Sinitic for "iron" in Balto-Slavic

« previous post | next post »

[This is a guest post by Chris Button]

There are a couple of brief suggestions in Mallory & Adams' Encyclopedia of Indo-European Culture (1997:314;379) that the Lithuanian word geležis and Old Church Slavonic word želežo for "iron", which following Derksen (2008:555) may be derived from Balto-Slavic *geleź-/*gelēź- (ź being the IPA palatal sibilant ʑ), could possibly have a Proto-Sino-Tibetan association.

The Old Chinese word for 鐵 "iron" may be reconstructed as *ɬə́c which certainly looks promising in that regard. Since the Balto-Slavic form appears to be an isolate in Indo-European, whereas 鐵 *ɬə́c belongs to an extensive word family connected to shininess, most directly in this case with 錫 *sɬác "tin" following a proposal by Schuessler (note the ə/a ablaut), the direction of the putative loan must be into Balto-Slavic rather than into Chinese.

Readings



54 Comments

  1. Sally Thomason said,

    February 15, 2019 @ 11:47 am

    This proposed Sino-Tibetan origin for the Balto-Slavic word for "iron" doesn't look "certainly promising" to me. First there's the big problem that the putative S-T source word lacks the first syllable of the B-Sl word (and B-Sl doesn't have a prefix that could account for the extra syllable). The post's Old Church Slavic word for "iron" is mistaken: the second sibilant is z, not "zh" [memo to self: learn, finally, to put diacritics into comments and email]. I'm not up on current research on the history of the Balto-Slavic lexicon, but Vasmer's etymological dictionary of Russian links this word to Homeric Greek khalkos 'copper, bronze' (vol. 1, p. 416). Vasmer adds that Meillet, Meillet-Vaillant, and Mikkola say that it's a loanword 'from an unknown eastern language', so the idea that it's a loanword isn't in itself bizarre — especially since it's apparently monomorphemic and therefore too long to be a plausible Indo-European root. But that doesn't make Button's proposed source more plausible. Also, iron isn't shiny. It *is* hard, and at least some other sources link the word to words for hardness.

  2. Philip Taylor said,

    February 15, 2019 @ 12:03 pm

    Sally, the ugly appearance of "copper, bronze … or d" was probably caused by use of a dollar-symbol, which I believe trips this forum's renderer into maths mode.

  3. Victor Mair said,

    February 15, 2019 @ 12:34 pm

    I fixed it. The problem was caused by a special beginning single quote mark before "copper" and before "from" (just after "loanword"). Took a while to figure out exactly what caused the distortion. Once I did, though, all it took was for me to remove those two tiny, slanted single quotation marks and replace them with perpendicular ones.

  4. Chris Button said,

    February 15, 2019 @ 1:42 pm

    @ Sally Thomason

    First there's the big problem that the putative S-T source word lacks the first syllable of the B-Sl word (and B-Sl doesn't have a prefix that could account for the extra syllable).

    I think I did Mallory & Adams a slight disservice by not including the Proto-Sino-Tibetan form *qhleks that they cite. The form seems to come from Kun Chang's (1972) paper on PST "iron" although it is not referenced. It should also be noted that Kun Chang includes Miao Yao and Kam-Tai in his "PST" (by the way, Schuessler's dictionary includes Proto-Kam-Sui *kʰlit⁷ and Proto-Viet-Muong *khăc).

    The Old Chinese rhyme *-əc in 鐵 *ɬə́c comes from earlier *-əkʲ which ultimately comes from PST *-jək.

    The post's Old Church Slavic word for "iron" is mistaken: the second sibilant is z, not "zh"

    Sorry, my mistake. Mallory & Adams have *želě̀zo and I had moved the diacritic over one space.

    Vasmer's etymological dictionary of Russian links this word to Homeric Greek khalkos…

    Mallory & Adams note "similarities" there too.

    Also, iron isn't shiny. It *is* hard, and at least some other sources link the word to words for hardness.

    Actually iron is shiny (the ever reliable wikipedia describes it as "lustrous metallic with a grayish tinge") and the connection with 錫 *sɬác "tin" is hard to deny (its phonetic 易 originally depicted a celestial body emitting rays from which an extensive word family has developed – no it is not a picture of a "lizard" as some sources claim!!). However, it does turn black and some languages also make connections in this regard (e.g. Japanese kurogane lit. "black metal", or the use of Sanskrit śyāmá "black" for "iron") with which we can also include Chinese 驖 "black horse" on which see Wang Li's dictionary (1980:469).

  5. Victor Mair said,

    February 15, 2019 @ 2:22 pm

    From Peter Golden:

    This is extremely interesting: 鐵 tiĕ LH thet OCM lhêt or lhît (Schuessler, Etym. Dict. : 497; Schuessler, Minimal Old: 227 [20-9b,c] ). It seems like a bit of a stretch to get to Slavic (Old Slav. желѣзо, Mod. Russ. железо, Ukr. залізо, Belarus. жалеза not зелезо as Vasmer has it, and ultimately to Balto-Slavic *geleź-/*gelēź- although his connection of the latter with Homeric Greek χαλκός is not totally out of the picture. Usually, Slavic z, ž = g in other IE langs., cf. zlato/ zoloto “gold” (Pers. zar); žena “wife” (Gr. γυνή “woman", cf. Eng. queen), Pers. zan “woman;" Russ. zelënyj (зелёный) “green” Avestan zairi etc.

    Some Turkologists would derive Turk. temür “iron” from 鐵 tiĕ (see entry in Э.В. Севортян, Этимологический словарь тюркских языков. Общетюркские и межтюркские основы на буквы “Б”, “Г”, и “Д” (Москва: Наука, 1980): 188-190), but this is far from certain.

  6. Jerry Friedman said,

    February 15, 2019 @ 3:21 pm

    Chris Button: Actually iron is shiny (the ever reliable wikipedia describes it as "lustrous metallic with a grayish tinge")

    That description refers to the pure metal, which is not what you get from meteors or by smelting. I can't tell you what the iron known in Old Chinese times would have looked like.

    Hijack: The English word "iron" might be linguistically amusing because for many speakers, it refers to metals "on either side" of steel, with more carbon or with less. That's a simplification of both usage and metallurgy, though. (And some metallurgists seem to classify pure iron as "mild steel".)

  7. Axel Schuessler said,

    February 15, 2019 @ 3:50 pm

    I agree with Sally Thomason as far as I can follow her (I am not familiar with Slavic). Compared to other metals, iron is not exactly strikingly shiny, no one wears an iron wedding ring. Iron is certainly not shinier than some other metals.
    Tiě 鐵 has no Chinese etymology as far as I can tell, it is a Wanderwort. The homophone 'black horse' 驖 seems to be the same etymon, i.e. an 'iron-colored' horse.
    As to the question of what the oracle bone graph for 易 depicted: sometimes I have the discouraging impression that the interpretation of such graphs is like a Rorschach test, it says more about the mind of the viewer than the graph. (A plausible explanation seems to be Kenichi Takashima's). And when looking for jewelry, tin 錫 would not be at the top of the list for its shininess either.

  8. David Marjanović said,

    February 15, 2019 @ 4:33 pm

    Intriguing. Are there similar words anywhere geographically between Sinitic and Balto-Slavic? Ideally some that could account for the *ge- part, which cannot be a prefix within Balto-Slavic?

    What's Tocharian for "iron"?

    may be reconstructed as *ɬə́c

    That may be so, but to the best of my knowledge your Old Sinitic reconstruction is not yet published, so it is hard to evaluate. Now, I couldn't do that anyway; but, just for comparison, what do published reconstructions (notably Baxter & Sagart 2014) have for 鐵?

    memo to self: learn, finally, to put diacritics into comments and email

    Just copy them in from the character map. That's what I do.

    Windows: Start > All Programs > Accessories > System Programs > Character Map. After you've brought it up once, it'll show up in the start menu.

  9. Chris Button said,

    February 15, 2019 @ 4:51 pm

    @ Peter Golden

    It seems like a bit of a stretch to get to Slavic…

    Old Chinese *ɬə́c (with some evidence for a velar prefix/pre-syllable in cognates elsewhere) strikes me as pretty close to Proto-Balto-Slavic *geleź-/*gelēź-, but I'm by no means a Balto-Slavic specialist!

    @ Axel Schuessler

    As to the question of what the oracle bone graph for 易 depicted: sometimes I have the discouraging impression that the interpretation of such graphs is like a Rorschach test, it says more about the mind of the viewer than the graph.

    Yes, I addressed that issue recently here:

    http://languagelog.ldc.upenn.edu/nll/?p=41538#comment-1559654

    A plausible explanation seems to be Kenichi Takashima's

    Takashima says “sunlight coming from the half-visible sun or even moon” which is exactly what I’m going with.

    Compared to other metals, iron is not exactly strikingly shiny…

    No it's not. But it is surely related to 錫 *sɬác "tin" which does belong to a word family with a semantic field of "shiny/emission". I'll post some of the word family later tonight when I have more time.

    It might also be noted that the shininess, or lack thereof, of "iron" also has no bearing on the phonological questions here. The only reason I introduced it was to give evidence for a possible direction of the putative loan.

  10. Victor Mair said,

    February 15, 2019 @ 6:28 pm

    I started to write a comment on yì 易 ("change; easy") and was determined to keep it brief. Despite my best efforts, it has swiftly ballooned to a size that is too large to include as a comment here, so I will make it into a separate post within a day or two.

  11. Axel Schuessler said,

    February 15, 2019 @ 7:30 pm

    About the graph 易: I thought Takashima once suggested that the oracle bone graph showed liquid poured from one vessel into another, the idea of "change". But I may misremember.
    About the color of iron, at least one person in ancient China calls it 'black metal': Xu Shen in his Showen jiezi.
    A hypothetic-deductive reconstruction for tiě is not in Baxter&Sagart 2014, but Sagart 1999 has OC *hlit.

  12. AntC said,

    February 15, 2019 @ 8:37 pm

    After the two earlier lengthy threads, I am by now hugely suspicious of any attempt to link a single word sound-alike/sense-alike. In the case of early Chinese, we can add into the confusion look-alike of the Oracle bone characters.

    Two questions. (The first also applies to Chris B's single-word observation of the 10-character puzzle, which was a start but far from convincing.)

    1. Is it ever legitimate Comparative Method to claim a loan on the basis of a single word? As opposed to a pattern of multiple words with parallel sound changes.

    2. When/how did the Chinese acquire iron/smelting and associated technology? What other iron-related words are there in Chinese? Are their derivations/sources attested?

    Tutankhamun's meteorite iron dagger (C14th BCE) shows iron working is ancient, but when did it become commonplace? And where did it spread from?

    Wouldn't Chinese already have words for iron before the Proto-Slavic expansion (600 CE)?

  13. AntC said,

    February 15, 2019 @ 8:59 pm

    Ah, apologies: please ignore that question 2 (or at least turn it round the other way). Wouldn't the Proto-Balt-Slavs have already acquired iron/smelting before any likelihood of contact with Chinese? (acquired by ~500 BC in Eastern/Northern Europe, according to wp) Did the Avar state use iron/steel weapons, when they allied with the Persians and Slavs?

    Strewth the history is complicated that prompted the Slavs to expand: there would seem to be gazillions of languages/warrior states from which they could have acquired technology and words.

  14. Chris Button said,

    February 15, 2019 @ 9:57 pm

    @ David Marjanović

    Pulleyblank (1973 – "Word Families in OC") has *lhəkʲ. He then vacillates between *-c (1977-8) and -kʲ (1991). The only area where I fundamentally differ from Pulleyblank here is that I do not believe the palatal codas were original, but that there was rather a shift of **ɬʲə́k > **ɬə́kʲ > *ɬə́c that merged with **ɬʲə́t > *ɬə́c. Baxter & Sagart's online reconstruction is l̥ˤik (*-ik is assumed to have merged in their system with *-it, hence the *-t coda in Schuessler's Minimal OC).

    @ Axel Schuessler

    You're not mistaken. Takashima says the following in his Bingbian commentary (I'm omitting the inscriptional forms in the original text):

    p.153

    The bone graph can be interpreted as a depiction of what I take to be the sunlight coming from the half-visible sun or even moon — we should remember that it is no simple task to depict the meaning of a change from one state to another.

    p.174

    "The grapheme in yi 易 has been interpreted as the sun rays emerging from the half-visible sun or moon… This palaeographic interpretation may still be alright as is, but there seems to be a difference in the bone graph [here] and the [one] here; that is, the right-hand side of the latter could be the half-visible sun or moon, but that of the former could be a depiction of a basin (i.e., min 皿) and its content is being emptied out as indicated by the tipped basin. But the graphic distinction between [the one] and [the other] is so fine as to be difficult for the scribe to distinguish them. It seems more cogent that both forms may be taken as variants. That is, both forms are susceptible of interpreting the sun rays emerging from the sun and a basin being tilted for the liquid to pour."

    @ Victor Mair

    In anticipation of your post, sticking to the phonetic series of 易, we might compare (purely in terms of universal semantic evolution rather than any notions of common origin) 易, 剔, 剔, 惕, 埸, 賜, 錫 with yield, glabrous ("hairless"), gloat (earlier sense of "look furtively"), guild, geld ("payment, tribute"), gold (as generic "shiny metal") all from Proto-Indo-European *gʲʰel- "shine, emit".

  15. AntC said,

    February 15, 2019 @ 10:54 pm

    Ancient Greek had khalkos "ore, copper, bronze;" [Etymonline], as Sally noted. The word's meaning morphed so much that Latin brought in a new word for copper, from the island Kypros which was its main source.

    Ok so khalkos is metal-harder-than-copper/suitable for weapons. Hand that through a few languages and their warring states. I could easily see it turning up as *geleź-/*gelēź-/*želě̀zo iron. At least khalkos has two syllables, an /l/ in the middle and a velar initial. PST *qhleks could also be a cognate(?) Yes anybody can play this game.

    However high the evidential bar is set to reject that from-Greek hypothesis, I'm not seeing the bar set so high for the from-Chinese hypothesis.

    Latin 'ferrum' is of unknown origin, possibly from blah blah. Latin 'bronzium' is of uncertain origin (its meaning in antiquity included brass). English 'brass' is a mystery word. To borrow a gag from J K Galbraith: Philology "was invented to make Astrology seem respectable".

  16. Philip Taylor said,

    February 16, 2019 @ 4:11 am

    JKG via Ant C ("Philology 'was invented to make Astrology seem respectable'"). That truly brought a smile to my face. and in fact I am still laughing. Thank you !

  17. David Marjanović said,

    February 16, 2019 @ 6:07 am

    Thanks for the different reconstructions – they're all very similar, practically different notations of the same thing, so clearly the reconstruction is reliable enough to use in further argumentation.

    (The most conspicuous difference between them is whether the frontness feature is localized on the initial consonant, the vowel or the final consonant. Its presence somewhere in the syllable seems uncontested.)

    AntC, you've confused Proto-Slavic (~ 600 CE) with Proto-Balto-Slavic (some 2000 years earlier). Wiktionary confirms that the word is all over Baltic: Lithuanian geležìs, Latvian dzelzs, Old Prussian in a German-based spelling gelso.

    yield, glabrous ("hairless"), gloat (earlier sense of "look furtively"), guild, geld ("payment, tribute"), gold (as generic "shiny metal") all from Proto-Indo-European *gʲʰel- "shine, emit".

    Yellow (German gelb) is another one.

    But it's difficult enough to get the Greek word lined up with the Proto-Balto-Slavic word. Projecting the former back to PIE gives us a range of "roots" in zero-grade, *g(ʲ)ʰ(h₂)lk(ʲ)-, if I haven't overlooked anything. Projecting the latter back to PIE gives us *g(ʷ)(ʰ)elegʲ(ʰ)- with *e on both sides of the *l, with an overlapping range of first consonants, and with a similar but non-overlapping range of last consonants. (The Lithuanian tone indicates absence of a laryngeal if I'm not confusing things.) Concerning the last consonant, the Baltic versions with their short or absent second vowels seem to indicate *gʲʰ, while the Slavic one, with its long second vowel, seems to indicate *gʲ (Winter's law), but there was an active derivational process that involved vowel lengthening, so we'll need a specialist to tell us if there's an obstacle to a unitary Proto-Balto-Slavic form.

    The Greek word could be the zero-grade of *gʲʰel- followed by an unexplained *-k(ʲ)-. For the Balto-Slavic ones, a derivation from *gʲʰel- plus unexplained *-egʲ(ʰ)-… could work if we employ Weise's law (*KʲR, *KʷR > KR in Balto-Slavic and Indo-Iranian if the whole cluster belongs to the same syllable; start from around footnote 3 here).

    Half a sentence on Wiktionary sets up a PIE "*gʰelgʰ-, probably borrowed from Asia Minor"; on what evidence, I don't know, and note that final *gʰ is not reflected in either Greek or Balto-Slavic.

  18. David Marjanović said,

    February 16, 2019 @ 6:10 am

    could work if we employ Weise's law

    If we go to zero-grade and back, of course. Maybe a bit of Schwebeablaut would do it. *handwave*

  19. David Marjanović said,

    February 16, 2019 @ 6:11 am

    Oh, sorry:

    1. Is it ever legitimate Comparative Method to claim a loan on the basis of a single word? As opposed to a pattern of multiple words with parallel sound changes.

    Sure. The inference will just be much less robust.

    The Comparative Method is only science. It deals in probabilities, not in true vs. false, proven vs. unproven. Metaphysical certainty is not to be had.

  20. Alexander Vovin said,

    February 16, 2019 @ 6:44 am

    Interesting hypothesis, but, sorry I am not persuaded. There are several serious obstacles, I believe.
    1. Note that it is OCS желѣзo, not *жьлѣзo, which would be better fit for an Chinese (OC) cluster.
    2. As OC *lhʕik (a source of Siamese *hlek D1S, see the same artcile by Kun Chang that you cite) shows, OC has a voiceless onset, while OCS желѣзo points to a voiced one.
    3. Why does OC -k surface as OCS з? Why OC -i- (or -e-) is reflected as OCS -ѣ-? See also 1. above.
    4. Iron melting did not start in East Asia before 600 BCE. Although I always try to avoid absolute dating unless it is documented philologically, this seems to be a ittle bit too late for Balto-Savic (if this node in IE has ever existed).
    5. The absence of intermediaries s bizarre and needs to be explained.
    In sum, this looks like an accidental chance resemblance.

  21. Chris Button said,

    February 16, 2019 @ 7:45 am

    @ Alexander Vovin

    I reconstruct 鐵 **ɬə́kʲ > *ɬə́c. A shift of c > s elsewhere is not extraordinary and has typological parallels (could some intervocalic voicing have also played a role?). Separately, there is no evidence for a velar prefix (or pre-syllable as some may have it) in Old Chinese itself.

  22. Victor Mair said,

    February 16, 2019 @ 10:11 am

    From Mark Gamsa (in Riga):

    It's dzelzs in Latvian, by the way, which of course sounds a lot like the Russian zhelezo.

  23. Victor Mair said,

    February 16, 2019 @ 11:00 am

    At the conclusion of this comment, Chris Button directed these remarks to me:

    =====

    …sticking to the phonetic series of 易, we might compare (purely in terms of universal semantic evolution rather than any notions of common origin) 易, 剔, 剔, 惕, 埸, 賜, 錫 with yield, glabrous ("hairless"), gloat (earlier sense of "look furtively"), guild, geld ("payment, tribute"), gold (as generic "shiny metal") all from Proto-Indo-European *gʲʰel- "shine, emit".

    =====

    In the context of our current and recent discussions about the possibility of IE-Sinitic interactions, this strikes me as a radical departure from what everyone else has been doing or denying. I have read large portions of Chris's draft dictionary, and it is full of similar observations.

    No one has responded to what Chris has accomplished in the above quoted comment, in which he advances the concept of "universal semantic evolution rather than any notions of common origin" and gives an impressive group of seven related terms. If this methodology can stand, it is of profound consequence for future research on IE-Sinitic interactions.

  24. Victor Mair said,

    February 16, 2019 @ 1:09 pm

    From Tsu-Lin Mei:

    There is the following item in F.K. Li’s Handbook of Comparative Tai: “Iron” Siamese lek D1S; 8.2, 14.6 Proto-Tai *hlek. This is a loan from Chinese 鐡 *thik > *thit > thiet. Iron Age in Chinese is during the Warring States Period, 200 – 400 BCE, which is also the date of Chinese 鐡*thik and its loan into Proto-Tai. I am skeptical about Sino-Tibeto-Burman “Iron” for two reasons. There is no Tibetan cognate for Chinese *thik “Iron”, nor Burmese cognate. Two, the date is wrong. PST is dated to 5000-6000 BP. At that time, there is no mineral recognizable as Iron.

  25. Jonathan Smith said,

    February 16, 2019 @ 4:07 pm

    @Prof. Mair and Chris Button re: the seven suggested associations 易/yield, 剔/glabrous, 惕/gloat, 埸/guild, 賜/geld, 錫/gold: this level of congruence between Chinese "word families"-cum-phonetic series and IE roots would be remarkable even if we were talking about only this single instance, let alone the "large number of similar observations" that we are to understand are on offer in Chris's dictionary; bringing forward a large collection of such elaborate associations, but characterizing them as illustrative merely of "universal semantic evolution", would be… coy, at least. Why not ride or die? My 2¢ as always —

  26. Jonathan Smith said,

    February 16, 2019 @ 4:13 pm

    fine, six associations

  27. Victor Mair said,

    February 16, 2019 @ 7:15 pm

    From Tsu-Lin Mei:

    Sally Thomason’s note reminds me that I heard Kun Chang’s paper on PST “iron” in 1970, when the Sino-Tibetan conference was held at Cornell (that was before I joined the Cornell faculty). At the meeting Paul Benedict objected to Kun Chang’s inclusion of Kam-Tai as Sino-Tibetan. The Tai item at issue is what I cited from Handbook of Comparative Tai. P.S. At the 1970 S-T conference I presented on behalf of Jerry Norman our “Austroasiatics in Ancient South China”. It was a big hit. The paper was published in 1976 because Jerry Norman had written a long section on the Austroasiatic origin of Chinese Heavenly Stems and he was unwilling to part with it.

  28. Victor Mair said,

    February 16, 2019 @ 7:19 pm

    From Tsu-Lin Mei:

    Handbook of Comparative Tai has under “change, exchange” 7.6 (9) D2L Siamese lEEk, Po-ai liik. This a loan from Chinese 易*lik > jiak. Baxter & Sagart (2014) has OC reconstruction *lik for 易also . F.K. Li in making Handbook of Comparative Tai tried to weed out obvious Chinese loans, but there are still a few left, e.g. 鐡,易,鬲。I do not know how to reconstruct 锡。

  29. Chris Button said,

    February 17, 2019 @ 7:41 am

    … characterizing them as illustrative merely of "universal semantic evolution", would be… coy, at least.

    To be absolutely clear, unlike Pulleyblank, I do not believe there is any evidence to show that Old Chinese and Proto-Indo-European came from a common source. The phonology at the time-depths to which we can reconstruct is simply not there at all. There is, nonetheless, good evidence of interaction between speakers of languages of the two language families in certain items of vocabulary as Victor Mair is trying to bring to the fore. I am simply demonstrating through common semantic evolution that humans in different communities often draw similar connections when trying to make sense of the world around them.

    Having said that, what seems to have mainly led Pulleyblank down the line of thought toward a common origin, is that when one does not force the data or employ notational sleights of hand, Old Chinese and Proto-Indo-European both go back to a ə/a (usually treated as e/o in the latter) vowel system. Since /ə/ represents the default "syllable", we are then left with what is really no vowel system at all. This structural association, about which I believe Puellyblank was absolutely correct in theory (if not always in demonstration), seems to represent an underlying fact about human language regardless of how it actually came about.

    I am skeptical about Sino-Tibeto-Burman “Iron” for two reasons. There is no Tibetan cognate for Chinese *thik “Iron”, nor Burmese cognate. Two, the date is wrong. PST is dated to 5000-6000 BP. At that time, there is no mineral recognizable as Iron

    Mallory & Adams voice a similar concern regarding time-depth over the possibility of a Proto-Sino-Tibetan term meaning "iron". They are certainly correct. Separately, I can personally vouch for the lack of a Burmese cognate (at least I am unable to find one). I am not a specialist in Written Tibetan, but does the phonology of lcags suggest that it is unequivocally a (Chinese?) loanword?

    Separately, there is no evidence for a velar prefix (or pre-syllable as some may have it) in Old Chinese itself.

    Thinking about this again, it helps that 鐵 has a lateral onset since, as would naturally be expected, it is in combination with liquids and rhotics that we have evidence for such clusters in Old Chinese. Given the clear association with 錫 *sɬác, if we treat 鐵 *ɬə́c (< **ɬə́kʲ) as Chinese in origin, then the possibility remains that the hypothetical velar component was simply dropped in Old Chinese. Alternatively, the velar component attested in other loans could simply be a separate development that was then spread through further secondary loans.

  30. James Wimberley said,

    February 18, 2019 @ 8:01 am

    Sally Thomason
    To put diacritics in email and comments:
    1. Stop using a web-based interface like the Gmail default.
    Then, in increasing order of range:
    2. Use the smartphone virtual keyboard – hold down a letter and a small selection of Western European characters with diacritics and variant forms like "ß" will pop up.
    3. Install a local email client like Thunderbird. Use Insert>Special characters and symbols to access a much larger list of characters, basically everything for Western and Central European Latin alphabets.
    4. Compose in a full-fledged word processor program like Word or LibreOffice Write. This will offer a larger range of characters, including Greek, Cyrillic, Hebrew and mathematical symbols. You can SFIK get extensions for IPA and Asian languages but I'm getting out of my depth here. Or try copying from websites.
    To save time hunting in the massive table, you can print out a small table of the characters you commonly use with their ASCII short codes (+nnn), and stick it to the side of your desktop.
    5. Copy and paste your now enriched text into your email or comment. LL, as you would jolly well expect, accepts everything I've thrown at it. Other sites, you may not be so lucky. But you tried.

  31. Chris Button said,

    February 18, 2019 @ 9:26 am

    Regarding 鐵 tʰɛt < *ɬə́c "iron" and 錫 sɛjk < *sɬác "tin", I was half-expecting someone to challenge the role of the /s/ component in the latter (the ə/a alternation is of course pervasive across the lexicon – to cite an example with a *kɬ- onset as it pertains to the discussion here, one might compare 齒 tɕʰɨ' < *kɬə̀ɣʔ "tooth" and 杵 tɕʰuă < *kɬàɣʔ "pestle" with which purely in terms of common semantics may be compared "molar" and "mortar" from PIE *melh₂- "grind, crush").

    The problem with the s- cluster is that s is needed to account for the aspiration of onsets prior to Old Chinese as we find in Tibeto-Burman (e.g. st- > tʰ-, sp- > pʰ-, sl > ʰl- [ɬ-] etc.). As such, something like *sɬ- seems to imply two rounds of s- prefixation. While this is not impossible (e.g. Kuki-Chin languages show "superadded -s suffixation) under the effects of analogy), it is not particularly likely either. Although, generally in Chinese a connection of *ɬ- with a sibilant articulation is only attested in the palatalizing environment conditioned by "type-B" syllables marked here with a "grave" accent rather than "acute" (e.g. 聲 ɕiajŋ < *ɬàɲ), from an external typological perspective the broader merger of /ɬ/ with /s/ is well-known in for example Hebrew. Given that Chinese evolved as any natural language replete with variations (e.g. the unexpected, but not unnatural, palatalization of the -n coda in 因 *ʔə̀ɲ < **ʔə̀n, showing a "stretched out person 大 on a mat" undoubtedly related to 安 *ʔán in which we have the ə/a ablaut again and the semantics of which are comparable to the development of Welsh hedd "peace" from IE *sed- "sit", as attested by 恩 *ʔə́n), it might be better simply to reconstruct 錫 as *ɬác, to compare with 鐵 as *ɬə́c, with the notion that *ɬ- > s- somewhat exceptionally but again not unnaturally (we have a similar situation with 賜 too).

  32. David Marjanović said,

    February 18, 2019 @ 2:42 pm

    To put diacritics in email and comments:
    1. Stop using a web-based interface like the Gmail default.

    This is not necessary. I use the character map in all web-based interfaces (e-mail, blog comments, some webforms).

    somewhat exceptionally

    Do you mean as an irregular sound shift? Or have you been able to identify different dialects where one turned *ɬ- into *tʰ- while another turned it into *s-?

  33. Chris Button said,

    February 18, 2019 @ 8:49 pm

    Also, iron isn't shiny…

    This reminds me of one of my favorite semantic shifts in Chinese: the homophones 音 "sound" and 喑 "dumb, mute" are essentially the same word in origin in spite of having almost exactly opposite meanings. The connection is simply one of "inarticulate sound" (as opposed to 言 "speech" – the graphs of 言 and 音 only coming to be distinguished in the bronze inscriptions)

    Or have you been able to identify different dialects…?

    I would be very skeptical of any suggestion that actual dialects can be distinguished in Old Chinese. Having said that, as in any real language, there are cases where "non-standard" variant forms seem to have become the "standard" and then influenced their derivatives. In addition to 因 *ʔə̀ɲ < *ʔə̀n above, I mentioned a couple of others in some earlier LLog posts which I'll copy and paste here to avoid inputting more of those pesky diacritics :) …

    肘 *trə̀wʔ tr- is phonologically entirely reasonable but which is not "standard" (the phonetic is 九 *kə̀wʔ < *kʷə̀ɣʔ now deformed to 寸)

    般 *pán -n is phonologically an entirely reasonable consequence of bilabial dissimilation of the coda from the onset (which is actually attested in some later Chinese languages) but which is not "standard" in Old Chinese (the phonetic is 凡 *bàm now deformed to 舟).

  34. Chris Button said,

    February 18, 2019 @ 8:55 pm

    Re-posting the last part in quotes in which the formatting became messed up:

    肘 *trə̀wʔ < *kʷrə̀ɣʔ
    The shift of kr- to tr- is phonologically entirely reasonable but is not "standard" (the phonetic is 九 *kə̀wʔ < *kʷə̀ɣʔ now deformed to 寸)

    般 *pán < *pám
    The shift of -m to -n is phonologically an entirely reasonable consequence of bilabial dissimilation of the coda from the onset (which is actually attested in some later Chinese languages) but which is not "standard" in Old Chinese (the phonetic is 凡 *bàm now deformed to 舟).

  35. Anthony said,

    February 18, 2019 @ 10:03 pm

    [Cross-posted from another Language Log thread\

    Not meaning to derail this discussion, but we should note the passing of Professor Eric Hamp:
    https://en.wikipedia.org/wiki/Eric_P._Hamp

  36. Chris Button said,

    February 18, 2019 @ 10:47 pm

    The first also applies to Chris B's single-word observation of the 10-character puzzle, which was a start but far from convincing

    Incidentally we had another *ɬ- versus *s- situation in the 秀 EMC suwʰ (phonetic in 誘 EMC juw' < OC *lə̀wʔ ) component of 秀支 there.

    http://languagelog.ldc.upenn.edu/nll/?p=41519#comment-1559991

  37. R. Fenwick said,

    February 19, 2019 @ 12:20 am

    @Chris Button:

    However, it does turn black and some languages also make connections in this regard (e.g. Japanese kurogane lit. "black metal", or the use of Sanskrit śyāmá "black" for "iron") with which we can also include Chinese 驖 "black horse" on which see Wang Li's dictionary (1980:469).

    For what it's worth, Ubykh also does this: wəʨʷ’á "iron" etymologically represents "black/dark metal" (for –ʨʷ’a, compare gʲá(n)ʨʷ’a "coal", y(ə)ʨʷ’ạ– "on the ground", məʁaʨʷ’a– "black alder"), as opposed to wəɕʷá "copper", the "white/bright metal" (ɕʷa "white, bright", ɕʷə "to dawn", etc.); both arise from a Proto-Ubykh root * "metal" that is no longer attested alone, but present also in wəkʲ’ə́ "blacksmith".

    And I literally just now realise, for that matter: compare also English blacksmith!

  38. Victor Mair said,

    February 19, 2019 @ 8:27 am

    @R. Fenwick

    You will be happy to know that, within a week or so (when I find the time to bring together my notes), I will make a post on your beloved Ubykh.

  39. Jonathan Smith said,

    February 19, 2019 @ 11:47 am

    @ Chris Button "I am simply demonstrating through common semantic evolution that humans in different communities often draw similar connections when trying to make sense of the world around them."

    You often point to shared-root sets of words within OC and IE respectively — [OC1, OC2, … OC7] and [IE1, IE2, … IE7] — which you suggest share meanings 'A'-'G'. It is patently obvious, as proposed cases proliferate, that we cannot interpret these as illustrative only of parallel associations of the mundane sort that cause 'iron' to be derived from 'black', etc., etc., in unrelated languages. Thus the impression that you are tough to pin down :P

  40. Chris Button said,

    February 19, 2019 @ 1:12 pm

    @ R. Fenwick

    Ubykh…

    It gives me so much satisfaction just seeing all those "ə"s and "a"s :)

    as opposed to wəɕʷá "copper", the "white/bright metal" (ɕʷa "white, bright", ɕʷə "to dawn", etc.);

    Color associations seem to be very common with metals. In the case of 銅 *láŋʷ "copper" in Chinese, I find Schuessler's link with 彤 *lə́ŋʷ "red, vermilion" very persuasive (once again note the "ə/a" alternation)

    @ Jonathan Smith

    It is patently obvious, as proposed cases proliferate, that we cannot interpret these as illustrative only of parallel associations of the mundane sort…

    I'm not so sure about that; I think it is more reflective of our common humanity. It's also worth remembering that semantic shifts are not discrete occurrences but continuously evolving. I'm not saying that PIE and OC didn't come from a common source, I'm just saying that there is absolutely no linguistic evidence demonstrating that they did, unless one reads more into the structural ə/a association than I think is warranted (and fortunately we have phonological analyses of languages like Ubykh to show it is not some flight of fancy either, regardless of the time depth).

    Thus the impression that you are tough to pin down :P

    How so? I thought I was pretty explicit on where I stand.

  41. Jonathan Smith said,

    February 19, 2019 @ 1:58 pm

    @ Chris Button Because, across OC and IE in particular, you find "word families" displaying intricate structural congruencies which you deny reflect cognancy but become statistically fantastic if viewed as parallel development. Fair enough though if your view of the implications differs.

    In the above case in particular, the word yi4 易 doesn't mean anything like 'light' AKAIK. So I get concerned that the OC word-family as you perceive it is IE-inspired, consciously or sub-. In general, without beginning from attested semantics, it becomes tough to constrain our hypotheses.

  42. Chris Button said,

    February 19, 2019 @ 2:35 pm

    @ Jonathan Smith

    I'm not sure what you mean about an OC word family being "IE-inspired" since I am not suggesting any relationship in that regard. In fact, I would assume that many of these similar semantic extensions could be found in other language families outside of PIE and OC. There is a nice little article by David Wilkins called "Natural Tendencies of Semantic Change and the Search for Cognates" (1996) which is clearly very much inspired by Jim Matisoff's excellent work on Proto-Tibeto-Burman semantics (and beyond) in this area. I also don't think there is anything "statistically fantastic" about being able to find similar semantic shifts to Chinese somewhere in one of the numerous and well-documented attested Indo-European languages without implying cognancy (I do not just provide English examples, although I obviously preference them whenever I can – for example I note Swedish "glutta" rather than "gloat" in my dictionary entry for 惕 since the sense of "look furtively" is now obsolete in English).

    As for 易, the vast majority of cases in the oracle bones occur with 日 "sun" as 易日.

  43. Chris Button said,

    February 19, 2019 @ 2:53 pm

    I should probably add that the real reason for including the PIE comparanda is to avoid the speculativeness associated with a work such as Todo's "Kanji Gogen Jiten" (i.e. this semantic shift is probably reliable since here is another example where it has happened). A secondary, unstated, aim of the dictionary is to silence all those ə/a doubters by drowning them in a deluge of data across the OC lexicon without having to engage in the usual theoretical mumbo jumbo populated by cherry-picked examples…

  44. Chris Button said,

    February 20, 2019 @ 2:40 pm

    Incidentally we had another *ɬ- versus *s- situation…

    On the voiced side of things, this would also account for the parallel situation of OC *l- usually surfacing as EMC j- but sometimes also as EMC z- instead (e.g. 邪 with its two readings of EMC jia and zia from OC *làɣ)

  45. Chris Button said,

    February 20, 2019 @ 3:08 pm

    For the finickity, I should have specified in type-B syllables (as opposed to both A and B types for /s/)

  46. Chris Button said,

    February 21, 2019 @ 11:42 am

    Seems like I'm just talking to myself now :) Well, in any case…

    Old Chinese *ɬə́c (with some evidence for a velar prefix/pre-syllable in cognates elsewhere) strikes me as pretty close to Proto-Balto-Slavic *geleź-/*gelēź-, but I'm by no means a Balto-Slavic specialist!…

    Thinking about this again, it helps that 鐵 has a lateral onset since, as would naturally be expected, it is in combination with liquids and rhotics that we have evidence for such clusters in Old Chinese. Given the clear association with 錫 *sɬác, if we treat 鐵 *ɬə́c (< **ɬə́kʲ) as Chinese in origin, then the possibility remains that the hypothetical velar component was simply dropped in Old Chinese.

    I wonder if the velar component could derive from /ɬ/ > /xl/ in loans followed by some fortition? We do have evidence of the inverse: 虎 *ʰráɣʔ "tiger" (Schuessler has *hlâʔ) being an old loan from Mon-Khmer *klaʔ "tiger" as I noted on the "reindeer" thread here: http://languagelog.ldc.upenn.edu/nll/?p=41164#comment-1558720

  47. Victor Mair said,

    February 21, 2019 @ 1:02 pm

    Keep talking, Chris. We always learn from you, though you may not always get an overt response. Moreover, everything that you say goes on the record, and it's all part of a very large conversation involving many posts with numerous contributors and commenters.

    We're making gradual progress in understanding the dynamics of Eurasian language interaction, and you're playing a major role in our discoveries, for which thanks.

  48. Jonathan Smith said,

    February 21, 2019 @ 5:11 pm

    Yes I remain interested in the thread / the problems, but again it looks like we'll have to 不了了之 here which is fine :)

  49. R. Fenwick said,

    February 25, 2019 @ 4:22 am

    @Victor Mair: You will be happy to know that, within a week or so (when I find the time to bring together my notes), I will make a post on your beloved Ubykh.

    I sometimes fear my contributions to Language Log discussions must seem like I can't talk about anything else… I promise it's through nothing but my familiarity with the language, and the fact that as part of the broader Eurasian linguasphere it (and the rest of North-West Caucasian) often throws up intriguing comparanda in these sorts of etymological discussions. In any case, I'm very much looking forward to your post!

    @Chris Button: Color associations seem to be very common with metals. In the case of 銅 *láŋʷ "copper" in Chinese, I find Schuessler's link with 彤 *lə́ŋʷ "red, vermilion" very persuasive (once again note the "ə/a" alternation)

    Ooh, that's intriguing indeed. Is this ə/a ablaut very common in old Sinitic? Are there morphological functions to it? (Please forgive my utter ignorance – I'm essentially a complete novice to the historical linguistics of eastern Asia, though I'm learning a great deal through these comment threads.) If so, it's a fascinating addition to evidence for what appears to have been a typologically common vowel system among the protolanguages of Eurasia.

  50. Chris Button said,

    February 25, 2019 @ 9:17 pm

    @ R. Fenwick

    Ooh, that's intriguing indeed. Is this ə/a ablaut very common in old Sinitic? Are there morphological functions to it? (Please forgive my utter ignorance – I'm essentially a complete novice to the historical linguistics of eastern Asia, though I'm learning a great deal through these comment threads.) If so, it's a fascinating addition to evidence for what appears to have been a typologically common vowel system among the protolanguages of Eurasia.

    The ə/a ablaut pervades the Old Chinese lexicon. As you might expect, most reconstructions obscure it with supposedly more "natural" triangular vowel systems, but when handled objectively the evidence is incontrovertible. Edwin Pulleyblank tried hard to identify a morphological function to the ablaut but it is not very convincing so the phonological evidence is really all we have.

    Presumably as a result of the discomfort that such vertical vowel systems cause many linguists, I've read proposals that in languages like Ubykh such systems derive from earlier triangular ones as if that somehow renders them more "natural". I was wondering if you had any thoughts on the evidence behind such proposals?

  51. Eidolon said,

    February 26, 2019 @ 2:00 am

    @Chris Smith @Jonathan Smith

    Have either of you considered that, rather than cognancy or universal semantic evolution, what these shared semantic sets with no obvious phonetic etymology actually represent is sustained and widespread calquing? After all, if the Chinese had a word like 橘 "orange," and they heard about a culture that drank "orange juice," it would seem rather obvious for them to coin the calque 橘汁"orange juice" as a semantic and phonetic extension of Chinese 橘 and Chinese 汁. The result would show no signs of being a loan under standard etymological analysis, but would carry the total package with respect to its semantic *and* phonetic parallel with 橘 "orange". This example is contrived for the sake of simplicity, and I am not saying that Chris's 易 set is necessarily a case of it, but the tendency of calque groups to exhibit "intricate structural congruencies" is perhaps best illustrated, in recent history, by the wholesale translation of Western terms by the Japanese via Chinese morphemes, in which much of the semantic relationships embedded in the former's vocabulary were preserved in the latter's translation.

    There is no reason to believe that similar processes haven't been ongoing across Eurasia and beyond since ancient times, involving not just Sinitic and IE but an enormous variety of languages through vast, complex trade routes. The difficulty of detecting and pinning down sources and targets for calques, in this respect, could have obfuscated the degree to which cultures and languages could share ideas without either inheriting it from a common source or by borrowing it in a manner that is easy to detect – ie phonetically. But one highly plausible manifestation of such sharing, with respect to language, *would* be these seemingly "miraculous" semantic sets – and unlike phonetic loans, they would not have needed to have been spread by a significant population of bilinguals. A traveler could've, indeed, just passed along the germ of an idea – ie "oranges can be squeezed for their juice," and that would've been enough for the calque to be created and the semantic parallel established.

    Food for thought.

  52. Chris Button said,

    February 26, 2019 @ 8:28 am

    So, sticking with an example above, we have an Old Chinese root *ʔən / *ʔan, represented graphically by 因 / 安, centered around the concept of "sitting". The semantic extension to "peace" in the latter case happens to be paralleled by Welsh hedd from Proto-Indo-Eutopean *sed- "sit". The fact that we can find similar semantic extensions to Old Chinese somewhere within the numerous Indo-European languages (in this case it just happens to be Welsh) is surely not surprising. Semantic extensions do not grow out of thin air after all; they are simply logical outgrowths that require no suggestion of cognancy.

  53. R. Fenwick said,

    February 28, 2019 @ 4:14 am

    @Chris Button:
    The ə/a ablaut pervades the Old Chinese lexicon. As you might expect, most reconstructions obscure it with supposedly more "natural" triangular vowel systems,

    Hah! Yes, I can imagine so. Some linguists apparently want to just wish vertical vowel systems away so they don't have to deal with them… several years ago I added a specimen Ubykh text to the Wikipedia page on the language, and on revisiting it a little while ago, I saw that some well-meaning but clueless editor has added an ostensibly rephonemicised version, consisting of exactly the same text with a global replacement of orthographic -ʷa- -ʲa- -ʷə- -ʲə- sequences by -o- -e- -u- -i-. Sigh.

    but when handled objectively the evidence is incontrovertible. Edwin Pulleyblank tried hard to identify a morphological function to the ablaut but it is not very convincing so the phonological evidence is really all we have.

    That's fair enough. It's a bit the same in North-West Caucasian; though there's some evidence for ablaut's morphological role, it's hard to reconstruct systematically, and in many cases vowels differ between one branch and another with no clear justification at all. (Fortunately, the consonantal evidence is usually enough on its own – even the most conservative Proto-NWC reconstruction has to account for nearly 100 consonant phonemes.) What's the phonology of the Old Chinese consonant system like? I can see even from your copper/vermilion example that phonemic labialisation is reconstructible. Is that also pervasive throughout the consonantal inventory?

    Presumably as a result of the discomfort that such vertical vowel systems cause many linguists, I’ve read proposals that in languages like Ubykh such systems derive from earlier triangular ones as if that somehow renders them more "natural". I was wondering if you had any thoughts on the evidence behind such proposals?

    I do, but whether those thoughts will be useful is another question. :) Apologies in advance for the ramble.

    I suppose the relative rarity of vertical vowel systems (here I'll call them VVSs for short) could be seen as evidence for their "unnaturalness", at least in principle. I only know of maybe half a dozen other examples myself, and one of those – Wichita – qualifies only under a technicality (I'm glad to go into it in more detail if you're interested, but for the moment let's just say it's not the same kind of system). Elsewhere VVSs do seem to be the result of developments from earlier, more "natural" vowel systems. The others I'm familiar with outside of North-West Caucasian (Arrernte, modern Irish, Kaytetye, Margi, Marshallese) are more or less isolated developments in a single language or a single branch of a larger language family, though rather preliminary evidence out of Papua New Guinea suggests there may be another family-level cluster involving some or all of the Sepik languages. Usually they arise by a single-stage development by which the traditionally vocalic features [+front] or [+round] (or in the case of pre-Proto-North-West Caucasian, also [+front,round] – there's solid evidence for four stop series, *C Cʲ Cʷ Cᶣ, in Proto-NWC) are reassigned from the vocalic nucleus of the syllable to the consonantal periphery. The precursor system need not be strictly triangular, either; Proto-North-West Caucasian's precursor probably had a cubic system /i e ə a ö ü o u/ not unlike that of modern Turkish.

    With that said, for a few reasons I do think VVSs are just an occasional consequence of ordinary sound change: even if they're not common, I don't think they're inherently unnatural at all (no more so than, say, tone or labiodentals are). Firstly, all the others I know of are so widely scattered (Australia, Micronesia, Europe, Africa, and PNG) that they can't be explained by contact phenomena. Secondly, once such a system has formed, it's not in itself unstable; even if one ignores the uncertain Sepik situation, the /ə a/ system of Arrernte and Kaytetye is a common inheritance within Pama-Nyungan, and in NWC the /ə a/ system goes right back to the protolanguage, probably accounting for at least a couple of thousand years. Lastly, I just wonder if many who are loath to accept the naturalness of VVSs are simply misunderstanding their nature. The name refers only to the underlying, abstract, phonology. The phonetic vowels themselves are invariably little different from what you hear in languages with more pedestrian phonological systems; early Ubykh texts, for instance, regularly transcribe at least nine short vowel phones as well as a handful of long ones.

    So in short, perhaps it's my NWC bias coming through, but I have absolutely no issue with the reconstruction of a binary vowel system in a protolanguage. Isn't it becoming more and more accepted that the */e o/ system of Proto-Indo-European probably reflects earlier or underlying */ə a/ as well?

  54. Chris Button said,

    February 28, 2019 @ 2:59 pm

    @ R. Fenwick

    Thanks – I posted my reply here on the Ubykh thread:

    http://languagelog.ldc.upenn.edu/nll/?p=41964#comment-1561097

RSS feed for comments on this post