Origin of South Caucasian Languages.

The time and place of origin of South Caucasian languages,” by Alexander Gavashelishvili, Merab Chukhua, Kakhi Sakhltkhutsishvili, Dilek Koptekin, and Mehmet Somel (Sci Rep 13, 21133 [2023]), looks quite interesting, and it’s open access, so you can check it out freely. The abstract:

This study re-examines the linguistic phylogeny of the South Caucasian linguistic family (aka the Kartvelian linguistic family) and attempts to identify its Urheimat. We apply Bayesian phylogenetics to infer a dated phylogeny of the South Caucasian languages. We infer the Urheimat and the reasons for the split of the Kartvelian languages by taking into consideration (1) the past distribution ranges of wildlife elements whose names can be traced back to proto-Kartvelian roots, (2) the distribution ranges of past cultures and (3) the genetic variations of past and extant human populations. Our best-fit Bayesian phylogenetic model is in agreement with the widely accepted topology suggested by previous studies. However, in contrast to these studies, our model suggests earlier mean split dates, according to which the divergence between Svan and Karto-Zan occurred in the early Copper Age, while Georgian and Zan diverged in the early Iron Age. The split of Zan into Megrelian and Laz is widely attributed to the spread of Georgian and/or Georgian speakers in the seventh-eighth centuries CE. Our analyses place the Kartvelian Urheimat in an area that largely intersects the Colchis glacial refugium in the South Caucasus. The divergence of Kartvelian languages is strongly associated with differences in the rate of technological expansions in relation to landscape heterogeneity, as well as the emergence of state-run communities. Neolithic societies could not colonize dense forests, whereas Copper Age societies made limited progress in this regard, but not to the same degree of success achieved by Bronze and Iron Age societies. The paper also discusses the importance of glacial refugia in laying the foundation for linguistic families and where Indo-European languages might have originated.

And the introduction ends:

There is linguistic evidence that points either to possible structural relationship or to prolonged contacts between Kartvelian and Indo-European languages in the South Caucasus. This is supported by recently discovered genetic evidence of a ghost population in or near the South Caucasus, which acted as the link connecting the Proto-Indo-European-speaking Yamnaya with the speakers of Anatolian languages. In this context our findings will help reduce the search area for the homeland of Indo-European languages and provide more clarity about the nature of ties between Kartvelian and Indo-European languages.

Thanks, Dmitry!

Comments

  1. January First-of-May says

    When I saw “Sakhltkhutsishvili” I thought it had to be a typo for sure. But nope, looks like it is in fact the guy’s actual surname, in all its consonant cluster glory. Here’s his entry on Mathematics Genealogy Project.

    (Admittedly some of it is just transliteration; it would be Сахлтхуцишвили in Cyrillic, with “merely” four consonants in a row. A Google search for the Cyrillic spelling brings up a bunch of results about other people with the same last name, though apparently nothing on this particular guy.)

  2. jack morava says

    A distinguished Vietnamese colleague who studied in Moscow and first published under the Russian transliteration of the French transliteration of his name has now found it to have been backtransliterated into something … strangelooking…

  3. Yes, I was struck by that surname too. Here it is in Georgian: სახლთხუციშვილი.

  4. Is Svan really as divergent from the rest of Kartvelian as the time depth (7600 ybp) indicates? That’s more than, say, German would be distant from a present-day descendant of Hittite.

  5. სახლი is ‘house’; I don’t know what the -თხუც- part is.

  6. David Eddyshaw says

    Is Svan really as divergent from the rest of Kartvelian as the time depth (7600 ybp) indicates?

    Not remotely.

    Bayesian true-believers all seem unable to accept the basic truth that if your method produces nonsense results, you need to look for errors in your method.

    Svan does do the Old Irish thing of omitting every other vowel and letting all the resulting chaos stand in all its glory.

  7. Trond Engen says

    I’ve read the paper. It’s combining lingusitic, paleo-ecological and genomic data to outline the historical geography of Kartvelian/South Caucasian languages. That’s fine, and it’s interesting, but it doesn’t convince me that the split is that old, rather than due to a more recent diffusion across an ecological/cultural divide, e.g. with the introduction of mining. I suppose the bayesian clock is there to do the convincing, but …

    The remark on a possible IE (or should we reinvent Indo-Anatolian?) homeland is also interesting, but not based on any evidence presented in the paper.

  8. David Marjanović says

    No comments yet, I haven’t yet tried to find out how the dating was calibrated.

  9. Dmitry Pruss says

    Dates appear to be unrealistically old. They are based on two much more recent dates, in VIII-Xth c. (Old Georgian manuscripts and the suggested era of the Laz migrations), but when the reported dates are so much older than the dates used for “calibration”, it often boils down to the choice of the “oldest possible date”in the modeling process. Often, the older is the maximum allowed age, the older become the split estimates.

    Frankly I don’t think the conclusions of the paper are worth attention. The data and the details, like all the botanical species, looked more interesting to me.

  10. “To calibrate the clock, we applied 900 years BP to Old Georgian, which is the last seen date identified by linguists. A normal time constraint of 1200 −/ + 10 years BP was put on the Zan internal node—i.e. a prior for the split time between Laz and Megrelian (see the introduction). No time constraints were put on other internal nodes in the Kartvelian tree because reliable historical clues were not available to make any proper assumptions. Dated trees were then inferred under the strict clock model and the uncorrelated lognormal relaxed clock.”

  11. It seems pretty common for one of these computational methods to produce a reasonable-looking ancestral tree (but we typically had a relatively good idea of how that should look already), but time scales that seem way too big. That suggests there is something systematically wrong with the way the time depths are being extrapolated. Perhaps it is as simple as that the most-secure correspondences tend to be ones in which today’s surviving cognates are atypically similar. Such similarity, after all, makes those cognates the easiest to recognize. However, judging rates of divergence from particularly similar cognates would lead to underestimation of rates of linguistic divergence, which is equivalent to overestimation of the time depths of splits.

  12. Is there a reason to think Svan or non-Svan-Kartvelian replaced basic vocabulary at an abnormally high rate? Say, maybe, because Svan’s divergent phonology led to homophone avoidance?

    Given that plant and animal names are easily borrowed, you really need to assure the reader that proper historical linguistics indeed reconstructs species names to the protolanguage, before inferring that this or that plant existed in the proto-homeland.

    I’m no geneticist, but the PC analysis looks messy. There are evidently a lot more than two significant principal components representing the wide range of populations analyzed. Wouldn’t it make sense to, say, do a PC analysis of just South Caucasian populations, and then compare the endpoints to external populations? Something like that?

  13. David Marjanović says

    A normal time constraint of 1200 −/ + 10 years BP was put on the Zan internal node—i.e. a prior for the split time between Laz and Megrelian

    That makes the other calibration, 900 BP for the split between Zan and Georgian, irrelevant, so there’s just this one. I’m surprised, though, that such a tightly constrained calibration in such a small tree – with a maximum age, even if it’s a soft one! – still lets the root node drift out of sight.

    I suppose something exaggerated the (not inconsiderable) distance between Svan and Zan + Georgian. Let me guess: lexical data only, coded as presence/absence of each word-and-meaning combination?

  14. “Admittedly some of it is just transliteration;” – well, the Gaelic script is meant to facilitate reading such words….

  15. Trond Engen says

    Because I hadn’t read the supplements (and I still haven’t, but since we’re all punching now:), I backed out of saying that the genetic plot amounts to little more than handwaving, and that the comparison of paleo-ecology with archaeological culture also seems very coarse. Still interesting as an observation, but not convincing.

  16. They quote older estimates for the time depth of Kartvelian, based on glottochronolgy, ranging from 4000 to 5000 ybp (which agrees with the current study’s estimate when using a Dollo model, supplementary tables & figures, fig. S1). That is still pretty old.

    Let me guess: lexical data only, coded as presence/absence of each word-and-meaning combination?
    (Examining supplementary files 2 and 3:) A-yup!

  17. “Frankly I don’t think the conclusions of the paper are worth attention. The data and the details, like all the botanical species, looked more interesting to me.” – I sympathise to this approach.

  18. When one subtracts შვილი švili ‘child’ from სახლთხუციშვილი saxltxucʻišvili, then what is left, სახლთხუცი- saxltxucʻi-, looks like it might mean ‘house-priest’ (‘chaplain’?). The first element in the title looks like სახლთ- saxlt-, oblique stem of the archaic plural (სახლნი saxlni nom., სახლთა saxlta gen.) of სახლი saxli ‘house’. (If anyone knows more about the use of stems in Old/Middle Georgian compound formation, I would love to hear about it.) There seems to be a Georgian (an Old Georgian?) word ხუცი xucʻi ‘elder, priest, old man’(?). Compare the xucʻ- in ხუცური xucʻuri ‘clerical, the khutsuri script’.

    There is also a Georgian title სახლთუხუცესი saxltuxucʻesi (Wikipedia entry here), a high-ranking official in the statelets of late medieval/early modern Georgia (approximately ‘Lord Chamberlain’?). The formation of the title looks parallel to that of western European titles like mayor of the palace or majordomo. The second element would be Middle Georgian უხუცესი uxucʻesi ‘old man, elder, priest, cleric’, which looks like it has the Georgian superlative circumfix უ-… -ესი u-… -esi, affixed to ხუცი xucʻi (? ‘elder, priest, old man’; cf. πρεσβύτερος ‘presbyter’, from πρέσβυς ‘old man, elder’).

    New to me: this ხუცი xucʻi is perhaps of Iranian origin, according to the Wiktionary here. I will have to follow up on this for my own edification. Perhaps LH readers can find more. (Posting this now because I have something else I must do.)

  19. Turkic word for “woman”, Persian for “God”? Wow.

  20. That’s an impressive spread of descendants of xātun!

  21. David Marjanović says

    Persian for “God”?

    That’s actually pretty popular

  22. drasvi wasn’t talking about that but about xodâ.

  23. A quick google search shows that სახლთხუც-მოურავი (sakhltkhuts-mouravi) means “housekeeper”. Mouravi means “governor” and the first part is something like “someone who sits in the house”. Provisionaly, I translate sakhltkhuts as “house sitter” (joke!)

  24. David Marjanović says

    drasvi wasn’t talking about that but about xodâ.

    I know; just having fun with parallels.

  25. Wait, I don’t understand something about those parallels. Wiktionary says that the word is inherited from IE in the meaning “share, cut, wealth, luck” and is a semantic borrowing from Scythian in the meaning “God”.

    Meanwhile Wiktionary reconstruct both meanings “share” and “god” (via dispenser, patron) for Proto-Indo-Iranian and Proto-Iranian, and names both “share, (favourable) lot” and “god” for Avestan.

    Then what makes it a semantic loan? How one can posit a stage where mostly unattested Scythian has both meanings in the inherited word and unattested pre-Proto-Slavic has only one meaning in its reflex of the same word “because it has not borrowed the other yet”? Because Baltic?

  26. Y: Inasmuch as Svan has undergone radical, Old Irish-type sound changes (as David Eddyshaw put it upthread), we would definitely expect above-average vocabulary renewal/replacement as a result. Since the (much older than expected) date for the split of Svan from Proto-Kartvelian is calculated on the basis of vocabulary comparisons only, I am afraid I must join others in expressing EXTREME skepticism as to the reality of these results.

    In addition, echoing Y’s earlier comment about the gap between Svan and other Kartvelian languages being comparable to the gap between German and a modern-day form of Hittite: It would be impossible to demonstrate that Modern German is genetically related to a sizeable number of other Modern Indo-European languages (Off the top of my head: Welsh, Latvian, French, Armenian, Kurdish, Bengali) if we did not have older texts + close relatives allowing us to follow their history. On this basis I think that demonstrating the genetic affiliation of Modern German (assuming no other Germanic languages are known and no older texts in any Germanic variety exist) and a present-day form of Hittite, with no written data on Hittite and various intermediate forms between it and this hypothetical descendant, would probably be impossible.

    Which leads me to conclude that the very fact that Svan is unequivocally related to other Kartvelian languages is seemingly incompatible with the authors’ dates.

  27. David Marjanović says

    How one can posit a stage where mostly unattested Scythian has both meanings in the inherited word and unattested pre-Proto-Slavic has only one meaning in its reflex of the same word “because it has not borrowed the other yet”? Because Baltic?

    I think so.

    Borrowing the abstract property of polysemy is not unheard of. English realize means both “make real” and “understand”. German realisieren used to mean only the former, but has recently taken on the latter meaning as well, evidently as a calque from English even though the word as such is not borrowed from there.

    It would be impossible to demonstrate that Modern German is genetically related to a sizeable number of other Modern Indo-European languages

    That strikes me as overly pessimistic. It would be a lot more work that comparing Ancient Greek to Sanskrit, sure, but “impossible” is a strong word. After all, it wasn’t impossible, even in the 19th century, to demonstrate that Albanian is IE, even though its written records only go back to the 15th century, it has a thick layer of Latin loans (much like Brythonic and Basque), and the rest contains at least Greek and Slavic loans.

    On top of that, the last common ancestor of all dialects that are counted as German today must have spoken around the year 500. That’s a hundred years before Proto-Slavic, and only one or two hundred years after Armenian was first written.

  28. Borrowing the abstract property of polysemy is not unheard of

    Of course (though the mechanism of borrowing in this case can be or not different from massive exposure of many people to L2… and of course different from translation). Which of course, does not mean that every convergence in presence of contact is borrowing from a larger group to a smaller group.

    But what I mean is the possibility of a common inheritance of an arealism that appeared in a situation where speaking of “Iranian” or “Slavic” wouldn’t make much sense.

  29. Gavaselishvili et al.’s tree haz Laz and Megrelian separated by 1200 years, Georgian and Old Georgian split 900 ybp, these two branches having separated 2600 ybp, and that branch having separated from Svan 7600 ybp. So, roughly, Modern Hittite vs. Russian + Old Church Slavonic + Lithuanian + Latvian.

    (If you can’t visualize Modern Hittite, Hindi will do, at the cost of a couple of millennia.)

  30. “Modern Hittite vs. Russian + Old Church Slavonic + Lithuanian + Latvian.”

    Unless we try Bayesian phylogenetics on Russian and Hittite and obtain a much earlier date:))))
    No, seriously, if you compare it to IE, then what are our dates for IE?

    It does not look like a reliable reference (also IE languages conver a much larger area, there is a chance that Kartvelian remained mutually influencing system, like what was proposed for Berber)

  31. No, I can’t visualise Modern Hittite. But I can visualise it four-dimensional, if that will do! (just kidding)

  32. David Eddyshaw says

    Actually, it’s very like Welsh. (It’s remarkable what three thousand years can do.)

  33. (unrelated) Somehow I missed these two Nilo-Saharan names:
    Mimi of Decorse, also known as Mimi of Gaudefroy-Demombynes and Mimi-D
    Mimi of Nachtigal, or Mimi-N.

    PS, about the cluster whose transliteration fascinated everyone: obviously the problem (it is unreadable absolutely) is that all letters have ascenders)

  34. David Marjanović says

    I’ve read the paper now and learned a few interesting things.

    The best part is (brackets in the original):

    Acknowledgements

    We are grateful for the insightful discussions, feedback, and advice shared by the members of the Facebook Group: [lushnu nin svanuri ena], which popularizes Svan language, and its founder Rusudan Ioseliani. We thank Marina Beridze and Givi Karchava for helping to map the distribution of Kartvelian languages and improve meaning concepts as well as cognate sets for Georgian and Zan, respectively.

  35. The very beginning of the article: “Based on the reconstructed proto-words of several Eurasian language families, proto-Kartvelian is suggested to have emerged over 12,500 BP (Before Present standing for years before 1 January 1950), predating proto-Indo-European, proto-Uralic, proto-Altaic, proto-Inuit-Yupik and proto-Chukchi-Kamchatkan [1]. “

  36. (Before Present standing for years before 1 January 1950)

    rather like the (supposed*) british legal definition of “time immemorial” as “before richard the first”.

    (and no wonder there are odd ideas about methodology going around, if their “present” is over 70 years in the past)

    .
    * no idea what trivia collection i picked this up from; i’ve never asked a barrister or solicitor about it.

  37. (I don’t intent to bash the artcile. It’s just funny).

  38. January First-of-May says

    Before Present standing for years before 1 January 1950

    AFAIK this is a fairly standard scientific convention; we might have discussed it on LH before.

    EDIT: one such discussion.

  39. rozele : “Before Present standing for years before 1 January 1950” : that’s only for carbon dating.

  40. And what verb form I should use when discussing English as it [will be, would be, was…] spoken N years After Present?

  41. Jen in Edinburgh says

    If I’m remembering correctly, 1950 is the ‘present’ for carbon dating because it’s before nuclear testing changed the balance of carbon isotopes – but presumably you don’t carbon date a language…

  42. Stu Clayton says

    presumably you don’t carbon date a language…

    You genomidate it !

  43. Genomidation is a cross between gemination and intimidation.

  44. Radiocarbon dating became established in the 1950s. 1950 was the last nice round year before then.

  45. Jen in Edinburgh : You’re remembering correctly — “nuclear testing changed the balance of carbon isotopes”. That’s why we have “before present” — it means before the start of nuclear testing, pegged to 1950.

  46. David Marjanović says

    1950 was declared Present in 1949; one comment above the one linked to is one of mine with a link to the Wikipedia article “Before Present”.

    I really like the implication that this is the year 73 After Present. As the saying goes, “the future is already here, it’s just not evenly distributed”…

  47. “the future is already here, it’s just not evenly distributed” that went dark quickly. Cheer up a bit.

  48. Etymologically, the present is past.

  49. David Marjanović says

    that went dark quickly

    You’re projecting. Depending on the context, it can be a statement of optimism.

  50. David Eddyshaw says

    Yes, it doesn’t strike me as a particularly miserabalist saying, as such things go.

    I can come up with much more miserable sayings than that. But then, we Calvinists rather pride ourselves on our abilities in that direction.

  51. Then what makes it a semantic loan? How one can posit a stage where mostly unattested Scythian has both meanings in the inherited word and unattested pre-Proto-Slavic has only one meaning in its reflex of the same word “because it has not borrowed the other yet”? Because Baltic?
    My opinion on Slavic bog, which hasn’t changed in the past 10 years, is that it’s a loan from Iranian in both the “share” and the “god” meanings. It’s not attested in Baltic.
    (IIIRC, LH linked to my blog post back then and we discussed it here, but my wife is calling me to feed the cats, so I don’t have the time to search for that now.)

  52. I found the discussion at LH; it quickly moved on to the status of /*a/ as a phoneme in PIE.

  53. @Hans, I understand this version.
    What I don’t understand is the idea that Slavs and Iranians both inherited it in the meaning “share”, but “god” is Iranian invention: I see no reason why anyone would attribute this innovation to Iranians, or Indo-Iranians.

  54. Right now I am hearing something … unexpected.

    Someone is playing a Soviet handheld console. I haven’t heard these sounds since….
    It is 6 am. Winter. Darkness. Snow. I’m extremely curious and want to look out my window.

    (Likely, of course, someone just downloaded an emulator for a phone)
    PS or not? The timbre is reproduced suspiciously well….

  55. I see no reason why anyone would attribute this innovation to Iranians, or Indo-Iranians
    PIE had a word for “god”, *deywos. Indic kept it (Sanscrit deva-), as did Baltic (Lithuanian dievas). In Iranian, that word came to mean “demon”, and baga- replaced it in the meaning “god”. In Slavic, the old IE word lives on in the group of divъ “wonderful, astonishing”. So it’s Iranian that went through a religious reversal where the old gods became demons, not Slavic; that is the reason why it is usually assume that the innovation is Iranian. The only question then is whether Slavic loaned only a new meaning for an existing word, or the word with both meanings from Iranian.

  56. I see no reason why anyone would attribute this innovation to Iranians, or Indo-Iranians
    PIE had a word for “god”, *deywos. Indic kept it (Sanscrit deva-), as did Baltic (Lithuanian dievas). In Iranian, that word came to mean “demon”, and baga- replaced it in the meaning “god”. In Slavic, the old IE word lives on in the group of divъ “wonderful, astonishing”. So it’s Iranian that went through a religious reversal where the old gods became demons, not Slavic; that is the reason why it is usually assumed that the innovation is Iranian. The only question then is whether Slavic loaned only a new meaning for an existing word, or the word with both meanings from Iranian.

  57. @Hans, wow, thanks. I did not connect it to the situation with the pantheon…

  58. I noticed that Sanskrit has:

    1. “dispenser”, gracious lord, patron (applied to gods, especially to Savitṛ)
    2. Bhaga
    3. the sun
    4. mud (also bitumen)
    5. good fortune, happiness, welfare, prosperity
    6. dignity, majesty, distinction, excellence, beauty, loveliness
    7. love, affection, sexual passion, amorous pleasure, dalliance
    8. the female organs, pudendum muliebre, vulva

    Bhaga: (Hinduism) One of the Adityas, a god of wealth and marriage in Hinduism. In the Rigveda Bhaga is the god who supervises the distribution of goods and destiny to each man corresponding to his merits

    So I thought the implied shift is Indo-Iranian.

    [the female organs, pudendum muliebre, vulva: three very differfent things. Amateur gynaecologists often confuse them, but not lovers.]

  59. Amusingly, the Korandje word for “share” – with an impeccable Songhay etymology – is baɣa.

  60. David Eddyshaw says

    Yet another evident cognate is the Farefare baga “idols, fetishes.”

    The “female organs” word is obviously of a different origin. It is reflected in the Dagbani baɣa “pregnancies.”

  61. David Marjanović says

    Oh, I forgot about Winter’s law. A wholesale loan it is, then – restrictions to open syllables seem to be motivated by false exceptions.

  62. It is reflected in the Dagbani baɣa “pregnancies.”
    Which empahsises my point that one should not confuse “female organs” with “vulva”. As for pudendum it must refer to spiritual life of a woman. It would be silly to be ashamed of a body part….

  63. David Marjanović says

    …so of course Western culture taught young ladies in the 19th century to be just that. “When you take a bath, strew sawdust on the water so you may be spared the sight of your shame.”

    That’s what made Freud possible.

  64. I think “Western culture” is a man, and one who has not (unlike me*) spent his youth discussing details of human reproduction and associated matter with people of opposite gender. My female informants are surprisingly much less well informed about what the “shame” in question looks like than I expected based on general knowlege of flexibility of human body. Perhaps he does not even talk about such things with his wife (the Eastern culture?)

    (I also have doubts about his spiritual sight… else he would have known that sawdust won’t help, when it comes to spiritual vision)

    * The community of math students in Moscow might be among the least sexualised in terms of “what a boy and girl who like each other will do if you leave them alone in a room that has only one bed for a night”, but is clearly ahead of others in terms of curiousity.

  65. Whoever publicized the sawdust solution, surely did so out of love of their own ingenuity, rather than any trace of practicality. I mean, wet sawdust sticks to things.

  66. Well, today it is foam (and coloured water), in wild nature it is lilies.

  67. 1-Drasvi: echoing and amplifying Hans’ answer to your question/comment: “I see no reason why anyone would attribute this innovation to Iranians, or Indo-Iranians”: Actually, there is good evidence that (Early) Indo-Aryan was originally spoken to the South of (Early) Iranian, and that much of the territory which became Iranian-speaking had originally been Indo-Aryan-speaking (the term Indo-Aryan might seem odd, since these “Indo-Aryan” speakers had never set foot in India, and neither had their linguistic forefathers/ancestors, but such is the weight of inherited terminology…). Crucially, Zoroastrianism arose among Iranian speakers while they remained in contact with Indo-Aryan speakers, with the latter practicing their Hinduism-like religion, and as a result a great many Indo-Iranian shared terms relating to religion which Zoroastrians came to associate with non-Zoroastrians acquired a negative meaning…the local reflex of *deiwos, for instance.

    2-Sawdust in bathwater, in order for women not to see their own bodies? That is a new one: I had only ever heard of starch being used for this purpose.

    3-David Marjanović (your December 4, 4:19 pm comment): Albanian could be shown to be Indo-European (So could Armenian or the Baltic languages, I might add) because there already was a good basic idea of what Proto-Indo-European must have been like based on comparing Latin, Greek, Sanskrit and a few other languages abundantly attested in writing millennia ago.

    Here is a better way of thinking of the Kartvelian situation: what if the only living Indo-European languages were Greek, Albanian, Bulgarian and Romanian? What if, furthermore, we had older written records for Greek, going back to the fifth century AD (i.e. the date when Georgian is first attested in writing), but for none of the three other Indo-European languages? Or indeed for ANY Indo-European language anywhere?

    I think it would be impossible to demonstrate the existence of the Indo-European language family under such conditions: ironically, I suspect that any such “demonstration” would involve misinterpreting contact phenomena (the massive Romance element in Albanian, the Slavic element in Romanian, Hellenisms…) as indications of a genetic relationship.

    (My reasons for thinking proving a genetic kinship between the above languages would be impossible involve my own attempts -which I believe I once mentioned here at the hattery-at reconstructing Indo-European using Modern language data only. It can indeed by done, but you need to start with branches of Indo-European with a lot of language/dialect diversity: on the basis of this you can obtain a sufficiently clear image of Proto-Romance, Proto-Germanic and Proto-Slavic to show that those three are definitely related: but only by relating these reconstructions to the more isolated branches -Albanian, East Baltic, Armenian- can you show that all of the latter are Indo-European: a direct comparison involving any or all those three branches with a single Slavic, Germanic, Iranian, Indo-Aryan or Romance variety will not be enough, I am afraid!)

    Things would be even worse with a living language deriving from Hittite. I stand by my original statement: I cannot believe in the authors’ dates, because if they were true Kartvelian would not be identifiable as a language family.

  68. Starch!?
    Were they bathing in kissel?

    (And seriously, maybe the idea was that other people should not see someone’s pudendum-schmudendum?)

  69. David Marjanović says

    my own attempts -which I believe I once mentioned here at the hattery-at reconstructing Indo-European using Modern language data only

    You’ve mentioned them, yes, but you’ve never elaborated more than here… if you’ve published and insist on remaining anonymous here, I’d greatly appreciate an e-mail; you can find mine through my publications 🙂

    ironically, I suspect that any such “demonstration” would involve misinterpreting contact phenomena (the massive Romance element in Albanian, the Slavic element in Romanian, Hellenisms…) as indications of a genetic relationship.

    Perhaps more likely is that the (comparatively few) inherited similarities would be ascribed to contact as well!

    branches of Indo-European with a lot of language/dialect diversity

    Albanian, Armenian and East Baltic are such, actually; but the first two in particular seem underresearched in this respect. A Proto-Albanian reconstruction has been made, though; IIRC it’s supposed to be the state of the language around the end of Roman times.

  70. David Marjanović says

    (And seriously, maybe the idea was that other people should not see someone’s pudendum-schmudendum?)

    No, this is about taking a bath alone in your own bathtub. Upperclass Victorians were fucked up.

    Never mind a lady’s ankle – there were young men getting aroused at the sight of a table leg…

  71. David:

    1-Alas, I have published nothing on the subject of reconstructing Indo-European with Modern data: will notify you when/if I do (A recent change of fortune in my life means it is much likelier now that I will publish it sometime in the future).

    2-The Proto-Albanian reconstructions I have seen, much like Mitxelena’s reconstruction of Proto-Basque, make VERY heavy use of the evidence given by Late Latin/Early Romance (LLER) loanwords in establishing phonological changes, their chronology, and sometimes that of some morphological changes as well (The fact, for example, that Late Latin/Early Romance neuter nouns are normally neuter in Albanian, when neuter nouns borrowed from Early Slavic became masculine, gives us a nice indication when neuter nouns in Albanian became a subset of non-productive, morphologically irregular nouns -which is what they remain in Albanian today).

    When it comes to how far back we could reconstruct Proto-Albanian with no use made of LLER (or Greek or Slavic…) elements, I think the last common ancestor of the two major Albanian dialect groups (Gheg and Tosk) must have lost its unity in 1000 AD or thereabouts (That was Eric P. Hamp’s belief, if memory serves).

    Furthermore, without LLER elements to guide us many aspects of the Proto-Albanian we could reconstruct would be far less secure than our real-life reconstruction of Proto-Albanian. For example, Gheg intervocalic /n/ corresponds to Tosk (on which the standard is based) intervocalic /r/, and LLER indicates that the former must have been the Proto-Albanian phoneme, since LLER loans with intervocalic /n/ have /r/ in Tosk: cf. Albanian “vrer”, which now means “gall, bile”, is from Latin “venenum”, with both intervocalic /n/ phonemes rhotacized.

    But without knowledge of LLER loanwords and their history, several reconstructions would be possible to account for this Gheg/Tosk intervocalic /n/-/r/ regular correspondence (a */nr/ cluster or the like would be a possible reconstruction, for instance).

  72. John Cowan says

    The split of Zan into Megrelian and Laz is widely attributed to the spread of Georgian and/or Georgian speakers in the seventh-eighth centuries CE.

    I wondered what that meant: perhaps that one or the other language had been heavily influenced by Georgian, perhaps. But no, it’s that the central part of the Zan dialect continuum was displaced by Georgian occupiers, thus allowing northern Zan and southern Zan, no longer in contact, to drift apart until they were Mingrelian and Laz respectively.

  73. Etienne: even so, wouldn’t, say, German + Bengali + Albanian + Armenian + French give hints of their genetic relation with some basic vocabulary (‘mother’, three’, ‘heart’, vel sim) which escaped the phonological blender, at least in some languages?

  74. David Eddyshaw says

    Certainly, I think you could demonstrate beyond reasonable doubt that modern Welsh was part of the family on the basis of shared core vocabulary: the relevant sound correspondences are less opaque than with German or Armenian (a low bar …)

    What would be difficult-to-impossible, though, would be to demonstrate morphological evidence of the relationship to someone who (quite reasonably) objected to a purported proof of relationship based on shared vocabulary alone, even “core” vocabulary (which in Welsh actually includes Latin loans for pretty basic concepts. Like “child.”)

    I mean: present indicative of “love” (a completely regular verb) goes (in Very Literary Welsh) caraf, ceri, car, carwn, cerwch [kerux], carant. Doesn’t look too promising … (incidentally, the -t in carant has not actually been pronounced for half a millennium or more.)

    And no trace of any cases. The fact that the most productive Welsh plural continues the PIE u-stems (of all things) wouldn’t help, either.

  75. What would be difficult-to-impossible, though, would be to demonstrate morphological evidence of the relationship to someone who (quite reasonably) objected to a purported proof of relationship based on shared vocabulary alone, even “core” vocabulary (which in Welsh actually includes Latin loans for pretty basic concepts. Like “child.”)

    If proving relationships were what linguists are doing, one angel with a message “all languages are related, wallah” would be enough, and you all would be screwed. At the very least you would have to change your occupation. In reality such a message (we believe angels) would hardly change anything.
    Because proving to each other relationships is simply not what linguists do.

  76. I think this habit of pretending that this is what linguistics is about seriously harms. (sadly, even addition of this comment did not material… er, digitalise? the edit button)

  77. unrelated:

    “Корабли военного флота разделены на 14 классов: 1) линейные корабли; 2) броненосные крейсера; 3) крейсера; 4) эскадренные миноносцы; 5) миноносцы; 6) миноноски; 7) заградители; 8) подводные лодки; 9) канонерские лодки; 10) речные канонерские лодки; 11) транспорты; 12) посыльные суда; 13) яхты; 14) портовые суда.”

    In English: its a list of vessels of imperial Russian Navy, and torpedo boats are listed with masculine and feminine suffixes (perhaps this has to do with that “ship” is masc. and “boat” is fem.)

  78. Following up on სახლთხუციშვილი saxltʻxucʻišvili and the possible Iranian connections of ხუცი xucʻi, ხუცესი xucʻesi ‘elder, priest’:

    As far as I can gather, the shorter Georgian form xucʻi is nowadays mostly used among rural mountain communities as the term for the community member officiating (conducting the sacrifice of a ox, repeating traditional formulaic prayers, etc.) in ceremonies centered around the ხატი xati, the traditional shrine. What about the possible Iranian origin of xucʻi? Here is Ilya Yakubovich on Vasily Abaev’s treatment of the relationship between Ossetic xwycaw ‘god’ and Georgian xucʻesi ‘priest’ in Abaev’s Историко-этимологический словарь осетинского языка (Historical and Etymological Dictionary of the Ossetic Language):

    (2) In spite of the emendations that Abaev made in the dictionary in reference to his early works, he still sometimes appears biased towards the Caucasian origin of certain Ossetic words. One of the best examples is xwycaw ‘god’. It is tempting to link it to Parth./MPers xwadāy, Bactr. χοαδηο, Sogd. xutāw (all “lord”), etc., which go back to *hva-tāw(y)a ‘autocrat’; but there are two difficulties. Firstly, the development *t > c is irregular; one would expect Oss. *xwydaw. Secondly, this word is not attested in Old Iranian, and so it was suggested that it represents a calque from Hellenistic Greek autokrátōr ‘autocrat’. The question of possible contacts between Scythians/Sarmatians and other Iranians at that time remains unsettled, and Abaev was inclined to minimize the likelihood of such contacts. Therefore, he compared Oss. xwycaw ‘god’ with Lezgin xucar ‘god’, Georg. xucesi ‘priest’, and similar words; and he accepted at most the secondary influence of Iranian cognates (IV, pp. 255-56). Caucasian words, however, do not solve formal problems (the Ossetic suffix -aw remains unexplained), and in addition it is difficult to establish the common source of these alleged cognates (Lezgin and Georgian belong to two genetically unrelated groups). If a contamination is indeed involved, it is logical to assume that *xwydaw underwent the influence of Caucasian forms, not vice versa.

    Abaev’s own treatment of these words is here (beginning in the middle of page 255). Abaev also cites Chechen цӀу c’u ‘божество’ (here specifically ‘idol, pagan deity’?) as another word that he considers might be connected to xucʻesi. But Yakubovich does not even bother to mention the possibility of an Iranian origin for Georgian xucesi ‘elder, priest’ (or for Lezgin xucar ‘god’, for that matter). Yakobovich has more to say about Iranian *hva-tāw(y)a on p. 478 in his 2003 review article of the H. W. Bailey Gedenkschrift, available here.

  79. David Eddyshaw says

    Who, exactly, has been claiming that linguistics is all about proving language relationships, drasvi?
    I don’t think I’ve ever encountered such a claim.

    The section you cite is nonsense even if the writer meant just “historical linguistics.” Even if we had a revelation that all natural languages were related (as indeed they might be) the only interesting part is showing how they are related, anyway.

    The Oti-Volta languages are quite certainly all related to one another. I do not feel that this renders my work superfluous.

  80. I think this habit of pretending that this is what linguistics is about

    You’re the only one who’s pretended that.

  81. David Eddyshaw says

    It’s certainly true, as I think drasvi is implying, that borrowing and areal effects are very interesting in their own right (Lameen is our expert on that.)

    Indeed, you can’t do comparative linguistics without taking such things properly into account, whether you find them interesting in themselves or just regard them as getting in the way of the real work.

    It works both ways, in fact: historical reconstructive work can be exactly what alerts you to the fact that a form found in two languages must have been borrowed by one or both languages, and can’t be a common inheritance. (The “money/cowries” word is a case in point in Oti-Volta; examples in Indo-European languages are too numerous to mention.)

  82. ” the only interesting part is showing how they are related, anyway.”

    DE, yes.

  83. I think Ruhlen said something like that, that the business of historical linguistics is figuring out family trees (which later meant the details of partitioning the branches of Proto-World). The whole Greenbergian mass comparison enterprise relied on the idea that borrowing of basic vocabulary was exotic and mostly negligible. At least until Thomason and Kaufman’s book came out, that was not a rare idea.

  84. @DE, the claim that linguists are “proving relationships” or determining whether langauges A and B (Nilo-Saharan) are “related or not” is false.

    And it is NOT good for a science when the goal of linguist’s work is wrongly described.

    The Oti-Volta languages are quite certainly all related to one another. I do not feel that this renders my work superfluous.

    Yes. But did I say that?
    What I said is that your work is not about proving to others that the Oti-Volta languages are related, AND work of other linguists also is not about proving to others that langauges are related.

    I’m saying that linguists keep using an extremely confusing term and this habit is harmful.

  85. I think you are the one who is confused, not “linguists.”

  86. predating proto-Indo-European, proto-Uralic, proto-Altaic, proto-Inuit-Yupik and proto-Chukchi-Kamchatkan

    Somewhat of an Arson Murder Jaywalking list of accomplishments: the first two are the usual levels of “old family”, the third would be very old if it was real, and even Proto-ChK might be in the PIE / PU ballpark (but currently fairly hard to tell due to poor data, little work and many Chukotkan loanwords in Kamchatkan). Proto-Eskimo however (poor Sireniki, always gets elided in the name of political correctness — or if a Sireniki First hypothesis holds then PIY would be distinct and still younger) has been approximated to be at a not too impressive maybe two thousand years, plausibly younger than all sorts of subbranches-of-subbranches like North Germanic or North Finnic.

  87. @J Pystynen: currently fairly hard to tell due to poor data

    Is Chukchi/Kamchatkan documentation that bad? I was under the impression that there is a lot of literature on Chukchi, and some newer work on the other languages.

  88. I think you are the one who is confused, not “linguists.”

    LH, you are wrong.

    Moreover, the fact that such description is false is demonstably true.
    As I said: a divine revelation that all langauges are related would change almost nothing in linguists’ work. Because linguists’ work is NOT about demonstrating that languages are related.

    The only thing that can be seriously discussed is whether this false description is a “harmlessly sloppy description” or “harmfully sloppy description”. I believe the latter.

    As for “linguists”, the term which I’m criticising here is used by exactly linguists. So no quotemarks.

  89. Peter Grubtal says

    David Marjanović
    “sawdust” – can you give us a source for what you say.

  90. @DE, there is a label, and there is reality it is applied to.

    It seems you’re thinking that I have some objections to this reality (what is a linguist’s work).
    No, your and mine views on this reality are same. I’m objecting to the label.

    The exact words used to describe what linguists do, not the work to which they refer (and which you imply when uttering these imprecise words).

  91. I was under the impression that there is a lot of literature on Chukchi

    Chukotkan is decently well documented and reconstructed, Kamchatkan however is not. A lot of this probably due to people for long ignoring diversity in the sources and treating it only as a single Itelmen language. Kind of similar to the situation David notes for Albanian and Armenian. (For that matter, even more closely similar to the situation in Eskaleut, where we have well-reconstructed Proto-Eskimo, but the more distantly related Aleut remains much more poorly studied and thus much about Proto-Eskaleut remains in the air too.)

  92. David Eddyshaw: if you mean we could demonstrate that Welsh is Indo-European (solely on the basis of lexical evidence) in an alternate universe where no other Celtic languages existed, no Middle or Old Welsh writings had survived (or ever existed), but where the rest of the family (living languages + dialects and historical documentation) was as well-known as it is in the real world, then I partly disagree: not only could we prove the Indo-European affiliation of Welsh, I think that demonstrating that (much if not most of) Welsh inflectional morphology is Indo-European in origin would be feasible.

    Conversely, in an alternate universe where (if I may repeat myself) Welsh, Latvian, French, Armenian, Kurdish and Bengali are the only existing Indo-European languages, with there being no historical documentation for any of them, I do not believe it would be possible to prove that they are all related, that Welsh is related to any of them, indeed that any two of them are related. We would have some tantalizing clues here and there, things suggestive of a genetic relationship (some of the vocabulary items mentioned by Y upthread, for instance), but not enough for said relationship to be proven.

    All: it seems to me that the entire point of demonstrating a genetic relationship lies in this demonstration being an indispensable first step in reconstructing the entire history of a language, and if you do not know what the relatives of a language are, if you therefore (unless you have old written data) have no means of knowing what the language was like in earlier times, then you have no way of knowing what morphemes or structures or phonemes were borrowed over the course of the history of the language. When you DO have such information, then you can discover wonderful instances of pre-modern cross-cultural contact, language spread and the like which are all too often more surprising, more counter-intuitive, indeed more wonderful, than what most writers of fiction could produce (The Indo-Iranian loanwords in various Uralic languages, the Celtic loanwords in Proto-/Early Germanic, the spread of such widespread language families as Austronesian and Indo-European, the origins of Romani speakers in Northern India, the northern origins of the Navaho and of other Southern Athabaskan groups…and the list goes on).

    J Pystynen: Such a situation, where a great deal of diversity on the ground (despite its potential importance in getting a more accurate picture of the Proto-language) is ignored, is sadly common: even in as well-studied a language family as Romance, much of the internal diversity of varieties such as Corsican, Gascon and Franco-Provençal remained unknown and little-studied for a very long time, and as a result many older reconstructions of various Late Latin/Early Romance varieties are in far more dire need of revision than outsiders suspect.

  93. “it seems to me that the entire point of demonstrating a genetic relationship lies in this demonstration being an indispensable first step in reconstructing the entire history of a language”

    @Etienne, this is very close to my point.

    It is not the same because there is no such thing as “demonstrating a genetic relationship”.
    “Demonstrating a genetic relationship” means (it may “imply” something else, but it means this) that an extra-linguistical proof of proto-World would suffice for you… which is not the case.

    Replacing it with “demonstrating a [chronologically] close relationship” (as in MRCA between 12000-20000 BP – or an earlier range if you like) won’t really help – it is not clear what exactly time means to linguists.
    Replacing it with “demonstrating a [linguistically] close relationship” would reqiure a clarification of what “linguistically close” means. And it seems it means something tautological: demonstating a demonstrable relationship. And this is my point.

    I think what we are aspiring to “demonstrate” (‘the indispensable first step’) is that some [known, identified] shared elements are inherited from MRCA.

    We can’t be fully confident for each individual element, but we can reach a certain level of confidence for a list of features: “similarity of most Semito-Hamitic languages (maybe not Omotic) is hardly by chance or due to areal influence. So let’s start working on it and elaborating details”

  94. David Eddyshaw says

    And it is NOT good for a science when the goal of linguist’s work is wrongly described

    Indeed not, but nobody actually seems to describe it thus … (certainly not the Chomskyites …)

    It’s not true, of course, that linguists never show that languages are genetically related. Obviously they do. As Etienne has just been telling us, it’s hardly obvious that Welsh is related to Bengali: the proof that it is related is due to linguists, and nobody else. In fact, nobody but a linguist can do more than guess at such a relationship: their opinions on the matter are irrelevant.

    You seem to be saying that linguists never show that languages are related, but that is so evidently not the case that I think I must have misunderstood you.

    Of course, if you work on the assumption that all human natural languages are in fact genetically related (which is possible, but likely to be beyond proof or refutation), then the question becomes, strictly speaking not “are these languages related?” but “how closely are these languages related, and in what way, exactly?”

    Is that what you have in mind?

  95. “Indeed not, but nobody actually seems to describe it thus …”

    DE, why?

    What I am criticising here is the label “demonstrating relationship” (or worse, “determining whether A and B are related or not”). It is a very common label. You used it yourself.

    All I’m saying is
    (1) this label is inaccurate – as a matter of fact,
    AND
    (2) this specific inaccuracy is not harmless – which is not a matter of fact, but my opinion.

    It’s not true, of course, that linguists never show that languages are genetically related. Obviously they do.

    I agree! But it is not the goal of their work. What a linguists aspires to demonstrate in your Welsh example, is, as I said, the hypothesis that certain roots, morphemes or structures (if not each individually, then most in a list) are inherited from the MRCA.

    It indeed implies that there was a CA at all (which is the same as “are genetically related”). But in isolation such a fact that there was a CA at all would be nearly useless for a linguist. This fact would immediately follow from proto-World, it can possibly follow from some extra-linguistical arguments, but per se changes nothing for you.
    Basically, when you describe a linguist’s contribution by “demonstrated a relationship”, it is pars pro toto, but in this case the part in question is of negligible importance for science, and using it as the label obscures the constructive (and much more important!) totum.

    It is as if we were habitually sayuing that evolutionary biologists are “demonstrating relationship between humans and monkeys”. Yes, of course, such a relationship is assumed by (or conversely follows from) their publications. But it is not what modern biology is about.

  96. David Marjanović says

    Somewhat of an Arson Murder Jaywalking list of accomplishments:

    This time, by “proto-X emerged” they mean “the X branch split from whatever its closest known relative might be” – thousands of years earlier before “proto-X” in the sense of “the last common ancestor of all attested X” was spoken.

    …and they take that one paper with its tree of Eurasiatic for granted.

    “sawdust” – can you give us a source for what you say.

    A very good TV series I watched some 25 years ago. Sorry. 😐

    We would have some tantalizing clues here and there, things suggestive of a genetic relationship, but not enough for said relationship to be proven.

    Here I must once again protest that “proof” is for mathematics and booze in freedom units. Science doesn’t deal in such absolutes any more than the Jedi claim to. It can only deal in probabilities.

    In particular, “do I find this convincing, y/n” is not a useful question.

  97. David Marjanović says

    demonstating a demonstrable relationship

    Demonstrating that a particular relationship is demonstrable – by demonstrating it exists. That isn’t tautological.

    (1) this label is inaccurate – as a matter of fact,

    Unlike biologists, historical linguists are not used to taking for granted that all languages are related. They feel they not only need to figure out what the tree of the languages they work on looks like, but also that it exists in the first place.

  98. David Eddyshaw says

    What I am criticising here is the label “demonstrating relationship” (or worse, “determining whether A and B are related or not”). It is a very common label. You used it yourself

    OK. Guilty. Pace DM’s sensible caveat about absolute certainty being unattainable in real science, I would say that disbelieving in the genetic unity of Indo-European (or Oti-Volta) is on the same epistemological level as disbelieving that the earth is round. To disbelieve such things means either that you are unaware of the evidence, or unable to follow the reasoning, or have some hangup that means you are unwilling to believe it and are not at all interested in the evidence or the arguments anyway.

    In that sense, yes: Demonstrable. As much as anything is “demonstrable” in this sad sublunar sphere. Many linguistic relationships are beyond reasonable doubt. Thanks to linguists. The fact that others need more work, or are controversial, or may be forever beyond our reach does not mean that we don’t know anything at all. (The statement that “we don’t really know anything” is true only on some wholly uninteresting and forever sterile cod-philosophical level, where it’s actually tautologous.)

  99. Here I must once again protest that “proof” is for mathematics and booze in freedom units. Science doesn’t deal in such absolutes any more than the Jedi claim to. It can only deal in probabilities.

    In particular, “do I find this convincing, y/n” is not a useful question.

    @DM, I agree with you and no, I don’t think using the word “proof” helps, but this part is understandable.
    People are not very comfortable with probabilities.
    It is easier to divide everything in two groups: 1. reliable 2. unreliable. And then rely on elements of the former group and be cautious with the latter. So when people say “proof”, they mean “we assign it to the former group”.

    Of course it is better to work with probabilities.

    What I’m criticising is “relationship”.

    Unlike biologists, historical linguists are not used to taking for granted that all languages are related.

    Yes. But IF we knew this, this fact (per se interesting!) would have negligible effect on actual works of linguists. And that’s what matters.
    I’m not saying that it is unimportant, but it is unimportant for their work.

    Even a more specfic fact, say: “Nilo-Saharan has a common ancestor within 15ky” would not help much. It would mean you maybe will stop looking for close relationships with other families (but presently most work on NS is about elaborating histories of its parts and not about attempts to include a part of it in another family) but otherwise it will change nothing for you. Just nothing.

    I propose to distinguish between
    (a) constructive contributions (identifying actual inherited material, elaborating on its structure in time – as in “to construct”)
    (b) abstract “proofs” which only prove the fact of relationship (my angel, wallah!).

    My claim is that what linguists do is constuctive all the way down (and up to the starting point). And this is GOOD news for linguists. It is just better to discuss and think of this work in these terms, without calling it “proofs of relationships” as if it were (b).

  100. David Eddyshaw says

    I propose to distinguish between
    (a) constructive contributions (identifying actual inherited material, elaborating on its structure in time – as in “to construct”)
    (b) abstract “proofs” which only prove the fact of relationship (my angel, wallah!).

    This is not possible.

    To talk of “inherited” material already assumes that the relationship between the compared words is genetic, and the demonstration that a genetic relationship is the most “parsimonious” hypothesis to account for the resemblances between the languages concerned depends on establishing many relationships of this kind and showing that the segments etc correspond regularly. These are two aspects of the exact same endeavour, and one cannot stand without the other.

    As with any scientific hypotheses, the initial hypothesis, given that induction is impossible, must come from prescientific things like hunches, mass comparison, or even non-linguistic facts (“these people are neighbours and claim the same origin, so their languages may be related.”)

    But it only becomes scientific linguistics when you start attempting to find regular correspondences. If you can’t, then you need to reject the hypothesis of relationship; the more you succeed, the more confident you can reasonably be that the languages really are genetically related. You can be wrong: you may have sloppy methods, or you may be accurately reconstructing the Iranian loans in Armenian. But that’s science: and the only way of establishiing with reasonable confidence that two languages are genetically related is the comparative method. All others, to date, are just snake oil.

    In other words: your (a) is meaningless without your (b); and your (b) can only be established by doing your (a). Dialectics!

  101. “Demonstrated genetic relationship” is an abbreviation. The most linguists can do is argue that the evidence — the shadow on the cave wall — looks like that of genetically related languages.

  102. Y and other hatters: but the reverse is not true however: linguists can demonstrate beyond a reasonable doubt (DBRD) that two or more languages are genetically related, and this has indeed been done for a huge number of languages, but there is no way we can DBRD that any two or more languages are NOT genetically related at some time depth.

    To echo something Joseph Greenberg wrote and which I agree with, you cannot disprove that Welsh is related to Cree. What you can do is DBRD that Cree belongs to a family (Algic) which Welsh does not belong to, and conversely that Welsh belongs to a family (Indo-European) which Cree does not belong to. But all this does is push the unknown further back in time: if Algic and Indo-European in fact both derive from an older proto-language, then Welsh and Cree are indeed genetically related.

    But if we fail to DBRD that Algic and Indo-European are related, that does not mean that they are certainly not genetically related: there is no way for us to DBRD that Algic and Indo-European are UNrelated.

    Let us suppose that Algic and Indo-European really shared a common ancestor (let’s call it Indo-Algic). Perhaps the passage of time, language contact, or both these factors mean that there are too few traces (or possibly none remain) in Proto-Algic and Proto-Indo-European for us to detect this genetic relationship (leaving aside the matter of the uncertainties surrounding the features of both reconstructed proto-languages: perhaps the existence of Indo-Algic would be DableBRD if we had an accurate and complete image of both proto-languages).

    In which case the REALITY of the genetic relationship of Cree and Welsh (and indeed of all Algic and Indo-European languages with each other) would be undetectable and unknowable to us today. And the above thought experiment applies to all pairs/groupings of genetically “unrelated” languages on the planet (leaving aside Sign languages, Esperanto and other artificial languages…).

  103. I regret to report that Xerîb’s extremely interesting comment from 7:22 am has been lost in moderation until now, so please check it out if you’re interested in Caucasian and Iranian etymology.

  104. To talk of “inherited” material already assumes that the relationship between the compared words is genetic,

    @DE, yes. But I propose to distinguish between (a) and (b) in our discussion, because I need a name for (a), namely “constructive”.

    Of course, when you demonstrate that a certain list of features is inherited from the MRCA (my(a)), it does imply that there is a CA (my (b)).
    And when you publish a paper it implies that some writing system (and paper) exists.

    What I’m saying is that our knowlege of (b) itself is of negligible importance for a linguist’s work (just our knowlege of the fact that paper exists and English can be written on it) – and that it is a bad idea to use (b) as a label for (a) – or for (a and b). As I said, by this you obfuscate the set of actual problem which linguistic community aspires to solve (which go so far beyond “proving a relationship” that knowing that two languages are related without knowing more simply changes nothing for you).

    As for whether (b) in turn implies (a) I am not sure.

    It is certainly possible to imagine (b) without (a) and I think it is also possible to have good reasons to think that a certain land was colonised by exactly one small group, in that case all languages there will be very likely to be related.

    ____
    It is also common among linguists (and no one else) to let confusing “proving relationship” degrade into utterly absurd “are these two languages related?” and even “these two languages are not related”. It is simply impossible to know that two languages are not related at present.

    @Etienne, why – a situation when there IS an IE langauge which changed so much that we can’t recognise it is possible.

  105. It is also common among linguists (and no one else) to let confusing “proving relationship” degrade into utterly absurd “are these two languages related?” and even “these two languages are not related”.

    Could you provide an example or two of this common phenomenon?

  106. languagehat, I will.

    Thank you for pointing at Xerîb’s comment!

  107. J.W. Brewer says

    Just to expand on what the “family” shorthand in Etienne’s example means.

    We have not disproven a possible relationship between Cree and Welsh, but we have proven (subject to future revision in the light of new evidence etc.) that any possible Cree-Welsh relationship that might become detectable in the future would be both:

    a) Meaningfully more distant than the relationships between Cree and all other currently-classified-as-Algic languages; and

    b) Meaningfully more distant than the relationships between Welsh and all other currently-classified-as-IE languages.

    Now “distant” here is maybe something that’s hard to precisely specify or quantify, but it stands for more than a figure of speech, I should think. And now I’m wondering if Cree has any loanwords from either English or French that are of more distant Brythonic (or other Celtic) origin …

  108. @JWB, that is true, but this “distant” might not be phylogenetical distance (“branched off earlier”).

    Languages, especially in case of massive contact can change substantially and I think at the time depth of IE … Well, let’s say so: it is possible that an IE language won’t contain any recognisable IE grammatical morphemes.

    The question is “basic vocabulary”. Long-rangers (the Moscow school in particular) assume that many basic IE roots will be there (perhaps less so in case of language mixing), though sound correspondences can still be tricky.

  109. J.W. Brewer: If by “distant” you mean “distant in time”, then I agree entirely. Obviously if Welsh is Indo-European and Cree Algic, a common (Indo-Algic, as I called it upthread) ancestor would mean that Welsh and Cree are related to one another, but less closely than Welsh is to other Indo-European and Cree to other Algic languages. This is no different than the fact that Welsh is more closely related to Irish than it is to English or Russian: this means that the shared common ancestor of Welsh and Irish (AKA Proto-Celtic) is unambiguously younger (=more recent in time) than the shared common ancestor of Welsh + Irish, English and Russian (AKA Proto-Indo-European, minus Anatolian).

    But if by “distant” you mean typologically distant from one another (whether we are looking at the phonology, phonotactics, morphosyntax, vocabulary or lexical/phraseological semantics, or all of the above), then I strongly disagree: quite apart from the fact that languages can and do heavily influence one another (With the practical consequence that, for example, L1 anglophones will find French a much easier foreign language to master than Icelandic, despite the latter and not the former being a Germanic language just like English), it is also a fact that languages can be typologically similar to one another without there being any explanation beyond plain ole’ coincidence: colloquial spoken French, for example, is typologically very similar to an Algonquian language in its polypersonal marking on the verb and resulting free word order (A fact a former Innu-speaking student of mine thought was VERY funny), despite this not being due to French influence upon Algonquian or vice-versa.

    And being a later, changed form of a given language does not guarantee that a feature of the proto-language lost over time might not, coincidentally, reappear in (a) daughter language(s). For example, as one scholar pointed out, the distinction between the active and the medio-passive set of personal verb endings (AKA the parasmaipada/ātmanepada distinction, if I may use the terminology used by the world’s first known linguist) of Sanskrit has been lost in later Indo-Aryan (replaced by analytical constructions with auxiliary verbs), but Romani, in grammaticizing some of these auxiliaries and turning them into verbal suffixes, has basically become Sanskrit-like again, typologically (and thus recreated the parasmaipada/ātmanepada distinction).

    Thus, if Indo-Algic existed, it may well have been (in some respects) Cree-like, Welsh-like, Mandarin-like or Navajo-like (or perhaps all of the above, according to different features): assuming for a moment that Indo-Algic existed (for the sake of argument), we would have no way of knowing.

  110. Thank you @Hat for pointing back to @Xerib’s 7:22 comment, AMOT:

    … so it was suggested that it represents a calque from Hellenistic Greek autokrátōr ‘autocrat’. The question of possible contacts between Scythians/Sarmatians and other Iranians at that time remains unsettled, …

    Looking on the following debate, I’m puzzled why folks are trying to solve relatedness amongst languages by purely inside-Philology criteria. We can say Cree is not related to Welsh because there was no contact between the speakers until modern times. (There might have been pre-history relatedness as in out-of-Africa; no amount of speculative reconstruction is going to establish that to above crank level, let alone DBRD.)

    Contact (in the archaeological/non-linguistic record) doesn’t ‘prove’ relatedness; but lack of contact proves non-relatedness. innit?

  111. Contact (in the archaeological/non-linguistic record) doesn’t ‘prove’ relatedness; but lack of contact proves non-relatedness. innit?
    Firstly, as you said yourself in the sentence before that, it’s rather that a lack of recent contact may make showing relatedness impossible. And then, not all cases are as clear-cut as Cree and Welsh; e.g., Dene-Yeniseian looks quite plausible although the speakers of Yeniseian are on a different continent and thousands of miles away from their cousins in North America, and their last contact must have been millennia ago. Before the relationship was demonstrated not so long ago, most linguists would have dismissed it based on something like your “no contact” heuristic.

  112. David Eddyshaw says

    It depends on how far back in time you are prepared to push your putative protolanguages. After all, the forbears of the Cree did come from Asia in the sort of time-scale that long-rangers invoke quite happily as time depths for their imagined protolanguages.

    For my part, while I agree in principle (who could not?) that real science cannot provide certainty (DM) or conclusively disprove that any two random languages once shared a common ancestor (Etienne), I must say that protolanguages inaccessible to any currently available rigorous methodology have a pretty tenuous claim on existence, and any presumption of their existence seems to have no testable consequences at all: unless we develop wholly new rigorous techniques for studying such questions, of which there is currently no sign at all, despite some loud claims to the contrary.

    Entia non sunt multiplicanda …

  113. Precise knowlege (based on extra-linguistical data) also has no testable consequences at all.

    We are simply not in the situation

    “if these two languages are related, THEN these two morphemes are also related”.

  114. David Eddyshaw says

    Precise knowledge ,,, also has no testable consequences at all

    I think we must be working with different understandings of the words “precise” and “knowledge.”

    If I know for certain (as certain as I can be in this imperfect world) that my daughter does not currently live in Aberdeen, I can save myself quite a bit in train fares going to see if she’s in. That seems quite a testable consequence to me; if I was stupid enough to try to test it.

    It makes no sense to say that two morphemes are genetically related in two languages which are not genetically related. It’s just incoherent. On the other hand, if two languages are genetically related, you are in a much better position to declare that two morphemes aren’t related. If you don’t know whether the languages are related or not, you really have no way of telling either way. All you can do is say whether they seem similar or not.

    This business of “certainty” is a great big red herring in this discussion. Yes, we cannot achieve certainty by the scientific method. But we can be reasonably confident that many of our assertions would only be disputed by someone who was either ignorant of the basis of the assertion or ideologically unwilling to accept it regardless of the evidence. I’m getting tired of adding little imaginary footnotes every time I say “demonstrable.”

  115. Well, it is not unlike the present situation with Afro-Asiatic.

    Everyone believes that it is real. And? We have an outline of proto-Semitic and hopefully, when we have proto-Chadic and proto-Cushitic (and Omotic?) and maybe some other proto-languages (borrowings) we will be able to say more.

    It makes no sense to say that two morphemes are genetically related in two languages which are not genetically related. ” – DE, I’m not arguing with this. But how do you (not historians!) use the abstract fact that two lnaguages are related?

    In my view, linguists work with elements of language (morphemes and not only). Molecules and not species if you like. They start from identifying shared features and sieving through them. Then (and only then) they say that such and such things can hardly be explained with convergence or borrowing. From this of course follows the relationship of languages.

    But this fact is improtant to historians. What matters to you is correlated history of elements.
    So what I propose is placing this relationship of languages to a secondary position (of a copnsequence).

  116. David Eddyshaw says

    And?

    The consequence of believing that AA is real (a belief based squarely on linguistic evidence, and actually rather counterintuitive to anyone approaching the matter on “racial”, cultural or geographic lines) is that it means that there is a point in comparing all these languages, because it may lead to concrete results. It’s a pointer to the direction of potentially productive research: an entirely real-world consequence of the belief.

    The consequence of believing in proto-Algowelsh would be years of frustration: this is so obvious to any serious scholar that nobody does.

    Hard cases make bad law, as the lawyers say. There are plenty of unsolved problems in comparative linguistics, and there have been many wrong turnings and much leaping to conclusions without adequate evidence. Happily, science doesn’t have to be perfect to be real science: it just needs to be systematically prepared to admit error and discard hypotheses which have outlived their usefulness.

    In my view, linguists work with elements of language (morphemes and not only). Molecules and not species if you like. They start from identifying shared features and sieving through them. Then (and only then) they say that such and such things can hardly be explained with convergence or borrowing. From this of course follows the relationship of languages.

    But this fact is improtant to historians.

    I see what you mean (I think), but in fact comparative linguistics doesn’t just create pretty protolanguages which just sit there in Plato-space being transcendent: any protolanguage with any real claim on credibiliy will shed light on the modern languages descended from it, explaining oddities and irregularities.

    (I was just recently looking at how the Buli language palatalises (as I thought) some proto-Oti-Volta velars before front vowels but not others, an awkward anomaly I had vaguely hoped might be handwaved away by dialect mixture or something: I finally realised that I had reconstructed the protolanguage wrong in the relevant respect, and this quite unexpectedly explained how two very irregular verbs in Nawdm got that way. Buli and Nawdm are basically at opposite ends of Oti-Volta …)

  117. ” is that it means that there is a point in comparing all these languages, because it may lead to concrete results.”

    Yes, of course. It is just that Orel & Stolbova and Ehret’s dictionaries look like “let’s reconstruct proto-Chadic and Cushitic first” (but again, Stolbova works exacly on proto-Chadic and I don’t know Chadic well enough to say anything about her contribution).*
    There are interesting comparisons between individual branches.

    * and if a reconstruction heavily relies on data from other branches, it is rather a reason to worry… It is not like we all are tired of all those reconstructions that ignore other branches.

  118. David Eddyshaw says

    I suspect we are basically in agreement over the sins of (some) long-rangers, and probably over the sadly rather similar sins of some famous names in AA comparative work. I strongly agree with what I take to be your view on the necessity of doing adequate reconstruction of subgroups before overenthusiastically leaping into trying to reconstruct the Ultimate Protolanguage (whether AA or “Niger-Congo.”)

  119. It makes no sense to say that two morphemes are genetically related in two languages which are not genetically related

    On the contrary, it makes a lot of sense to define genetic relationship of morphemes to include loaning (be preserved over loaning; that’s really nothing other than the fact that X did, in fact, come from Y). This may seem unintuitive in a well-studied milieu like modern Europe, but is kind of a necessity whenever we have some material amongs a set of languages that is readily comparable, but we do not know the involved languages’ history in any real detail. To continue on the Afroasiatic theme, all the Omotic–Cushitic comparisons collected in previous research (including not just lexicon but quite a bit of grammatical material) did not suddendly become non-etymologies when people decided that Omotic is not a member of Cushitic and perhaps not; they would merely call for reinterpretation as loans in either direction, or maybe, at best, now common Proto-Afroasiatic inheritance rather than merely Proto-Cushitic.

    And it’s really this kind of data that establishes that there “is a point in” comparing languages, at least if we’re not a priori defining the process of “comparison” as something that has to give a positive result of “yep, they’re related all right”, to the exclusion of other options. Something like Proto-Uralic vs. Proto-Indo-European have this in good amount (there is no reasonable way to chalk all of it just to coincidence I think) even if it’s still not obvious if it adds up to a relationship. Something like Welsh vs. Cree of course don’t.

    There are even then still some long-shot hopes that, in some large enough family, even languages that pairwise share nothing in common, or nothing above the level of random noise, could end up demonstrable as related by the help of numerous other, intervening relatives. The thought experiment like comparing just modern French and modern Armenian really is just this, isn’t it? if we also have Latin and Greek and Persian and etc. then at that point the relationship indeed is demonstrable.

    (But then I would not swear off the cuff that Franco-Armenian must be undemonstrable — even between just the two of them there still are many readily observable similarities that do result from common descent like argent ~ arcatʰ ‘silver’, araire ~ arar ‘plough’, lune ~ lusin ‘moon’; and don’t make me dig up the numerals or the pronouns! Plenty of other cognates also exist to the effect of vin ~ gini ‘wine’ that are not instantly obvious but where additional work could perhaps still at least suggest a relationship eventually. Getting any kind of a good reconstruction of PIE out of just these is of course hopeless but that wasn’t the task, was it?)

  120. @LH, so you asked for an example.

    “His idea is that one needs to find levels of similarity and also to find out which languages are not related”
    (about Greenberg’s method).

    Bybee, Language Change.

  121. ‘rather counterintuitive to anyone approaching the matter on “racial”, cultural or geographic lines’

    Well, geographically their distribution matches rather well some sort of “extended MENA”, that is a set of regions with known contacts*. What is absent is…
    … West Africa! At least parts of it (starting from Senegal) clearly do belong to the area and were in contact with Morocco, but south of Touaregs-Zenaga-Arabic-Hausa we don’t have much of AA. Unless there is AA substrate there.

    * a strange sort of contacts which makes Sahara work like a sea, facilitating both movement along the shore and across it:/

  122. Speaking of Armenian, I learned that Meillet’s famous PIE *dw > Arm. rk rests on a total of only four etymons (plus several derivatives of *dwo); and that erku is pronounced /jerˈku/.

  123. Peter Grubtal says

    @David Marjanović

    I remember someone (I think it was Bernard Levin) many years ago asking for genuine Victorian sources for the story about the table legs having to be covered out of prudishness. IIRC nobody came up with anything.
    Like the even less plausible sawdust story, I suspect they were invented in the 20th C. to make us feel good about ourselves.

  124. Table legs are curvy. But I am too monogamous to notice them:)

    In a bath it does often happen that your groin is somewhere in the middle of your field of sight. But I doubt that public hair on mons pubis is so offensive. A Victorian man of course could experience an eureka when staring at his own penis. That proves (one more time) the well known fact that men are smarter than women.

  125. “an eureka”? You must be thinking of the Russian pronunciation. In English it’s pronounced with initial /j/. (What happened to the etymological h- when English borrowed it? Nobody really knows. Previously at Language Hat: (H)EUREKA.)

    A barely post-Victorian man looking forward to a bath

    foresaw his pale body reclined in it at full, naked, in a womb of warmth, oiled by scented melting soap, softly laved. He saw his trunk and limbs riprippled over and sustained, buoyed lightly upward, lemonyellow: his navel, bud of flesh: and saw the dark tangled curls of his bush floating, floating hair of the stream around the limp father of thousands, a languid floating flower.

    but I don’t think that’s a eureka.

  126. @ktschwarz, rather quasi-Latin.

    I picked English rather than learned it, and for a long time (more than a decade, less than two) I only read and almost never wrote, spoke or listened to it.

    I subvocalise English words in all sorts of ways:) Sometimes it is just “how I would explain the spelling to another Russian” because I wanted to remember the spelling, other times it is something rather random. It is quite likely that there was a period when I subvocalased “laugh” as [la:.u:.g.x] (just to remember it) if so, this reading has been updated. Eureka is spelled differently from Russian evrika, so quasi-Latin.

    Once (2018) I tried to use this word in a conversation, realised that I have no idea what it sounds like in English and stumbled, but my interlocutor was not a native English speaker and pronounced it in her own way.

  127. January First-of-May says

    I subvocalise English words in all sorts of ways:) Sometimes it is just “how I would explain the spelling to another Russian” because I wanted to remember the spelling, other times it is something rather random. It is quite likely that there was a period when I subvocalased “laugh” as [la:.u:.g.x] (just to remember it) if so, this reading has been updated. Eureka is spelled differently from Russian evrika, so quasi-Latin.

    Same – and occasionally it even appears to extend to spoken pronunciation…

    I think I’ve mentioned some of the particularly weird cases on LH before. Spelling is indeed a major contributor, as are Russian cognates (in the shared-loan sense).

  128. About the particular Georgian compound type mentioned in my first comment above and possibly seen in the element სახლთხუცი saxltʻxucʻi of Sakhltkhutsishvili, with სახლთ saxlt- apparently being the truncated form of სახლთა saxlta, archaic genitive plural of saxlti ‘house’…

    There is an account of Georgian nominal compound formation in chapter 8 of Bolkvadze and Kiziria (2023) Georgian: A Comprehensive Grammar. The following is their treatment of the compound type in question, on p. 145 (visible here on Google Books, I hope):

    Compounds consisting of two nouns, the first of which has the archaic genitive case plural marker -თა in full or truncated form: გულთამხილავი (გულთა + მხილავი) omniscient (hearts’ + seer); ზეთამზე (მზეთა + მზე) Mzetamze, the name of a village in Georgia (suns’ + sun); ხუროთმოძღვარი (ხუროთ + მოძღვარი) architect (carpenters’ + leader). In some of such compounds the particle -თა may be truncated: კაცთმოძულე (კაცთ + მოძულე) misanthrope (men’s + hater); კაცთმოყვარე (კაცთ + მოყვარე) philanthrope, kind person (men’s +lover).

    შენ ნამდვილი გულთამხილავი ხარ.
    You are (a) truly omniscient (person).
    არსაკიძე მე-12 საუკუნის ქართველი ხუროთმოძღვარი იყო.
    Arsakidze was a 12th-century Georgian architect.
    ამ კაცთმოყვარე ადამიანს ყველა პატივს სცემს.
    Everybody respects this kind man.

    About the diachrony of this formation… Kevin Tuite ‘Early Georgian’, in The Ancient Languages of Asia Minor (2008), has an account of the plural types of Early Old Georgian (5th–8th centuries). He discusses this type on p. 151 (visible here on Google Books, I hope):

    4.2.2 Plural marking

    Early Georgian has two structurally distinct means of marking nominal plurality. By far the most frequently used is the synthetic n-/t- plural declension. The n-rectus-plural suffix is limited to the absolutive and vocative, and may be historically related to the plural absolutive suffix of the verb. The single oblique plural morpheme -t-(a) can represent the dative, genitive, or ergative cases; the instrumental and adverbial do not appear to have had distinct plural forms in this declension (cf. the instrumental with plural reference in Mt 15:8: eri ese bag-ita mat-ita p’at’iv mcems “these people honor me with their lip(s)”).

    The agglutinative eb- plural suffix, followed by the case endings of the singular declension, appears only a couple of dozen times in the Early Georgian corpus, sometimes in conjunction with n-/t- plural nouns: brm-eb-i da q’ruv-n-i (Mt 15:30) “the blind (eb-plural) and the deaf (n-plural).” While there is no evidence of a semantic distinction between the two plural morphemes in Early Georgian, only n-plural nouns can control plural agreement in the verb and within the noun phrase, whereas eb-plurals are syntactically singular: rabami kw-eb-i ars “what large stones there are (lit. is)” (Mk 13:1).

  129. Great stuff, and your Google Books links work fine!

  130. Slavic bog

    When the translation improves on the original… Luke 12:21:

    οὕτως ὁ θησαυρίζων ἑαυτῷ, καὶ μὴ εἰς Θεὸν πλουτῶν

    ⱅⰰⰽⱁ ⱄⱏⰱⰻⱃⰰⱗⰻ ⱄⰵⰱⱑ · ⰰ ⱀⰵ ⰲⱏ ⰱ҃ⱏ ⰱⱁⰳⰰⱅⱑⱗ

    тако събираѩи себѣ · а не въ б҃ъ богатѣѩ ·

    tako sŭbirajęi sebě, a ne vŭ bogŭ bogatěję

    (‘So is he that layeth up treasure for himself, and is not rich toward God’.)

    After reading the Slavic version the first time in OCS class, I skipped down the lecture hall stairs, out the building, and around campus and town, constantly repeating it.

  131. Lars Mathiesen (he/him/his) says

    Google Books works much better in the US than in Europe. I have tried a lot of the links that Hat says work fine, and in 9 out of 10 cases I get the “This page is not part of the accessible content or you have exceeded the number of pages available to you in this book” screed. (I’m currently in Madrid and for some reason I get it in French!?: Cette page ne fait pas partie de la section consultable ou vous avez dépassé le nombre de pages que vous êtes autorisé à consulter pour ce livre.)

    So please, do not assume that because a Google Books link works for you, that everybody else can see it. (GB are especially nasty about this; I do sometimes get links that want to direct me to the Danish or EU version of something, but it doesn’t exist). Archive.org always works.

  132. David Eddyshaw says

    Google books also mysteriously intermittently seems to think that I am French. I’ve no idea how that came about.

    [EDIT: just tried Xerîb’s link, which actually worked fine, showing the whole thing: complete with a popup offering to translate the page from French …]

  133. Google very correctly assumes that you all are foreigners.
    (Someone from China could claim that I’m a foreigner too. I would respond: you’re a foreigner yourself!)

  134. Men are from Mars, women are from Venus. Send them all back!

  135. David Marjanović says

    Google Books will show you different parts of a book from different top-level domains (country codes), and also different parts from the same TLD in different weeks.

    I suspect they were invented in the 20th C. to make us feel good about ourselves.

    I hope so.

    EDIT: just tried Xerîb’s link, which actually worked fine, showing the whole thing: complete with a popup offering to translate the page from French …

    It is in French. Xerîb didn’t cut &hl=fr from the URL, so it’s officially in French.

  136. It does seem plausible to me that such a brochure could be published, you can’t overestimate human stupidity.

  137. What brochure?

  138. Sorry, it was an inaccurately truncated version. I intended to say, I would not be surprised if someone published a brochure with lifehacks for maidens (and housemaids) that contained it etc…. but chose to be less verbose and somehow the longer version collapsed into “such brochure”

  139. I would not be surprised if someone published a brochure with lifehacks for maidens (and housemaids) that contained it etc….

    Contained what?

  140. The advise (I don’t know if “the advise” is good English for it is not countable, while recommendation and tip are somewhat different: I mean sovet) to strew sawdust on the water in order to hide pubic hair which God in his infinite wisdom planted on the mountain of Venus to hide the ravine with running brooks of [here my imagination fails me]….

    P.S. to the comment below:
    Thank you! Yes, I meant “the advice”: I know about c/s at abstract level, but it is something I usually have to correct after having typed the wrong letter:/ In this case I was distracted by the difference in usage stemming from (un)countability.

  141. Oh! I’d forgotten all about that. These threads move fast… (And yes, “the advice” — with -c- in the noun — is perfectly good English.)

Speak Your Mind

*