. Although the spelling is misleading, gʰ in the verb, with superscript h, is one phoneme, different from gh₂ in the noun, with a regular h₂, which are two phonemes.

Michael Weiss:

Ved. duhitar- is certainly cognate with daughter but it is not 'the one who milks'. The root 'milk' which actually means something more like 'produce' is *dheugh- with two voiced aspirates. The word for daughter only has a voiced aspirate in Indo-Iranian (Ved. duhitár-, OAv. dugədā where the form shows the effects of Bartholmae's Law and points to PIr. *dugdar- < *dhughdhar < *dhugh-tar- ) due to the apirating effects of the following laryngeal *dhugh2ter-. Greek θυγάτηρ shows that the final velar was unaspirated.

Hannes Fellner:

Here is the lemma from the Nomina im Indogermanischen Lexikon (pp. 144-149). The Sanskrit word is of course cognate. The meaning of the original root is debated (see the first footnote [copied in the appendix below]. The connection to the root which means 'to give milk' in Sanskrit is phonologically problematic and the meaning is a special Indian development of the original root meaning.   

Andrew Ollett:

You should probably check with a "real" historical linguist, but a few observations:

    1. Kinship nouns often end in *-ter (the Sanskrit pitr̥ type)
    1. Agent nouns often end in *-ter as well (although Sanskrit generally has a type that shows a different vocalism, namely *-tor, so there are formal differences between pitr̥ and kartr̥)
    1. Because of that similarity it's possibly to analyze some kinship nouns as agent nouns (pi-tr̥ "the one who protects," duh-i-tr̥ "the one who milks," mā-tr̥ "the one who is honored" according to the Uṇādisūtras).
    1. None of those analyses, however, are very convincing, because of formal problems (agent nouns usually have full grade in the root, unlike pi-tr̥ and duh-i-tr̥; there is no root bhrā for bhrā-tr̥, and the root mā suggested for mā-tr̥ doesn't have a suitable meaning; the root duh "to milk" already has an agent noun, namely dōgdhr̥, although there are laryngeal issues involved here…), but I suppose it's possible to explain all of them away, with special pleading.
    1. There are other nouns of kinship ending in *-er/-or (svasr̥-) that aren't accounted for if kinship nouns are derived from agent nouns.

Jeremy Rau:

Unfortunately that etymology doesn't hold – although it's often found in the literature. Vedic duhitar– is in fact exactly cognate with the word for daughter in the other IE languages, but like most IE kinship terms doesn't have a clear etymology. In any event, the roots of Ved. duhitar– < *dhugh2ter- and duh– 'milk; give milk' < *dhugh– are fundamentally different, though in Indo-Iranian (and Indo-Aryan) where doh- took on the meaning 'milk; give milk,' it's of course very likely they were connected by folk etymology.

The best and most reliable source for all Indo-Aryan and Indo-Iranian etymology is Manfred Mayrhofer. 1992-2002. Etymologisches Woerterbuch des Altindoarischen. I-III. Heidelberg. It's really an excellent book.

Georges-Jean Pinault:

1) Vedic duhitar/duhitr- does not mean 'one who milks'. The relationship with the verbal root duh- 'to milk' and further derivatives from the same root is merely a MIRAGE, du to a phonetic evolution proper to Indo-Aryan. The PIE form of 'daughter' (as noted below) is incompatible with the reconstruction of the root *dheugh 'to milk' (actually, the original meaning was 'to obtain a product'). The latter, the verbal root, had a palatal dorsal voiced aspirate stop 'gh, while the former, 'daugher' has a velar tectal voiced stop. Moreover, from the derivational point of view, 'daughter' IS NOT an agent noun of any root.

2) Yes, duhitṛ is cognate with "daughter".

Melanie Malzahn:

The daughter word is one of the best-attested kinship terms (even attested in Anatolian) and sure daughter and duhitar- are cognates. The PIE reconstruction is *dhugh2tér-/ dhugh2tr´- , though looking for an internal etymology displaying a social role is no longer fashionable in linguistics: it is definitely not to be connected with the "milk"-verb.

Stephanie Jamison:

On Vedic for "daughter", duhitṛ, allegedly meaning "one who milks", I don't even know of any folk etymological play with that. Inter alia, it would 1) not be seṭ so it shouldn't have an -i-; 2) have a full-grade root; 3) be feminine. Result of these three would be dogdhr-ī 'milker (fem.)', which I think is attested.

Yes, duhitṛ is cognate with "daughter".  The *g has just been aspirated by the laryngeal in a slightly aberrant fashion and then Grassmann's Law has deaspirated the initial.

Brian Joseph:

Yes, duhitṛ is indeed cognate with English daughter.  Second, there are some phonological issues with the Sanskrit form, in that while *dhugh2ter is needed for Greek θυγάτηρ, it only more or less works for the Sanskrit form, since we see both the vocalization of the laryngeal (giving the -i-) and the development of *g’H2 (voiced plain stop + laryngeal) to *g’h (voiced palatal aspirate) – a development also seen, probably, in mahi ‘great’ (= Greek μέγα – I say “probably” because Hittite mekki suggests that there were i-stem forms of this adjective so the -i- of mahi could be a real -i- and not a reflex of a vocalized laryngeal).  But there are issues in other languages too, in that Lithuanian duktė has a “centum” reflex –k– for the guttural in this word, while other languages, e.g. Slavic and Armenian, point to a “satem” reflex (but then Lithuanian has several such centum reflexes where a satem outcome would be expected).

The Greek cognate of the root √duh- is τυγχάνω, which means more or less ‘happen by chance’, suggesting to me that the original meaning of the root is something like ‘yield’, and the Sanskrit meaning of ‘milk’ (also ‘yield’) is a specialization of ‘yield’, rather than ‘yield’ being a generalization of ‘milk’.  I kinda suspect that the “milker” interpretation of duhitar- is a folk-etymological connection within Sanskrit rather than a reflection of the PIE derivation.

Douglas Adams:

As far as I know, there is NO evidence that duhitar– was ever used as a word for ‘milkmaid’ or the like. Where that meaning is given in dictionaries, such as Monier-Williams great late nineteenth century dictionary, it was basically a suggested etymological meaning based on the fact that duhitar– looks much like an agent noun to the Sanskrit root duh– ‘give milk’ [intr] and ‘milk’ [tr]. 

PIE dheuĝh (whence Sanskrit duh-) is attested also in Germanic (Gothic daug ‘is useful’) and Greek teukhe/o- ‘produce.’ It’s easy to see that a verb for ‘be productive’ or the like when applied to a cow could come to mean ‘give milk’ but not vice versa. 

Most (e.g., Mallory and Adams) reject, or are at most agnostic about, any connection of PIE *dhuĝh̥2ter- ‘daughter’ and *dheuĝh ‘be useful.’ Birgit Olsen is an exception. In the very recently (this month) published Kin, Clan and Community in Indo-European she argues for the connection and makes the case that *dhuĝh̥2ter– originally meant ‘producer, preparer’ or the like (NOT ‘milkmaid’). While the smart money is not prone to bet against Birgit, in this case I resist following her lead. First, the phonological connection is difficult/complex for reasons I won’t go into here (but note that *-ĝh– is not the same as *-ĝh̥2-), and (2) there is no evidence anywhere I think, certainly not in Indo-European, that a kinship term like ‘daughter’ ever comes from an agent noun meaning ‘doer of something.’ If the inherited word is replaced, it’s always by some ‘birth term,’ e.g., ‘full brother’ > ‘brother’ (e.g., Spanish hermano or Greek adelphós) or by some term meaning ‘in-born’ (to the family) (e.g., in Old Irish).

Don Ringe:

Vedic duhitár- does NOT mean 'one who milks'! That's a 19th-c. myth that was exploded generations ago.

The Vedic 'milk' root is PIE *dhewgh-, which originally meant 'produce'; that's why Homeric Greek teúkhein means 'make, fashion', OE dēag means 'is useful', etc.

By contrast, PIE 'daughter' was *dhugh2tér-; that's guaranteed by Greek thugáter- (accent by analogy with 'mother', in which it was generalized from the vocative). It has nothing to do with milking for two reasons: (1) the *dhewgh- root did not mean 'milk' in PIE, and (2) 'daughter' isn't derived from that root anyway; the second consonant is *g, not *gh. So why does Sanskrit have an aspirate in that position? Simple: the following consonant, PIE *h2, regularly aspirates preceding stops in Indo-Iranian.

Yes, daughter is cognate; there are also Celtic, Italic, Balto-Slavic, and Armenian cognates.

Now I can see that what started out to be a seemingly simple and charming etymology turns out to be something else altogether.



First footnote for lemma *dȹugh2tér-/*dȹugh2tr ́- f. 'Tochter'1 IEW 277, EIEC 147f.

[VHM:  ȹ = superscript "h".  A few other special symbols have been corrupted here, for which see the link to Nomina im Indogermanischen Lexikon above (Fellner).]

Laryngalreflex sichtbar fortgesetzt nur im Ai., Gr. und Toch., Keltib., möglicherweise auch im Anat.; dagegen Formen ohne Laryngalreflex im Iran., Arm., Slav., Balt., Germ., Lat. und Gall.

Laut SCHMIDT, KZ 87 (1973) ist lautgesetzlicher Schwund des Laryngals an zweiter Stelle in einer Konsonantengruppe der Struktur -CHCC- bei nachfolgendem Akzent bereits in grundsprachlicher Zeit vorauszusetzen. Das würde auf die obliquen Kasus zutreffen – für die jeweiligen Fortsetzer in den idg. Einzelsprachen könnten dann Verallgemeinerung der starken Kasus (ai., gr. usw.) oder der schwachen Kasus (iran., arm. usw.) eines ablautenden Paradigmas angenommen werden.

Zum Akzent: angesetzt wird ein hystrodynamisches Paradigma, doch zeigt keine der Einzelsprachen ein eindeutiges Bild, was, wenn man SCHMIDTs Theorie akzeptiert, zu verschiedenen Reflexen des interkonsonantischen Laryngals in einem Paradigma geführt haben müßte. Dazu ferner HACKSTEIN, HS 115 (2002), 1ff.

Zu verschiedenen etymologischen Anknüpfungspunkten s. z. B. SZEMERÉNYI 1977, 19ff, der gegen die alte Anknüpfung an die ani -Wurzel *dȹegȹ- 'melken' in ved. duhé 'gibt Milch' usw. ist (LIV 148f. sub * dȹegȹ- 'treffen'). Er sucht Anschluß an Formen wie got. gadauka- m. 'Hausgenosse' und dauhts f. 'Gastmahl, Bewirtung' und geht von einer Vorform *dȹug- aus, die im Iir. mit Metathese zu *dugȹ- umgebildet worden sei. Als Suffix sei *-ter- abzutrennen, die verbleibende Form *dȹugǩ sei als VSg. eines ablautenden ā-, bzw. eh2-Stammes zu verstehen, der zum NSg. umgedeutet sei. Semantisch sei die Tochter folglich "the person who prepares a meal."

Der von SZEMERÉNYI abgelehnte Anschluß an *dȹegȹ- wird in neuerer Literatur z. B. wieder von PÂRVULESCU, IF 98 (1993), 55ff vertreten. Er geht von einer Grundbedeutung 'produzieren, machen, arbeiten, schaffen' aus, was letztlich zur Tochter als 'Dienerin, Arbeiterin' führe und weitere Parallelen in idg. Einzelsprachen aufweise.

Ähnlich auch TREMBLAY 2003, 86ff, der das Tochter-Wort für ein Nomen agentis hält (aller- dings ohne Motionssuffix) und den Laryngal analogisch aus den Wörtern für Vater usw. be- zieht, bzw. durch "Reimwortbildungen" nach den übrigen Verwandtschaftstermini auf *-ter-.


Selected readings


[Thanks to Asko Parpola]


  1. Twill said,

    November 18, 2020 @ 7:33 am

    It particularly strikes me how, granting the spurious "milker" etymology, it is totally unwarranted and unlikely to assume an association with milk for a basic kinship term would refer to that of an animal's. Surely the mother's milk is a little thicker than cow's milk!

  2. Philip Taylor said,

    November 18, 2020 @ 1:01 pm

    It is far too long since I last suckled mother's milk to remember how thick (or otherwise) it was, but is it also possible that you (Twill) are being led astray by the current predilection for skimmed milk ? I drink only full-fat Jersey milk, and I suspect (though cannot prove) that it is more-or-less on a par with human (mothers') milk in terms of consistency …

  3. Martin Schwartz said,

    November 18, 2020 @ 2:15 pm

    I'm with the above Indo-Europeanists' consensus, but elementary clarifications may be helpful to non-Indo-Europeanists: The stem of the suffix of the Indo-Iranian word for 'daughter" (and similarly
    'mother', 'father', 'brother') is *-t(a)r-, but the nominative
    *-tā. Further the schwa between g and d of the Avestan word
    merely reflects a recitation feature; the word is phonemically
    /dugdar-/. The Iranian here differs from the Old Indic in their respective treatment of the laryngeal; it disappears in Iranian,
    but becomes -i- in Old Indic.
    Martin Schwartz

  4. Terry Hunt said,

    November 18, 2020 @ 2:21 pm

    I suspect, Philip Taylor, that Twill was (rather creatively) adapting the well-known metaphor "blood is thicker than water", rather than commenting literally on the relative viscosities of the two varieties of mammary secretions concerned.

  5. Philip Taylor said,

    November 18, 2020 @ 2:57 pm

    You may well be correct, Terry (that was a thought that had not entered my mind) but if the milk to which the word "daughter" putatively refers, why would female be singled out ? Both sons and daughters suckle at their mothers' breast, but in many societies only the females milk the cattle.

  6. Jerry Friedman said,

    November 18, 2020 @ 8:48 pm

    I'm struck, as I have been before here, by the way some experts are certain and some express doubt.

    For the convenience of anyone who's interested, the origin of the Sanskirt verb duh- for 'milk, give milk' is said to be

    *dheugh- (Jay Jasanoff, Michaël Peyrot (who spells it *dheugh-), Michael Weiss, Georges-Jean Pinault, Douglas Adams, and Don Ringe (who spells it *dhewgh-)), meaning 'produce' or 'be produced', with a Greek cognate teukhe/o Joshua Katz discusses the reasons for doubting that this word—I think it's this word—is the origin of *dhugh2-ter, the source of English daughter.

    *dhugh- (Alexander Lubotsky, Jeremy Rau), meaning 'fit, be fitting'.

    The source of Greek tukhano 'happen by chance', the original word probably originally meaning 'yield' (Brian Joseph).

    Hope I got that right, including the superscripts.

  7. Jerry Friedman said,

    November 18, 2020 @ 8:51 pm

    Darn it, no superscripts. The h's are supposed to be superscripts in dhugh- and in Michaël Peyrot's spelling of dheugh-

  8. Bathrobe said,

    November 18, 2020 @ 8:52 pm

    whatever the original meaning of p- and mā- may have been (some think of baby language, almost world-wide spread, note Chinese!)

    My understanding is that 爹 diē 'father' and 娘 niáng 'mother' are the older terms in Chinese. 爸 and 妈 are relatively recent.

    In Japanese, on the other hand, my understanding is that 父 chichi 'father' and 母 haha 'mother' do date back to the older forms titi and papa.

    (I am not a historical linguist. The above are fairly rough and ready characterisations.)

  9. Bathrobe said,

    November 18, 2020 @ 8:54 pm

    Does this work? dhugh.

    If there are no superscripts, then obviously not.

  10. Jerry Friedman said,

    November 18, 2020 @ 9:46 pm

    Bathrobe: I tried to use the <sup> tag. "View source" shows no such tag in my comment and nothing in yours other than "dhugh". I guess the software strips those tags out of comments.

  11. Hiroshi Kumamoto said,

    November 18, 2020 @ 9:46 pm

    The book of Oswald Szemerényi on IE Kinship terms referred to above (actually one half of the book called Varia 1977) is available for download in pdf:


    Click on the link (High res. or Low res.) at the upper left.

  12. Chris Button said,

    November 18, 2020 @ 11:03 pm

    @ Bathrobe

    Regarding 娘 and 爹:



  13. Philip Taylor said,

    November 19, 2020 @ 3:28 am

    I don't know where the superscripts were intended, but if (a pure guess) the first "h" of Bathrobe's "dhugh" was intended to be raised, then one could achieve that effect by using U+02B0 : MODIFIER LETTER SMALL H, as in dʰugh.

  14. Jerry Friedman said,

    November 19, 2020 @ 10:12 am

    Philip Taylor: I'm not Bathrobe, but thanks.

    I'm sure Bathrobe was trying to help me raise the h's I was talking about, which are the same ones that are raised in some of the linguists' comments in the original post.

  15. Philip Taylor said,

    November 19, 2020 @ 10:39 am

    Jerry F — "I'm not Bathrobe" — I did not think you were ! But it was Bathrobe who wrote :

    Does this work? dhugh.

    If there are no superscripts, then obviously not.

    and therefore Bathrobe whom I sought to help.

  16. Jerry Friedman said,

    November 19, 2020 @ 5:18 pm

    Yes, but since it helped me too, I wanted to thank you even though I wasn't the person you were addressing.

  17. Andrew Usher said,

    November 19, 2020 @ 11:07 pm

    It might be some help to seeing the reconstruction if we knew how this Sanskrit 'h' was pronounced – /x/, /ɣ/, /h/, /ʔ/, or maybe nothing? I don't think everyone can be expected to know that, but I'm sure it's been reconstructed.

    k_over_hbarc at yahoo.com

  18. Andreas Johansson said,

    November 20, 2020 @ 1:17 am

    Sanskrit intervocalic 'h' is [ɦ], as I understand it.

  19. Philip Taylor said,

    November 20, 2020 @ 5:48 am

    So dʱugh ? How should that be represented in pure IPA, Andreas ?

  20. Andreas Johansson said,

    November 20, 2020 @ 8:16 am

    @Philip Taylor:

    The root *dhugh- (also cited as dheugh-) is neither Sanskrit nor does it involve an intervocalic 'h', so I believe you're somehow confused, but FWIW the mainstream reconstruction of PIE phonology would have it come out as something like [dʱugʱ] (plus affixes to turn it into an actual word rather than a bare root).

  21. Andrew Usher said,

    November 20, 2020 @ 8:28 am

    I considered listing [ɦ] as a possibility, but did not as I take it to be simply a predictable variant of [h]. So it apparently descends from an earlier 'voiced aspirate', which lost the plosive element.

  22. AJD said,

    November 20, 2020 @ 10:08 am

    Can anyone explain why the PIE *g (not *gh) in *dhugh2ter produced *h in Germanic, not *k? Does it voice-assimilate to the *t after the loss of the laryngeal but before Grimm’s Law?

  23. Philip Taylor said,

    November 20, 2020 @ 11:38 am

    Thank you, Andreas. I confess that I could not see why the raised dʱugh in "dʱugh" was considered inter-vocalic, but assumed that that was simply ignorance or stupidity on my part. As to *dhugh not being Sanskrit, that completely escaped my notice — I was clearly assuming a linear flow of question—answer, and failed to realise that there were discontinuities in the sequence.

  24. Jerry Friedman said,

    November 20, 2020 @ 11:06 pm

    Andreas Johansson: So you're saying that *dʱugʱ- meaning 'to fit, be fitting' and *dʱeugʱ- meaning 'produce' or 'be useful' are the same word with different choices of reconstructed spelling and meaning? As you may have noticed, I totally missed that.

  25. Andrew Usher said,

    November 21, 2020 @ 1:26 pm

    Yes, if those are the reconstructions, they would be different 'grades' of the same root, and the meanings are clearly related.

    The question of why Germanic has /h/ in the 'daughter' root does deserve asking and devoicing seems entirely plausible; it does not have to have been before Grimm's Law any more than the voicing of Verner's Law does (as the actual sounds concerned were, post-Grimm, [x] and [ɣ]). It might be a clue that there is no Germanic sign of the extra syllable that Sanskrit has i.e. the laryngeal was necessarily deleted or merged.

  26. Andreas Johansson said,

    November 22, 2020 @ 2:45 pm

    @Jerry Friedman:

    As Andrew Usher says, they're different grades of the same root.

    So what's "grades"? Basically different forms of the same root, specifically differing in what vowel they have, occuring in different contexts. The phenomenon may be seen in the inflection of English strong verbs, like sing~sang~sung, but it was far more pervasive in PIE than it is in English.

  27. Jerry Friedman said,

    November 23, 2020 @ 12:24 am

    Thanks, Andrew Usher and Andreas Johansson.

  28. Moonfriend said,

    November 23, 2020 @ 2:35 am

    Why am I now hankering for some دوغ?


RSS feed for comments on this post