"Lord of Heaven" in ancient Sino-Iranian

« previous post | next post »

[This is a guest post by Chris Button about xiān 祆 (usually defined in English as:
Ahura Mazda, god of the Zoroastrians 
From: Shuowen Jiezi, circa 2nd century AD
Xiān: húshén yě. [Pinyin]Xian is the god of the foreigners.
The two components of the 祆 glyph are shì / ("show, reveal, manifest; spirit") and tiān ("sky, heaven, celestial").
Although hugely important in the history of religions in China, the etymology of xiān 祆 is highly elusive.  Through close attention to the phonology of the glyph and its components, Chris aims to ferret out the source of a possible loanword.]
I've been pondering over 祆 EMC xɛn "Ahura Mazda, Zoroastrianism" for a while and its possible relationship with 天 EMC tʰɛn "heaven" (compare 忝 EMC tʰɛmˀ with 天/祆 as phonetic in the top half). 

Pulleyblank 1983 (originally going back to a 1962 comment) suggests the EMC xɛn ~ tʰɛn alternation reflects dialectal divergence from an OC lateral fricative ɬ-. In 1995, he changed the lateral fricative to a velar fricative xj- and noted the palatalization of the -m coda to -n.
I favor a reconstruction of 祆 EMC xɛn as OC xjə̯m, but I don't accept a dialectal relationship with 天 EMC tʰɛn OC tʰə̯ɲ since 天 appears to have 丁 EMC tɛjŋ OC ta̯ɲ as phonetic in the oracle-bone script (note the schwa/a ablaut).
It occurs to me that xjə̯m for 祆 perhaps reflects a loanword instead that just became associated/confused with 天 on account of some semantic and phonological proximity. One source of the loan could be a compressed form of what became variously Ahuramazda, Ormazd, Hormoz, etc. Alternatively, if we are willing to go back a little further, xjə̯m could perhaps reflect an intermediary stage still retaining some features of Indo-European h₂ems- that was the source of just the Ahura "lord, deity" component.


Selected reading


  1. Martin Schwartz said,

    May 28, 2024 @ 4:02 pm

    I can't say if Early Middle Chinese could have
    derived the reconstructed form in question from contraction a Middle Iranian form of Old Iranian Ahura Mazdā, but the second alternative
    (with us going back "a little further") would have EMC borrow a Proto-Indo-European etymon for 'to beget' as the designation
    of the god of Zoroastrianism. Moreover, while the latter etymon explains
    Hitt. hassu- 'king' (cf. the parallel of Eng. "king" < PIE √g'enh1 'to beget'), that (Old Avestan ahu- pej. 'lord' and) Vedic ásura-, OAv. ahura-
    'lord' derive from PIE *√h2ems is quite uncertain phonologically,
    morphologically, and sematically.

  2. Chris Button said,

    May 28, 2024 @ 4:15 pm

    @ Martin Schwartz

    can't say if Early Middle Chinese could have
    derived the reconstructed form …

    I'm talking about a possible Old Chinese (OC) xjə̯m, not Early Middle Chinese (EMC)!

  3. Chris Button said,

    May 28, 2024 @ 4:23 pm

    So a proposed OC xjə̯m gives EMC xɛn.

    I was following Guus Kroonen's (2013) proto-Germanic dictionary for the PIE form, but appreciate any guidance there.

  4. Victor Mair said,

    May 29, 2024 @ 2:47 pm

    In the last sentence of the o.p., Chris cited Indo-European h₂ems-.





    From AHD 5th ed., IE roots appendix:


    Spirit, demon. Oldest form *h2e/onsu‑, colored to *h2a/onsu‑.

    Aesir; Asgard, from Old Norse āss, god, from Germanic *ansu‑.
    Suffixed zero-grade form *n̥su-ro‑. Ahura Mazda, Ormazd, from Avestan ahura‑, spirit, lord.

    [Pokorny ansu‑ 48.]


    Wikipedia, "Dunhong":

    Victor Mair noted the word kaelum in the Tocharian languages of the Tarim Basin, meaning "sky" or "heaven" (and therefore related distantly to the Latin caelum) which may have been the basis of qilian.

    Victor Mair, "Reflections on the Origins of the Modern Standard Mandarin Name 'Dunhuang,'" in Li Zheng et al., eds., Ji Xianlin Jiaoshou Huadan Jinian Wenji (Essays for the eightieth birthday of Professor Ji Xianlin), vol. 2, p. 933.


    Cf. xiān 祆 ("Ahura Mazda, god of the Zoroastrians")

    Old Sinitic

    (Zhengzhang): /*qʰliːn/


  5. Chris Button said,

    May 29, 2024 @ 3:29 pm

    This is Kroonen's take from p.32 on the PIE form:

    "… which in view of CLuv. ḫamsa/i- c. ‘grandchild’ must be reconstructed as h₂ems- (Kloekhorst 2008: 293)."

    The Kloekhorst reference seems to be "An etymological dictionary of the Hittite inherited lexicon."

    But, regardless, a contraction of Ahuramazda, Ormazd, Hormoz, etc. may be the better of the two options.

  6. David Marjanović said,

    May 29, 2024 @ 4:24 pm

    xjə̯m could perhaps reflect an intermediary stage still retaining some features of Indo-European h₂ems- that was the source of just the Ahura "lord, deity" component.

    There is indeed evidence that Proto-Iranian still had a [h] derived from *h₁ and *h₂. While completely gone as such in attested Avestan (both Old and Young), it is still present in a number of words in Modern Persian (as /h/ or /χ/ without known rhyme or reason; loans from other Iranian languages are presumably involved). Please burrow through the open-access work of Joachim Kümmel, I shouldn't do that this late at night.

    However, OC *xjə̯m strikes me as a bad match for (all or the first syllable of) Proto-Iranian *hasura- or for even Pre-Proto-Indo-Iranian *hm̩suro-.

  7. Martin Schwartz said,

    May 29, 2024 @ 8:06 pm

    Indo-Iranian asura (> OIr. ahura-) has the basis mg. 'lord' (in the secular sense, secondarily as a religious epithet); not 'spirit'. It is reflected
    (with mg. 'lord') in Uralic words retaining the *-s-.
    For Vedic ásura- 'a lord', see Wash Hale, Asura. Btw the Middle Persian
    form is Ōhrmazd (-mizd), not "Ormazd".
    Martin Schwartz

  8. Chris Button said,

    May 29, 2024 @ 10:17 pm

    @ Martin Schwartz

    Thanks for that. I think you've effectively ruled out a case for the second proposal with a putative/contentious *h₂ems-.

    So, returning to the first proposal, we have an internally reconstructed OC form *xjə̯m and a meaning of "Ahura Mazda". The idea that *xjə̯m represents a contraction of some form of Ahura Mazda (such as Ōhrmazd) remains tantalizing.

  9. Jonathan Smith said,

    May 30, 2024 @ 8:59 am

    The most serious discussion I see of the Chinese word at issue is in Coblin (1994), "Remarks on some early Buddhist transcriptional data from northwest China"; see Section 2.2 (pp. 155-156), "The Word Tian 天 in Early Northwest Chinese."

    Gist is that in these early 4th c. C.E. materials, the character "天" is used to transcribe Indic language syllables like (Sanskrit) -har(d)- or perhaps more directly (Pali) -hi(d)-, (Gāndhārī) -he(d)-.

    See also Coblin (1994), "A compendium of phonetics in Northwest Chinese" p. 341, where the author further points out that this "northwest dialect pronunciation of 天" is likely the same item written "祆" in early dictionaries (XEN or HEN or whatever in Middle Chinese transcriptions.)

    That's as good as it seems to get re: philology. This word is not "Old Chinese" — not in early texts AFAIK, definitely not "internally reconstructed." It doesn't end in -m. Etc, etc. It's unhelpful to refer repeatedly to an "OC *xjə̯m" in the presence of specialists in other areas as if it's a thing.

    Of course it is totally possible, perhaps even likely, that there's no etymologcal relation to modern Ch. tiān 天, relatives of which replaced the gutteral onset item in the west very early. And of course the general idea of borrowing from (e.g.) some Iranian language remains totally plausible.

  10. Chris Button said,

    May 30, 2024 @ 2:40 pm

    …definitely not "internally reconstructed …

    It doesn't end in -m.

    So the internal evidence (however far that extends) supports the idea that it does. Words in the phonetic series of 忝 show that to be the case.

    One can of course choose to discount that evidence as not compelling enough if it contravenes a proposed OC reconstruction system.

    But the palatalization of OC -jə̯m to EMC -ɛn here attests a regular sound change in my Pulleyblankian rendition of OC,

    Compare say 年 ᶮʣjə̯m and 稔 ᶮʣjǝmɁ

    Of course it is totally possible, perhaps even likely, that there's no etymologcal relation to modern Ch. tiān 天

    The oracle-bone evidence is not unassailable, but it is pretty compelling in my (and others') opinion. And it suggests phonological and semantic confusion with original 天 and 天/祆.

    And of course the general idea of borrowing from (e.g.) some Iranian language remains totally plausible.

    And some contracted form of "Ahura Mazda" would be a good place to look for a candidate.

  11. Chris Button said,

    May 31, 2024 @ 6:51 am

    And so the following could be proposed as a possibility:

    天 EMC tʰɛn OC tʰə̯ɲ perhaps from Tengri

    祆 EMC xɛn OC *xjə̯m perhaps from Ahura Mazda

  12. Chris Button said,

    June 1, 2024 @ 7:34 am

    The oracle-bone evidence is not unassailable…

    And if we do choose to reject it, we could simply have:

    天 EMC tʰɛn OC ɬjə̯m (Compare 忝 EMC tʰɛmˀ OC ɬja̯mɁ

    祆 EMC xɛn OC xjə̯m

    The x- pronunciation would then just be a non-standard pronuciation that for some reason became established. Variation between ɬj- and xj- is hardly untoward.

    And so perhaps loanwords have nothing to do with any of it?

    Although the sudden ascendancy of 天 to oust 帝 with the Zhou conquest of the Shang does need some explanation. Did the Zhou take Tengri (天 tʰə̯ɲ) to new heights? (-ɲ goes back to pre-OC -ŋʲ, although tʰ- would not in this case go regularly back to pre-OC st- and perhaps reflects the loan origin instead)

  13. Chris Button said,

    June 1, 2024 @ 8:06 am

    And then there is the question of whether 天 even meant "heaven" in the oracle bones, Fowler's UBC dissertation has a good discussion about this.

    I wonder if 天 really has an original sense more like "realm" and then fits into the extensive 町, 政, 城 …. etc. word family (note that all have the phonetic 丁 in their earlier forms).

  14. Chris Button said,

    June 1, 2024 @ 1:34 pm

    For now, I'm going with:

    天 EMC tʰɛn OC tʰə̯ɲ (native to OC)

    祆 EMC xɛn OC *xjə̯m ~ ɬjə̯m (onset variation as a result of the loan)

    @ Martin Schwartz

    What could we be looking at phonologically between Old Persian and Middle Persian Ōhrmazd?

  15. Chris Button said,

    June 2, 2024 @ 3:32 pm

    And to be clear, OC ɬjə̯m would give EMC tʰɛn just like OC tʰə̯ɲ would– hence the later confusion (while the variant OC xjə̯m would give EMC xɛn and remain distinct).

RSS feed for comments on this post