The Indo-European Controversy: An Interview.

George Walkden at New Books in Language:

Who were the Indo-Europeans? Were they all-conquering heroes? Aggressive patriarchal Kurgan horsemen, sweeping aside the peaceful civilizations of Old Europe? Weed-smoking drug dealers rolling across Eurasia in a cannabis-induced haze? Or slow-moving but inexorable farmers from Anatolia?

These are just some of the many possibilities discussed in the scholarly literature. But in 2012, a New York Times article announced that the problem had been solved, by a team of innovative biologists applying computational tools to language change. In an article published in Science, they claimed to have found decisive support for the Anatolian hypothesis.

In their book, The Indo-European Controversy: Facts and Fallacies in Historical Linguistics (Cambridge University Press, 2015), Asya Pereltsvaig and Martin Lewis make the case that this conclusion is premature, and based on unwarranted assumptions. In this interview, Asya and Martin talk to me about the history of the Indo-European homeland question, the problems they see in the Science article, and the form that a good theory of Indo-European origins needs to take.

At the site you can hear the hour-long interview. Thanks, Trevor!

Update. Discussion of some of the issues raised in the (very long) thread below continues at Eli’s Indo-European Etymology Blog.

Comments

  1. David Marjanović says:

    Ooh, nice. I’ll have to listen to it this evening.

    Together, the recent papers on the genetics of the Corded Ware/Battle Ax people make it highly likely that the non-Anatolian branch of IE expanded with the brown-eyed, lactose-tolerant hordes of the Yamnaya culture.

  2. Greg Pandatshang says:

    I had a brief conversation with Asya Pereltsvaig, one of the authors, at Razib’s blog, following up on this conversation (not that I contributed a lot of insight here, but she had some comments):

    http://www.unz.com/gnxp/open-thread-july-26th-2015/

    Pereltsvaig’s blog, http://languagesoftheworld.info/ is pretty rad.

  3. David Marjanović says:

    From your comment there (the thread is closed):

    There was presumably some linguistic diversity in Anatolia at some point in the past. We might question whether it’s unusual for Indo-European to originate in a place so close to where a language as exotic as Hattic was spoken. However, I don’t think we really know enough about Hattic to say how different it was or wasn’t from PIE; and anyway I don’t know how we would know for sure that that level of diversity would be unusual in that time and place.

    The Hattic language is well enough known to tell that it was very different from (P)IE indeed (see the German version for a bit more on the grammar). But your point stands: Anatolia is really large, and the finding that it was linguistically diverse in the Bronze Age cannot come as a surprise under any hypothesis about the PIE homeland.

    According to one of the two genetics papers, the Yamnaya people represent a mixture of local Eastern Hunter-Gatherers = Ancient North Eurasians and immigrants from the south who were similar to present-day Armenians. The other paper didn’t investigate this issue in the first place.

  4. Thank you Hat. Can I ask a methodological question?

    (This isn’t a comment on the “computational tools”, but on the philologists’ groundwork that they relied on.)

    The discussion makes quite a bit of the ‘anomaly’ of the so-called Tocharian languages not fitting in to the models from the “innovative biologists”. (They don’t fit because the Tarim basin is so geographically remote from IE homelands, so the models provide no explanation how the languages/their speakers migrated there.)

    These languages are identified as Indo-European from their sound pattern. Furthermore there’s a claim that the A and B variants are not mutually intelligible. [Wikipedia]

    My methodological question is: how could anyone tell these languages are IE?

    According to Wikipedia, the Tocharian languages are extinct, known only from manuscripts. So how do philologists have any idea what they sounded like, particularly with enough accuracy to divine whether their vocabulary derives from PIE + consistent sound shift?

    OK you’re going to tell me “the comparative method”, hermeneutics, yada yada. How does this differ from the cabalists, or voodoo?

    (And I know that Tocharian is by no means the only language ‘family’ attributed as PIE but long extinct and known only from scraps of manuscripts).

  5. These languages are identified as Indo-European from their sound pattern.

    No, they’re identified as Indo-European the same way other languages are: by consistent patterns of phonetic and morphological relationships. If you scroll down the Wikipedia article to the Proto-Tocharian section, you’ll see a table with examples like “PIE *pḥ₂tḗr “father” > PToch *pācer > TB pācer, TA pācar” and “PIE *(h₁)eḱwo– “horse” > PToch yä́kwë > TB yakwe.” Any single apparent cognate might be pure coincidence (this often trips people up who encounter a familiar-looking word in a foreign language and leap to the conclusion it must be related to whatever language they’re familiar with), but if you have entire series of words where Toch. p corresponds to p in (say) Latin and Greek and a corresponds to a and so on, then the only logical conclusion is that those languages are related. There is no doubt whatever that Tocharian is an IE language; it was recognized as such almost immediately, unlike Albanian, which hid its IE origins under a heap of borrowings. If you want more information on how the comparative method works, marie-lucie has written some excellent comments explaining it which I am too lazy to try and find at the moment, but the Wikipedia article will give you a head start.

  6. The keys to Tocharian were the known scripts (Brahmi and Manichean) in which the texts were written, and the fact that they were evidently translations of Buddhist scriptures (though secular material has also been found for Tocharian B). With the scripts known, the pronunciation was known, nearly enough (at worst it represented a slightly older version of the language spoken at the time).

    Mere inspection showed that the Tocharian material was fairly close linguistically to the Pali originals, in a way that could not have been the case if Tocharian were a Turkic language, say. Given parallel texts, it was possible to work out the sound-changes separating Tocharian from Pali, and then triangulate back to PIE.

    Voodoo and cabalism are religious movements. The comparative method is the common core of all historical sciences, including evolutionary biology: the Museum of Comparative Zoology has been employing the method since its founding in 1859.

  7. marie-lucie says:

    AntC: These languages are identified as Indo-European from their sound pattern

    No competent linguist would identify a language as belonging to this or that family just from its “sound pattern”. You need resemblances of vocabulary (not always obvious because of sound changes) and especially morphology (eg noun and verb structure, pronouns).

    Perhaps by “sound pattern” you mean the pattern of sound correspondences between Tocharian and other IE languages, which can be detected from a comparative study of the common vocabulary. For instance, the Germanic family is identified partially by the pattern of initial consonant correspondences in otherwise almost identical words, such as the p/f correspondences in Latin pisc-is but Scandinavian fisk, German Fisch, English fish, and similarly for several other consonant groups. But that is not the only thing that makes this family part of IE. Similarly, there are probably some correspondence patterns which are typical of the Tocharian group.

    It is true that some languages are only attested by “scraps”, such as the inscriptions in some of the ancient languages of Italy, too scanty for decipherment, so that their affiliation is doubtful, but the Tocharian family, especially one of its two languages, is much better attested than those. Even if the manuscripts have deteriorated into “scraps”, there are enough of them to get a very good idea of the languages they are written in, especially since many of them are translations of texts known from Indian languages.

    As for how to determine the actual sounds of these languages, the documents are written in variants of Indian scripts which are known, so that they were able to be read before they were understood. A few of the sounds may have been hypothesized from correspondences and from the knowledge of the kind of sound-changes likely to occur in certain circumstances (such as word-initially, before some vowels, etc).

  8. Albanian, which hid its IE origins under a heap of borrowings

    Not its IE origins; that was always obvious (një, dy, tre, katër, pesë, gjashtë, shtatë, tetë, nëntë, dhjetë), and was nailed down by Bopp in 1854. What wasn’t obvious was that Albanian is a separate branch of IE. There were similar problems with Armenian, which was first thought to be an aberrant Iranian language, but actually has two layers of Iranian loans.

    (And yes, numbers aren’t 100% reliable: Chamorro now uses un, dos, tres, kuatro, sinko, sais, siette, ocho, nuebi, dies instead of hacha, hu-gua, tulu, fatfat, lima, gunum, fiti, gualu, sigua, manot. But still.)

  9. Greg Pandatshang says:

    David Marjanović,

    Right. Pereltsvaig then introduced additional details of the specific Anatolian-origin hypothesis which, she advised me, are incompatible with what we know about Hattic. I had not known about those additional details beforehand, so obviously I had nothing intelligent to add further. I continue to believe that the argument they originally presented in the interview didn’t work because it did not specify those details, but it seemed no more than pedantic to belabor that point.

  10. Michel Alexandre Salim says:

    The old Chamorran (sp.?) numbers look curiously similar to Javanese

  11. “The old Chamorran (sp.?) numbers look curiously similar to Javanese”

    Actually those languages are pretty closely related:
    https://en.wikipedia.org/wiki/Sunda%E2%80%93Sulawesi_languages

    As with any other expansion, the settlement patterns are kind of a jumble.

  12. MIchel,
    Here‘s a site that has compiled lists of the numbers 1-10 in a long list of Austronesian languages. You’ll see many resemblances to the old Chamorro and Javanese numbers. Most have not replaced all their numbers with borrowings, as Chamorro has, but many in the New Guinea area have replaced their numbers 6-10 with compounds: 5+1, 5+2, etc., or (near the Admiralty Islands) 10-1 for 9, 10-2 for 8, etc.

  13. marie-lucie says:

    Even though numbers (especially 1-10) may be cognate in most if not all of the Indo-European languages, and therefore seem to be part of basic vocabulary unlikely to be borrowed, there are many exceptions in the world, especially in small-scale societies. Modern Western societies are obsessed with exact numbers (age, weight, dates, population, temperatures, and other numerical measurements) but this was not the case in ancient societies. Where numbers are used especially for purposes of storage and exchange of goods (as for instance in Egypt or Mesopotamia), the words for them can travel long distances through a variety of societies along with the goods, even if typically the human bearers of the goods only travel relatively short distances between specific stopping points where exchanges take place (eg along the Silk Road or other traditional path). It is not difficult or rare for the average person to learn 5 or 10 words for numbers without learning the language these words originated in, any more than to learn the names of the “exotic” goods traded and transported (see for instance the fortunes of words for “coffee” and “tobacco” once these substances were circulated far from their countries of origin).

  14. Indeed. In fact I find it odd that numerals are so conservative throughout IE languages, which for much of their history were spoken by small societies of subsistence farmers.

  15. marie-lucie says:

    This suggests that indeed PIE did have those numerals. But some of them do resemble numerals in other languages unrelated to PIE, suggesting that PIE may have borrowed the numerals before splitting. I seem to remember that the PIE word for “7” may be such a borrowing. Some of you Hatters will know better than I do.

  16. Trond Engen says:

    I sometimes enjoy the idea that the Indo-European expansion established a Trans-Eurasian high-speed trade network. Numerals beyond three or four would have been an inherent part of the technology.

  17. The Uralic words for 7 (Finnish seitsemän, e.g.) are very similar to the Indo-European ones, but it’s usually assumed that the borrowings go the other way. Indeed, no number greater than 6 is reconstructable to Proto-Uralic.

  18. marie-lucie says:

    JC: Uralic borrowing from (P)IE is not incompatible with PIE (earlier) borrowing form some other language(s).

  19. The PIE numerals ‘three’, ‘six’ and ‘seven’ resemble the Semitic ones quite a bit, and I believe people have been looking into that since the 19th century. I have no idea where matters stand at present. A quick search finds, e.g., this on Paleoglot, of which I have no opinion one way or another. Note Kartvelian numerals are involved somehow as well.

  20. Yes, it is much the safest to express no opinion of Paleoglot.

  21. David Marjanović says:

    The comparative method is the common core of all historical sciences, including evolutionary biology: the Museum of Comparative Zoology has been employing the method since its founding in 1859.

    …No. Nowhere close.

    The name of the MCZ refers to comparative anatomy, a discipline founded by Cuvier at the beginning of the 19th century when evolution was just Buffon’s idle speculation. It is not a method like the comparative method of historical linguistics.

    Well into the 20th century, phylogenetics in biology (as opposed to linguistics) had basically no method at all. Rather than a science, it was an art. A reproducible method was first published in 1950 on the wrong side of the Iron Curtain, became widespread only in the 1980s and 90s, and still hasn’t reached the study of all branches of the tree.

    Language classification began as phylogenetics, as the tree model. In biology, classification came a full hundred years (Linnaeus) before Darwin and Wallace published their paper in 1858 and Darwin’s book finally came out in 1859; it shows, it still shows.

    Yes, it is much the safest to express no opinion of Paleoglot.

    That particular post has a few interesting ideas. But those ideas rest on a number of assumptions. For example, it seems pretty obvious that the Proto-Semitic *š wasn’t [ʃ] at all, but an alveolar [s], and that the PIE *s was the exact same thing. (Proto-Semitic *s most likely was [t͡sʰ].)

    The reconstruction of various pre-PIE stages (developed across several more posts) is intriguing, but the thing about internal reconstruction is that it’s almost unconstrained by anything other than the imagination. To reconstruct ancestral states, you need an outgroup – a close relative but not member of IE.

    Finally, complaining about markedness while simultaneously reconstructing a phoneme */sʷ/ is an odd thing to do. Even Abkhaz and Ubykh don’t have such a phoneme; Archi and Lezgian among the East Caucasian languages do, and so does Lao, but that seems to be all.

  22. @David Marjanović: The reconstruction of various pre-PIE stages … is intriguing, but the thing about internal reconstruction is that it’s almost unconstrained by anything other than the imagination.

    That is very true — but I for one would find it wildly fascinating to know what (if anything) the real experts think about the possible shape of such stages. But most of them prudently don’t share.

  23. Gary Moore says:

    Ultimately, the origins of Indo-European may be much farther east, in southern Siberia and possibly North America. An article published in journal Genetics in 2012 titled “Ancient Admixture in Human History” revealed that Europe experienced an admixture with a population related to Native Americans about 5,000 years ago, leading one blogger to speculate: “All that being said, what went on k)
    Arm (Cl) akn
    Arm achk
    Mod. Arm ačk’
    Avestan čama, čašman-, chashman (m > k)
    Old Persian čaša-, čašna
    Sanskrit akshân, akshi; chakshush, chaksu
    Vedic Skt ákṣi, cháksus
    Hindi ax
    Old Prussian ackis

    Mohawk okà:ra
    Oneida okáh(la)
    Anc Grk ophthalmós
    Grk *Her) oːy-ó-n {ᾠόν}
    Old Church Slavonic oko око
    Ukrainian oko, vusko
    Old Norse auga
    Swedish öga
    Latin oculus

    Siouan-like forms:

    Luwian tāwīs
    Lycian tawa
    Lakota ištá
    Osage ehtah
    Tutelo tasu
    Crow ishtá

    Hittite and the other Anatolian languages are an interesting case because they appear to incorporate a higher proportion of forms of Siouan origin. For example, Hittite form for ‘ear’ is istāman, istam-an- ~ istam-in-. More typically in IE languages, this form is derived from the *PIE *H₂ous-. The Hittite form is a bit of a mystery: “The root was lost in IE (connections with Ancient Greek and Avestan forms with the invariant meaning ‘an organ of perception’ or ‘a hole in the head’ are semantically unsatisfactory). …” (http://starling.rinet.ru/new100/ana.pdf) The Hittite form is a close match for the modern Lakota form for ‘eye’, ištá, suggesting the ‘organ of perception’ etymology.

    A model for Proto Indo-European is the Central American language Garifuna. The genetic heritage of the Garifuna people is only 15% Native American, but their core language is still essentially a composite of Arawan and Carib. Likewise, according to genetic studies, modern Europeans are only 5-18 % Native American but their North American heritage lives on in their languages, which pear to be a creole based largely on a lexicon of Paleo-Iroquoian and Paleo-Siouan languages which were probably brought to Siberia as a result of a back migration from North America.

  24. Gary Moore says:

    Unfortunately, a lot of my previous posting was truncated. To summarize, some linguists have noted parallels in both lexicon and syntax between the Native American Iroquoian and Siouan languages and Indo-European. Two papers describing similarities between Siouan languages and Indo-European were published in 1881 and 1882 by Andrew Williamson. More recently, in “The New Sound of Indo-European: Essays in Phonological Reconstruction” edited by Theo Vennemann, Pierre Swiggers comments ion Beekes’ comparison of the phonology of Proto Indo-European and the Salishan language Shuswap in the chapter titled “On (the nature of) PIE Laryngeals” :

    “Is there an implication of a generic relationship in the statement that “the sound system reconstructed for Proto-Indo-European could have arisen from the Shuswap system through losses”?”

    Indo-European appears to combine a lexicon to a very great extent based on archaic Iroquoian and Siouan with a sound system (especially laryngeals) partly based on Salishan languages. IE languages also retain fossilized versions of Iroquoian 3rd person singular prefixes (especially in forms for body parts) as well as the Siouan ‘wi-‘ and ‘wa-‘ noun prefixes (examples: *Proto Siouan *wiˈroːka ‘man’ (male) / *PIE *wiHrós ‘man’ (male) / Old Prussian wirakan.)

    “man (male)”: *PIE *wiHrós / *Proto Siouan *wiroka
    “man, hero” *PIE *H₂ner- / *Proto Iroquoian *-hniɹ- ‘be durable, be hard, be solid, be strong’ (See Latin neriōsus “strong”)

    #68 ‘horn’ onà:kara (Mohawk) { *-naɁkaɹ- } ‘antler, horn’ (*PNI) / *ḱr̥nom, *ḱerh₂(s),*ḱerh₂sr̥, *koru (*PIE)
    #71 ‘hair’ { -kiɁɹh- } (*PI) / *keres- (“rough hair, bristle”) (PIE*)
    #74 ‘eye’ okà:ra (Mohawk) / *h₃okʷ-, *h₃ekʷ- (*PIE)
    #75 ‘nose’ o’niónhsa (Mohawk) / hnéh₂s , *nā́s (*PIE)
    #76 ‘mouth’ ohsakà:ra (Mohawk) / aholi (Cherokee) / *h₁oh₁s- (*PIE) (Note: Basque: ‘aho’. Uhlenbeck used this form as a ‘probe’ for linking Basque to other languages.)
    #77 ‘tooth’ onawí:ra/onò:tsa (Mohawk) / *h₃dónts, *ǵembʰ- (*PIE) [Note: The Mohawk form is “hiding out” in the *PIE form: *h₃d-on(o)ts(a)-]
    #81 ‘leg’ (a)kohsinkoɁˈtaɁkeh (PNI *) / *kroksko (*PIE)
    #82 ‘knee’ kanigeni (Cherokee) (literally, ‘his knee’) / *g(e)neu- (*PIE)

    The possibility of a North American origin for IE appears to be supported by genetics research. From a recent article in Nature #522, 167–172 (11 June 2015) about the DNA of Bronze Age steppe populations:

    “Intriguingly, individuals of the Bronze Age Okunevo culture from the Sayano-Altai region are related to present-day Native Americans, which confirms previous craniometric studies. This finding implies that Okunevo could represent a remnant population related to the Upper Palaeolithic Mal’ta hunter-gatherer population from Lake Baikal that contributed genetic material to Native Americans.”

  25. David Marjanović says:

    Ultimately, the origins of Indo-European may be much farther east, in southern Siberia and possibly North America. An article published in journal Genetics in 2012 titled “Ancient Admixture in Human History” revealed that Europe experienced an admixture with a population related to Native Americans about 5,000 years ago

    Don’t read papers from 2012 without reading papers from 2015! 🙂 These “Ancient North Eurasians” (to whom the 24,000-year-old skeleton from Mal’ta in Siberia belongs), together with more southern people most similar to present-day Armenians, are the genetic components of the people of the Yamnaya culture. Most likely the ANE simply spanned all of Siberia and eastern Europe.

    leading one blogger to speculate:

    German Dziebel? 🙂

    Unfortunately, a lot of my previous posting was truncated.

    Simply try again after replacing every < and > by &lt; and &gt; so they won’t be interpreted as HTML code.

    To summarize, some linguists have noted parallels in both lexicon and syntax between the Native American Iroquoian and Siouan languages and Indo-European.

    So, where are the regular sound correspondences? So far you’ve only presented resemblances that are indistinguishable from random.

    And what do you think of the idea that PIE “tooth” wasn’t *h₃donts, but *h₁donts (so that the Greek o- is an epenthetic copy vowel), which would have been an active present participle of the verb *h₁ed- and would have originally meant “eating”, “biting”? Such nominalizations aren’t unknown in PIE; wind is “the one that makes [xw]”, *h₂w-e-nt-…

    The possibility of a North American origin for IE appears to be supported by genetics research. From a recent article in Nature #522, 167–172 (11 June 2015) about the DNA of Bronze Age steppe populations:

    That the ANE came from North America is neither implied by that quote, nor is it the simplest interpretation of the facts in that quote.

  26. Proto-Siouan is dated to 1000 B.C.

    Proto-Indo-European was spoken around 4000 B.C.

    Therefore, PIE origin of Proto-Siouan appears more likely than vice versa. (unless Proto-Siouans had time machine)

  27. Gary Moore says:

    The blogger who suggested the relationship between the Native American admixture in Europe and the advent of Indo-European was Razib Khan writing in Discovery Blogs (http://blogs.discovermagazine.com/gnxp/2012/09/across-the-sea-of-grass-how-northern-europeans-got-to-be-10-northeast-asian/#.VeKDZuntiS0). Khan writes:

    “But the bigger shock is that Europeans, and especially Northern Europeans, seem to have a substantial Northeast Asian component. From the nature of the prose I feel that the authors were definitely taken aback. They basically say so in so many words. In the process of resolving their confusion they skinned the cat every which way. And it does look to me that Northern Europeans are truly descended in part from a population which has affinities to the “First Americans.” I say this specifically because the Siberian samples they tested actually gave a weaker result than the South American Amerindians on the 3-population test. …

    All that being said, what went on <5,000 years before the present to reorder the European landscape? The answer may sound crazy, but I think the most probable explanation (even if it is unlikely) is something to do with the Indo-Europeans. …

    Though it strikes me as a bizarre conjecture, but I can’t help but imagine the rapid expansion of Indo-European populations into Europe, pushing into the peninsulas of the south. These people may have been a newly formed cosmopolitan mix of West Asians, Northern European Mesolithics, and Northeast Asians. I am at a loss to hazard a guess as to who the First American-like Northeast Asians were, though perhaps they were a western offshoot of the Kets? These people were then absorbed into a melange of tribes who themselves emerged from a synthesis between immigrant West Asian farmers and Northern Europeans. In shorthand: perhaps the Indo-Europeans were mongrels! …"

    It should be noted that distinctively Native American Y haplogroup Q-M3 has turned up in ancient DNA from western China as all as modern DNA from bulk testing of men in central and eastern Europe. Varieties of Y haplogroup Q () closely related the Native American varieties are also found in Scandinavia. Because YHG Q-M3 is widely believed to have originated in North America, its presence in Eurasia could be viewed as evidence of a back migration. Intriguingly, the concentration of YHG Q in Europe appears to correlate well with the areas of the Nordic Bronze Age and Hallstatt cultures.

  28. Gary Moore says:

    I was not suggesting that Indo-European was directly descended from Proto-Siouan. Rather, certain forms in Indo-European languages appear to be derived from an archaic ancestor of Proto-Siouan.

    As far as regular sound correspondences go, it appears that the proto language from which PIE was ultimately derived, like Proto Iroquoian, was lacking in bilabial consonants. For instance, the forms for ‘five’:

    w- /wh- > p- ‘five’ *Proto Northern Iroquoian (PNI) *hwihsk / *PIE *pénkʷe / Armenian ‘hing’ / Cherokee ‘hisg(i)’ (Note: the ’n’>’s’ correspondence is also attested in IE forms for ‘five’: Tocharian A päñ / Tocharian B piś; Greek pénte / Albanian pesë.)

    Similarly for ‘four’ :

    k- > kw- > m- ‘four’ *PNI *kaˈjeːɹih (Julian Charles) *kayeri (Lounsbury) / *PIE *kʷetwóres / Hittite ‘meyawes’ (Note that the -t- is omitted in some IE forms for ‘four’, particularly in Indo-Iranian languages Ex: Pashto calor, tsalór, Saka (Khotanese) cahora / Saka (Tumshuqese) cahari. Compare with modern Mohawk kaié:ri.)

    And ‘one’: *PIE *sem Cherokee saqui (or sawu) / Laurentian+ secada / Mohawk ens()ka / Oneida ós()kah / Seneca s()kat / Arikara as()co] / Pawnee us()k’-o, where () indicates a vowel deletion. (-m > -k-) (-e- > -a-) Note also how Caddoan forms also fall in line with Iroquoian. (Incidentally, there is an interesting parallel of ancient Greek form εἷς ‘heîs’ derived from this root, to the Wichita chí?ass.)

    PIE *leubh- “to care, desire, love” / Cherokee (u)lvkw(di) ‘he likes it, him’ / Oneida (-no)luhkw- love’ ( -bh- > -kw-)

  29. Gary Moore says:

    One of the most compelling pieces of evidence for an Indo-European / North American connection is the apparent cognate form for ‘dog’ in IE and North American languages. Based on the lack of evidence for the presence of domesticated dogs in North America before 11,000 years ago and the shared forms for ‘dog, horse’ between the language families, it can be reasonably inferred that the precursor language of Proto Indo-European most likely split from North American languages no earlier than this approximate date. Below are a few more forms for ‘dog, horse’ to the list from Siouan-Iroquoian-Caddoan complex:

    Crow: biška
    Hidatsa: wašúka
    Dakota: šų́nka
    Ioway: šų́ñe
    Osage: šǫ́ke
    Biloxi: čhǫ́ki
    Ofo: ačhų́ki
    Tutelo: čhų́ki
    Saponi: “chunkete”
    Cherokee: soqueli (‘horse’)
    Cayuga: só:wa:s
    Skiri Pawnee: asaaki
    Shuswap (Salishan) sewt –1. Slave (also: sesésu7t); 2. animal owned

    (Note: Crow bi- and Hidatsa wa- are noun prefixes.)

    Equivalents in Indo-European are:

    Russian: сука (súka) (‘female dog’, ‘bitch’)
    Polish: suka
    Sanskrit: ś(u)vā́ (śúnaḥ)
    Avestan spā (acc. spānǝm, pl. gen. sū̆nam)
    Middle Persian sak
    Kurdish sah
    Wakhi šač “dog”
    Old Church Slavic: suka “bitch (female dog)”
    Old Prussian: sunnis “dog”
    Lithuanian: šuo “dog”
    Armenian: շուն • (šun)
    Luwian ásùwa (‘horse’)
    Luwian zuwanis ‘dog’

    C. C. Uhlenbeck suggested a generic relationship between Eskimo-Aleut and Indo-European. In contrast, equivalent forms in Inuit-Yupik-Aleut are:

    qimmiq (Most inuit dialects including Inupiaq, Inuvialuktun, Inuktitut)
    qipmiq (Malimiutun inupiaq)
    qinmiq (Inuinnaqtun)
    qingmiq (Natsilingmiutut/Kivalliq)
    kimmik (Labrador Inuttut)
    qimmeq (Kalaallisut)
    qikmiq (Siberian Yupik)
    qimugta/piguta (Yup’ik)

    However, the Aleut form for ‘dog’, sabaakax, suggests a bridge between the Eurasian and North American forms: sa(baa)ka(x)

  30. Gary Moore says:

    Linguists are quick to discount claims of links between languages based on similarities in forms for ‘water’. However, the cross correlations for forms for ‘water’, ‘wet’, ‘lake/river’, ‘salt’, and ‘wind’ between IS and North American forms add a degree of confidence the the candidate cognate word for ‘water’ is not accidental.

    Iroquoian/Salishan

    *PrIro *awe ‘water’
    *PrIro *-entar- / *-otar- ‘lake’ (literally – ‘to be a lake’)
    Nottoway awa ‘water’
    Nottoway kanatariya ‘lake’
    Mohawk kaniá:tare ‘lake’
    Cherokee vdali ‘lake’
    Cherokee gadulida ‘wet’
    Cherokee ama ‘water’
    Cherokee vdali ‘lake’
    Mohawk oniá:tare ‘river’
    Mohawk ohné:ka /
    ohné:kanos ‘drinking water’
    Mohawk ionà:nawen ‘wet’
    Mohawk ówera ‘wind (breeze)’
    Cherokee u²no²le ‘wind, air’ (r > l)
    Oneida owe·lá
    Mingo kææha’
    Suswap líq’wem ‘to spill/sprinkle water’
    Suswap liq’wt ‘liquid spills over the edge’
    Suswap séwllkwe ‘water’

    *PIE *h₂ekʷeh₂ ‘water’
    Persian tar ‘wet’
    Kurdish terr ‘wet’
    Persian wɔb ‘water’
    Kurdish aw ‘water’
    Persian daryåche ‘lake’
    Persian daryå ‘sea’
    Mazandarani derka ‘river’
    Urdu samandar ‘sea’
    *Pr Celtic awa ‘river’
    *PIE ḱewero- ‘wind’
    *PIE *wleik- “to flow, run.”

    Indo-Iranian roots: tar, terr = ‘wet’ dar-/der- ‘body of water’

    Hittite wa-a-tar ‘water’ is a compound form consisting of (a)wa + tar
    *PIE *wod-or ‘water’ compound of (a)w + odar?

    ‘Salt’

    Persian namak ‘salt’
    Cherokee a:ma ‘salt’
    Tajik namak ‘salt’
    Cherokee ali ‘sweat’
    *PIE *sal- (1) “salt”
    Latin alumen “alum,” literally “bitter salt,” cognate with Greek aludoimos “bitter” and perhaps with English “ale”
    Grk háls (halós)
    *PC salano- ‘salt’

    Considering similar Japanese, Ainu, and Dene forms suggests a possible Northeast Asian sprachbund.

    Japanese kawa ‘river’
    Japanese âme ‘rain’ (-m- > -w-)
    Ainu wakka ‘water’
    Ainu pet ‘river’ ( ‘w’ > ‘p’ ?)
    Ainu to ‘lake’
    Navajo tó ‘water’
    Navajo tooh ‘river, lake’
    Ainu réra ‘wind (breeze)’

  31. —However, the Aleut form for ‘dog’, sabaakax, suggests a bridge between the Eurasian and North American forms: sa(baa)ka(x)

    sabaakax is borrowed from Russian.

    The language had extreme degree of Russian impact which almost killed it at the time (two hundred years ago, not 11 thousand years before present)

    Last page of my Aleut dictionary lists words such as

    zaavtrikax- breakfast
    zaavtrikal – to have a breakfast
    zanavisxix – curtain
    zaymil – to borrow
    Xliibaax zaymida – borrow me some bread
    ziilitix – vest
    ziirkalax – mirror
    zuulutix – gold

  32. So you mean that, for “salt” and “water”, Italian preserved the s- and -k-, lost in the American languages?

  33. Florian Blaschke says:

    David: Brown-eyed?

    https://en.wikipedia.org/wiki/Kurgan_hypothesis#Genetics

    Of course, the appearance of Slavic-speaking natives of Eastern Europe is still generally like that, fair and light-eyed, so the idea that the ancient steppe dwellers by and large looked the same is hardly unexpected. From what I’ve read, even though the steppe expansion left a much more significant genetic footprint than formerly believed, I understand it had surprisingly little impact on the general appearance of European natives, so Neolithic steppe dwellers had a typical modern Eastern European appearance, Pre-IE Northern Europeans generally looked “Nordic”, Pre-IE inhabitants of Northwestern Europe had what we think of as “Celtic” looks, and Pre-IE Southern Europeans in the late Neolithic looked by and large like typical modern Southern Europeans. Which is almost an anticlimactic conclusion (and not quite what “Aryan master race” ideologues seemed to have in mind, just in case you were wondering).

    It’s not unexpected, though, if the expansion was brought about largely by young immigrant males taking local wives (and then propagating wave-like, Kulturkugel-fashion, so that the genetic influence became more and more diluted away from the steppe, especially after the end of the Bronze Age) and establishing themselves as part of the local upper class, as is a plausible mechanism for the language spread. Ancient Europe was certainly a melting pot, much like Hungary, Anatolia or Latin America, without the immigrants even crowding out the natives in most regions (the best example for Überfremdung and Umvolkung in the modern world is ironically North America, where the dirty furriners are whites of European origin).

    Palaeolithic Europeans were a different matter, however, according to https://en.wikipedia.org/wiki/Cro-Magnon#Physical_attributes, even they were anything but uniformly brown-eyed. So I’m not sure what David is referring to here.

    Ötzi was brown-eyed. But then, Ötzi was apparently unconnected to the Yamnaya culture and instead to the first agriculturalists in Europe, who appear to have immigrated from Anatolia via the Danubian corridor, and who may have well imported a language themselves, but certainly not Proto-Indo-European.

    As for Gary Moore, I think he should keep out of the historical linguistic business and stick to what he is good at. Since the only Gary Moore I’m familiar with is unfortunately departed, I assume it’s not guitar wizardry – although I see now there’s a namesake wielding the axe for Mother’s Finest. I mean, I’ve been amused by the curiously Indo-European-ish appearance of the Lakota word for dog myself before, but it’s your garden-variety false cognate – neither under a direct inheritance nor borrowing scenario, the sound laws work out. And Aleut sabaakax is clearly a Russian borrowing.

  34. it’s your garden-variety false cognate

    Yup. Amazing how hard it is for people to accept the prevalence of coincidence.

  35. Late Neolithic people of southern Europe were dark skinned. They lacked genes for light skin present in current Europeans.

  36. Florian Blaschke says:

    SFReader: Huh, really? Must have missed that one. Interesting. Is the evidence from ADNA only or do we have evidence from artistic depictions (with the usual caveats about possible non-naturalistic features of depictions) pointing in that direction, too?

    Of course, that would only confirm the general picture – if steppe admixture was really responsible for the lightening of the skin of the natives of Southern Europe. Some of them are still relatively dark-skinned, but not exactly brown, of course.

    Steppe migrations can even have led beyond the historical domain of Indo-European languages – even in the Middle East, especially in the Levant, a lot of people look relatively European, with fair skin, light eyes and relatively light hair. In Israel, European admixture in the Ashkenazi-descended Jewish population is the obvious explanation, but I’m not sure if the same explanation works for the Lebanon, for example. Indo-European-speaking steppe immigrants could have been assimilated (back) to (local) Semitic languages in the Bronze or Iron Ages.

    Certainly a pushback against Indo-Europeanisation happened in various parts at various times and the resulting picture is quite complicated and messy. In Finland – according to http://kielievoluutio.uta.fi/lib/exe/fetch.php?media=bedlan2011_heikkila.pdf – the situation was actually inverted, in that Indo-Europeans belonging to the Corded Ware horizon and still on a Neolithic level of technology were assimilated by Uralic-speaking immigrants who already possessed Bronze-Age technology.

    This is a useful reminder that just because in the historical period the indigenous languages of a region are non-Indo-European does not mean that Indo-Europeanisation never happened there (for example in parts of the Caucasus region: it’s quite possible that the area of modern Georgia was once largely Indo-European-speaking). Indo-European migrations were not an irresistible wave steamrolling everything else; they were just one among countless similar movements in human history, despite their uniquely far-reaching consequences for the history of Europe and large parts of Asia. It’s also a useful antidote to Social Darwinist master-race and supremacist nonsense: Indo-European language and culture did not always win out everywhere (nor did their genes), and Indo-European culture was not even always and everywhere (technologically) superior in the first place.

  37. even in the Middle East, especially in the Levant, a lot of people look relatively European, with fair skin, light eyes and relatively light hair.

    At least a bit is due to the British and their bints.

  38. ^ I’ve usually heard this phenomenon ascribed to the Crusaders, though that sounds suspiciously neat to me.

  39. David Marjanović says:

    The blogger who suggested the relationship between the Native American admixture in Europe and the advent of Indo-European was Razib Khan

    Ah, thank you. Keep in mind, though, that science has moved on – and Khan’s views with it. In particular, start at the headline “European man, made and unveiled” a few screens down in this post by Khan from 2015.

    w- /wh- > p- ‘five’ *Proto Northern Iroquoian (PNI) *hwihsk / *PIE *pénkʷe

    That’s one example. I asked for regular sound correspondences.

    k- > kw- > m- ‘four’ *PNI *kaˈjeːɹih (Julian Charles) *kayeri (Lounsbury) / *PIE *kʷetwóres / Hittite ‘meyawes’

    …Where to even start!

    *kʷetwóres ~ *kʷétwores isn’t some kind of indivisible blob with three syllables. It has an etymology within IE. And it cannot be reconstructed to PIE in the widest sense; the Anatolian root (Hittite miyu-, meyu-; Cuneiform Luwian mawa-) is a completely different, unrelated one.

    As with several other sound changes you propose, [kʷ] changing into [m] is hard to imagine, especially in a language that retains [kʷ] “elsewhere” as Hittite and Luwian did.

    Equivalents in Indo-European are:

    These are indeed cognate with each other – and are derived from a PIE word that began with [kʲ], not with [s] or any other fricative. The reconstruction is *ḱwō (nominative), *ḱwon- (stem).

    Didn’t you notice that you only cited “satəm” forms?

    (And you mixed Luwian asuwa- in; it does not belong, it’s from *h₁eḱwos… zuwanis does belong.)

    Linguists are quick to discount claims of links between languages based on similarities in forms for ‘water’.

    No. Linguists are quick to discount claims of links between languages based on similarities in any single word. You need systematic correspondences across all the data.

    *PIE *h₂ekʷeh₂ ‘water’

    No: 1, 2, 3.
    …but it hardly matters, because the forms you cite under that heading aren’t cognate with that anyway!

    Latin alumen “alum,” literally “bitter salt,”

    No. There’s no trace of “salt” in that word.

    Japanese kawa ‘river’
    Japanese âme ‘rain’ (-m- > -w-)

    Wait. The same language does and does not shift *m to w? What’s the conditioning environment? If it’s the following -a vs. -e, why does the syllable we exist in the same language?

    David: Brown-eyed?

    As Razib Khan put it in the link at the top of this comment:

    Finally, in this panel for pigmentation they included a major SNP in [the] OCA2-HERC2 region. This locus is famous for being involved in blue-brown eye color variation, explaining 75% of the variance, and also exhibiting the third longest haplotype in the European genome. Naively projecting from these SNPs one could credibly argue that the ancient hunter-gatherers of Europe at the beginning of the Holocene were dark-skinned and blue-eyed! The Bronze Age European samples, which in this case are biased toward Northern Europeans, had a range of genetic variation equivalent to modern Southern Europeans. The people of the steppe did not seem to have blue eyes at all.

    75 isn’t 100, but it’s reasonably close. 🙂 “They” are Allentoft et al. (2015), “Population genomics of Bronze Age Eurasia”, one of the papers linked to from higher up in Khan’s post.

    Late Neolithic people of southern Europe were dark skinned.

    Not so much; but the Mesolithic hunter-gatherers all over Europe were – and they had blue eyes (again with the caveat that 75 isn’t 100).

  40. even in the Middle East, especially in the Levant, a lot of people look relatively European, with fair skin, light eyes and relatively light hair.

    Like the Solluba, aka Ṣleb. Their name has been folk-etymologized to mean they are descendants of the crusaders (cf. صَلِيب ṣalīb ‘cross’), but the word goes back to Akkadian times. Blench thinks they descend from a pre-Semitic Arabian population. Their language, if they have or had a distinct one, has not been recorded.

  41. Consistent patterns of phonetic … Proto-Tocharian … table with examples …

    Thank you Hat, marie-lucie, John Cowan and others. I was not wishing to discredit the comparative method in general. But all of your replies referred to the sounds of languages. With the Germanic family (that marie-lucie mentions) we have modern descendants we can listen to.

    With Tocharian there are no descendants. How can we be so sure how it sounded? John says With the scripts known, the pronunciation was known, … [marie-lucie makes a similar point.] But look at the variety of sound-values given to Latin script across the languages of Europe. In particular look at the higgledy-piggledy spelling in English. We can attest that English is Germanic by listening to it (and preferably ignoring the script).

    Is there strong archaeological evidence to corroborate the IE migration story of Tocharian? How did they get there? Where did they stop on the way? Did they bring the script with them? https://en.wikipedia.org/wiki/Tocharians seems full of speculation.

    What is the linguistic evidence against (for example) a hypothesis that Tocharian was an isolate (like Basque), and one day somebody turned up with a Brahmi script (and a Brahmi ear) and wrote down what they thought they heard?

  42. AntC: Even if the sound of English, or French, is unknown, a literal, Latinate or Italianate pronunciation would still firmly place the languages as Germanic and Romance.

  43. David Marjanović says:

    But look at the variety of sound-values given to Latin script across the languages of Europe. In particular look at the higgledy-piggledy spelling in English. We can attest that English is Germanic by listening to it (and preferably ignoring the script).

    Using the spelling and ignoring the sounds actually makes such comparisons easier, because the spelling is more conservative: it fits Middle English better than what has become of that.

    This applies to changes elsewhere, too. Wind is spelled the same in German, but while English has kept the pronunciation unchanged, German has turned w into [ʋ] or [v] depending on the region.

    What is the linguistic evidence against (for example) a hypothesis that Tocharian was an isolate (like Basque), and one day somebody turned up with a Brahmi script (and a Brahmi ear) and wrote down what they thought they heard?

    The evidence is in the words and the grammar!

    Language families don’t generally have distinctive sound systems, as you seem to believe. Basque is a case in point: it uses the exact same sounds Don Quixote used, with practically the same allophony even – it just arranges them into completely different words and sentences.

    (BTW, most or all surviving Tocharian writings must have been written by native speakers. The texts are sometimes long, and the spelling system is too consistent to be a pile of ad-hoc workarounds; this includes a consistent way to represent a vowel the Brahmi script couldn’t originally deal with.)

  44. David Marjanović says:

    Indeed, Tocharian had undergone some pretty drastic sound shifts from PIE. The last table in that section explains what a ś /ɕ/ is doing in the Tocharian B word for 5: it’s regularly derived from *-kʷe, so it cannot be directly cognate to the /s/ found in Iroquoian.

  45. primary archaeological suspect is the Afanasevo culture in south Siberia dated circa 3300 BC. It was carried by population of European extraction (previous inhabitants in the area were Mongoloid).

    the closest analogies were found in contemporary Yamna culture of the Pontic-Caspian Steppe (same guys who brought Indo-European to Europe)

    it should be added that the Afanasevo were only the first of many steppe cultures of European origin in the east. The Indo-European languages (Iranian, Tocharian and possibly Indo-Aryan as well) were displaced from eastern steppes only in the 1st millenium AD – by expansion of Turkic speaking tribes.

  46. Bravo, David, what words well put-together!

  47. Gary Moore says:

    SFReader write

    “sabaakax is borrowed from Russian.”

    Thanks for the input. This time was not central to my thesis, but it is nice to know. This still leaves the central issue of why so many of the North American and Indo-European forms for ‘dog’/’horse’ are so close. These forms are not likely to be derived from a wanderwort because they seem to only appear in IE and North American languages.

    Minus273 wrote:

    “So you mean that, for “salt” and “water”, Italian preserved the s- and -k-, lost in the American languages?”

    Deletion of the initial s- occurs fairly often in IE languages.

    The Proto-Celtic form for ‘river’, ‘awa’ is essential the same form as in some Iroquoian languages, and means ‘water’ in Nottoway, an Iroquoian language, while the *PC word for ‘water’ is ‘*akwā-‘. One Cherokee form for river, ‘equoni’ is close to *PIE *h₂ekw-eh₂- ‘river’ plus the locative particle ‘-ni’ . Lyle Campbell (1997), citing (Haas 1965) lists “Proto-Siouan *qʷaʔ / *qoʔ ‘water'” In Cherokee, the form for ‘water’, ‘ama’, is the product of a fairly recent sound shift from Proto-Iroquoian ‘awe’. (In fact, it is one of the few forms in Cherokee that contains ‘m’.) Forms for ‘water’ in IE show a wide variety of sound shifts in IE: for instance, Dacian/Thracian ‘apa’/’aba’. The shift to ‘m’ is attested in Greek potamos, ‘river (flowing water)’.

    To reiterate a key point, note that the forms for ‘lake/river’ and ‘wet’ also appear to be correlated.

    David Marjanović wrote:

    “No. There’s no trace of “salt” in that word.”

    Or is there? It’s plausible that *PIE ‘sal-‘ ‘salt’ shares a deep etymology with PIE root for ‘bitter’ ‘al-‘/. The two can be linked by addition/deletion of initial ‘s’.

    Also –

    “Wait. The same language does and does not shift *m to w? What’s the conditioning environment? If it’s the following -a vs. -e, why does the syllable we exist in the same language?”

    This is not a core of my hypothesis, but only a suggestion that IE shares an areal pattern with languages of Northeastern Asia/North America. For what it’s worth, though, the Japanese forms are probably based on borrow words from Ainu, so a specialist in these languages would have to sort that out.

  48. Gary Moore says:

    David Marjanović wrote –

    “Indeed, Tocharian had undergone some pretty drastic sound shifts from PIE. The last table in that section explains what a ś /ɕ/ is doing in the Tocharian B word for 5: it’s regularly derived from *-kʷe, so it cannot be directly cognate to the /s/ found in Iroquoian.”

    That wasn’t my point. I was indicating that the Armenian form could plausibly be mapped to Cherokee. In any case, the Armenian form demonstrates a correlation between ‘p-‘ to ‘h-‘ (and ‘hw-‘).

    Speaking of Tocharian, the forms for ‘old’ in the Toch A and Toch B differ considerably from the more typical forms based on *PIE *sen- ‘old’ which bears an obvious resemblance to equivalent forms in Dene-Yeniseian languages: ex. Ket sīn, Tlingit shaan (of people) and Navajo sání. However, the form for ‘old’ in Tocharian B closely resembles its equivalent in Iroquoian languages both phonetically and morphologically:

    Toch B ktsaitstse ‘old’ ( = -ktsai- + -tstse )

    Compare with Iroquoian equivalents:

    Mingo kekëhtsi (-KËHTSI- Verb Root.) katkëhtsistha’ ‘to become old’ -at- (Middle prefix, -këhtsi- /someone is old/, -st- Base suffix. -atkëhtsist- )

    Oneida: -kstʌ- ‘old, aged’; -kstʌhaˀ- ‘to become old’

    *PNI * { -kẽhtsi- } ‘be old’

    Indo-Iranian also appears to share yet a third form for ‘old’ with Mohawk, another Iroquoian language:

    Persian kohne ‘old’
    Kurdish kone ‘old’
    Gujarati kıhan ‘old’
    Mohawk aká:ion/iokaiòn:’on ‘old’

  49. Trond Engen says:

    Gary Moore: [various correspondences]

    You can find stunning lookalike words in any two languages. What people try to tell you is that it takes regular correspondences across a large enough part of the lexicon to rule out chance. Even better are regular correspondences of morphology, and the more complex the morphology, the better. Multiplicating probabilities brings the nominal threshold down.

  50. Gary Moore says:

    Trond Engen wrote:

    “You can find stunning lookalike words in any two languages. What people try to tell you is that it takes regular correspondences across a large enough part of the lexicon to rule out chance. Even better are regular correspondences of morphology, and the more complex the morphology, the better. Multiplicating probabilities brings the nominal threshold down.”

    Some of what got cut off in my earlier post dealt with parallels in syntax between IE and North American families:

    “In the 20th century, linguist C.C. Uhlenbeck used Siouan as a reference when analyzing syntax of early Indo-European languages, and more recently Brigette Bauer in her book “Archaic Syntax in Indo-European: The Spread of Transitivity in Latin and French” has noted parallels between syntax in early IE languages and the active-patient system of many North American languages.”

    As for regular correspondences, as I noted the key is shifts towards bilabialization of consonants. More recently, I have been focusing on correspondences between laryngeals because maintaining the existing scheme of laryngeals in PIE seems to be a major concern of Indo-Europeanists. There are a lot of forms that contain consonant clusters in *Proto-Iroquoian, *Proto-Northern Iroquoian and *PIE that seem to show related patterns’

    *PrIro *-Ɂnjõːhs- } “nose” / *PIE *nas-, hnéh₂s “nose” Ɂn > hn; hs- > h₂s

    *PrIro *-hniɹ- ‘be strong’ / *PIE *h₂ner ‘man’ -hn > h₂n-

    *PrIro *- õːnh- ‘be alive, life, live’ / *PIE *h₂enH₁- “to breathe” – õː > h₂e-; -nh- > nH₁

    *PrIro -hnhõtho- ‘feed, put in the mouth’ / *PIE *h₁oh₁s ‘mouth’ hõ > h₁o (Oneida -shuw- ‘hole’)

    *PNI -iɁtaɹ- / *PIE *h₁eh₂ter- ‘fire’ -Ɂt -> h₂t-

    *PrIro *-hnùːkòːl- / *PIE h₃nogʰ- ‘claw’ -hn- > h₃n-

    *PNI * -(a)hõht- ‘ear’ / *PIE *h₂ous- “ear” -(a)hõ- > h₂o-

  51. Split of Proto-Iroquoian is dated to 3800 years before present.

    Split of Proto-Indo-European occured 6000 years ago.

    You can’t compare them directly. There is a distance of 2200 years between them.

    Remote ancestor of Proto-Iroquoian circa 6000 BP likely looked and sounded very different from what you cite.

  52. David Marjanović says:

    This still leaves the central issue of why so many of the North American and Indo-European forms for ‘dog’/’horse’ are so close.

    But they’re not, as I just demonstrated. They’re random lookalikes due to a sound shift that happened within IE.

    Branches where this shift didn’t happen also have descendants of this *ḱwon-, for example Greek kyōn and English hound. (Except I don’t know where the -d comes from.)

    Deletion of the initial s- occurs fairly often in IE languages.

    But this* doesn’t happen at random. It happens across the board or not at all.

    * Actually a shift from [s] to [h], sometimes followed by general loss of [h].

    The Proto-Celtic form for ‘river’, ‘awa’

    This reconstruction seems to be contradicted by one of the links I posted. Have you read them yet?

    *PIE *h₂ekw-eh₂- ‘river’

    Oh, had my long comment not yet made it out of moderation when you posted this?

    The shift to ‘m’ is attested in Greek potamos, ‘river (flowing water)’.

    That would be the only word in the whole Greek language where anything had shifted to /m/.

    This is now how sound shifts work. And pot- is neither a prefix, nor does it mean “flowing”.

    It’s plausible that *PIE ‘sal-’ ‘salt’ shares a deep etymology with PIE root for ‘bitter’ ‘al-’/. The two can be linked by addition/deletion of initial ‘s’.

    Why and how would the same language both lose and not lose a sound in the same word?

    I was indicating that the Armenian form could plausibly be mapped to Cherokee.

    For what? You’re trying to relate PIE to Proto-Iroquoian, not Armenian to Cherokee.

    In any case, the Armenian form demonstrates a correlation between ‘p-’ to ‘h-’

    [p] > [pʰ] > [ɸ~f] > [h] is a common sequence of sound shifts the world over. You don’t need to establish that it’s possible.

    However, the form for ‘old’ in Tocharian B closely resembles its equivalent in Iroquoian languages both phonetically and morphologically:

    Yeah, it resembles it. I can’t work out regular correspondences, however; the sounds are in different orders and stuff…

    What are you trying to suggest? That Tocharian borrowed the word from farther east? Even for that the correspondences aren’t regular enough.

    has noted parallels between syntax in early IE languages and the active-patient system of many North American languages

    Well, yes. Most languages in the world are either nominative-accusative, ergative-absolutive, or active-stative. With so few possibilities, it’s inevitable that different languages pick the same ones independently.

    There are a lot of forms that contain consonant clusters in *Proto-Iroquoian, *Proto-Northern Iroquoian and *PIE that seem to show related patterns’

    That list gives one example for each of the correspondences you’re trying to establish as regular.

    Strangely, they all concern whole clusters instead of individual sounds. If I take them apart and look into them, I find a lot of questions:
    – So PIr *hs- corresponds to PIE *h₂s, *-hn to *h₂n- and *-(a)hõ- to *h₂o-. This looks like *h corresponds to *h₂. But then you claim that *-hn- corresponds to *h₃n-, not to *h₂n- as you said a few lines above, and you further say that *hõ corresponds to *h₁o, not to *h₂o- as you state a few lines below; also, *-nh- is to correspond with *nH₁. What is it, then?
    – This could of course very easily be explained if PIr *h simply represents a merger of the ancestors of PIE *h₁, *h₂ and *h₃… except that this totally fails to take into account your claims that *-Ɂt corresponds to *h₂t- and *Ɂn corresponds to “*hn” (where “h” means “I cannot tell whether this was *h₁, *h₂ or *h₃”).

    In short, you have two PIr phonemes corresponding to three PIE phonemes seemingly at random, everything-to-everything.

    Sequences of perfectly regular mergers, or mergers and splits, can indeed produce such patterns. But you haven’t even tried to propose any. You have a lot of work to do.

    (Your use of “>” is misleading. As used by historical linguists, it’s an arrow that indicates that the form on the left is ancestral to the one on the right. With PIr being much younger than PIE, that isn’t possible.)

  53. Bravo, David.

  54. A waste of time and effort in this particular case, I fear, but others will learn from it. You and marie-lucie have a clear expository style and endless reserves of patience that I can only envy.

  55. SFReader wrote –

    “Split of Proto-Iroquoian is dated to 3800 years before present.

    Split of Proto-Indo-European occured 6000 years ago.

    You can’t compare them directly. There is a distance of 2200 years between them.”

    Remote ancestor of Proto-Iroquoian circa 6000 BP likely looked and sounded very different from what you cite.”

    I work with what I’ve got to work with. Not much progress has been made in reconstruction of Native American languages beyond this time frame. This model may be related to the Macro-Siouan Hypothesis, which remains unproven. Analysis of data suggests that even if the Siouan and Iroquoian languages are indeed generically related, the precursor of IE likely diverged from North America languages after the Siouan-Iroquoian split. (BTW – the Macro-Siouan Hypothesis was first formulated by Latham, who was mainly an Indo-European scholar.) Anyone who has read Wallace Chafe’s work, “The Caddoan, Iroquoian, and Siouan languages” is aware of Williamson’s two papers linking Siouan (specifically Dakota) to Indo-European. Williamson’s papers were in response a paper linking Siouan to “Turanian” (“Ural-Altaic”):

    “Professor Roehrig, in his able paper on the Dakota, points out some very interesting analogies to Turanian languages. Others might be added. These similarities are chiefly in features common to I. E. and Turanian. …”

    I think that the actual split probably coincided with the 8.2 Kiloyear Event, when a sudden drop in temperatures combined with a rise in sea levels resulted in extensive population displacements. It appears that their may have been a crossing by a population or populations related to modern Caddoan, Iroquoian, and Siouan speakers out of the Americas which passed through Dene-Yeniseian speakers on the way in. In this model, Indo-European would have the same relationship to the languages of the hypothetical Macro-Siouan family that Yeniseian has to Dene languages of North America. (By way of background, two eminent 18th century ethnologists placed the urheimat of the Iroquoian people in the Pacific Northwest: Lewis H. Morgan claimed that the Iroquois were from the Puget Sound region and Horatio Hale cited the Columbia River Valley. Either location could have placed them in contact with Salishan speakers.)

    Some of the differences between Anatolian and the other branches of IE might be related to a different mix of forms of “Paleo-Siouan” and “Paleo-Iroquoian”. Specifically, the kinship terms of Anatolian look more like Siouan, and also the Hittite form for ‘wheel’ – ḫurki- may be related to *Proto-MVS Mississippi Valley Siouan *-ˈwrĩ’ ’twist, turn’ as well as *PIE *h2wrgí-lo- *‘of turning/twisting’←*h2w(o)rgi- ‘turning’. This could go a long way to explaining why the split between Anatolian languages and the remainder of IE appears to be more ancient than it really is. (The form for ‘wheel’ in the other IE languages is derived from *PIE *kwel- ‘‘spin, rotate’, which can be plausibly linked to Cherokee root ‘-gwal-‘ by relatively simple and straightforward sound shifts. Ex. Proto-Tocharian *kwə́kwlë (‘chariot, wagon’) / Cherokee ‘dagwalela’ (‘car, wagon’).

  56. languagehat wrote:

    “It’s your garden-variety false cognate

    Yup. Amazing how hard it is for people to accept the prevalence of coincidence.”

    Except that both the North American and IE forms probably share a related etymology. The North American forms are probably related to the Proto-Northern Iroquoian root *{ (-na)hs(vowel deleted)kw- } ‘(to be) domestic animal, prisoner, slave’, where the prefix (-na) indicates state of being. On the IE side, both the form for ‘horse’ and ‘dog’ is likely related to *PIE *sokw-yo- ‘follower, companion’ or *PIE *sekw- ‘follow’. The analogy is with Latin ‘sequeli’ – ‘band of followers’ (i.e., retainers or servants). Note that the Cherokee form for ‘horse’, ‘sogweli’, follows this pattern closely, with the addition of the noun ending ‘-li’. (The name of the famous Cherokee scholar ‘Sequoyah’ is a variant of this form.)

    In North American languages, when horses were adopted, typically the form for ‘dog’ was transferred to the new animal, and IE appears to have followed this pattern. Probably both the IE forms for ‘horse’ and ‘dog’ were formed from and earlier form that simply meant ‘dog’ by deletion of the initial ‘s-‘ or ‘sV-‘.

    BTW – If you looked at the candidate cognates I posted, you will notice that they constitute a fair portion of Pokorny’s Semantic Field 4, “Body Parts and Functions” – and I only listed possible matches for forms included on the Swadesh list. This is only a small part of the likely matches that I’ve identified. In all, I’ve identified about 50 to 60 candidate cognates out of the 207 item Swadesh List. There are an unusually large number of these “coincidences’ between IE and North American languages – well above the 10% figure typically cited for false cognates.

    One elegant feature of the hypothesis is that is provides “missing” etymologies in IE daughter languages:

    PI * { -hsnuːɹiɁ } ‘be fast, be quick’
    German ‘schnell’ ‘fast’
    Cherokee ‘gatsanula’ ‘fast’
    Oneida ( -snole- ) ‘fast’ (e.g. yosno·lé ‘it’s fast’)

    Ancient Greek: ” takho-, comb. form of takhos “speed, swiftness, fleetness, velocity,” related to takhys “swift,” of unknown origin.” Proto-Iroquoian ‘to run’, *-takh-.

  57. Is there anything, any conceivable piece of information, that would lead you to suppose that your theory might be mistaken?

  58. Gary Moore: The 10% figure quoted for false cognates was actually first proposed by Joseph Greenberg, with the two caveats that 1-This is the *lower bound* of the percentage of false cognates that will appear between two unrelated languages, which 2- Have *a similar number of phonemes*.

    Proto-Iroquoian is reconstructed by your own source, Julian Charles, as having had a total of eleven consonant phonemes only. As a result finding a Proto-Iroquoian root which bears some similarity with an Indo-European root whose meaning is broadly comparable in meaning, and with no attempt at finding regular sound correspondences (as David Marjanović has shown above) is thus very easy, indeed too easy, and, unfortunately, as a result, proves nothing.

    If you disbelieve this allow me to suggest to you the following exercise: start looking for Proto-Iroquoian roots which look like Modern English words/roots. You may be in for a surprise.

  59. David Marjanović says:

    endless reserves of patience that I can only envy

    It’s not like I could do otherwise.

    Like so:

    On the IE side, both the form for ‘horse’ and ‘dog’ is likely related to *PIE *sokw-yo- ‘follower, companion’ or *PIE *sekw- ‘follow’.

    …Excuse me. Do you read my comments?

    How could *ḱwon- or *h₁e*ḱwo- be related to *sekʷ-/*sokʷ- (same root, different ablaut grade)?

    In North American languages, when horses were adopted, typically the form for ‘dog’ was transferred to the new animal, and IE appears to have followed this pattern. Probably both the IE forms for ‘horse’ and ‘dog’ were formed from and earlier form that simply meant ‘dog’ by deletion of the initial ‘s-’ or ‘sV-’.

    1) Languages in North America had a very good reason for deriving “horse” from “dog”: there weren’t any horses in the Americas for a long time. When they were introduced, it made obvious sense to name them after the dogs whose functions they partly took over.

    2) Wait. Are you saying that *ḱwon- and *h₁e*ḱwo- are related to *sekʷ- by random deletion of *s, random inconsistent deletion of *e, and random transformation between the unitary labialized velar plosive *kʷ and the consonant cluster *ḱw (that’s the palatalized velar plosive [kʲ] followed by the labiovelar approximant [w])? Because if so, you can relate anything to anything with minimal effort. Reminder: if you can explain everything, you have explained nothing.

    In all, I’ve identified about 50 to 60 candidate cognates out of the 207 item Swadesh List. There are an unusually large number of these “coincidences’ between IE and North American languages – well above the 10% figure typically cited for false cognates.

    Regular sound correspondences. You need them.

    German ‘schnell’ ‘fast’

    But the verb schnellen refers to jumping, jerking movements. That’s why the click beetles are called Schnellkäfer.

    Ancient Greek: ” takho-, comb. form of takhos “speed, swiftness, fleetness, velocity,” related to takhys “swift,” of unknown origin.” Proto-Iroquoian ‘to run’, *-takh-.

    But Ancient Greek /kʰ/ came from PIE *gʰ and *ǵʰ, not from any consonant clusters – no, not clusters with “laryngeals” either (unlike in Indo-Iranian).

    which 2- Have *a similar number of phonemes*.

    PIE: /p t kʲ k kʷ/ is 5; adding /d gʲ g gʷ/ gives 9 ([b] may only have existed as an allophone of /p/); adding /bʱ dʱ gʲʱ gʱ gʷʱ/ makes 14 already. /w j l r m n/ makes 20, /s/ and *h₁, *h₂, *h₃ makes 24. That’s more than twice as many consonant phonemes as Proto-Iroquoian had.

    (It has been suggested that *h₁ was actually two separate phonemes, /ʔ/ and /h/. I’m not qualified to evaluate this; I don’t even know on what evidence this was suggested. Anyway, even without a /b/, we could have 25. – On the other hand, it has repeatedly been suggested that the distinction between the palatalized /kʲ gʲ gʲʱ/ and the plain /k g gʱ/ didn’t exist, putting us back down to 21 or 22; that, however, would make it really hard to explain certain distinctions in Albanian and Luwian, and certainly puts up a challenge to explaining why /gʲ/ was regularly borrowed into Proto-Uralic as /j/.)

  60. start looking for Proto-Iroquoian roots which look like Modern English words/roots

    It’s indeed what he did. He looked not for lookalikes in *PIE, or any single Indo-European language, but across all the daughter languages, thus greatly multiplying the resulting “cognates”.

  61. It has been suggested that *h₁ was actually two separate phonemes, /ʔ/ and /h/. I’m not qualified to evaluate this; I don’t even know on what evidence this was suggested.

    I’m not sure there is any evidence; I think it’s just that some people feel PIE should have had both of those phonemes. (For what it’s worth, I tend to think so, not that I can give any terribly cogent reasons.) That leaves your choices as basically either “h1 = /ʔ/ and /h/” or the “h1 = /ʔ/, h2 = /h/” (unpopular but I think possible; in Hebrew, /h/ patterns with the pharyngeals in causing epenthesis of [a] etc., so it’s not too big of a problem to assume PIE */h/ could color [e] to [a]).

  62. Actually, English dog and Latin can-is are perfect cognates, if we posit that in Latin *d>k, *o>a and *g>n. If the sound changes seem too odd, consider that Armenian also has some weird sound changes. Or invoke metathesis, together with the more reasonable *d>n and *g>k.

  63. Origin of English dog according to Piotr Gąsiorowski

    http://www.academia.edu/1499785/The_Etymology_of_Old_English_docga

  64. I was just toying with absurdity. If you presume unconfirmed sound changes you can relate anything to anything.

  65. (Of Tocharian as an IE language) …it was recognized as such almost immediately, unlike Albanian, which hid its IE origins under a heap of borrowings.

    Don’t you mean Armenian? In Albanian, most borrowings are Latin and Slavic. But even in Armenian the proportion of evidently non-IE loans in its old vocabulary isn’t high. Before 1875 (the annus mirabilis of IE studies) Armenian was generally regarded as an Iranian language (that is, an IE one, though from the wrong branch). What makes both of them a bit special is their long separate histories, during which sound changes not shared with any other group have accumulated, producing highly exotic regular correspondences. For example, Albanian sy ‘eye’ and Armenian erku ‘two’ are perfectly good IE words; they just don’t look it.

  66. (French øj ~ zjø ‘eye’ and o ‘water’ are rather nice IE words; they just don’t look it.)

  67. Not sure about o though: *akwā is restricted to a small cluster of western branches. But you are right. French is every bit as odd as Albanian, only more familiar to most people.

  68. David Marjanović says:

    I’m not sure there is any evidence; I think it’s just that some people feel PIE should have had both of those phonemes. (For what it’s worth, I tend to think so, not that I can give any terribly cogent reasons.)

    I’ve found that in many positions where *h₁ is reconstructed, either [ʔ] or [h] is really hard to pronounce while the other is no problem… of course they could have been allophones even so…

    “h1 = /ʔ/, h2 = /h/” (unpopular but I think possible; in Hebrew, /h/ patterns with the pharyngeals in causing epenthesis of [a] etc., so it’s not too big of a problem to assume PIE */h/ could color [e] to [a]).

    [h] indeed colors vowels in some languages. But assuming that *h₂ was [χ] explains a lot more things, for instance its written representations in Anatolian languages.

    Or invoke metathesis, together with the more reasonable *d>n and *g>k.

    Obligatory joke about how God made dog in his mirror image.

  69. Obligatory joke about how God made dog in his mirror image.

    And then god unproblematically became cat in Germanic (via the *a/*o merger and Grimm’s law).

  70. For what it’s worth, though, the Japanese forms are probably based on borrow words from Ainu, so a specialist in these languages would have to sort that out.

    Just for the record, there’s absolutely no reason to suppose that Japanese /kawa/ or /ame/ are borrowed from Ainu (even if you very generously expand the definition of “Ainu” to “any language (i) in the same family as the direct ancestor of Ainu, and (ii) spoken on the Japanese archipelago after the arrival of an ancestor of Japanese”).

    It’s not impossible that /kawa/ and /ame/ ultimately derive from a language that was ancestral to Ainu, through borrowing or common descent, but I can’t recall ever seeing a plausible argument (with regular correspondences, etc.) to that effect.

  71. But assuming that *h₂ was [χ] explains a lot more things, for instance its written representations in Anatolian languages.

    Yes, that’s certainly true — there are Ugaritic spellings of Hittite proper names which write the Hittite fricative as a velar, though Ugaritic had /h/.

  72. David Marjanović says:

    Cats certainly are autotheists…

    there are Ugaritic spellings of Hittite proper names which write the Hittite fricative as a velar[/uvular]

    Oh, good to know!

  73. Seconding what Matt said about the Japanese case, and adding that this is a good example of why knowledge of a language’s internal phonetic and orthographic history is necessary for comparative work.

    For all of Japanese history until the last fifty-odd years, kawa “river” was written not かわ, but かは, reflecting the origin of the present /w/ in something like a bilabially fricative /h/ between two /a/ vowels (in other intravocalic environments it was regularly dropped). The /h/ in turn is thought to derive from the lenition of an original /p/. So, /kapa/ to /kaha/ to /kawa/. There are of course ways to get from “m” to “p”, but the mechanism has to be explained, and systematically demonstrated, and the link to ame justified by more than the semantic appeal to both words being about forms of water.

  74. Trond Engen says:

    (French øj ~ zjø ‘eye’ and o ‘water’ are rather nice IE words; they just don’t look it.)

    Both øj and o: look Scandinavian, though.

  75. Greg Pandatshang says:

    Anyone have an idea of how atypical PIE is in its rigorous contrast of /kʷ/ with /kw/? I don’t recall seeing that elsewhere (such as in any of the daughter languages).

  76. David Marjanović says:

    contrast of /kʷ/ with /kw/

    Has *kw /kw/ actually been reconstructed? I – for what that’s worth – am only aware of *ḱw /kʲw/.

  77. Greg Pandatshang says:

    Boy howdy, you’re right, that’s a Joker-grade boner on my part. On the other hand, ḱ might be phonetically /k/ if we could imagine some other explanation for the Uralic data, so perhaps there was /kw/. I suppose, if /kw/ contrasting with /kʷ/ is really atypical, that by itself could be a (weak) argument against identifying ḱ as /k/.

    Assuming that ḱ is in fact /kʲ/, well, can we think of other examples of languages contrasting /kʷ/ with /kʲw/?

    Given the much greater frequency ḱ of relative to k (and yet we know for sure that in some IE languages, ḱ does not behave like a plain velar), I suppose mostly everyone finds it hard to believe that there wasn’t something interesting going on. I’m not attached any particular theory of when and how the interestingness happened. I wonder if early PIE might have had a voiceless velar only and voiced palato-velars only, like in Arabic … with later complexities caused by different dialects’ attempts to level that out.

  78. David: Has *kw /kw/ actually been reconstructed?

    Yes, for example initially in *kwah₂t- ‘ferment, bubble, turn sour’. The exact shape of the root is somewhat uncertain; the reconstruction above is the one I favour. The Lexicon der indogermanischen Verben has *kwath₂-, but such a form makes little sense to me, since it doesn’t account for the attested reflexes. I tentatively accept Birgit Olsen’s “laryngeal preaspiration” hypothesis, which predicts alternations like *kwatʰ-V-/*kuh₂t-C-/*kwah₂t-C- from underlying //kweh₂t-//, which is exactly what we see here: cf. OCS kvasъ, vъ-kyse (hence of course English kvass), Ved. kváth- ‘boil’, Goth. ƕaþiþ ‘foams’.

    There may also be examples of medial *-kw- at morpheme junctures, e.g. (possibly) Greek lákkos ‘pond, cistern’, if from *lakw-o- (the lake, loch root).

  79. David Marjanović says:

    …Fascinating.

    Given the much greater frequency ḱ of relative to k (and yet we know for sure that in some IE languages, ḱ does not behave like a plain velar), I suppose mostly everyone finds it hard to believe that there wasn’t something interesting going on.

    “Everything is the way it is because it got that way.”
    – D’Arcy W. Thompson (1917): On Growth and Form [a book about development biology]

    PIE, as usually reconstructed, had an odd consonant system and an odd vowel system. How could it have gotten that way?

    Start from a language with only one row of velars, rather than three, and a boring five-vowel system. Then blame the features of the vowels on the nearest consonants:
    */ki/ > */kʲə/
    */ke/ > */kʲə/
    */ka/ > */kə/
    */ko/ > */kʷə/
    */ku/ > */kʷə/

    …and suddenly the palatalized and the labialized vowels are each twice as common as the plain ones.

    It can’t have been quite that simple. But sound systems with such frequency distributions do exist. In Ubykh, the plain velars were so rare that they ended up merging into the palatalized ones, rather than the other way around! (Attested Ubykh had plain velars again – in loanwords.)

    In short, it’s expected that *ḱ was more common than *k.

  80. Greg Pandatshang says:

    Aw heck, I still think that counts as something interesting going on.

  81. marie-lucie says:

    DM: Start from a language with only one row of velars, rather than three, and a boring five-vowel system. Then blame the features of the vowels on the nearest consonants:
    */ki/ > */kʲə/
    */ke/ > */kʲə/
    */ka/ > */kə/
    */ko/ > */kʷə/
    */ku/ > */kʷə/

    …and suddenly the palatalized and the labialized vowels are each twice as common as the plain ones.

    Surely you mean this as an example of what not to do.

  82. David, I never saw your reply to my objections when you brought up this theory a couple of years ago. If I understand correctly, the idea is this:

    1. Some stage of pre-PIE had 5(+) vowels and a single dorsal series. It had ablaut patterns, but (I’m not clear about this part) only some roots participated in them.

    2. Frontness and roundness of vowels were reinterpreted as due to adjacent consonants, so that next to a front vowel you got a palatalized consonant and next to a round vowel you got a labialized consonant; these features were phonemicized into the consonants and away from the vowels, with some vowel mergers occurring as a concomitant. (I’m not sure what the specific mergers are supposed to have been.)

    3. All(?) consonants now came in plain, palatalized, and labialized sets, but the distinctions were subsequently done away with except in the case of the dorsals.

    4. /a/ still survived from the pre-reorganization stage, but it later unrelatedly changed into (or merged into?) /o/.

    This all seems like a really high price to pay in terms of hard-to-verify conjectures just to explain something that isn’t actually all that weird to begin with (the three dorsal series). I mean, nothing here is impossible, but it doesn’t seem preferable to a hundred other stories that could be concocted about pre-PIE sound systems with a bit of imagination. It posits a lot more weirdness than it accounts for, and comes with its own difficulties (e.g. **/k/ splits into */kʲ/ and */kʷ/ which then presumably alternate in ablaut paradigms, but don’t get leveled?).

    Maybe I’m misunderstanding parts of the theory, and presumably Nostraticists claim to have comparative evidence supporting it, whose value will depend on how much you buy into Nostratic, but it looks like a flight of fancy to me.

  83. Greg Pandatshang says:

    David Marjanović can, of course, speak for himself, but I took his hypothesis as neither an example of what to do nor what not to do, but simply as another brief outline of a hypothesis that cannot be decisively proven or disproven because it butts up against the limits of our data. I suppose that’s a flight of fancy in some sense. If someone came up with a hypothesis that explained all the data in a straightforward way and didn’t create any of its own difficulties, then we could say that there’s a good chance it’s true. But, failing that, a hypothesis that is simple in some respects but requires some amount of epicycles to work might be true, but it’s hard to be very certain about it. I still think my /g/-fronted-first-like-in-Arabic idea is cute, but that’s definitely a flight of fancy with not a whole of lot falsifiability going for it.

  84. Gary Moore says:

    “Yeah, it resembles it. I can’t work out regular correspondences, however; the sounds are in different orders and stuff…

    What are you trying to suggest? That Tocharian borrowed the word from farther east? Even for that the correspondences aren’t regular enough. …”

    Look again: the Tocharian B form and the Mingo/Seneca form also share morphological features as well as a similar sound pattern: the base suffix ‘-tstse- vs ‘-st-‘.

    I don’t think this was a borrow word: rather, I think that Tocharian retained the original form and more common *PIE form, ‘*sen-‘, ‘old’ was most likely borrowed at some point from Dene-Yeniseian languages.

    No, I have not worked out regular correspondences yet to the degree that neogrammarians would like. I am still in the “collecting forms lists” phase, but based on what I am seeing, there is a strong case for the deep affinity of IE with North American languages. I am hoping that someone with more time, resources, and specialized skills (and who can get paid for it) will take a crack at working out regular correspondences.

    And BTW, the Tocharian A form for ‘man, person’ is ‘enkwe’. In *Proto Northern Iroquoian, it’s ‘-õkweh’ (‘be a person’) Tuscarora ‘-ək̃ wɛh’, Cayuga ‘-õkweh’, Seneca ‘-ɔk̃ weh’. and Mohawk ‘ón:kwe’. I’m not just posting random associations: there are patterns to these correspondences.

    According to a recent Chinese paper describing the results of aDNA studies of remains in Xinjiang: “… Y chromosome haplogroups of Hunnu remains included Q-M242, N-Tat, C-M130, and R1a1. Recently, we analyzed three samples of Hunnu from Barköl, Xinjiang, China, and determined Q-M3 haplogroup. Therefore, most Y chromosomes of the Hunnu samples examined by multiple studies are belonging to the Q haplogroup. Q-M3 is mostly found in Yeniseian and American Indian peoples, suggesting that Hunnu should be in the Yeniseian family. …”

    Because Q-M3 is widely believed to have originated in North America, it’s presence could be interpreted as evidence of a back-migration from the Americas. While this finding in no way directly proves that IE languages are derived from North American languages, there can be no doubt that early IE speaking populations were in contact with populations descended from North Americans.

  85. Gary Moore says:

    David Marjanović commented:

    “PIE: /p t kʲ k kʷ/ is 5; adding /d gʲ g gʷ/ gives 9 ([b] may only have existed as an allophone of /p/); adding /bʱ dʱ gʲʱ gʱ gʷʱ/ makes 14 already. /w j l r m n/ makes 20, /s/ and *h₁, *h₂, *h₃ makes 24. That’s more than twice as many consonant phonemes as Proto-Iroquoian had.”

    The model does not suggest that *PIE is a direct lineal descendant of the ancestral form of Proto Iroquoian (which, for the sake of convenience, I refer to as “Paleo-Iroquoian”). Rather, *PIE may have been derived from a creole language that arose out of an ancient sprachbund of the Pacific Northwest Coast region that preceded the current Northwest Coast Sprachbund. This ancient sprachbund apparently involved ancestral versions of Iroquoian, Siouan, and Salishan languages, and this creole predecessor of *PIE borrowed features from all three families. There appear to have been at least two principal lexifers, Paleo-Iroquoian and Paleo-Siouan, with Paleo-Iroquoian being the primary. The sound system, as Beekes has suggested, could have been derived from a reduced subset of “Paleo-Salishan”.

    Basque is widely acknowledged to have incorporated fossilized pronominal prefixes, and IE languages appear to have them too. One indicator that *PIE is a creole is the apparent presence of fossilized pronominal prefixes such as Iroquoian “Mohawk-like” ‘o-‘ and “Cherokee-like” ‘a-‘ preserved in forms for body parts as well as some other forms. (In fact, it provides another way of classifying IE languages, like the “Satem vs. Centum” division, with Anatolian with its Siouan-derived forms for some body parts as the exception.) Roots of many of the IE forms in daughter languages appear to based on the third-person singular version of Iroquoian forms, as though the ancestral “pre-Proto Indo-Europeans” had difficulty in learning the complex Iroquoian system of pronoun prefixes and stuck with the third-person form, adapting an alternative grammar from what may have been a substrate language.

    An analogy to the proposed model is the Garifuna language of Central America. While core language is still essentially a composite of Arawkan and Carib, through a series of historical accidents, the speakers of Garifuna are genetically mostly African.

    BTW – Here is a sign in the language that Beekes suggested could have provided the sound inventory for Proto Indo-European. Does the form remind anyone of an alternative word for “to be motionless” in Indo-European languages?

    https://en.wikipedia.org/wiki/File:Bonaparte_-_Secwepemctsin_stop_sign.jpg

  86. Rather, *PIE may have been derived from a creole language that arose out of an ancient sprachbund of the Pacific Northwest Coast region that preceded the current Northwest Coast Sprachbund.

    Reconstructed PIE looks like anything but a creole language.

  87. marie-lucie says:

    Gary Moore: based on what I am seeing, there is a strong case for the deep affinity of IE with North American languages.

    You are not the first person to mention a resemblance of SOME North American languages to IE. The great linguist Edward Sapir proposed a classification of the 60-odd North American language families recognized in the 19th century into 6 superstocks or “phyla”. One of these was the “Penutian phylum”, comprising approximately 15 families of the West Coast, from California to British Columbia. Sapir did not base his classifications on any items of common vocabulary or even on common or at least widespread pronouns, but on general morphological structures which reminded him of IE rather than of any other North American languages. For example, several of the “Penutian” families have complex verb formations including vocalic ablaut (as in English sing, sang, sung).

    I don’t have the relevant texts at hand but the descriptions of the “phyla” should be easily findable on the internet, including the one(s) comprising Iroquoian, Siouan and others that you have been quoting.

    I should add that Sapir’s “phyla” were not generally accepted by other linguists and that “Penutian” in particular is considered doubtful. The main reason for the non-acceptance is that most linguists researching the families and phyla in question have been looking at vocabulary items, forgetting that the “comparative method” uses phonological correspondences between the words of languages already considered to be related, and that relatedness is most easily detectable through common morphology (both structure and individual morphemes). Actually there are a fair number of morphological resemblances within the “Penutian phylum”, although much more work is needed to determine its structure and even its membership, before considering what it could itself be related to.

  88. marie-lucie says:

    But the fact that some Penutian languages have some structural resemblances with some IE languages does NOT mean that the two groups should be considered related. Morphological structures, although more varied than syntactic ones, can resemble each other without the languages they appear in being related. The opposite is also true: once related languages can lose some of their features in such a way that the relatedness is no longer discernable.

  89. @David Marjanović

    “Together, the recent papers on the genetics of the Corded Ware/Battle Ax people make it highly likely that the non-Anatolian branch of IE expanded with the brown-eyed, lactose-tolerant hordes of the Yamnaya culture.”

    Just to clarify about the genetics of the Yamnaya. It looks like they were actually NOT lactose-tolerant. Surprise, surprise!

  90. Piotr Gąsiorowski commented:

    “Reconstructed PIE looks like anything but a creole language.”

    Note that I stated that “*PIE may have been derived from a creole language”.

    With all due respect to Professor Anthony, there are some features that IE inherited from its predecessor that point to its ultimate origin from a creole:

    1) IE forms appear to be derived from two different language stocks: Iroquoian and Siouan. Consider the fact that there are a number of cases of multiple *PIE forms for the same item such as the first person pronoun, the numeral ‘1’, ‘water’, ‘fire’, ‘man, person’, etc. I have found that in many of these cases, one *PIE form appears to correspond to an Iroquoian form and the other to a Siouan form.

    2) The precursor of *PIE shows drastic reduction in elements from its hypothetical parent lexifer. Pronoun prefixes in IE forms derived from Iroquoian, when present, are reduced to the third person singular only. There are a few examples other than body parts already listed in earlier postings. Here is another example:

    *PIE ḱewero- ‘wind’

    This form looks very much “in family’ with Iroquoian equivalents, taking into account the ‘r’ > ‘l’ shift.

    *PNI * { -wɹ- } ‘air, wind’
    Mohawk ówera ‘wind (breeze)’
    Cherokee u²no²le ‘wind, air’
    Oneida owe·lá ‘wind, air’
    Mingo kææha’ ‘wind to blow’

    The ‘ka- ‘ prefix is the 3rd person inanimate pronoun prefix for C-Stem Class Verb Bases in Mingo, and the *PIE form appears to have retained this element from the “Paleo-Iroquoian” parent language.

    To cite some examples in daughter languages for an alternative form for ‘head’

    Mohawk ohsò:kwa ‘head’
    Cherokee asgoli ‘head, peak’ (also uska – ‘detached head, skull’)
    Toch B āśce ‘head’ (Dropped noun ending)
    Hittite sag ‘head’ (Dropped pronoun prefix and noun ending)
    Old Norse skalli “a bald head, skull’ (retained noun ending but dropped pronoun prefix)
    Swedish skulle

    As can be seen, the IE forms are simplified from the Iroquoian forms.

    3) Anthony also remarks that repeated forms are typically used in creole languages as an intensifier. An instance in *PIE is the non-Anatolian word for ‘wheel’ *kwékwlo-s, which is formed by reduplication of the base *kwel- ‘turn’. In this case, reduplication is used to indicate repetitive action. In turn, *PIE *kwel- can be plausibly mapped to its Cherokee equivalent, ‘-gwal-‘, by relatively simple and straightforward sound shifts. Compare Proto-Tocharian form *kwə́kwlë ‘chariot, wagon’ to Cherokee dagwalela, ‘wagon, car’ (Feeling). Comparable forms in Proto Northern Iroquoian given by Charles Julian are * { -kaɹhateni- }, * { -kaɹhatenj- } ‘roll, turn around’ ( limited to Iroquoia ) and PNI * { -kaɹhatho- }, * { -kaɹhathw- } ‘invert, roll over, turn over’. (Of course, the resemblance to common IE forms such as ‘car’ and ‘cart’ are remarkable.)

    Although some may object to these proposed correspondences above, claiming that Native Americans ‘did not have the wheel’, it should be pointed out that hoop and stick games involving rolling disks or hoops on edge were widely popular in North America, so Native Americans were indeed familiar with the concept of the wheel.

  91. Note that I stated that “*PIE may have been derived from a creole language”.

    OK. PIE looks like anything but a language “derived from a creole”. If there’s anything creoles typically lack (or almost lack), it’s an elaborate system of inflections and complex, semantically and formally opaque derivational rules. PIE abounds in both. Creoles also tend to have small and typologically unremarkable phoneme inventories: can you show me a creole language with more than a dozen stop phonemes?

    Lookalikes don’t count as cognates until you can show that they display robust, systematically recurring phonological correspondences and morphological homologies. Random “root” resemblances are easy to find between any pair of languages.

  92. PIE accentual mobility, too, is too difficult for second-language speakers to handle.

  93. I repeat my earlier question to Gary: Is there anything, any conceivable piece of information, that would lead you to suppose that your theory might be mistaken?

  94. It would be a waste of time and space to point out all the errors in Gary’s data. I’ll just give one example: there are several Hittite words that could be translated approximately as ‘head’ — first and foremost, haršar (oblique stem haršn-) — but there’s no Hittitesag” ‘head’. SAG is a sumerogram employed instead of the corresponding Hittite word. Gary, will you start arguing now that Sumerian is also an Iroquoian-based creole?

  95. Gary, will you start arguing now that Sumerian is also an Iroquoian-based creole?

    Now that you mention the possibility…

  96. I think there is some miscommunication here (Sigh. Doctors make the worst patients, and linguists can’t communicate…): Gary seems to be using “creole” with the meaning “lexically mixed language”. Now, while some creoles do indeed have mixed vocabularies, this is the exception rather than the rule: most creoles derive the bulk of their lexicon from a single language: most Jamaican Creole lexemes come from English, most Haitian Creole lexemes come from French, and so on. Conversely, of course, most languages with mixed vocabularies are not creoles.

    Piotr: While I do not believe that Gary has a case, I must point out that creole phoneme inventories can be quite rich: Chinook Jargon (A pidgin which nativized at least once) had a substantially richer consonant system than Proto-Indo-European, with fifteen stop phonemes (Which is quite banal in North American Pacific Northwest languages, of course…).

    Gary: while I do not believe you have a case, I at least am trying to be fair. I had pointed out upthread that since Proto-Iroquoian had a mere eleven consonant phonemes, the odds of finding a coincidental look-alike between Iroquoian and Indo-European were disturbingly high (Incidentally, another language with a very poor consonant inventory is Basque, which, as Luis Michelena (AKA Mitxelena) pointed out, did make finding alleged “cognates” of Basque words in other languages very, very easy).

    More broadly, unless you show that the similarities you have listed between Indo-European on the one hand and various Native American language families (including Iroquoian) *cannot have arisen by chance*, then I’m afraid you are wasting everybody’s time. including your own. And our cyberhost’s question (which I believe deserves an answer) as to what it would take for you to accept that your theory is wrong goes to the heart of the matter: as Stephen Jay Gould once wrote, “unbeatable systems are Dogma, not Science”.

  97. Chinook Jargon (A pidgin which nativized at least once) had a substantially richer consonant system than Proto-Indo-European, with fifteen stop phonemes (Which is quite banal in North American Pacific Northwest languages, of course…).

    I stand corrected, thanks. 🙂

  98. marie-lucie commented:

    “You are not the first person to mention a resemblance of SOME North American languages to IE. The great linguist Edward Sapir proposed a classification of the 60-odd North American language families recognized in the 19th century into 6 superstocks or “phyla”. One of these was the “Penutian phylum”, comprising approximately 15 families of the West Coast, from California to British Columbia. Sapir did not base his classifications on any items of common vocabulary or even on common or at least widespread pronouns, but on general morphological structures which reminded him of IE rather than of any other North American languages. For example, several of the “Penutian” families have complex verb formations including vocalic ablaut (as in English sing, sang, sung). ”

    BTW, Siouan languages also exhibit a form of ablaut, but of the final vowel. See this paper by Robert Rankin:

    http://udel.edu/~pcole/fieldmethods2010/CatchingLanguageFelixKAmekaAlanCharlesDenchNicholasEvans2006book/The%20interplay%20of%20synchronic%20and%20diachronic%20discovery%20in%20Siouan%20grammar-writing%20Robert%20L.%20Rankin.pdf

    In my earlier posts, I had also noted two papers describing possible links between Siouan (Dakota) and IE by Andrew Williamson, published in 1881 and 1882, respectively. This papers unfortunately were published in local American journals and received no visibility to European scholars.

    The idea of an explicit link between the Iroquoian languages and Indo-European is not new: the Jesuit missionaries – the same people who pointed out the parallels between Sanskrit and Indian languages to the classical languages of Europe – remarked on the resemblance of Iroquoian languages and Ancient Greek early on. Father Joseph-François Lafitau, who noted strong similarities between words in languages of ancient Anatolia and equivalents in Iroquoian languages, remarked: “For besides those which I have mentioned, I can cite more of them which, without alteration, are purely Huron and Iroquoian; and others which having all the construction and flavor of these languages, can be found in them with slight changes.” Lafitau may have been exaggerating, but the resemblances are often striking, as in the case of Hittite (unknown to Lafitau) takku ‘if’ and Mohawk tóka’. ‘if’. Speaking of his near contemporary, Fr. Jean-Andre Cuoq, Horatio Hale wrote: “The resemblances of these Indian languages to the Greek struck him, as it had struck his illustrious predecessor, the martyred Brebeuf, two hundred years before.” The idea that Iroquoian languages were in some way derived from ‘a form of ancient Greek’ was taken seriously in the 19th century, and the pioneering linguist Albert Gallatin wrote a paper comparing the middle voice in Ancient Greek and Iroquoian, and one scholar, Giles Yates, even presented a paper proposing correspondences between Ancient Greek and Iroquoian sounds, according to H. R, Schoolcraft, who was himself skeptical of the proposal. Unfortunately, speculation centered around the fanciful vision of some Bronze Age transatlantic migration and not the broader issue of the possibility of a very ancient relationship of Iroquoian languages to the general IE family. However, Thomas Jefferson, who was a serious student of Native American languages, suggested that North America may have been a source of Eurasian languages, and as previously noted, some major IE scholars have remarked on possible connections between North American languages and Indo-European. With advances in population genetics that appear to show evidence for a back flow of Native American populations into Eurasia, possibly reaching even as fall as eastern and central Europe, such links can no longer be peremptorily ruled out. By the era of proposed for the split of the predecessor of IE from North American languages – that is, about 8,000 years ago – North America had been settled for at least 5,000 years and possibly longer, so there is no valid scientific reason for preferring a west-to-east migration from Siberia to North America over an east-to-west migration from Alaska back into Eurasia by that time.

  99. Etienne commented:

    “Gary: while I do not believe you have a case, I at least am trying to be fair. I had pointed out upthread that since Proto-Iroquoian had a mere eleven consonant phonemes, the odds of finding a coincidental look-alike between Iroquoian and Indo-European were disturbingly high (Incidentally, another language with a very poor consonant inventory is Basque, which, as Luis Michelena (AKA Mitxelena) pointed out, did make finding alleged “cognates” of Basque words in other languages very, very easy).”

    One thing that give me a high degree of confidence in many of my proposed correspondences is that the forms not only merely monomorpheme ‘sound-alikes’, but have similar structure. For example, the *PIE form for ‘leg’, ‘’krosko’ with *Proto Northern Iroquoian (*PNI) ‘kohsinko?t’ ‘leg’/’knee’/’ankle’:

    1 2 3 4 5 6 7
    *PIE k ro k s – ko –
    *PNI k o h (k>h) s in ko Ɂt

    As you can see, the *PIE form closely resembles its *PNI counterpart, with possibly some limited loss (?) of morphological elements. To emphasize, this is not merely an isolated example: there are plenty in the core IE lexicon. Correspondences between Native American and *PIE and other IE languages are not, as Schoolcraft remarked, a few “disjointed examples in the lexography” – there are high densities of candidate correspondences in many semantic fields – especially body parts, numbers, and personal pronouns. Moreover, many correspondences display distinctive patterns which suggest that *PIE retained vestiges of an earlier polysynthetic morphology involving incorporated pronouns.

    Another example is *PIE *H₁dónt- and *PNI * oˈnoɁtsaɁ ‘tooth’

    0 1 2 3 4
    H₁d ó n t-
    oˈ noɁ ts- a

    As can be seen, while the *PIE form appears to have a prefix of unknown origin, it is otherwise is quite similar to the *PNI form structurally.

    Etienne also commented:

    “Piotr: While I do not believe that Gary has a case, I must point out that creole phoneme inventories can be quite rich: Chinook Jargon (A pidgin which nativized at least once) had a substantially richer consonant system than Proto-Indo-European, with fifteen stop phonemes (Which is quite banal in North American Pacific Northwest languages, of course…).”

    This brings up an interesting point: the Pacific Northwest region of North America has been particularly productive of trade languages and pidgins, and not just the Chinook Jargon. As Beekes has pointed out, the *PIE sound system could have been derived from a reduced set of the sound system of the Salishan language Shuswap.

    During the early stages of European settlement, many Europeans in the Pacific Northwest adopted Chinook Jargon and even spoke it as the primary home language. This could provide a peaceful model for how europeanoid steppe dwellers came to speak a language borrowed from populations derived from Native Americans on the eastern fringe of the steppes.

    The Pacific Northwest is of course linguistically diverse. Whether or not Siouan or Iroquoian populations were once part of this mix is impossible to definitively prove, but for what its worth, two of the most eminent Iroquoian scholars of the 19th century, Horatio Hale and Lewis Hunt Morgan, placed the urheimat of the Iroquoian languages in the Pacific Northwest.

  100. languagehat commented:

    “I repeat my earlier question to Gary: Is there anything, any conceivable piece of information, that would lead you to suppose that your theory might be mistaken?”

    The theory that PIE is somehow autochthonous and arose from a west Eurasian linguistic isolate is questionable, and I think that’s the real issue here. The model suggesting links between North American languages and PIE falsifies the theory that IE is a monogenic language. The distinctive features of IE probably arose as independent language innovations or borrowings after the split from North American languages which most likely took place around 8,000 BP, and 6,000 or 5,000 years BP when the IE “baseline” was established.

    As I have pointed out, I am not the only person to suggest the likelihood of links between IE and North American languages. Accomplished linguists have separately suggested links between IE (specifically, Greek) and Iroquoian, and IE and Siouan, and all I have done is to synthesize these hypotheses and place them in the context of recent genetic findings.

    The smug attitude of IE scholars towards IE-American connections is illustrated by Carl Buck’s snide remarks in “Hittite an Indo-European Language?”:

    ” It would not be difficult to pick out items here and there which would seem to discredit the author’s understanding of Indo-European relations or his sense of perspective. But to my mind that would only becloud the issue, which must rest in the main on those resemblances which I have summarized above.

    Not, of course, on any one of them. Not, for example, on kuis ‘who,’ and the adverbs kuwabi, etc., for one might as well prove the Indo-European affinity of a North American Indian language which shows an interrogative stem qa- (satem-language!, cf. Skt. ka-) in the adverbs qama ‘how,’ qaxba ‘where,’ etc. (Handbook of American Indian Languages, pp. 674, 676). …”

    Note how Buck never questions whether or not the Native American language referenced could be potentially related to IE: he automatically assumes the correlation was absurd simply because it was an “Indian language” without doing any further analysis. If he had dug further, he might have found:

    Avestan: a- ‘he, this’ / Cherokee a- ‘he/she/it’ (3PS pronominal prefix)
    Avestan: dah- ‘gift’; dā- ‘give, put’; dāna- ‘offering’
    Cherokee: adanedi ‘give’; adanehi ‘giver’; adanelv ‘given’; adahnehdi, adanedi ‘gift’
    Avestan: gam ‘go’ / Cherokee -ega ‘go’
    Avestan: kō, ka, ko ‘who’ / Cherokee gago ‘who’
    Avestan: kat ‘what’ / Cherokee gado ‘what’
    Avestan: aθaham; verb; 1st person singular imperfect indicative active ‘speak, say, declare’ /
    Cherokee: adiha ‘he’s saying it’

    While this list of potential correspondences by itself does not “prove” a generic relationship between IE Avestan and Iroquoian Cherokee, one has to wonder why linguists like Buck never bothered to delve into a possibility that they were, in fact, somehow related. Sure, the forms in the list possibly could all be false cognates, but at this point, that proposition is beginning to strain credulity.

    The hypothesis at least answers some of the key questions that has bedeviled historical linguistics: what are the antecedents of Indo-European? Where are the living linguistic cousins of Indo-European? These questions could be answered now; they were mysteries because linguists were looking on the wrong continent.

  101. I have to say, Gary, you didn’t really answer the question in this most recent comment. Rather it’s more just a restatement of your general thesis.

  102. Exactly. You seem unable to comprehend what the question even means. A theory that is not falsifiable is not a theory but a dream.

  103. Another example is *PIE *H₁dónt- and *PNI * oˈnoɁtsaɁ ‘tooth’

    0 1 2 3 4
    H₁d ó n t-
    oˈ noɁ ts- a

    As can be seen, while the *PIE form appears to have a prefix of unknown origin, it is otherwise is quite similar to the *PNI form structurally.

    The PIE word does not contain a prefix. The root here is *h₁ed- ‘eat, take a bite’, and *-(o)nt- is a suffix (forming present pasticiples and agent nouns). Since the accent in this word was shifted to the suffix, the root occurs in the zero grade. The etymology of ‘tooth’ is thus transparent, and it’s clear that the word was formed in PIE by derivational processes productive in that language. Of course it doesn’t mean that before the PIE period people had no teeth, but any predecessors of *h₁d-(o)nt- must have gone extinct or changed their meaning. The alternative IE word for ‘tooth’, *ǵombʰo- (cognate to English comb) is also deverbal. I’m surprised you don’t seem to be aware of all that.

    But never mind. Now show how the constituents of *h₁d-(o)nt- map onto North Iroquoian morphemes. What is the North Iroquoian root? What does it mean, and how is that meaning modified by affixes?

  104. During the early stages of European settlement, many Europeans in the Pacific Northwest adopted Chinook Jargon and even spoke it as the primary home language..

    Do you have a source for this? Pacific Northwest history used to be one of my hobbies, and I do not recall ever coming across any claims that a significant number of early European settlers switched to Chinook Jargon as their primary language.

  105. marie-lucie says:

    A large number of Europeans on the West Coast learned CJ and it was the primary medium of communication between them and the aboriginal people. But it is unlikely that it became the “primary language” among Europeans although non-English settlers might also have used it with other local Europeans. It may also have been the home language of some mixed race families.

    CJ is known to have been learned as a first language by some children in the two very large reservations in Oregon where people of four or five native groups speaking as many distinct languages were resettled together by the American government, so that CJ became the lingua franca. Under these conditions, intermarriage was common and spouses often only had CJ as their common language, and that’s what their children learned first.

  106. Gary:

    1-I must third Brett and Marie-Lucie’s point/question: I too have never heard of Chinook Jargon becoming the home language of any significant number of European settlers, a reference would be very much appreciated.

    2-Your latest round of data does not improve your case, I’m afraid! By comparing any daughter of Proto-Iroquoian (such as Cherokee, which by the way only had twelve consonant phonemes) with any daughter of Indo-European (such as Avestan) you are making the task of finding similarities a lot easier. I can’t help but note that if you replace Avestan with Proto-Indo-European forms the match with Cherokee forms becomes much poorer…

    3-I must also echo David D.: you are basically avoiding answering the questions Piotr, our cyberhost and I had brought up. If you wish to know what kind of evidence I would find convincing in demonstrating a genetic relationship, I refer you to my comments on Proto-Romance on this Language Hat thread from two years ago:

    http://languagehat.com/an-anatolian-script-mystery/

  107. David Marjanović says:

    Reminder:

    As long as you can answer the question “if I were wrong, how would I know?” all the way down, you’re doing science; as soon as you can’t do that, you’re not doing science anymore.

    ====================================

    Lactose tolerance: I need to read the papers again.

    Scenario for why PIE had nine velar plosive phonemes and the plain ones were rarest: Deliberately oversimplified in that I wanted to show only the principle behind this. West Caucasian languages don’t have a single vowel phoneme either, they have a minimum of two, and there’s no reason to think that ablaut is a PIE innovation out of nowhere… I recommend the second part (pages 306–320) of this paper and pages 6–8, 11 and 19 of this presentation which will be familiar to some

    ====================================

    Bedtime.

  108. David Marjanović says:

    No, I have not worked out regular correspondences yet to the degree that neogrammarians would like. I am still in the “collecting forms lists” phase, but based on what I am seeing, there is a strong case for the deep affinity of IE with North American languages. I am hoping that someone with more time, resources, and specialized skills (and who can get paid for it) will take a crack at working out regular correspondences.

    How can you have a strong case if you don’t have any regular correspondences?

    And BTW, the Tocharian A form for ‘man, person’ is ‘enkwe’. In *Proto Northern Iroquoian, it’s ‘-õkweh’ (‘be a person’) Tuscarora ‘-ək̃ wɛh’, Cayuga ‘-õkweh’, Seneca ‘-ɔk̃ weh’. and Mohawk ‘ón:kwe’. I’m not just posting random associations: there are patterns to these correspondences.

    But you’re not trying to relate Tocharian A to Northern Iroquoian. You’re trying to relate IE as a whole to Iroquoian as a whole.

    Unless you can reconstruct the PIE ancestor of this Tocharian A form, and the Proto-Ir ancestor of this PNIr form, you’ve got nothing. 😐

    Maybe you can. But I see one difficulty right away: PIE appears not to have had any words that began with a vowel.

    Do the necessary work, and then come back.

    According to a recent Chinese paper describing the results of aDNA studies of remains in Xinjiang: “… Y chromosome haplogroups of Hunnu remains included Q-M242, N-Tat, C-M130, and R1a1. Recently, we analyzed three samples of Hunnu from Barköl, Xinjiang, China, and determined Q-M3 haplogroup. Therefore, most Y chromosomes of the Hunnu samples examined by multiple studies are belonging to the Q haplogroup. Q-M3 is mostly found in Yeniseian and American Indian peoples, suggesting that Hunnu should be in the Yeniseian family. …”

    I’m completely baffled that you quote a paper without citing it.

    And what does “a Chinese paper” even mean? There’s no such thing as national genetics.

    Mind you, I’m not upset by the conclusion in your quote. 🙂 It has long been suggested, and quite convincingly, that one of the who knows how many languages spoken in the Xiongnu confederation was Yeniseian.

    The idea of an explicit link between the Iroquoian languages and Indo-European is not new: the Jesuit missionaries –

    In hindsight, all that can be said of their idea is “nice try”. They lived long before historical linguistics had become a science.

    Creationists love to point out that just about every biologist and “philosopher of nature” was a creationist right up until the year 1858.

    As Beekes has pointed out, the *PIE sound system could have been derived from a reduced set of the sound system of the Salishan language Shuswap.

    Beekes’s point wasn’t that PIE came from the Pacific Northwest, it was that it came from a language that had a sound system rich in ejective consonants. Such sound systems are found today in languages from the Pacific Northwest, from the general Caucasus area, and from southern Africa.

    Shuswap is simply an example of a language with a very large consonant system.

    The model suggesting links between North American languages and PIE falsifies the theory that IE is a monogenic language.

    You can’t falsify a hypothesis with another hypothesis. You can only falsify it with facts that it doesn’t fit.

    The question becomes: is your idea that the lexicon of PIE has two origins more parsimonious than the idea that it has a single origin?

    So far, the answer seems to be “no”.

    As I have pointed out, I am not the only person to suggest the likelihood of links between IE and North American languages. Accomplished linguists have separately suggested links between IE (specifically, Greek) and Iroquoian, and IE and Siouan, and all I have done is to synthesize these hypotheses and place them in the context of recent genetic findings.

    And you haven’t even noticed that they all contradict each other. You can’t, for example, suggest a link between specifically Greek and Iroquoian without either proposing that Greek isn’t an IE language or that Iroquoian is an IE subbranch.

    Also, there were accomplished biologists before 1858. They were still wrong.

    Avestan: a- ‘he, this’ / Cherokee a- ‘he/she/it’ (3PS pronominal prefix)
    Avestan: dah- ‘gift’; dā- ‘give, put’; dāna- ‘offering’
    Cherokee: adanedi ‘give’; adanehi ‘giver’; adanelv ‘given’; adahnehdi, adanedi ‘gift’
    Avestan: gam ‘go’ / Cherokee -ega ‘go’
    Avestan: kō, ka, ko ‘who’ / Cherokee gago ‘who’
    Avestan: kat ‘what’ / Cherokee gado ‘what’
    Avestan: aθaham; verb; 1st person singular imperfect indicative active ‘speak, say, declare’ /
    Cherokee: adiha ‘he’s saying it’

    While this list of potential correspondences by itself does not “prove” a generic relationship between IE Avestan and Iroquoian Cherokee, one has to wonder why linguists like Buck never bothered to delve into a possibility that they were, in fact, somehow related.

    Oops! You did it again.

    This time you’re suggesting a relationship between specifically Avestan and specifically Cherokee even though you actually want to suggest a relationship between IE as a whole and Iroquoian as a whole.

    You need to compare PIE and PIr.

    If you go back to the PIE ancestors of the Avestan forms you cite, many of the similarities just evaporate. For example, Avestan is an Indo-Iranian language, which means that Avestan a represents a merger of PIE *e, *o (in closed syllables) and the rare *a. I’ll try for a few of yours…

    Avestan: dah- ‘gift’; dā- ‘give, put’; dāna- ‘offering’
    PIE: *doh₃- “give, put”; *doh₃-no-m “offering”

    Avestan: kō, ka, ko ‘who’
    PIE: *kʷi-s “who” (animate)

    Avestan: kat ‘what’
    PIE: *kʷi-d “what” (inanimate)

    Avestan: aθaham; verb; 1st person singular imperfect indicative active ‘speak, say, declare’
    – I don’t know by heart where that comes from. However, Iranian (including Avestan) θ between vowels comes from PIE *th₂; Iranian h comes from *s; Iranian a almost always comes from *e or *o; and seeing as PIE words may not have been able to begin with a vowel, you’re probably missing *h₁ or *h₂ at the beginning of the word, less likely *h₃.

    Now show me the Proto-Iroquoian ancestors of the Cherokee forms you quoted…

    Sure, the forms in the list possibly could all be false cognates, but at this point, that proposition is beginning to strain credulity.

    I’m sorry, that’s nothing but funny.

    The hypothesis at least answers some of the key questions that has bedeviled historical linguistics: what are the antecedents of Indo-European? Where are the living linguistic cousins of Indo-European? These questions could be answered now; they were mysteries because linguists were looking on the wrong continent.

    …No, for the most part they weren’t and aren’t looking at all.

    Those who have been looking for relatives of IE have been looking in the geographical vicinity: Uralic and Etruscan* are the candidates I personally find most promising, West Caucasian and a large grouping of language families including lots of north-central Eurasian families other than West Caucasian have also been suggested.

    The problems lie elsewhere. IE has been researched with a great effort for the last 200 years, with the help of ancient documents going back four thousand years; that is very much the exception. The Uralists have to make do without any writings older than one thousand years; they’re still not sure where to put the root of the Uralic tree (between Samoyedic and everything else, or between an eastern branch – Samoyedic and Ugric – and everything else?), and they’re still discovering heaps of new cognates every decade. So, if you want to compare PIE to Proto-Uralic, you’ll find that Proto-Uralic is much more nebulous than you’d like. The Etruscan language is poorly known and poorly understood. If Uralic plus something else is the closest relative of IE, the possibilities multiply, and the state of the research worsens, as generally does the state of the data known to science; I’ll just mention the Altaic controversy.

    * Or rather the “Tyrsenian languages” as a whole, of which Etruscan is the best known one by far.

  109. George Gibbard says:

    The form aθaham ‘I said’ is actually Old Persian, not Avestan. The Avestan root is saŋh- (http://avesta.org/avdict/avdict.htm), corresponding to śaṃs- in Sanskrit (http://spokensanskrit.de/index.php?script=HK&beginning=0+&tinput=zaMs&trans=Translate&direction=AU), and the Indo-European root must be *kens- which AHD has as the source of Latin cēnseō ‘I assess, estimate, judge’. So at this point not looking close to Cherokee adiha.

  110. And of course the a- of aθaham is a tense marker (the augment), while the a of adiha is, if I’m not mistaken, a person marker. One more example of the many things that can go wrong when you treat inflected forms as your comparanda without regard to their morphological structure.

  111. David Marjanović says:

    So, aθaham is from PIE *h₁e-ḱens-mi? I’m becoming more and more curious about the Proto-Iroquoian ancestor of adiha.

  112. Presumably rather *(h₁)e-ḱens-e-m, thematic imperfect.

  113. Presumably rather *(h₁)e-ḱens-e-m, thematic imperfect.
    Presumably rather *(h₁)e-ḱens-o-m?
    Avestan: kō, ka, ko ‘who’
    PIE: *kʷi-s “who” (animate)

    Avestan: kat ‘what’
    PIE: *kʷi-d “what” (inanimate)

    But the Avestan forms clearly go back to the o-stem variants, *kʷo-s and *kʷo-d.

  114. Presumably rather *(h₁)e-ḱens-o-m?

    Yes, of course. Is there a historical linguistics equivalent of the Law of Prescriptivist Retaliation?

  115. Actually, instead of *(h₁)e-ḱens-o-m, shouldn’t we reconstruct *ḱens-o-m? As I recall, the augment, being found in Greek, Armenian and Indo-Iranian only, is assumed to be a late post-Proto-Indo-European innovation which left geographically peripheral languages (Celtic, Latin, Tocharian, Anatolian languages…) unscathed.

  116. Well, since we’re talking specifically about the preform of aθaham, I think the augment has to be there.

  117. David Marjanović says:

    But the Avestan forms clearly go back to the o-stem variants, *kʷo-s and *kʷo-d.

    Yes, except I wasn’t sure those existed, beyond knowing Latin quod. I have never systematically studied this whole subject. :-]

  118. Well, even English who and what (as well as German wer and was) reflect *kʷo-s and *kʷo-d.

  119. *kʷo-d erat demonstrandum.

  120. What’s the definition of IE languages anyway? It seems a bit circular to me. IE languages are descendants of PIE. PIE is the common ancestor of all IE languages.

    So if a newly discovered language, extinct or not, is shown to definitely be related to IE, but branced off before any of the current branches, does that change what PIE is?

  121. Trond Engen says:

    IE-ness is a matter of classification, and that makes it circular by definition.

    The situation you describe happened when Hittite was discovered to be IE. In some ways it confirmed the reconstruction. They famously provided written evidence for the existence of the laryngeals. In others it didn’t and led to a refined model. The PIE gender system is one example.

  122. What’s the definition of IE languages anyway? It seems a bit circular to me. IE languages are descendants of PIE. PIE is the common ancestor of all IE languages.

    The definition is genetic. You can define PIE formally as the most recent common ancestor of, say, Hittite and Vedic, and the IE family as PIE plus all languages descended from it. This definition is not circular. If, for example, one day someone gathers enough evidence that IE and Uralic are related via a common ancestor, the definition will not hyave to be revised.

    To 19th century scholars PIE was the most recent common ancestor of all languages known at the time to be related to Sanskrit, and IE was that language plus all its descendants (or, equivalently, the “crown group” of the family: all its extant representatives and their ancestors back to the most recent common one, plus all its descendants). However, it didn’t occur to anybody to formalise this definition and codify the terminology. (presumably it was regarded as too obvious to require clarification). Today we know that the Tocharian and Anatolian languages are more distantly related to the IE languages in the 19th-c. sense than the latter are to each other. That’s why we continue to use the old term for a more encompassing taxonomic unit (also because the position of Anatolian and Tocharian wasn’t quite clear from the start). “Indo-Hittite” somehow didn’t catch on. It would be nice to have established terms for the “Neogrammarian IE” node and for a larger group that includes all languages more closely related to Sanskrit than to Hittite (it includes Tocharian). My personal preference if for “Neo-Indo-European” (the crown group) and “Core Indo-European” (Neo-IE + Tocharian), but there are other terms in circulation as well.

  123. My personal preference if for “Neo-Indo-European” (the crown group) and “Core Indo-European” (Neo-IE + Tocharian)

    That sounds terrible to me (sorry!). “Neo-Indo-European” sounds like it should mean some newer form/version of PIE, and “Core Indo-European” sounds like it should be just the central group of languages, not “plus” anything.

  124. Hah! There’s no pleasing everyone.

  125. Hat is right. How can the core be bigger than what it’s the core of?

  126. But what did you intend the “core” to mean?

  127. Or what Keith said while I was typing.

  128. The “core” of the family after you peel off Anatolian.

  129. Piotr Gąsiorowski can answer for himself, but his classification seems to be like this

    PIE
    / | \
    Languages closely related to Sanskrit Tocharian Anatolian
    =
    Neo-IE
    \_______________________________________/
    = Core IE
    \_____________________________________________________/
    =IE

  130. I agree that Neo- may sound confusing if it refers to the old notion of IE, but in biological phylogenetic nomenclature “Neo-” often refers to a crown taxon (the part of a larger taxon which contains all its living representatives).

    “Core IE” is not my private term. Some authors use it like me for the non-Anatolian branch, while for others their “Core” = my “Neo-“. Don Ringe (From Proto-Indo-european to proto-Germanic) calls IE without Anatolian “North IE”, and North IE without Tocharian — “West IE” (now, this is confusing, because his West IE includes Indo-Iranian, Armenian and Balto-Slavic, among others). He further divides West IE into Italo-Celtic and “Central IE” (Greek + Armenian + Albanian + Balto-Slavic + Indo-Iranian + Germanic). I’m not happy either with this family tree or the accompanying terminology.

  131. Wow, all that terminology is awful! I hope they can figure out some more satisfying terms.

  132. To make matters worse, I’ve also heard “nuclear IE” used for Piotr’s “Neo-IE”. It’s confusing because the same people who use it also use “core IE” for “nuclear IE” + Tocharian, and I find it impossible to remember which is the core and which is the nucleus.

  133. Truth to tell, “Indo-European” is awful too. The hyphen suggests that we have two subfamilies, Indic and European. To make matters worse, Anatolian, Armenian and most of Iranian are neither Indian nor European, and there are many non-IE langauges in both Europe and (especially) India. If we drop the hyphen, the Indoeuropeans could be interpreted as Europeans of Indian descent (the Roma?). “Indogermanic” was still worse. Time to rethink it all? 😉

  134. J. W. Brewer says:

    I thought the rationale for indogermanisch was supposed to be about the geographical distribution of the subfamilies (before 1492 etc): Indic-speaking regions to Germanic-speaking regions and all points in between (German wikipedia glosses it as „von Indien bis Germanien“), not implying that Indic and Germanic were the only two pieces?

  135. I know how it was motivated, but what about the Celtic and Romance (Neo-Italic?) languages?

  136. Florian Blaschke says:

    1) Where does the claim that the steppe people did not have blue eyes at all (isn’t that a quite strong claim to make?) come from? I mean, who proposed it and on what evidence? The Wikipedia entry I linked to says (with ref) that many steppe people had blue or green eyes, which makes the opposite claim surprising.

    2) Regarding Chinook Jargon being spoken among European-descended white North Americans, see the wonderful anecdote about the two ladies mentioned at https://en.wikipedia.org/wiki/Talk:Chinook_Jargon#Note_on_comment_re_usage_by_.22American_leaders.22_and_.22residents_of_Vancouver.22

    3) Old Irish loch (Proto-Celtic *loku-) and Latin lacus are indeed cognate and related to Greek lákkos, but English lake is unrelated; it is neither borrowed from Latin or Romance, nor is it cognate (don’t forget the Germanic sound shift!), but has an independent origin. (The Wikipedia article on false cognates lists various pretty cute and instructive examples, several of which I have added myself, including a Proto-Indo-European/Proto-Algonquian comparison.)

    4) For a plausible venture into Pre-Proto-Indo-European vocalism, see https://web.archive.org/web/20130124063132/http://www.indogermanistik.uni-freiburg.de/seminar/pers/kuemmel/umat/idgphon.pdf

    I’m still disturbed by the vocalism of PIE, regardless of whether Kümmel’s hypothesis is correct. If most instances of high vowels (/i/ and /u/) can or even must be explained away as allophonic with the semi-vowels /j/ and /w/, most vowels by far are the ablaut vowel //e ~ o ~ Ø ~ a?//, which does seem to mean that PIE had only a single common vowel, and other vowels turn up only marginally or in vocabulary which is suspicious of being borrowed, onomatopoeic or elementary (such as *atta “father”, which looks un-PIE and is apparently derived from baby-talk – it might not even be of PIE age, and can be recreated repeatedly in descendants, much like German Pappe vs. English pap, which does not show the effects of the High German sound-shift but does not seem to be a Low German borrowing either). Even the reduplication /i/, which cannot be related to a zero grade of a diphthong with /j/, seems to alternate with the reduplication /e/ on a more abstract level. Kümmel’s Pre-PIE system with two vowels /a/ and /a:/ distinguished by quantity rather than quality is even weirder typologically. But you don’t even need it to acknowledge that the PIE vowel system is odd. So the idea that an originally significantly larger vowel system was collapsed to a basically one-vowel or two-vowel system (perhaps under the influence of West Caucasian?) is attractive and makes sense intuitively. If Uralic is really the closest kin of IE – or related at all, of course – this would help us understand why it is so different typologically.

    5) Considering the vagaries of subgrouping in both Indo-European and Uralic, I’m prepared (like the LIV) to be generous in what I accept as “reconstructible to PIE/PU” – with varying probabilities attached. The requirement that an etymon absolutely must be attested in an outlying group (Anatolian and Tocharian for PIE, Samoyedic for PU) to even be considered for inclusion in a list of proto-language reconstructions seems excessive to me (and, as argued in the preface of the LIV, we need to anticipate future progress – it’s always possible that a cognate turns up eventually after all). If, for example, a verbal root is attested in both Baltic and Greek, or Celtic and Indo-Aryan, or Saami and Hungarian, and there is no reason to suspect that it is secondary somehow, it is highly probable that it was part of the proto-language and justified to treat it as a proto-vocable. If it is only attested in a single branch, but there is no particular reason to suspect secondary formation, borrowing or what have you, the possibility of it being inherited from the proto-language is of course much more uncertain, but for practical purposes it could still be justified or useful in some contexts to treat it as a proto-vocable – with appropriate reservations, caution and disclaimers. Any individual case is to be judged on its own terms.

    As for Anatolian and Tocharian, I would also like to point out that the matter is still not settled and I don’t find it acceptable to treat “Indo-Hittite” as fact. In the realm of phonology, for example, there are no clear non-Anatolian common innovations, and in the field of verbal roots, Anatolian isn’t really distinctive, either. So the honest approach, IMHO, is to admit that we just don’t know for certain what the relationship between Anatolian, Tocharian and PIE really is like. A possibility I don’t see given as much thought as it deserves (IMHO, of course) is a kind of compromise between the traditional comb-like and the “Indo-Hittite” model: Anatolian and Tocharian did split off early and lost contact with the remainder of IE, but only after internal differentiation within PIE had already started, turning it into a dialect continuum, like the relationship between Icelandic/Faroese and the remainder of North Germanic, which wasn’t homogeneous either anymore when Icelandic/Faroese split off. So common innovations within the “core” IE dialect continuum that did not reach Anatolian (and in some cases not even Tocharian) were still possible. (This means that strictly speaking, Anatolian and Tocharian are on the same level as Indo-Iranian, Balto-Slavic, Albanian, Armenian, ?Graeco-Phrygian, ?Italo-Celtic and Germanic.)

    6) For the PIE word for “tooth”, see now http://www.martinkuemmel.de/liv2add.html#_Toc392257927

    7) I’m surprised to hear that the steppe people of the Yamnaya culture weren’t lactose tolerant yet, but then, “milk” doesn’t even have a solid PIE reconstruction (“to milk”, however, has).

  137. J. W. Brewer says:

    The relative northwesternmostness (or whatever you want to call distance from Bengal) of Germanic v. Celtic probably depends on your century of reference. If you’re willing to treat Iceland as part of the “natural” range of Germanic rather than a colonial expansion that’s too recent to count, Germanic clearly wins. Supposedly both “Indo-Celtic” and “Tocharo-Celtic” have been proposed as alternatives, although I’m not sure how non-facetiously.

  138. J. W. Brewer says:

    OK, the dark horse could also win if you’re willing to take the dubious-but-arguable position that if Germanic gets credit for Iceland, Romance gets credit for the Azores. More recently, of course, IE has become so globally dispersed as to not have coherent “end points.” The antipodes of the conjectural Urheimat on the Ukrainian steppes from which it may have begun its expansion is someplace in the part of the South Pacific that is almost entirely devoid of islands, very roughly equidistant between Germanic (New Zealand) and Romance (Chile).

  139. Why don’t we just rename PIE Pre-Proto-Ukrainian? I can’t imagine any objections to that.

  140. Piotr, I’d be curious to hear why you dislike Ringe’s division of West-Neo-Nucleo-Kernular IE into Italo-Celtic vs. all the rest.

  141. …in the field of verbal roots, Anatolian isn’t really distinctive, either.

    As regards verb stems, however Anatolian is rather special in completely lacking the “simple thematic presents” like *bʰér-e/o-. Tocharian shows a few of them as a category “in the making”. In the rest of IE, they are the dominant type of present stems.

    Just a few other telling innovations, off the top of my head:

    Anatolian has no -eh₂-feminines (not even anything vestigial corresponding to them), and arguably no feminine gender as a grammatical category. This sets it apart from Tocharian and the rest.

    The *TK metathesis is shared by all IE branches except Anatolian and Tocharian.

    The absence of the aorist/present distinction in Anatolian seems to be due to a branch-specific loss, but there are cases where pairs of verb roots with the same meaning but different lexical aspect are segregated between Anatolian and non-Anatolian. Consider, for example, the words for “drink”: the original present root survives in Anatolian but the aorist root is lost (leaving at best some doubtful traces). In non-Anatolian, the present root is still present in Tocharian but disappears elsewhere (except for a few fossilised derivatives), while the aorist root survives and gives rise to a new (reduplicated thematic) present in the “crown group”.

    All these facts are, in my opinion, more consistent with successive splits (and a nested hierarchy of the family tree) than with areal diffusion.

  142. Piotr, I’d be curious to hear why you dislike Ringe’s division of West-Neo-Nucleo-Kernular IE into Italo-Celtic vs. all the rest.

    I’d like to see some real support for it (other than, say, lexical innovations). My gut feeling is that Germanic belongs with Italo-Celtic, and whatever it has in common with Balto-Slavic can be plausibly explained by contact. If, on the other hand, the satem shift and RUKI rule in Indo-Iranian and Balto-Slavic are due to areal convergence rather than common origin, why wasn’t Germanic affected by either of these changes despite being so close to Balto-Slavic? The position of Greek, Armenian and Albanian is more problematic. The treatment of the dorsal series in Armenian and Albanian is quite unique and different from “satemisation” proper. So the family tree as I imagine it would feature two large taxa, (Italic + Celtic + Germanic) and (Indo-Iranian + Balto-Slavic) plus three or more (Thracian? Dacian? Phrygian?) small branches closer to the root. But all these deeper branchings are hard to resolve. Of course family trees have limited validity if there’s a lot of secondary convergence going on.

  143. Why don’t we just rename PIE Pre-Proto-Ukrainian? I can’t imagine any objections to that.

    Sergey V. Lavrov is already preparing an angry note of protest.

  144. Is that all? I was hoping for little green men at my door.

  145. Florian: English lake is unrelated

    Which English lake? OE lacu ‘stream, watercourse, gully’ is probably native and related to leccan ‘moisten’, but the Middle English word for ‘lake’ was at least influenced semantically by Latin lacus and French lac. Why shouldn’t it simply be a Latinate loan falling together with a native word meaning something rather different (running, not stagnant water)?

  146. According to Florin Curta, Proto-Slavic never existed.

    Common Slavic was just a mixed language which arose in 7th century as lingua franca of Avar Khaganate.

  147. marie-lucie says:

    Florian Blaschke: Chinook Jargon spoken by two ladies in New York

    Thank you for linking to the Wikipedia Talk section on CJ.

    Judging from this very informative section, it seems to me that while many European settlers, traders and priests on the Northwest Coast (in both Canada and US) indeed learned and spoke CJ quite fluently, the anecdotes about them speaking it among themselves refer to this use in the presence of strangers in order NOT to be understood (I imagine that this would explain its use by the “two ladies” in a New York hotel). Other uses were as a source of euphemisms, or just for fun.

  148. According to Florin Curta, Proto-Slavic never existed.

    It’s like saying that Old English never existed — and maybe even English in general has never existed, because:

    (1) The Angles, the Saxons and the Jutes were distinct ethnic communities with little or no sense of common identity (at least before the unification of “England” under the rule of Wessex, which however did not extend over all West-Germanic-speaking inhabitants of Great Britain. Scandinavian presence in the north only complicated the picture.

    (2) At no point in its history has English been a uniform language; Old English started out as a cluster of related but different Ingvaeonic dialects not much different from pre-Old Frisian and (continental) Saxon; then it became partly bastardised via contact with Old Norse.

    (3) Middle English (also extremely non-uniform) was a kind of lingua franca under Norman rule, in a linguistically mixed society.

    (4) Modern English is such a mess.

    What would it take for a language to be a “100% real” entity? We know that languages are useful abstractions whose “realness” varies depending on the circumstances, and we know how to live with it.

  149. David Marjanović says:

    Where does the claim that the steppe people did not have blue eyes at all (isn’t that a quite strong claim to make?) come from?

    It’s not a strong claim to make. The famous mutation rs12913832 in the gene OCA2-HERC2, which explains 3/4 of the distribution of blue vs. brown eyes, is quite recent; it hasn’t been that many millennia since nobody in the world had this mutation, and therefore most likely blue eyes, at all.

    I quote from the famous recent paper on the subject:

    66 authors with delightful names (first: Morten E. Allentoft, last: Eske Willerslev): Population genomics of Bronze Age Eurasia. Nature 522: 167–171.

    Page 171:

    Temporal dynamics of selected SNPs The size of our data set allows us to investigate the temporal dynamics of 104 genetic variants associated with important phenotypic traits or putatively undergoing positive selection³³ (Supplementary Table 13). Focusing on four well-studied polymorphisms, we find that two single nucleotide polymorphisms (SNPs) associated with light skin pigmentation in Europeans exhibit a rapid increase in allele frequency (Fig. 4). For rs1426654, the frequency of the derived allele increases from very low to fixation within a period of approximately 3,000 years between the Mesolithic and Bronze Age in Europe. For rs12913832, a major determinant of blue versus brown eyes in humans, our results indicate the presence of blue eyes already in Mesolithic hunter-gatherers as previously described³³. We find it at intermediate frequency in Bronze Age Europeans, but it is notably absent from the Pontic-Caspian steppe populations, suggesting a high prevalence of brown eyes in these individuals (Fig. 4).

    Fig. 4a shows the frequency of rs12913832 in purple: it’s at 0 in the Paleolithic sample, at 1.00 in the Mesolithic European sample, maybe 0.55 in the Neolithic European sample, maybe 0.4 in the Bronze Age European sample, at 0 again in the Bronze Age Steppe sample, again at about 0.4 in the Bronze Age Asian sample, and at 0.25 or just above in the Iron Age Asian sample. Among modern populations, it’s at 0 in Africa, about 0.3 in southern Europe, maybe 0.85 in northern Europe, maybe 0.15 in South Asia, 0 in East Asia and about 0.3 in America.

    On goes the quote:

    The results for rs4988235, which is associated with lactose tolerance, were surprising. Although tolerance is high in present-day northern Europeans, we find it at most at low frequency in the Bronze Age (10% in Bronze Age Europeans; Fig. 4), indicating a more recent onset of positive selection than previously estimated³⁴. To further investigate its distribution, we imputed all SNPs in a 2 megabase (Mb) region around rs4988235 in all ancient individuals using the 1000 Genomes phase 3 data set as a reference panel, as previously described¹². Our results confirm a low frequency of rs4988235 in Europeans, with a derived allele frequency of 5% in the combined Bronze Age Europeans (genotype probability > 0.85) (Fig. 4b). Among Bronze Age Europeans, the highest tolerance frequency was found in Corded Ware and the closely-related Scandinavian Bronze Age cultures (Extended Data Fig. 7). Interestingly, the Bronze Age steppe cultures showed the highest derived allele frequency among ancient groups, in particular the Yamnaya (Extended Data Fig. 7), indicating a possible steppe origin of lactase tolerance.

    (Obviously, “lactase tolerance” at the end of the quote is a confusion of “lactose tolerance” and “lactase persistence”.)

    I don’t understand why both 5% and 10% are given for Bronze Age Europeans, and thus the difference between figures 4a and 4b. Apparently the latter represents a larger sample of individuals, where just the region around the LCT gene was investigated rather than the entire genome?

    Fig. 4a apparently puts the frequency of rs4988235 at 0 in Paleolithic Europeans. Astoundingly, it implies 0.15 in Mesolithic Europeans; this is not mentioned in the text at all. The data point is shared with that for another allele, so maybe there’s a mistake. Anyway, it’s at 0 again in Neolithic Europe. Bronze Age Europe has the mentioned 0.10, and the Bronze Age Steppe, Bronze Age Asia and Iron Age Asia samples are all at 0. Today, it’s at 0 in Africa* and East Asia, 0.25 in southern Europe, 0.7 or so in northern Europe, about 0.15 in South Asia and 0.2 in America. Brown-eyed, lactose-intolerant hordes, then?

    * Many Africans are lactose-tolerant, but that’s due to two other mutations (one eastern, one western).

    Fig. 4b shows apparently the same frequencies for the modern populations as Fig. 4a. But for the ancient ones, the numbers differ: 0 for Paleo-, Meso- and Neolithic Europeans, the mentioned 0.05 for Bronze Age Europeans, about 0.20 for the Bronze Age Steppe sample, a bit above 0.1 for Bronze Age Asians and 0.25 for Iron Age Asians. A fifth of the brown-eyed hordes was lactose-tolerant?

    It looks to me like the allele rs4988235 comes from Yamnaya and spread both east and west while the selection for it continued everywhere.

    Extended Data Fig. 7 splits up the samples from Fig. 4 and shows the frequencies of rs4988235 according to the method used for Fig. 4b. Both Western and Eastern Hunter-Gatherers are at 0, and so are the two Neolithic samples; among Bronze Age Europeans, “Rem”, “Bb”, “Une” and “Sca” are at 0, but “Hu” is at maybe 0.05 and “Cw” above 0.15; in the steppe, “Oku”, “Sin” and “Androv” are still at 0, but “Arm” is at maybe 10, “Afan” slightly higher, and “Yam” towers at 0.30; of the further Bronze Age samples, “Mezh” is at a whopping 0.50, “Karasuk” at about 0.1, and “AfGo” is at 0, as is the Iron Age sample from the Altai. The modern American samples, it turns out, are Mexican, Puerto Rican, Colombian and Peruvian.

  150. David Marjanović says:

    From the last paragraph of the paper proper:

    The enigmatic Sintashta culture near the Urals bears genetic resemblance to Corded Ware and was therefore likely to be an eastward migration into Asia. As this culture spread towards Altai it evolved into the Andronovo culture (Fig. 1), which was then gradually admixed and replaced by East Asian peoples that appear in the later cultures (Mezhovskaya and Karasuk).

    It goes on to repeat the suggestion that the Afanasievo culture is ancestral to the much later, much more southern Tocharians.

  151. From the AI Koans:

    A cocky novice once said to Stallman: “I can guess why the editor is called Emacs, but why is the justifier called Bolio?” Stallman replied forcefully: “Names are but names. ‘Emack & Bolio’s’ is the name of a popular ice cream shop in Boston-town. Neither of these men had anything to do with the software.”

    His question answered, yet unanswered, the novice turned to go, but Stallman called to him: “Neither Emack nor Bolio had anything to do with the ice cream shop, either.”

    This is known as the ice-cream koan.

  152. Ow!

  153. In particular, Austroasiatic is not spoken all over Southern Asia, nor is Afroasiatic spoken all over Africa and Asia.

  154. @Piotr Gąsiorowski: How much evidence is there about the Jutes as a distinct ethnic and linguistic group, actually? The last I heard, the information about them was pretty thin.

    @David Marjanović: Eye color is weird, because the degree of pigmentation is human eyes is governed by three separate but very similar genes. They are three copies of the same original gene, duplicated through unequal crossing over. So there are seven discrete hues of human eyes, depending on how many positive alleles (from 0 to 6) an individual has. (The colors are light blue, dark blue, green/gray/hazel, light brown, dark brown, light black, and dark black. The two shades of blue and two shades of brown are easy to distinguish if you look carefully. The blacks are more difficult, both being quite dark). No single mutation (unless it occurred prior to the genes being duplicated) can therefore dominate the variance in eye color. A new mutation in one of the genes can move the population at most two steps along the spectrum (and often only one—a population with only brown or darker eyes cannot get blue eyes through a single mutation).

  155. I am the result of a blue × green mating, and have blue eyes with green flecks in them.

  156. How much evidence is there about the Jutes as a distinct ethnic and linguistic group, actually? The last I heard, the information about them was pretty thin.

    The well-documented Kentish dialect of Old English was very distinct, different in many respects from both the Anglian varieties (Mercian and Northumbrian) and West Saxon. Anglo-Saxon tradition attributes its origin to Jutish colonisation. The Jutes who had settled in Hampshire and the Isle of Wight were subjected to ethnic cleansing as Cædwalla of Wessex annexed those areas in 686 (according to Bede). There is no particular reason to question those traditions, though the Jutes are a little more enigmatic than the Angles and the Saxons. This is partly due to the accidental similarity of their tribal name to the OE pronunciation of the Scandinavian Gauts (Ġēatas), with whom they tended to be confused already in the Middle Ages.

  157. Yah, it’s de Yats an’ de Yutes again…

  158. Don’t forget the Eotenas and the ettens!

  159. @John Cowan: It’s obvious just from looking at most people’s eyes that there is a lot of spatial variation in the pigment levels between different portions of the iris. This precise expression of the color is also shaded a bit (in two senses!) by other genetic and environmental factors. This level of variability is not unexpected when you realize that the actual number of melanin molecules present in the iris is (relatively speaking) quite small, and so the fractional variation is pretty large.

    However, the fact that your eyes seem to be mixture probably has nothing to do with your parents being separately blue and green.

  160. David Marjanović says:

    However, the fact that your eyes seem to be mixture probably has nothing to do with your parents being separately blue and green.

    Yeah, I’m such a product too, and all four of us have either blue or green eyes.

  161. David Marjanović says:

    I’m still disturbed by the vocalism of PIE, regardless of whether Kümmel’s hypothesis is correct. If most instances of high vowels (/i/ and /u/) can or even must be explained away as allophonic with the semi-vowels /j/ and /w/, most vowels by far are the ablaut vowel //e ~ o ~ Ø ~ a?//, which does seem to mean that PIE had only a single common vowel,

    That would be at least two. The difference between *e and *o was phonemic in that it was used to distinguish words all on its own, like the nominative *nokʷts (“night”) from the genitive *nekʷts (Hittite na-ku-zi, ne-ku-zi, IIRC).

    Anyway, I’m trying to come up with a theory, which is mine, on how the PIE vowel system came to be, taking both Kümmel’s idea and this, which at first sight contradicts it.

    […] can be recreated repeatedly in descendants, much like German Pappe vs. English pap, which does not show the effects of the High German sound-shift but does not seem to be a Low German borrowing either).

    The most impressive examples are picken and pecken, the verbs for what birds do with their beaks.

    Picken has an additional, unrelated, and just as onomatopoeic meaning in Austrian dialects: “glue/stick”.

    For the PIE word for “tooth”, see now http://www.martinkuemmel.de/liv2add.html#_Toc392257927

    Oh. Fascinating!

    “milk” doesn’t even have a solid PIE reconstruction (“to milk”, however, has)

    Blessed are the cheesemakers.

    Piotr, I’d be curious to hear why you dislike Ringe’s division of West-Neo-Nucleo-Kernular IE into Italo-Celtic vs. all the rest.

    I’d like to see some real support for it (other than, say, lexical innovations).

    Ringe’s argument, and apparently only argument, is that the passive in *-r (Anatolian, Tocharian, Italo-Celtic) could only have been replaced by the one in *-i a single time. I don’t know where he takes that certainty from; right here chez Hat, Hans linked to two works on Google Books a few months ago that tried to derive the Italo-Celtic *-r from something other than the ending seen in Anatolian and Tocharian, but the interesting parts weren’t included in the preview. 🙁

    My gut feeling is that Germanic belongs with Italo-Celtic

    Phonological evidence:
    1) Dybo’s law: long vowels (both original and from compensation for lost laryngeals) were shortened if they came before the originally stressed syllable.
    2) *TT clusters changed into the entirely new phoneme */sː/, even though PIE didn’t have phonemic consonant length at all.

    I also wonder if Douglas Kilday is right and Kluge’s law (plosive + *n under Verner conditions > */pː/, */tː/, */kː/), rather than being limited to Germanic, is shared with Italo-Celtic; but that idea needs a lot more work.

    The treatment of the dorsal series in Armenian and Albanian is quite unique and different from “satemisation” proper.

    I knew that about Albanian, and so does Wikipedia; but Armenian?

    The Angles, the Saxons and the Jutes were distinct ethnic communities

    It has been suggested that “Saxon”, rather than an ethnonym, was a job description (“pirate in the North Sea”) much like “Viking” a few hundred years later. I’ll dig up the paper(s) at some point, it’s really late at night. There’s no such doubt about the Angles and the Jutes, however.

  162. marie-lucie says:

    *TT clusters changed into the entirely new phoneme */sː/, even though PIE didn’t have phonemic consonant length at all.

    Is this supposed to have been a direct change, or to have gone through an intermediate stage *TS, thus *TT — *TS — *SS (later S:) which seems more likely? (Sorry, I forgot how to make arrows properly – I had written it down but lost the note).

  163. @Marie-Lucie

    An [s] was inserted already in PIE between any two dental stops followed by a vowel (actually the rule was a little more general, and a sibilant was inserted also in the so-called “thorny” clusters like /tk/, with further branch-specific complications. Anyway, the realisation of /t+t/ was *tst, and there are phonotactic reasons to analyse the epenthetic *s as a separate segment rather than part of an affiricate, at least in PIE. The development of this *tst in Germanic, Italic and Celtic is quite odd (whatever the intermediate stage) and unlikely to have affected them independently. Some propose a similar outcome for Albanian, but the Modern Albanian reflex (s) must have developed from an earlier affricate that merged with the result of the palatalisation of * before front vowels, and with the reflex of *ḱw — presumably some kind of labialised alveopalatal affricate, but at any rate nothing to do with *ss.

  164. David Marjanović says:

    So. I have about an hour; let’s see what I can write down and where it takes me.

    One idea I have is to start with Kümmel’s idea that *e – *o used to be a length contrast, but to interpret that as */ə əː/ rather than */a aː/ in order to accommodate the evidence (presented by Piotr) that *a and also existed in PIE. If I simply ignore for the time being – Kümmel seems to think that this could be possible in the nominative singular; I have no idea about verb roots, however –, then this gives us a pre-PIE system with four vowels, */ə əː a aː/, of which the first two were much more common than the last two, and ablaut was a matter of length alone.

    Ignoring length for the moment, this and the multiplicity of velar plosives might have come from a six-vowel system */a e i o u ə/: all except */a/ (PIE *a) merged as */ə/ (PIE *e), while */e i/ palatalized the preceding (and perhaps the following?) consonant in the process, */o u/ labialized it, and the original */ə/ did neither. This latter phoneme would explain why both *e and *a seem to occur next to plain velars.

    What about length then? Perhaps it, and thus ablaut, is an IE innovation derived from stress. This would mean it’s unrelated, except perhaps as a vaguely areal feature, to the ablaut seen in Afroasiatic and, from what very little I’ve read, Kartvelian and West Caucasian. (Otherwise, it could have been retained from Proto-Nostratic, as Bomhard has in fact proposed; the other Nostratic branches, which may form a larger branch of their own, would then have lost it). West Caucasian isn’t supposed to be Nostratic, but has a fair number of similarities with IE that point to noteworthy amounts of contact.)

    Or length could be inherited. Vowel length is not (anymore) reconstructed for Proto-Uralic, but it is for Proto-Altaic (by people who hypothesize a Proto-Altaic language in the first place, it goes without saying).

    The other idea I have is to not ignore PIE ; after all, Piotr’s presentation doesn’t. 🙂 That gives us a system with three “full” and two “reduced” vowels: *ā, *ē, *o, *a, *e, from earlier */aː eː oː a ə/. This immediately reminds me of the northwestern African vowel system (shared by the local Arabic “varieties” and the local Berber languages), which has three “full” and a single “reduced” vowel, ends up as a, i, u, e/è/é in French transcription and a, i, u, ə in more scientific transcriptions, and has default phonetic values somewhere around [æˑ iˑ uˑ ɪ] (with plenty of allophony). Both in Arabic and in Berber, the reduced vowel results from a merger of three historically separate short vowels, while the full vowels are the historically long ones. Next door*, in the Tuareg languages which belong to the Berber family, only two of the original three short vowels have merged, leaving two reduced vowels that are transcribed ə and ă. Just like how *h₂ turned adjacent *e into a in PIE, several consonants turn adjacent ə into ă in Tuareg.

    In PIE, and *o certainly were not [iː uː], but more like [ɛː ɔ(ː)]. Perhaps they were pushed away by the syllabic allophones of /j w/? Really surprisingly far away?

    I have to go now; I’ll think about whether the two ideas can be merged, because the first doesn’t predict and the second predicts neither that *a and were so rare nor why there were palatalized and labialized velars nor why *a was more common but still far from universal next to the plain velars.

    * Relatively speaking. Old joke about asking for directions in the desert: “go straight ahead, and turn left in two weeks.”

  165. Marie-Lucie: > is no problem, you can just use it. To write < write & l t ; without spaces. The mnemonic is “Less Than”.

  166. marie-lucie says:

    Piotr: Thank you for the clarification. So there was indeed an intermediate stage, though tst rather than <ts.

    JC: Thank you! “And Less Than” makes perfect sense.

  167. David Marjanović quoted in regards to lactase persistence:

    “To further investigate its distribution, we imputed all SNPs in a 2 megabase (Mb) region around rs4988235 in all ancient individuals”

    And concluded:

    “It looks to me like the allele rs4988235 comes from Yamnaya and spread both east and west while the selection for it continued everywhere.”

    For everyone who is not familiar with what it means to impute genetic data, this means that the missing nucleotides in the ancient DNA sequences were “filled in” based on statistics from modern data.

    I’m not sure how this got past the reviewers, but the actual mutations present within ancient haplotypes can not be statistically implied from modern samples, in the absence of any actual direct ancient association.

    This means that at maximum, the lactase persistence alleles we’re present at about 5% in Yamnaya, but could actually be 0% from a different reading of the same data.

    Also, the exact same mutation is present in individuals with R1a and R1b Y chromosome haplogroups in India and even ‘older’ ones among the unusually R1b cattle pastoralists of Northern Africa.

    This says to me that the mutation is much older than the Yamnaya. It seems to have spread at a very low level with some pastoralists ancestors.

    Some of their steppe relatives also founded the Bell Beaker culture in western Europe, and this seems to have been even more important in its expansion within Europe.

  168. David Marjanović says:

    For everyone who is not familiar with what it means to impute genetic data, this means that the missing nucleotides in the ancient DNA sequences were “filled in” based on statistics from modern data.

    Oh.

    Have you found anything redeeming in the supplementary information?

  169. David Marjanović says:

    I wrote:

    I have to go now; I’ll think about whether the two ideas can be merged, because the first doesn’t predict and the second predicts neither that *a and were so rare nor why there were palatalized and labialized velars nor why *a was more common but still far from universal next to the plain velars.

    The only idea I can come up with is to start from the second idea and postulate that the full vowels *ā, *ē, *o were somehow descended from earlier */a ə ɨ/, a West-Caucasian-like system which could have been produced by a collapse where frontness and roundedness were transferred from the vowels to the consonants. But why *[ə ɨ] would rotate to become *[ɛ ɔ] – and even in which direction: *[ə ɨ] > *[ə ʉ] > *[ɛ ʊ] > *[ɛ ɔ] or *[ə ɨ] > *[ə ɪ] > *[ɵ e] > *[ɔ ɛ] – is beyond me.

    Now we need external cognates to test any of this.

    Earlier, I wrote:

    It has been suggested that “Saxon”, rather than an ethnonym, was a job description (“pirate in the North Sea”) much like “Viking” a few hundred years later.

    One thing I have appears to be:

    Elmar Seebold (2013): Die Aufgliederung der germanischen Sprachen. NOWELE: North-Western European Language Evolution 66: 55–77.

    Google Scholar can’t find more than the citation, though, and the pdf I have – which Google can’t find either – has the right page numbers but no other indication of belonging to a journal; it’s even written in the tone of an introductory lecture. Anyway, the interesting part here starts with the oldest sources about the invasion of Britain, both from the mid-6th century; Gildas consistently called the invaders Saxones, which apparently meant “Germanic pirates” in contemporary sources, while Procopius said that “Brittia” was inhabited by Britons, Angles and Frisians, complicated by the fact that he called Britain Brettanía elsewhere. Next is Bede, who in 731 first called the invaders Anglorum sive Saxonum gens – one people with two names – before inserting a mention of the Jutes.

    If the sources are taken at face value as they are preserved, Saxones go way back into antiquity. Apparently, however, that name was so prominent in later times that it was inserted when illegible names were copied or when unspecified “barbarians” or “pirates” were mentioned as occurring in the later Saxon area; very similar things seem to hold for the Franci. The oldest mention of Saxones that is “clearly reliable” is in a work of Julian the Apostate and refers to the years 353–356, and apparently means “pirates”. And that, Seebold goes on, fits in with the Saxon origination story written down much later by the noble Saxon Widukind of Corvey. It says that the Saxons arrived in later Saxony by ship and took over by trickery and betrayal; they had large knives, “as the Angles still carry them today according to the way of the ancient people”, and these knives are nostra lingua called sahs. Even earlier, Rudolf of Fulda had expressly stated that the Saxons were descended from the Angles (whom he located, however, in Britain).

    When the later Anglo-Saxons referred to themselves as a whole, they called themselves Angles all the way to King Alfred of Wessex. Outsiders, however, consistently spoke of Saxons (and still do in Welsh and Gaelic).

    Beda’s mention of the Jutes is the only one in any early source (excepting of course those who copied the whole story from him), apart from the saxonibus euciis in a letter from the Frankish king Theudebert I to Emperor Justinian. Then, there are two Old English sources: a poem called Widsith, which says that an otherwise unknown Gefwulf ruled the Jutes while Fin(n) ruled the Frisians, and Beowulf, where Finn’s people are called Jutes four times – Frisians are mentioned three times in the same passage, but it’s not clear if they’re also ruled by Finn.

    Taking all this together, Seebold proposes that the Frisians were at one point conquered by sea warriors from northern Jutland. That would explain why their names are both synonymous and not synonymous, why Frisian and English are “so closely connected”, and why the presumably Frankish source Procopius had mentioned Frisians invading Britain. It could also explain what’s going on with the runes: shortly after the invasion, special English runes appear in Britain that have no precursors among “the normal Norse runes”; the Frisian runes, which first appear a bit later, are closely related to the English ones; and the next most similar rune shapes are “special developments” found on bracteates from northern Jutland and southern Scandinavia.

    Then Seebold spends half a page on what I would call a fun fact: the Franks, first mentioned in 291 and coming out of nowhere; “the hints in Latin sources point, if anywhere, to the Frisians and their neighbors (especially the Chamavi). On the other hand, there is some evidence (far from proof) that the Franks were earlier called Hugones or similar, and in that designation many have sought a reflex of the name of the Chauci who disappear from the picture of history in about the 2nd century. […Wow, that would require that Verner’s law hadn’t acted yet.] Where they were, the Saxons are found later.” Now the scenario: in the 2nd/3rd century the Saxons push out or conquer the Chauci, and the Frisians are pressured by the Jutes. At the same time, the Franks show up, connected to Frisians and Chamavi as well as to the Chauci. Perhaps the Franks, the Free Ones, were those Frisians and Chauci who were not conquered by the Saxons and Jutes.

    Another half-page treats another fun fact: the Danes who took over Jutland and are first mentioned in the mid-6th century (Jordanes and Procopius, the former stating that they are descended from the Swedes and kicked out the Heruli). Given the lack of a tribe called Danes in Scandinavia, perhaps Dane was a nom de guerre like Saxon before it and Viking after it. “Thus, when Hygelac/Chochilaicus is sometimes called a Geat, sometimes a Dane, this need not be a contradiction: he is king of the Geats; but when he assaults Frisia, he’s a Dane.”

    Old Saxon was part of the continental Germanic dialect continuum, Old Frisian was not. Seebold explains this by postulating that when the sea warriors partly left for England, the superstrate influence ended in Saxony, and adstrate influence from the south and east came in; Frisian, on the other hand, was isolated in the extreme west, and only the name “Frisians” came back after not being mentioned for several centuries.

  170. Interesting stuff, thanks for passing it along. I remember respecting Seebold back when I read him for my IE studies.

  171. Names are only names, but there are unquestionable linguistic differences between the southern (“Saxon”) and northern and Midland (“Anglian”) varieties of Old English, and a few of them are fundamental enough to go back to the “continental” period. To quote Bede:

    From the Jutes are descended the people of Kent, and of the Isle of Wight, and those also in the province of the West Saxons who are to this day called Jutes, seated opposite to the Isle of Wight. From the Saxons, that is, the country which is now called Old Saxony, came the East Saxons, the South Saxons, and the West Saxons. From the Angles, that is, the country which is called Anglia, and which is said, from that time, to remain desert to this day, between the provinces of the Jutes and the Saxons, are descended the East Angles, the Midland Angles, Mercians, all the race of the Northumbrians, that is, of those nations that dwell on the north side of the river Humber, and the other nations of the English.

    Bede’s credibility should not be overestimated, but he was a pretty careful historian for his time. He relied more on written sources (monastic annals, foreign historians) than folk memories (which did not extend into the fifth century anyway). His account makes a lot of sense linguistically.

  172. David Marjanović says:

    Fun thing is, Seebold quotes from that very passage, but just the part about the Angles leaving an empty country behind – “de illa patria, quae Angulus dicitur, et ab eo tempore usque hodie manere desertus” – and only in the chapter on the Danes. The linguistic diversity of Old English is only mentioned at all when it comes to the name of the Jutes: their land is eota lond in at least one unspecified manuscript of the OE translation of Bede, and Seebold calls that form “probably Mercian”, while Manuscript B has ytena lond, which is “probably Late West Saxon”.

    …How, actually, did lond turn back into land? It can’t be some kind of Dutch loan…???

    a few of them are fundamental enough to go back to the “continental” period

    Which ones? Arguments from the assumed speed of evolution make more sense in linguistics than in biology, but I still feel uneasy about this.

  173. Which ones? Arguments from the assumed speed of evolution make more sense in linguistics than in biology, but I still feel uneasy about this.

    (1) Different operation of “Anglo-Frisian brightening” (the fronting of *a), which was blocked in Anglian (Mercian/Northumbrian) in some environments in which West Saxon and Kentish had it, e.g. before dark /l/. This must be an old difference, not a secondary retraction of the vowel in Anglian, since velar palatalisation did not take place in Anglian before this vowel. That’s why today’s English has cold, calf etc. (from Anglian cald, calf with OE /ɑ/), rather than *cheald, *chalf, which would have regularly developed out of West Saxon ċeald, ċealf (with [æɑ] < *æ). Note that palatalisation itself is old — older than i-umlaut.

    (2) The Anglian raising of West Germanic *ā, which merged with *ē in prehistoric times, hence A ē corresponding to WS ǣ. (A similar change in Kentish, however, is independent and later: the vowel was still low in early Kentish texts). This must be older than the monophthongisation of *ai and i-umlaut, since the umlauted reflex of *ai is A ǣ, not ē.

    One striking consequence of the latter difference is the outcome of the ‘cheese’ word in Mercian (ċēse, the ancestor of the modern form) and Late WS (ċӯse, ċīse. Here, Proto-WGmc. *ā became Proto-Anglian *ē and remained stable, while Proto-WS *ǣ underwent a whole series of changes restricted to that dialect: after the palatalisation of the initial *k the resulting alveopalatal consonants caused it to diphthongise, yielding something like *ēæ, which merged with the i-umlauted reflex of Germanic *au into Early WS īe (phonetically [iːy]) > Late WS ӯ, ī.

    There’s also some morphological stuff to consider, but the phonological differences are more evidently old.

  174. David Marjanović says:

    Impressive.

    Note that palatalisation itself is old — older than i-umlaut.

    Do you think it’s inherited from Proto-Anglo-Frisian or a later areal feature?

  175. Palatalisation works differently in English and Frisian. For example, the diphthong *au became primitive OE æu, later ēa [æːɑ], causing a preceding velar to palatalise, whereas in Old Frisian *au was monophthongised to ā with no palatalising effect (cf. OFris. cāp [kɑːp] vs. OE ċēap [ʧæːɑp]). Still, whether brought from the continent or developed areally in Britain, palatalisation must have been one of the oldest specifically English changes. Even umlaut predates the oldest written attestation of OE, leaving very little time for still older changes, including palatalisation, its grammaticalisation through the loss of final vowels that triggered it (as in gōs/gœs) and its spread all across Old English dialects.

    The rising of WGmc. *ā and the non-brightening of *a before word-final or preconsonantal *l connect Anglian with Frisian. It seems as if some “polymorphisms” existing in the Proto-AF dialect continuum had been inherited by Old English.

  176. It’s notable that umlaut exists in both the languages of Britain.

    consistently spoke of Saxons (and still do in Welsh and Gaelic)

    Also Breton Saoz and Finnish/Estonian Saksa(maa), now meaning ‘Germany’.

  177. It is notable that Angles dwelling “north side of the river Humber” also occupied entire south-eastern part of modern Scotland, including its largest city of Edinburgh.

    Scots language is a direct descendant of early medieval Anglian dialect and lowland Scots are actually Angles by origin.

    Therefore, Scots language should be properly called English and Scotland should be renamed England.

  178. Finnish/Estonian Saksa(maa), now meaning ‘Germany’.
    these knives are nostra lingua called sahs

    Finnish for “walnut” is saksanpähkinä, and the word for “scissors” is sakset, a borrowing from Swedish (sax).

  179. David Marjanović said:
    Have you found anything redeeming in the supplementary information?

    Overall, it is a very interesting paper, including the supplement. Of course there are clearly multiple gaps in our knowledge.

    The sampled Yamnaya have almost completely R1b Y-haplogroups, while Corded-Ware are largely R1a. This means that although the genomic resemblance is strong, there may not be a “father and son” relationship between these cultures. Or there was a strong founder effect from an unsampled Yamnaya group.

    In the next few years we should hopefully get a flood of new ancient genomic data on various ancient cultures throughout Eurasia and the Middle East. Hopefully it will align much more strongly with linguistics.

    There will have to be higher quality data as well, which seems to be possible. And a few parent and child pairs would yield fantastic information, but nearby burials almost always seem to be unrelated individuals so far.

  180. David Marjanović says:

    Palatalisation works differently in English and Frisian.

    Could this be a sequence along a tree? Perhaps palatalisation happened first while *au was still intact, then the Frisians wandered off (perhaps from Denmark to the land of the ancient Fresones) and eventually monophthongised *au, then Pre-English decided to raise *au after all, creating new occurrences of a front vowel behind unpalatalised consonants, and palatalised again?

    The Slavic palatalisations come to mind: two ran their course, then *ai became a new front vowel, which triggered the third round except in Novgorod.

    The sampled Yamnaya have almost completely R1b Y-haplogroups, while Corded-Ware are largely R1a. This means that although the genomic resemblance is strong, there may not be a “father and son” relationship between these cultures. Or there was a strong founder effect from an unsampled Yamnaya group.

    Good to know, thanks.

  181. then Pre-English decided to raise *au after all, creating new occurrences of a front vowel behind unpalatalised consonants, and palatalised again?

    To front the starting point of the diphthong rather than raise it. But note that i-umlaut actually produced a new generation of phonetically front vowels (fully phonemicised before the 7th century) which no longer triggered velar palatalisation. Cyning < *kuningaz, gœs ~ gēs < *gōsiz etc. had initial velars, not palatals.

  182. David Marjanović says:

    To front the starting point of the diphthong rather than raise it.

    Oh, of course.

    i-umlaut actually produced a new generation of phonetically front vowels (fully phonemicised before the 7th century) which no longer triggered velar palatalisation

    …So my whole scenario would have needed to happen well before that time. I’ll need to think about it when I’m less tired.

  183. David, Piotr: here’s another scenario: what if there had been a Pan-Ingveonic shift of short /a/ to /æ/, even when this /a/ was the first element of the diphthong /au/ (thus, the diphthong would have become /æu/), and Frisian had then turned the initial element of the diphthong back to /a/ because of the influence of the following /u/?

    The main advantage I see with this scenario is that you could assume that the same shift from /æ/ to /a/ took place in Anglian and Frisian before final/pre-consonantal /l/: /l/ and /u/ are quite close in terms of point of articulation, after all.

    So if we assume a universal fronting of short /a/ to /æ/, what we later see is that:

    1-West Saxon and Kentish keep the original state of affairs.

    2-Anglian, in common with Frisian, centralizes /æ/ to /a/ before final/pre-consonantal /I/, and

    3-Frisian, unlike any variety of Old English, centralizes /æu/ to /au/.

    One could thus offer a unified explanation for Frisian: it centralizes all instances of Ingveonic /æ/ when pre-consonantal/final /l/ or /u/ follows, and within the Ingveonic dialect continuum the innovation is found in Anglian, but only under the first conditioning factor, not the second.

    Thus making Anglian intermediate between Frisian and West Saxon/Kentish. Which, oddly, fits with another scholar who sees the dialect diversity of Old English as being due to multiple migrations, with West Saxon/Kentish going back to an older and Anglian to a more recent migration from the continent:

    https://openaccess.leidenuniv.nl/bitstream/handle/1887/1889/344_058.pdf?sequence=

  184. David Marjanović says:

    Ooh, back-and-forth sound changes! I like those. 🙂 I also like the paper, except I need to read up on what Ringe thinks of the reflexes of PIE ; I remember he defends a back-and-forth sound change for this one, as does Piotr above.

  185. David Marjanović says:

    My gut feeling is that Germanic belongs with Italo-Celtic

    Phonological evidence:
    […]
    2) *TT clusters changed into the entirely new phoneme */sː/, even though PIE didn’t have phonemic consonant length at all.

    Well. Ringe (2006) has a sentence on this:

    Italic and Celtic show the same outcomes of these PIE clusters, but it seems clear that the changes were parallel developments rather than historically shared changes, if only because an intermediate stage is clearly attested in Gaulish (at a time when Latin had long completed the process).

    In keeping with the purpose of his book, but unfortunately for my purposes, he didn’t say what that intermediate stage was and didn’t cite any source.

  186. marie-lucie says:

    David: About: *TT clusters changed into the entirely new phoneme */sː/ : I thought there must have been an intermediate state *tts, which Piotr corrected to *tst (see above).

    Ringe: Italic and Celtic show the same outcomes of these PIE clusters, but it seems clear that the changes were parallel developments rather than historically shared changes, if only because an intermediate stage is clearly attested in Gaulish (at a time when Latin had long completed the process).

    I guess Ringe must mean “parallel developments” in Germanic as well as Italic and Celtic, but I don’t see why the intermediate stage attested in Gaulish precludes a shared development in Italic and Germanic. In this I seem to agree with Piotr’s statement above, that The development of this *tst in Germanic, Italic and Celtic is …. unlikely to have affected them independently.

  187. David Marjanović says:

    I guess Ringe must mean “parallel developments” in Germanic as well as Italic and Celtic, but I don’t see why the intermediate stage attested in Gaulish precludes a shared development in Italic and Germanic.

    Because Italic and Celtic are more closely related to each other than to Germanic. Any development shared by Italic and Germanic but not Celtic could perhaps be a borrowing that was made in darkest prehistory, but cannot be a common inheritance from a common ancestor.

  188. marie-lucie says:

    I thought of that explanation, but borrowing of a feature of pronunciation implies actual contact, so is there evidence of geographical closeness of Italic and Germanic at some point?

  189. David Marjanović says:

    No.

  190. Not so fast. There are a few words in the Ringe/Tarnow data set (though I am not up to picking them out of the thicket today) that indicate contact between Proto-Italic and proto-Germanic. It is also quite probable that proto-Italic was originally spoken in Central Europe rather (or as well as) in Italy.

  191. marie-lucie says:

    Thanks JC. The plot thickens! It makes sense, since Italic seems to have largely displaced non-IE languages in Italy (I don’t mean just Etruscan which seems to have been intrusive at a later date).

    Where can I access the Ringe/Tarnow dataset?

    Meanwhile, I recently acquired the new Beekes volume on comparative IE, but I can’t find anything there on the things discussed here! The book has an index of forms in the various languages, but not an index of topics.

  192. David Marjanović says:

    It is also quite probable that proto-Italic was originally spoken in Central Europe rather (or as well as) in Italy.

    Well… on the northern side of the Alps, there was the Celtic La Tène culture; before that, there was the Hallstatt culture, which is also considered Celtic, though I don’t actually know why; before that, what people spoke is anyone’s guess, but Hallstatt starts pretty early. As soon as we can say something about Germanic, its southern limit was pretty far north. From the height of the Celtic expansion there’s the Celtic kettle of Gundestrup in Denmark…

  193. @David: But the usual story is that it was the Cimbri who went to Serbia and nicked the kettle from the Scordisci before the Romans sent them packing back to Himmerland. Gold and silver travels widely.

    Of course the Cimbri may have brought some Celts back as thralls, but 1) it’s hard to prove and 2) it doesn’t really count anyway.

  194. David Marjanović says:

    …You know what? I really, really should have checked Wikipedia first.

    Still, maps like this still put a lot of distance between Germanic and Italic long before the Celtic expansion.

  195. m-l: Somewhere in this thicket of papers is the raw data, but it’s not reasonably indexed. There are “updated” versions of this that aren’t based on Ringe’s professional cognacy judgments, so accept no substititutes

  196. @ David: The Germanic- Italic parallels have been known for a long time. E.g., Krahe & Meid’s “Germanische Sprachwissenschaft” (Vol. 1, Sammlung Göschen 1969, pp 15-17) lists over 30 lexical and morphological isoglosses that are shared only by Germanic and Italic. I’m not going to list them all, but I’ll quote this (my translation):

    Especially characteristic is that not only nouns, but also a number of verbs (frequently up to details of the conjugation class) is only Germanic and Italic, e.g. Latin du:cere “draw, lead” – Goth. tiuhan, OHG ziohan “draw”; Lat. cla:ma:re “shout, cry” – OHG hlamo:n “roar”; Lat. tace:re “be silent” – Goth. thahan, OHG dage:n “be silent”; Lat. sile:re “be taciturn” – Got. ana-silan “become silent”.

    The similarities are of a kind that can have arisen in a still relatively undifferenciated IE; so both languages may even have shared a neighbourhood before the Proto-Germans moved to Southern Scandinavia.

  197. marie-lucie says:

    JC, thanks for the diagram, but I am fearful of embarking on a search of the thicket! I guess that Ringe is the historical linguistic “anchor” in the enterprise led by mathematicians etc. From what I can gather of that type of work, it relies far too much on (a smallish amount of) vocabulary, ignoring the importance of morphology.

  198. David Marjanović says:

    lists over 30 lexical and morphological isoglosses that are shared only by Germanic and Italic.

    The lexical ones could easily, I suppose, be cases where Celtic once had the words in question and then lost them. After all, except for the rather innovative Island Celtic branch, Celtic is very poorly attested. I wonder if the morphological ones could be similarly overwritten by later Celtic changes.

  199. There are several amazing lexical isoglosses connecting Italic with Slavic. They are only explicable through early contact:

    Lat. pastor : PSl. *pastyrь ‘shepherd’ (note that Slavic uses *-tel-, not *-tor- in occupational terms).

    Lat. anser : PSl. *gǫserъ ‘gander’ (note that the initial /g/ rules out borrowing from Latin, no matter how early).

    Lat. secūris: PSl. *sekyra ‘axe’ (the root *sek- ‘cut’ is widespread, but the suffix *-ūr- is unusual and the chances of its appearing independently twice to form words with the same meaning in different branches are extremely low).

  200. I’ve read somewhere that the so called “Italic” glosses in Slavic are actually Venetic.

    The Veneti were Indo-European people found all over Europe – from Italy to Poland and their language may have been a separate branch of IE family, but obviously closely related to Italic.

  201. m-l: My rule of thumb is that if Don Ringe, whose historical linguistic credentials[*] are impeccable, is not involved in the mathematical work, it may as well go into the oubliette at once. Accept, as I say, no substitutes, no matter how many mathematicians or biologists (with a bow to David M) or astrologers are involved. He can easily be wrong, but his theories are never “not even wrong”.

    Part of the list. The sound-changes (all of them highly conditioned or parts of chains) are on p. 113-14 (physical pp. 55-56 of the PDF). The morphological replacements (things like “How is the medio-passive marked: suffix A, suffix B, non-morphological?”) are on pp. 117-18 (physical pp. 59-60 of the PDF). Just a few of the 370+ lexical items are on pp. 121-23 (physical pp. 63-65 of the PDF). Intervening pages map arbitrary “character numbers” like “1=absent”, “2=present” to particular languages.

    [*] His Stammbaum: August Schleicher > August Leskien > Ernst Fraenkel > Werner Winter > Warren Cowgill > Donald Ringe.

  202. The Veneti were Indo-European people found all over Europe – from Italy to Poland

    Russian Wiki identifies them as Slavic (at least the Northern ones), and of course that’s how Finns or Estonians call Russians – Vene / Venelainen

  203. Venice, I suppose, is also Russian?

  204. David Marjanović says:

    If only we knew anything about all those Veneti/Venedae beyond their names…

    Part of the list.

    Ooh, thanks!

  205. I would venture a guess that Veneti means “immigrants”

  206. One of these ancient indo-european genetics papers was just updated with new genomes a couple of days ago. It is looking very interesting for the Indo-Iranian branch.

    Of the 6 samples extracted from Srubna culture sites for whom a Y-DNA hapogroup could be tested, all belonged to haplogroup R1a, and four of them to subclade R1a-Z93, which is most common among modern-day Indo-Iranians.

  207. @David Marjanovic

    ANE is an Amerindian marker in the Old World by definition. Read Raghavan et al. 2013 to see that MA-1 is closer to modern Amerindian populations than it is to any modern Eurasian population. Nothing has changed since Patterson et al. 2012. It’s an anti-Amerindian bias that keeps people from consistently referring to it according to what it really is, namely “Ancient Amerindian.”

    Re: Gary Moore’s ideas, yes, they look like chance resemblances, but all long-range linguistic proposals (save may be Yeniseian-Na-Dene, although lexical support for it is weak, too) look like they are based on look-alikes or borrowings. We know what a proven first-order language family looks like, but we don’t know what a proven macrophylum looks like. Nostraticists claim that they are establishing sound laws for Nostratic but nobody’s buying it, so “regular sound correspondences” are not enough to make people convinced. They are as subjective as the look alikes. This is because comparativism does not have a strong methodology – first order language families benefit from the objective recency of divergence and the availability of ancient written attestations but the methodology of recovering linguistic kinship in the absence of those objective perks is not well-developed. Historical linguists just got spoiled by 19th century successes, which they attributed to their “methodology,” and in the 20th century they got completely diachronically lazy and switched to synchronic linguistics for the most part. However, considering that there’re now Amerindian-Eurasian whole-genome markers, search for linguistic connections between the Old World and the New World along the lines suggested by the distribution of “ANE” in the Old World should be able to intensify. And Gary Moore is doing what he can here – pulling data together for everyone to look at.

  208. David Marjanović says:

    Read Raghavan et al. 2013 to see that MA-1 is closer to modern Amerindian populations than it is to any modern Eurasian population.

    Reference, please?

    We know what a proven first-order language family looks like, but we don’t know what a proven macrophylum looks like.

    Please don’t feel singled out by the following:

    This whole whining about “but it’s not proven” is just ridiculous. Only creationists and historical linguists ever use that word anymore when talking about science. Science deals in varying amounts of evidence, not in proof.

    Nostraticists claim that they are establishing sound laws for Nostratic but nobody’s buying it

    That has a bunch of reasons, but it mostly boils down to “nobody else is even looking at the evidence”…

    And Gary Moore is doing what he can here – pulling data together for everyone to look at.

    Noise is not data. Seriously, you’re talking about someone who didn’t even notice he was comparing the kentum and the satem forms of the same word to different language families.

  209. @David Marjanovic

    “Reference, please?”

    I gave you one. If you don’t know yet the original paper on ancient Mal’ta DNA, just google it: “Raghavan”,
    “Mal’ta” 2013. Ok, fiine, here it is: http://www.nature.com/nature/journal/v505/n7481/full/nature12736.html.

    “This whole whining about “but it’s not proven” is just ridiculous. Only creationists and historical linguists ever use that word anymore when talking about science. Science deals in varying amounts of evidence, not in proof.”

    So, you decided to lower the scientific bar as a way to differentiate science from creationism?? Sorry, you are on your own here.

    “That has a bunch of reasons, but it mostly boils down to “nobody else is even looking at the evidence”

    It’s an excuse. I personally looked at the evidence within one lexico-semantic class and its pretty poor (more here: http://anthropogenesis.kinshipstudies.org/2015/10/the-current-state-of-nostratic-theory-or-a-psychoanalytic-reading-of-a-russian-utopian-idea/ with a further reference). Nostraticsts can’t even reproduce the dictionary meaning of a word correctly. Everything gets edited to make it easier for them to arrive at some sound laws of their choosing. My point is a methodological one, though, namely that sound correspondences don’t guarantee that the evidence of linguistic kinship is robust. They are a world of its own and there’s a lot of subjectivity there.

    “Seriously, you’re talking about someone who didn’t even notice he was comparing the kentum and the satem forms of the same word to different language families.”

    Yes, amateurs make mistakes that trained linguists don’t. It’s like speaking with a foreign accent. But I’d rather have an amateur do lexical comparisons between first-order language families, than have professional linguists waste their time on that. Greenberg compromised his reputation precisely because he was chasing after long-distance ghosts for the last 20 years of his life. Professional linguists should work on methodology, theory, etc. In astronomy, amateurs have been very useful to the academic community precisely because you can’t monitor so much data out of just a few observatories.

    @Gary Moore

    I got interested in your attempt to link Hitt meyu ‘four’ to Narrow IE (NIE) *kwetwo:r. Currently there’re no sound laws that can justify the connection. However, we do have a few cases in IE when m- (and other sonorants) unexpectedly appear in the initial position. Arm merk ‘naked’ is sometimes considered a reflex of IE *negwno-/*negwro-: merk < megwro- (typical Armenian metathesis) < *negwro- (distant dissimilation). Same for Avest magna 'naked' (< *nagna-). In Greek, gymnos 'naked' (< *nogwno-) shows that segment-level rearrangement can affect a labiovelar as well (*kw- in NIE *kwetwo:r). This is in addition to the dissimilation m ~ n.

    I don't have any similar m-examples for Hittite but Hitt does show unexpected l- in a number of words: laman 'name', lissi 'liver' (if related to NIE *yekwr-, but see Arm leard, OHG lebara).

    So hypothetically Hitt meyu-, CLuw mauwa- could have arisen by dissimilation from *weiwo-. *kwetwo:r is a complex form, so it's possible that it started with PIE *kwe- to which a suffix was later added after Anatolian had split off. Lat quattuor shows an unexpected reflex of *h2, which is not attested in other NIE forms for 'four', so it's possible that PIE had *kweH- 'four'. NIE *kweH- and Anat *weiwo- is as much as we can get in terms of bringing the two disparate daughter forms together. Now, we do have the example of Arm vec' and OPruss ushts 'six' in which the expected *sw- (PIE *swek's-) is replaced by w-. There's no good explanation for this development but apparently it happened. So, again, as a single case example you could postulate the loss of k- from PIE *kweH- to arrive at *weH- in Anatolian.

    But considering that the above analysis operates with models based on attested exceptions to rules rather than rules themselves and it still leaves some phonetic developments unexplained, I would just leave Hitt meyu- alone as presently unrelated to PIE *kwetwo:r.

  210. So, you decided to lower the scientific bar as a way to differentiate science from creationism?? Sorry, you are on your own here.

    Not at all. He’s saying that in science, we proportion the strength of our beliefs to the strength of the evidence for them. We don’t arbitrarily divide hypotheses into “proven” and “not proven” like a Scots jury (which, to be sure, has different objectives).

  211. @John Cowan

    “He’s saying that in science, we proportion the strength of our beliefs to the strength of the evidence for them. We don’t arbitrarily divide hypotheses into “proven” and “not proven” like a Scots jury (which, to be sure, has different objectives).”

    Indo-European, Austronesian, Siouan are proven, Nostratic or Amerind or Indo-Paciifc are not proven. It’s not arbitrary at all. There’s something interesting going here, though. There’s a semantic difference between “laws” in historical linguistics and “laws” in jurisprudence. “Laws” in historical linguistics were originally modeled on “natural and physical laws” in contradistinction to man-made legal “laws”. You are trying to divorce linguistic laws from natural laws, thus lowering the bar, but then you justify this epistemological move by referring to the “laws” of jurisprudence as a poor ideal model for linguists to follow. It’s either a bad metaphor on your part, a fundamental misunderstanding of the scientific method, or a belief that linguistic “laws” are different from both “natural” and “legal” laws (so that they are not human-made but neither they are given in nature, but then what ARE they?). Which one is it?

    I would agree, though, that there’s a moment in time when evidence is being accumulated. During this time, people assess whether there’s enough evidence to exclude alternatives, or alternatives should still be in the running for truth, and they hold their beliefs commensurate with the quality of the evidence mounted in favor of one model vs. the other. But it’s a “race” of sorts meaning that there must be a movement forward toward the goal of establishing “laws” as in “natural laws” that make a theory “proven.”

  212. I am very far from wanting the things you think I want. We are basically on the same side, but you put things more rigidly than I can wholeheartedly agree with. Sometimes in the historical sciences (including both biology and historical linguistics), we never do accumulate enough evidence: look at <a href="https://en.wikipedia.org/wiki/Messapian_language&quot;Messapian, for instance.

    But that doesn’t mean that any hypothesis is ever confirmed beyond the reach of any future refutation (which is the meaning of proven), nor does it mean that hypotheses are worthless because they are not yet accepted by all. The problem with Nostratic is bad data; the problem with Amerind is bad data compounded by bad methodology; and Indo-Pacific isn’t even a hypothesis, just a wastebasket taxon.

  213. @John Cowan

    I’m with you on that. I do sometimes come across as more rigid than others. Agree on everything you just wrote. Thanks for the Messapian link.

  214. It’s always good when problems dissolve into sweet, sweet harmony!

  215. David Marjanović says:

    Oh, Raghavan et al. (2013) is just the original paper on the genome from Mal’ta? Why didn’t you say so? I happen to have the pdf. 🙂

    So, how do you interpret its figure 2, what do you make of it in the light of the recent papers on Neolithic European and Bronze Age Eurasian genetics, and what do you mean by “closer”?

    So, you decided to lower the scientific bar as a way to differentiate science from creationism??

    What… no. Nothing can be proved. I can’t even prove I’m not the solipsist, and I’m pretty sure you can’t either. Certainty, as John Cowan explained, comes in degrees; only the Sith deal in absolutes… and the creationists… and far too many historical linguists.

    Yes, amateurs make mistakes that trained linguists don’t. It’s like speaking with a foreign accent. But I’d rather have an amateur do lexical comparisons between first-order language families, than have professional linguists waste their time on that. Greenberg compromised his reputation precisely because he was chasing after long-distance ghosts for the last 20 years of his life.

    Greenberg compromised his reputation because he used “mass comparison”, which is fine for proposing hypotheses, but incapable of testing them. If he had used the comparative method instead, he’d have proposed fewer hypotheses, but his reputation would be a lot better now.

    How does it help to bring in people who not merely aren’t professionals, but haven’t learned the comparative method to begin with (whether in university courses or on their own)?

    Professional linguists should work on methodology, theory, etc.

    What – not on actual phylogenetic hypotheses?

    @Gary Moore

    I got interested in your attempt to link Hitt meyu ‘four’ to Narrow IE (NIE) *kwetwo:r.

    I already linked to a much more parsimonious explanation of *kʷetwóres ~ *kʷétwores.

    Now, we do have the example of Arm vec’ and OPruss ushts ‘six’ in which the expected *sw- (PIE *swek’s-) is replaced by w-. There’s no good explanation for this development but apparently it happened. So, again, as a single case example you could postulate the loss of k- from PIE *kweH- to arrive at *weH- in Anatolian.

    Oh no, that does not follow. There are tons of cases of “s mobile”, but this never extends to *k to the best of my knowledge; worse, you’re quietly implying that PIE didn’t have a phoneme /kʷ/, only a consonant cluster /kw/ – when in fact there’s pretty good evidence that it had both, and that it had /kʷ/ in the “4” word.

    Indo-European, Austronesian, Siouan are proven, Nostratic or Amerind or Indo-Paciifc are not proven. It’s not arbitrary at all.

    It’s clinal. Is Afro-Asiatic “proven” or “not proven”? Is Siouan-Catawban? Is Siouan-Catawban-Yuchi? Is Sino-Tibetan? Is Dené-Caucasian (pdf on comparative morphology here)? Is Totozoquean? Is Toto-Zoque-Chitimacha? Is Indo-Iranian? Is Balto-Slavic? Is Italo-Celtic? Drawing a line through this continuum is arbitrary; it can’t help being arbitrary.

    You are trying to divorce linguistic laws from natural laws, thus lowering the bar

    Neither nor. We’re divorcing natural laws from the rather useless concept of proof.

    (But then, I wouldn’t speak of natural laws outside of relativity and quantum physics anyway; those elsewhere all just follow from them.)

    the goal of establishing “laws” as in “natural laws” that make a theory “proven.”

    …That’s not Science Theory As We Know It. *headshake*

    Finally:

    I personally looked at the evidence within one lexico-semantic class and its pretty poor (more here: http://anthropogenesis.kinshipstudies.org/2015/10/the-current-state-of-nostratic-theory-or-a-psychoanalytic-reading-of-a-russian-utopian-idea/ with a further reference).

    That post is about something quite different: the aftereffects of the Iron Curtain in historical linguistics (very, very similar to those I see in paleobiology). The Russians – I overgeneralize only slightly – have no money for international journals or to go to international conferences (never mind hosting international conferences!), so they publish in their inbred in-house journals that apparently aren’t even properly peer-reviewed (peer review being a quite recent development in the West, after all). On top of that, many of them (more so in Moscow, less so in St. Petersburg, it seems – another leftover cultural divide…) don’t know English well enough that reading international literature or going to conferences would do them any good. Conversely, almost nobody in the West can so much as read Russian, so the institutions don’t carry Russian in-house journals in the first place. And never the twain shall meet, even before politicians with a persecution complex descend upon the situation. It’s quite horrible.

    Rant over, I’m off to bed.

  216. @David Marjanovic

    “what do you mean by “closer”?

    Shared drift as measured by f3 stats. See Fig. 1c in Raghavan.

    “what do you make of it in the light of the recent papers on Neolithic European and Bronze Age Eurasian genetics.”

    The ANE gradient decreases from east to west from north to south and from “refugia areas” (such as the Caucasus) to “spread zones.” It’s present all over West Eurasia, including Basques, and I think it’s Paleolithic heritage in Basques. I think ANE back migrated to South Siberia from the steppe or the Caucasus area (along with mtDNA X2 lineage) during the Bronze Age.

    ” Nothing can be proved. I can’t even prove I’m not the solipsist, and I’m pretty sure you can’t either. Certainty, as John Cowan explained, comes in degrees;”

    No, it doesn’t come in degrees. There’s a time factor involved in making hypotheses proven or falsified (as I explained above), but there’s no way around the notion of “proof,” be it the social or the natural sciences.

    “Greenberg compromised his reputation because he used “mass comparison”, which is fine for proposing hypotheses, but incapable of testing them. If he had used the comparative method instead, he’d have proposed fewer hypotheses, but his reputation would be a lot better now.

    How does it help to bring in people who not merely aren’t professionals, but haven’t learned the comparative method to begin with…”

    An answer to your question is in your own statement: amateurs can do ‘mass comparison’ and professionals should R&D the comparative method. And they should exchange results with each other.

    “I already linked to a much more parsimonious explanation of *kʷetwóres ~ *kʷétwores.”

    I don’t buy etymologies that “explain” one etymon from a whole lexico-semantic class (numerals in this case) by pointing to some kind of notion that lies outside of that class (as in an arbitrary verbal root, as in *mei- ‘measure’ for Hitt meyu- or *kwet- ‘group in pairs’). Numerals, kinship terms, etc. form systems and etymologizing individual items can only come from an understanding how these classes of words evolved as systems.

    “There are tons of cases of “s mobile”, but this never extends to *k to the best of my knowledge.”

    I’m not sure it’s an s-mobile case here, and the whole phenomenon is still a bit enigmatic… Some authors (G & I) reconstruct a single, compact phoneme *sw- in the form for ‘six’. But in any case, I was just thinking through Moore’s hypothesis. It’s apriori attractive because it’s the most parsimonious of all: if meyu and *kwetwo:r are related, then the problem is solved. I just couldn’t find enough evidence to support it.

    I also happen to think that PIE may have had a palatolabiovelar phoneme that can account for such forms as Skrt asva vs. lat equus/Gk hippos ‘horse’; Gk the:ros vs. Slav *zve:ri ‘beast’.

    “It’s clinal. Is Afro-Asiatic “proven” or “not proven”? ”

    Proven. In some of your other example it’s again a time factor that needs to be factored in. But all of them need to be tested in terms of “proof” and be open to be falsified and alternative hypotheses advanced in their stead if needed. The Italo-Celtic node is a special case because it’s a problem internal to a proven language family. It is interesting (in the light of recent ancient DNA data, to your earlier point) that the only two proven nodes downstream from the proven IE node are the easternmost Balto-Slavic and the Indo-Iranian nodes.

    “That’s not Science Theory As We Know It.”

    This is nonsense to me. Sorry.

    “That post is about something quite different…”

    yes, a more relevant reference is in the bottom of the post (http://www.kunstkamera.ru/index/science/books/books/algebra_rodstva_11/).

  217. marie-lucie says:

    David: The Russians – I overgeneralize only slightly – have no money for international journals or to go to international conferences

    I have been more or less regularly attending a major international conference (ICHL) for a number of years and the presenters include more and more linguists from Russia and other countries previously behind the Iron Curtain. A good sign!

  218. marie-lucie says:

    German Dziebel: But I’d rather have an amateur do lexical comparisons between first-order language families, than have professional linguists waste their time on that. … In astronomy, amateurs have been very useful to the academic community precisely because you can’t monitor so much data out of just a few observatories.

    I would not compare those “amateur linguists” to “amateur astronomers”. Some of the latter, armed with limited equipment (and some relevant knowledge of astronomy), have indeed contributed data to the professionals because they have been able to devote time and energy to long term, consistent observations while the professionals were busy with different agendas, many of them requiring extremely sophisticated equipment. In this respect the amateurs are acting like unpaid assistants to the professionals, who appreciate their contributions. An amateur astronomer eager to contribute to Flat Earth or Hollow Earth studies would not re similarly appreciated by professionals.

    By amateur linguists I don’t mean people like some of the Hatters who are extremely knowledgeable about linguistics. Professionals are those who are expected to do research, present at conferences, etc where their findings and opinions will be tested by “peers”. Unfortunately it does not always mean that they are more knowledgeable about some aspects of the discipline. But most “amateur historical linguists” start basically from scratch, with no training and sometimes an axe to grind. As an example, a few years ago I stumbled across a website run by a Spanish retired officer who was engaged in proving to his own satisfaction that Spanish was derived from Basque rather than from Latin. The proof? words such as “comandante”. So gathering data in the absence of some relevant qualifications (however acquired, but not a priori contradicting those of the professionals) is often a waste of everybody’s time and energy if the data turned out to be unreliable or even unusable for various reasons.

    Professional linguists should work on methodology, theory, etc.
    The accepted methods in historical linguistics were established and refined over 150 years or so, through the study of actual documents in the various languages and extrapolation from the actual data towards reconstruction of the common ancestor (and intermediate stages). Concentrating on methodology and theory with less and less attention to actual (not made up) data has not worked very well in “synchronic linguistics”, especially syntax.

    Greenberg compromised his reputation precisely because he was chasing after long-distance ghosts for the last 20 years of his life.

    Theoretically, the methods for “long-range comparison” should be the same as for “short range”, but working in first-order families (eg Romance, Celtic) where resemblances are obvious is not the best training for working in second-order (eg IE) and even potentially third-order (eg Nostratic) groups. First-order families show resemblances in both vocabulary and morphological features which are usually obvious even to untrained persons: a French speaker finds a lot that is familiar (though not identical) in Spanish or Italian, but hardly anything recognizable in German or Russian apart from a few identical loanwords. Greenberg attempted to go almost directly from first to nth order, collecting a lot of vocabulary and a few pronouns (and making a lot of mistakes while doing so). Resemblant vocabulary even in closely related languages can be misleading if the words are not analyzed and if their participation in paradigms (such as sets of verb forms) is ignored.

    Greenberg should have used the comparative method
    The comparative method is a method of reconstruction of the common ancestor and only secondarily a method of confirming (or infirming) a proposed genetic classification. It works best when used with closely related languages, since classification (at least in general terms) is not a problem (eg one might argue about the exact degree of relationship between some Italian or German dialects, but not about the fact that they are Italian or German). One can collect huge amounts of data in hundreds of not fully classified languages, but even if those data can be considered reliable, using the comparative method to reconstruct their potential common ancestor (impossible if the languages are not yet classified) in order to classify them better would be an insurmountable task, even if one’s intuitions about their relationships (or other linguists’ classifications) could be assumed to be correct.

    Greenberg’s Amerind classification

    By now everyone knows that except for Eskimo and Na-Dene (basically Athapaskan), ALL of the other languages of North and South America are “Amerind” according to Greenberg. Of course Amerind is not a giant melting pot, there are many families and subfamilies listed under that name. Greenberg recognized that he had basically accepted current classifications and just put them together, but he disagreed with a few. Here are a few comments about his North American classification (reprinted in Wikipedia), specifically for Western North America:

    – in California the groups “Penutian” and “Hokan” are put together; I have been working with Penutian languages for years and I can assure you that these two groups are VERY different from each other. The only reason they are there together is that since they are both major families of California, for some years there was an annual meeting called the “Hokan-Penutian conference”, which accepted papers about the two families without any suggestion that the languages were at all related. One might as well have heard of a conference in Russia called “Slavic-Caucasian” and from there conclude the two groups had a common ancestor.
    – Another large group is Uto-Aztecan (with languages in the US and Mexico); this group was originally classified with another, much smaller group in the SW US called Tanoan, and Sapir put them together as “Aztec-Tanoan”, a classification that is now disputed (rightly in my opinion). Later, Uto-Aztecan was considered by Swadesh (a well-known “lumper”) to belong to a larger “Macro-Penutian” group. I think this potential grouping has a lot of merit although much more work is needed to substantiate (or not) the relationship. But Greenberg considers “Hokan-Penutian” and “Aztec-Tanoan” as belonging to two distinct major subclassifications of “Amerind”.
    – As for Zuni, a neighbour (though unrelated) of “Tanoan”, Greenberg places it together within “Penutian”, following an older article which was actually the result of a hoax.

    Together with the numerous errors of various kinds in the published data, is it any wonder that specialists in Amerindian (NOT “Amerind”) languages do not take Greenberg’s classification seriously?

  219. @Marie-Lucie

    “I would not compare those “amateur linguists” to “amateur astronomers”.”

    Some amateur astronomers, just like amateur linguists, are knowledgeable, others are not. There’s no difference here. Professional astronomers seem to be better than professional linguists in terms of their ability to utilize the amateur community to advance science. Professional linguists tend to stare at amateurs as if the latter were exotic animals and to gloat at their “mistakes.” Professional linguists tend to wade through thousands of languages and tens of thousands of forms on their own and its hard to marry quantity with quality. Consequently, currently none of the megaphyla have a proven status despite decades of research.

    “The accepted methods in historical linguistics were established and refined over 150 years or so,”

    I would emphatically disagree. Those methods need to evolve and one of the reasons for long-range failures is precisely because the methodology (which as you rightly point out should be the same as that used to prove first-order language families) is weak and antiquated. You can’t easily see methodological flaws in the study of first-order families because of the sheer amount of valid cognates that haven’t really changed much. But those very methodological flaws come out into the open in long-range comparisons. The major flaw (and I wrote several articles in Russian to justify this critique for Indo-European languages) is that cognate sets are poorly defined and the criteria for the inclusion/exclusion of items into a cognate set are subjective. (Similarity in sound and meaning is not an objective criterion to group items into a valid cognate set as they are historically variable.) Sound laws and reconstructions that derive from cognate set grouping procedures are hence incomplete and sometimes wrong. When there’s a need to provide an etymology for a reconstructed etymon, two look-alike etymons are typically grabbed and forced into a comparison. (As in Hitt meyu- vs. NIE *mei- ‘measure’). One of them is used a source for the other one. Some linguists accept those etymologies, others reject them. Then another etymology based on look-alikes is proposed and so on and so forth for decades and decades. As a result, there are a lot of forms without good etymologies, forms that are declared isolates, presumably “local” isoglosses, forms with phonological anomalies, etc. Whole lexico-semantic classes (such as Indo-European numerals or kinship terms) are basically left without good etymologies.

    The implications of those methodological flaws are likely considerable. Just one example: PIE *krd- ‘heart’ shows up as *srudice in Slavic and sirdis in Lithuanian. So, the velar is presumably palatovelar (PIE *k’rd-) because the so-called satem languages (named so precisely because they turn up palatals where Greek, Latin, etc. have plain velars) show s- and not k-. But then there’s another cognate set represented, e.g., by Gk rhadix and Lat radix ‘root’. The Greek form presupposes *s (< *sradiks), which is presumably not the same /s/ as the satem /s/ < PIE *k' but PIE spirant *s. But now the full morphology of this root is unmistakably the same as the full morphology of Slav *srudice and the meanings are fully compatible ('heart', 'middle', 'core', 'root'). So, if we put the two conventional sets together and treat it as one single cognate (super)set, the reconstruction will be radically (pun intended) different.

    Other examples can also be adduced to show that satem *s and PIE * is one and the same PIE phoneme. And not two, as the methodology devised 150 years ago wants us to believe! That's where we need professional linguists. At the same time, the searching for potential Eskimo-Khoisan parallels can be done by a host of amateurs.

    I wholeheartedly agree with you on your critique of lumping efforts in the New World. And thanks for your "insider" perspective on that!

  220. Went to German Dziebel’s links and found on his page this interesting exchange:

    Alexei Kassian:
    14.01.2015 в 1:07
    > Вы правы, я не очень связан с российской наукой

    Герман, извините, но вы тут развели такую активность по защите А.Клесова, что все-таки надо бы и вас представить публике. Я по долгу службы знаю, кто вы такой, а коллеги, наверное, нет.

    Вы находитесь в пограничной зоне между лингво-фриком и любителем. Т.е. иногда вы делаете что-то, что пригождается в научной работе (я видел ссылки на вашу д/б терминов родства), хотя обычно это всё бесполезное и фрическое.

    Несколько лет назад вы пытались овладеть сравнительно-историческим языкознанием. Оказалось, что быстро это ну никак не получается, увы.

    Я сейчас заглянул на ваш сайт и увидел, что вы сдвинулись в область междисциплинарных исследований (геногеография, лингвистика, археология). Это понятное желание, т.к. для междисциплинарных исследований вроде как требуются менее глубокие знания.

    Но это обманчивое впечатление, Герман. Для современных междисциплинарных исследований нужно соавторов классных специалистов из нескольких областей, а не надерганные цитаты из разрозненных публикаций.

    > You’re right, I am not really connected with the Russian science

    German, I’m sorry, but you became so active in defense of A.Klesov that one needs to introduce you to the public. By work, I know who you are, but the colleagues probably do not.

    You are in the border zone between linguistic freak and an amateur. That is, sometimes you are doing something which comes in handy for scientific work (I’ve seen references to your database of kinship terms), but usually it’s all useless and freaky.

    A few years ago, you tried to master comparative linguistics. It turned out that it’s not possible to do it quickly, alas.

    Now I looked at your website and see that you have moved to interdisciplinary research (genogeography, linguistics, archeology). It is an understandable desire, because interdisciplinary studies probably require less profound knowledge.

    But it is misleading, German. For modern multi-disciplinary research, it is necessary to have as co-authors experts from several areas, it is not sufficient to copy-paste quotes from disparate publications.

  221. Reading further, I learned that German Dziebel is founder of a new science of idenetics (иденетика), later renamed to gignetics (гигнетика).

    I haven’t been able to determine the exact subject of this new science, but apparently both idenetics and gignetics reach conclusion that humankind originates from North America.

    Definition of idenetics from 2001 book by German Dziebel is quite baffling:

    Иденетика – область системных эмпирических исследований поведения биологических особей как взаимных знаков (сообщающихся коммуникантов). Строящаяся на теоретическом и практическом материале этнолингвистики, когнитивной антропологии, семиотики и этносоциологии родства, она призвана исследовать процесс идентификации системой (человеком, социумом) коммуникативно-активных элементов окружающей ее социальной и природной среды. Под «коммуникативно-активными элементами» подразумеваются не только человеческие особи, но и потенциально любой действительный или воображаемый объект, если он сообщает субъекту свойство или сознание активности в отношении себе подобных.

    Idenetics – the area of ​​systematic empirical studies of the behavior of biological individuals as the mutual characters (communicating communicants). Built on a theoretical and practical material of ethno-linguistics, cognitive anthropology, semiotics and ethno-sociology of kinship, it studies the process of the identification by the system (human or society) of communicative and active elements of the surrounding social and natural environment. “Communicative-active elements” refer not only to the human individuals, but potentially to any real or imaginary objects, if they assign to the subject a property or consciousness of activity in relation to their own kind.

  222. @SFReader

    I’m probably the least freakish of all scholars. Alexei Kassian is an ignorant individual with homemade credentials splitting time between going through scientists’ dirty laundry and publishing something about the Hittite language. His racist social media activity in Russia has recently caused some of his colleagues to campaign for his dismissal from a teaching position at the RGGU in Moscow. He is an example of how freaks and amateurs, under the right circumstances, can become associated with academic institutions. BTW, further up there Kassian claimed that my father paid to have my Russian book published and also paid for my education at Stanford University. It’s just so funny it made me cry.

    Thanks for bringing it up, though, because Alexei Kassian is a good example supporting my point that it’s not so easy to tease apart freaks, amateurs and professionals. But may I ask you: why did you decide to put on a pseudonym, dig through my website and then translate a bizarre quote without even including my response to it?

  223. It gets even better!

    Later in the thread, Kassian accuses A.Klesov (defended by German Dziebel) of discovering new hominid species – rusanthropus which lived on Russian plain circa 700 000 before presen.

    Rusanthropus is claimed to be the direct ancestor of Russian nation!

  224. @@SFReader

    I would recommend that you translate my whole book into English and send it to me for corrections. Your English is somewhat misleading.

  225. @SFReader

    if you read carefully, you’ll notice that I never defended Klyosov, although he was right about molecular dating and other Russian scholars, who consider Klyosov a pseudoscientist, were wrong.

    I was defending the ethics of academic debates.

    But keep trolling the thread…

  226. “Just because the world’s biggest fool says it’s raining, doesn’t mean the sun is shining.” (If this is not also a Russian proverb, it certainly should be). However, one would do well to check.

  227. @John Cowan

    It’s not an English “proverb” either, you fool! 🙂

  228. Quite right, it’s not. But I use it anyway.

  229. Athel Cornish-Bowden says:

    I knew of Anatole Klesov/Klyosov back in 1972 when he and I. V. Berezin published a paper entitled “Use of integrated rate equations for the determination of the kinetic constants of enzyme reactions” in Biokhimiya 37, 170-183 (cited 36 times, including by me in 1975). That and his other papers in biochemistry are perfectly respectable (albeit not very exciting) pieces of work, and he is certainly not a pseudoscientist, at least so far as biochemistry is concerned. Until today I didn’t know of his work in dating fossils, and maybe he has strayed away from his real expertise.

  230. @Athel Cornish-Bowden

    Yes, that’s the Klyosov we are talking about. Here’s the scoop. He finished his academic career, worked for biotech in Boston (where I met him) and then got interested in molecular genealogies and the dating of divergence of molecular lineages (not fossils but Y-DNA). This became his hobby, but a hobby related to his primary area of expertise. Over the years he developed a new science called “DNA Genealogy,” which attracted a motley crew of science hobbyists in Russia (but no one in the U.S.), branched into archaeology, linguistics and other social scientific disciplines, founded a journal called “Advances in Anthropology,” published books on the origins of everybody from Slavs to all of modern humans, launched a Y-DNA sequencing lab in St. Petersburg and became a media personality in Russia brandishing his Harvard credentials and upsetting some established paradigms of thinking developed by academic scholars working in the social sciences and humanities. (E.g., his DNA-Genealogy showed that there was no significant influx of immigrants from Scandinavia in the 9-10th centuries A.D. that could justify the theory of the Western (“normanist”) origin of the Russian state.

    Last year, a team of academic scholars in Russia (population geneticists, archaeologists, craniologists and some linguists) published a letter in a web journal decrying Klyosov’s activity (outside of biochemistry) as “pseudoscience.” The letter was accompanied by a series of web posts by a Russian-born journalist Lebedev (based in Boston) entitled “Klyosov as a Science Porn Star” in which claims were made that Klyosov cooked up his Harvard credentials and earned too much money working for biotech. (The latter compromised him even more than the former in the eyes of financially insecure Russian academics.)

    So, a scuffle began. Klyosov defended himself out of a web property called Pereformat.ru, and I, in tandem with a head of biology lab in one of Moscow institutes, criticized the academic team for bad ethics, biases against Klyosov and poorly built arguments.

    The bad news for the Russian academic team is that Klyosov was indeed right about the mutation rate on Y-DNA lineages, and this had direct implications for the models of migrations of Slavic- and Indo-European-speaking populations. A Stanford team has recently revised the “molecular clock” first advanced by another Russian scholar, Lev Zhivotovsky (with whom I enjoyed a number of tasty lunches and thoughtful conversations when I was at Stanford), and the new “clock” is very close to what Klyosov proposed some 10 years ago.

    The bad news for Klyosov is that nobody is giving him any credit and that his attempt to create a new science of everything in prehistory (“DNA-Genealogy”) faces rejection even from such balanced scholars as myself. 🙂

  231. Thanks for digging that stuff up, SFReader, it makes things much clearer.

  232. @languagehat

    So, instead of deleting the spam from an anonymous troll, SFReader, you acquired a greater awareness of something?

  233. SFReader is not an anonymous troll, and the very fact that you feel entitled to characterize someone that way just because they disagree with you is telling.

  234. Please don’t generalize your distaste for pseudonymity to the Internet in general, German Valentinovich. A lot of people have a very long presence here, and it would be simply contrary to fact to call them anonymous trolls.

  235. @minus273

    It’s a fact that SFReader, at least in this thread, is an anonymous troll. Now you are telling me that s/he’s been a troll for a while. And this somehow gives him or her a lifelong right to not be referred to as such by a person with scientific credentials, a real name and meaningful contributions to the conversation.

  236. You have a right to refer to anyone by whatever insulting terms you like, and we have a right to judge you on that basis. Even if you have scientific credentials.

  237. @languagehat

    Here’s a Wiki definition of a “troll”: In Internet slang, a troll (/ˈtroʊl/, /ˈtrɒl/) is a person who sows discord on the Internet by starting arguments or upsetting people, by posting inflammatory,[1] extraneous, or off-topic messages in an online community (such as a newsgroup, forum, chat room, or blog) with the deliberate intent of provoking readers into an emotional response[2] or of otherwise disrupting normal on-topic discussion,[3]

    SFReader is a perfect example of a troll. No insults on my part. S/he’s just illustrating a dictionary entry.

    Can we now please return to the topic of this thread? Mrs. Tarpent (if I may), do you have any thoughts on my last post re: comparativist methodology?

  238. A troll in the usual sense is someone who enters an online discussion forum with posts designed to infuriate the regular participants and create drama, especially by reiterating a absurd standpoint and refusing to accept arguments to the contrary.

    SFReader does not do that. Everybody but you knows that because we are regular participants.

  239. Oh, sorry, I forgot to refresh before I posted. Well, since you know what a troll is already, you should be able to judge who the troll is by the amount of disruption and emotional response generated by the various posts in this thread.

  240. David Marjanović says:

    I haven’t even have time to properly read the latest comments. Just so much for the moment:

    1) Semantic fields should be studied as integrated systems – yes, generally, I agree, but there are exceptions. “40” in Russian is not remotely like “20, 30, 50, 60, 70, 80, 90” (which are all perfectly regular, as they are – together with “40” – in other Slavic languages); it’s a completely different word that used to designate a unit.

    2) Using a pseudonym is orthogonal to being a troll. To claim otherwise bespeaks a Mark-Zuckerberg-level failure to think things through.

  241. @Lars

    “A troll in the usual sense is someone who enters an online discussion forum with posts designed to infuriate the regular participants and create drama, especially by reiterating a absurd standpoint and refusing to accept arguments to the contrary.”

    This is an example of an absurd standpoint. I gave you a definition of a “troll” from a principal web source. You countered with a definition of your own making. The Wiki definition says nothing about a “regular” participant. Trolls can get regular if they are not banned. Who cares if a participant is regular. It’s more important if s/he is credentialized, curious, ethical, good listener, etc. SFReader interrupted a discussion (fact) by attempting to compromise a participant (inflammatory, deliberate, intentional) and shift it from a discussion about languages to a discussion about a participant and still another person who is not even part of this conversation or this topic (biochemist Klyosov).

    And languagehat should have thanked me, and not SFReader, for making a full account of the Klesov controversy openly and freely available in one place (my own website). Instead he chose to thank a troll.

  242. @David Marjanovic

    “Semantic fields should be studied as integrated systems – yes, generally, I agree, but there are exceptions. “40″ in Russian is not remotely like “20, 30, 50, 60, 70, 80, 90″ (which are all perfectly regular, as they are – together with “40″ – in other Slavic languages); it’s a completely different word that used to designate a unit.”

    It a valid counterpoint and counterexample.

  243. The notion that SFReader is a troll is absurd. And it’s ironic that anyone claiming such linguistic expertise would be unable to figure out the meaning of “troll.”

  244. @Brett

    “The notion that SFReader is a troll is absurd.”

    Why is it absurd?

    “unable to figure out the meaning of “troll.””

    It’s been figured out. You are just posting gibberish.

  245. a person who sows discord on the Internet by starting arguments or upsetting people,posting inflammatory,[1] extraneous, or off-topic messages in an online community (such as a newsgroup, forum, chat room, or blog) with the deliberate intent of provoking readers into an emotional response[2] or of otherwise disrupting normal on-topic discussion,[3]

    Yet there was no discord until your arrival. No one is inflamed but you. The messages you don’t like are not extraneous nor off-topic, and you don’t have to believe me, our host has already told you so. Finally, no one else has provided an emotional response, nor has normal on-topic discussion been disrupted. Nor is it likely to be.

    It’s not just that SFReader has been here for years and we know his strengths and weaknesses (he’s obviously not the perfect commenter, but who is?). It’s that all of us are singing the same tune here about him except you. Yet this tells you nothing. In short, the evidence is against you. Time to re-evaluate your theory.

  246. @John Cowan

    You are wrong and here’s why.

    “Yet there was no discord until your arrival.”

    Discord emerged only after SF Reader’s posts. Before it was a discussion. A discussion presupposes disagreements, otherwise what’s the point of debating.

    “No one is inflamed but you.”

    SFReader’s posts were inflammatory. This doesn’t mean that I got inflamed. I just had all the reasons to.

    “The messages you don’t like are not extraneous nor off-topic.”

    They were off-topic. Nothing in SFReader’s posts had anything to do with what David Marjanovic, Marie-Lucie and I were talking about. Or with what Gary Moore posted earlier (which attracted my attention).

    “The messages you don’t like are not extraneous nor off-topic.”

    People’s opinions are not evidence. You seem to be living on some kind of anti-science planet.

    “Time to re-evaluate your theory.”

    Why? It’s being proved (sic!) with every post. Since SFReader’s posts we are only talking about him and his posts. That’s a troll’s ultimate goal. And I’m willing to let it roll so you can see it with your own eyes.

  247. Troll kalla mik
    trungl sjǫtrungnis,
    auðsug jǫtuns,
    élsólar bǫl,
    vilsinn vǫlu,
    vǫrð nafjarðar,
    hvélsveg himins –
    hvat’s troll nema þat?

    They call me a troll,
    moon of the earth-Hrungnir
    wealth sucker of the giant,
    destroyer of the storm-sun
    beloved follower of the seeress,
    guardian of the “nafjord”
    swallower of the wheel of heaven.
    What’s a troll if not that?

  248. marie-lucie says:

    GD: Mrs. Tarpent (if I may)

    Please don’t..

  249. marie-lucie says:

    GD: back to comparative linguistics:

    PIE *krd- ‘heart’ shows up as *srudice in Slavic and sirdis in Lithuanian. So, the velar is presumably palatovelar (PIE *k’rd-) because the so-called satem languages (named so precisely because they turn up palatals where Greek, Latin, etc. have plain velars) show s- and not k-.

    Indeed.

    then there’s another cognate set represented, e.g., by Gk rhadix and Lat radix ‘root’. The Greek form presupposes *s (< *sradiks), which is presumably not the same /s/ as the satem /s/ < PIE *k' but PIE spirant *s.

    I don’t have any expertise on Greek anything, but it seems to me that Greek only admits “rh” as word initial, not plain “r”, or am I wrong? And this is not a cluster, but an aspirate (should be written with superscript h if possible with the font). So even if all or most occurrences of Greek “h” derive from PIE *s, that does not mean that “rh” is actually *hr, itself from *sr. I think there are other languages pointing to reconstructable PIE *sr, eg the ancestor of English stream. But rhad-iks corresponds to English root, not “*stroot”.
    Sorry I am not competent to go any further.

    the searching for potential Eskimo-Khoisan parallels can be done by a host of amateurs

    We need well-trained people familiar with many languages and with the kinds of strange things that can happen, not a host of amateurs who will look for lookalikes/meanalikes and find them in all the wrong places.

  250. @German, thanks for your opinion, but you seem to be in a minority with it. A minority of one. If that doesn’t make a bell ring, I have a shorter version:

    NO U!

  251. @Lars, “but you seem to be in a minority with it.”

    This is easy to explain: SFReader attacked only me. Had he tried to compromise everybody, he would have been a minority, not me. The problem is that the rest of you are supporting his unethical behavior only because he’s been around for a while. This is just forum politics. But it’d be great to see sort of objectivity and integrity here.

    @Marie-Lucie

    r is indeed aspirated in the beginning but also in the middle (rrh) of Greek words as in diarrhea.” This means that *s > hr in Greek and not *s > r as it happened in Latin (radix) where *s is always lost in the onset of a word before sonorants. But we do need to check the hypothesis against other phonetic environments. And we indeed find support for it in Gk hippos ‘horse’ (Lat equus), which I compare not just with Skrt asva but also with Skrt sapho- ‘hoof’ (Slav kopyto, with an ending just like in Gk hipotes). Not only do we have Gk h ~ Satem *s again, but also centum kw > p in Satem languages! Not only does the 19th century belief in satem languages disappears under a 21 century methodological scrutiny, but so does the 19th century belief in kentum languages! Instead, what we have is a positional conversion of velars into palatalovelars or into labiovelars on the PIE level (!) in response on some – admittedly still poorly understood – suprasegmental pressures that result in empirically attested doublets (comp. the mobility of aspiration in Gk thriks vs. trikhos ‘hair’). Now it becomes clear why even some Anatolian languages (Luwian) show satemization signs (HLuw azuwa ‘horse’), while even such heavily satemized languages as Slavic often show centum-like reflexes in forms where Sanskrit has a satem /s/ (kopyto ‘hoof’ above next to Skrt sapha-).

    I obviously have other examples to support both developments, but for it suffices to say that the methodological flaw that I’m correcting is that a generative, etymological connection has to be “baked” into the cognate set composition and not added after a cognate set had already been defined on the basis of subjective and ahistorical “similarity in sound and meaning” assumptions. The “old” approach leaves forms without good etymologies and generates many phonological anomalies (-h- in Skrt sapha- or s in Skrt asva), which linguists armed with 19th century methods tend to dance around inventing ad hoc developments to explain them.

    “We need well-trained people familiar with many languages and with the kinds of strange things that can happen, not a host of amateurs who will look for lookalikes/meanalikes and find them in all the wrong places.”

    Yes, I agree that some level of linguistic sophistication is required from amateurs to make them useful to the academic linguistic community (just like some genome bloggers know enough about genes and stats to make professional geneticists visit their websites and read their posts and look at their PCAs). I guess my thoughts on this matter were inspired by an experience in Merritt Ruhlen’s class at Stanford where he showed Greenberg’s method of multilateral comparison. And I thought: geez, you don’t have to be a trained linguist to do it, but you do have to have a lot of passion for the linguistic material. 🙂

  252. So why do you have both aśva- and śapha- in Skt?

  253. I guess my thoughts on this matter were inspired by an experience in Merritt Ruhlen’s class at Stanford where he showed Greenberg’s method of multilateral comparison. And I thought: geez, you don’t have to be a trained linguist to do it, but you do have to have a lot of passion for the linguistic material.

    That accounts for a great deal. Yes, multilateral comparison is easy enough. But it’s also useless as a method of choosing between hypotheses. At most it can generate hypotheses to work on, but it also generates a vast amount of junk hypotheses.

  254. an old troll says:

    The part which caused a strong emotional reaction in myself, German, was your post about poor Kassian snagged by the ugly flamewars, losing grant support, and threatened with loss of his academic position – and about your enthusiastic support for all these despicable developments. My gut reaction is that whenever there is such a campaign of concerted denunciations to the real-world employers, then it may be too dangerous to stay involved, for anyone can become the next target. It is very emotionally distressing, and it was you, German, who brought it up, and who made me break with my long tradition of posting at LH openly, and to switch to an anon handle 🙁

  255. @minus273

    I honestly don’t fully know yet. That’s why I wrote “admittedly still poorly understood.”

    Note that Avestan has aspa ‘horse’ which is an even more literal match for Skrt sapha- ‘hoof’. I would re-write Skrt sapha- as *sahpa- (perceptually likely identical) to align it with asva. The simplest explanation is that PIndAr *saskwo- dissimilated rightwards into asva and leftwards into sapha- in Indic. But I’m not sure if those were indeed /s/.

    In the light of Skrt sapha-, the always-problematic geminated -pp- in Gk hippos can be understood as -pp- < *-hp- as in *hihpos < *hihkwos.

    We need to pull in more material and look at a bigger picture. A small methodological change yields a butterfly effect across the whole corpus. But first thing first: build a cognate set correctly, build another one, then another one, etc., then derive a correct sound law from them. Just don't be scared by a doublet. Doublets happen, we just need to explain them phonetically.

  256. @John Cowan

    “Yes, multilateral comparison is easy enough. But it’s also useless as a method of choosing between hypotheses. At most it can generate hypotheses to work on, but it also generates a vast amount of junk hypotheses.”

    You are right. That’s why I think amateurs could generate hypotheses, while professionals could test them but also work on methodologies, etc, to ultimately make amateurs more effective. It was just sad to see such competent and rigorous people as Joe and Merritt get mired in hypothesis generation using a massive amount of often opaque and intricate material, while trying to do hypothesis testing and proving at the same time.

  257. It’s still quite hard to understand. For example, why is what you write as s pronounced as ś when it corresponds to Greek k‘s and as s when it corresponds to Greek aitches?

  258. marie-lucie says:

    GD: I guess my thoughts on this matter were inspired by an experience in Merritt Ruhlen’s class at Stanford where he showed Greenberg’s method of multilateral comparison.

    I second JC’s comment That accounts for a great deal.

    And I thought: geez, you don’t have to be a trained linguist to do it

    That’s exactly the attitude Ruhlen tries to foster (thereby downgrading the work of linguists) in his book on the origins of language (a misleading title). He takes the reader through some (carefully chosen) material, adding here and there “linguists know that (such and such a rule applies)”, something the reader could not possibly have thought up, and ends up saying “See? you don’t have to have studied all those languages, you can do better than those old fuddy-duddies!” (or words to that effect). His approach reminds me of what adults do when they are playing a game with small children and arrange to let them win.

    amateurs could generate hypotheses, while professionals could test them but also work on methodologies, etc, to ultimately make amateurs more effective

    This reminds me of Ives Goddard’s reply to someone asking him about why he was not debating Greenberg (about Language in the Americas): something like “Geographers don’t feel compelled to debate the Flat Earth Society”.

    such competent and rigorous people as Joe and Merritt

    Perhaps they appeared that way to students who did not know any better. “Rigorous” is not a term I would associate with either of them. (See for instance my comment on Greenberg’s Amerind subclassification in this thread a short while ago).

    That said, I agree with some of Greenberg’s criticisms about the state of historical linguistic methods, but he seems to have confused the situation among some (not all, as he suggests) linguists dealing with Amerindian (NOT “Amerind”) languages (former students of Mary Haas in particular) with the state of historical linguistics in general. In my opinion, there is nothing wrong with the traditional (IE-inspired) methods, except that they are often misunderstood and oversimplified (eg by the above-mentioned linguists). Reliance on “cognate sets” (dealing with lexical items only) while ignoring morphology is a major fault.

  259. But then there’s another cognate set represented, e.g., by Gk rhadix and Lat radix ‘root’. The Greek form presupposes *s (< *sradiks)

    It doesn’t, because all initial rs in Greek are written as aspirated (or voiceless — whatever exactly ῥ stood for). In this case there is plenty of evidence (including English wort) that it’s actually from PIE *wr-.

  260. @Marie-Lucie

    “I second JC’s comment That accounts for a great deal.”

    I guess I didn’t understand what JC means and now what you mean by that. I learned linguistics not from Merritt. But I did talk to Greenberg quite a bit and took Merritt’s class to have a first-hand knowledge of the way they worked. And I agree with you that Greenberg made some good criticisms of the comparative method.

    “This reminds me of Ives Goddard’s reply to someone asking him about why he was not debating Greenberg (about Language in the Americas): something like “Geographers don’t feel compelled to debate the Flat Earth Society”.”

    This varies by linguist. LC debated them quite a bit.

    “Rigorous” is not a term I would associate with either of them. (See for instance my comment on Greenberg’s Amerind subclassification in this thread a short while ago).”

    My choice of a descriptor may not be right. But Greenberg made a significant contribution to the study of typology, so I presume he was rigorous or not depending on the kind of data and methodology he worked with. Ruhlen has pretty good logical thinking (and he may have had a math degree before switching to linguistics).

  261. @TR

    “In this case there is plenty of evidence (including English wort) that it’s actually from PIE *wr-.”

    No, you are wrong methodologically. You artificially split cognate sets into partial ones and arrive at wrong reconstructions. That’s my whole point.

  262. ““This reminds me of Ives Goddard’s reply to someone asking him about why he was not debating Greenberg (about Language in the Americas): something like “Geographers don’t feel compelled to debate the Flat Earth Society”.”

    This varies by linguist. LC debated them quite a bit.”

    I would even say that Goddard debated a bit too little and LC debated a bit too much.

  263. Then I don’t understand your point. What is gained by discarding a perfectly good cognate set like this one, based on close semantic relationship and regular sound changes for which there’s plenty of other evidence?

  264. @TR

    I’m not discarding them but the set you have in mind is a subset of a real, natural, historical set. The two sets (the HEART set and the ROOT set) were misanalyzed as two separate, unrelated sets. Systematic morphological, semantic and phonetic resemblances between the two have been overlooked. Phonetic problems (e.g., Skrt hrd ‘heart’ that can’t be derived from PIE *k’rd but it is undoubtedly part of the HEART set) have been ignored or underestimated. The loss of s- in some forms within that set needs to be explained. I hypothetically think that there was some kind of dissimilatory process that involved /s/ and “laryngeals.” A laryngeal is reconstructed in the ROOT set, which looks like *wreH2d-.

  265. marie-lucie says:

    GD: Greenberg made a significant contribution to the study of typology, so I presume he was rigorous

    Nobody questions Greenberg’s significant contribution to typology. Classification and reconstruction are another matter.

    I agree with you that Greenberg made some good criticisms of the comparative method.

    I was not criticizing the comparative method (as practiced and refined by traditionally trained historical linguists), but a flawed, simplified interpretation of the CM as practiced by some Amerindianists, which Greenberg appeared to confuse with the full traditional method. Incidentally, this confusion shows that he was not himself trained in the method (something abundantly confirmed by his and Ruhlen’s “Amerind cognate sets” and reconstructions).

  266. @Marie-Lucie

    “I was not criticizing the comparative method (as practiced and refined by traditionally trained historical linguists), but a flawed, simplified interpretation of the CM as practiced by some Amerindianists, which Greenberg appeared to confuse with the full traditional method.”

    OK, thanks for the clarification. I personally think that Greenberg was close to truth when he argued that classification precedes reconstruction. He just misapplied this idea. Let me explain. I also believe that classification precedes reconstruction and my critique of the comparative method identifies an error precisely at the classification stage. But unlike Greenberg, and the difference here is crucial, I think that it’s cognate sets that are the minimal units of classification and not languages.

  267. I’m not discarding them but the set you have in mind is a subset of a real, natural, historical set. The two sets (the HEART set and the ROOT set) were misanalyzed as two separate, unrelated sets.

    So the same PIE root gave rise to one set of reflexes that look like they reflect *wr- and another set that look like they reflect *k’r-? And somehow *s- is involved too (I don’t understand this part)? Maybe you have more data or arguments that you haven’t adduced, but unless you have a good solution to the phonological difficulties, I don’t see how you can declare so confidently that these are the same cognate set rather than what they look like, i.e. two unrelated sets.

  268. marie-lucie says:

    GD: I too agree that classification precedes reconstruction. You can’t even think of reconstructing a proto-language, the ancestor of a language group, unless you have good reason to believe that the group does consist of related languages. It is relatively easy to group languages which are indeed closely related (but unusual phonological correspondences or the presence of a substantial number of borrowings can obscure the relationship if “cognate sets” is all you look at). It is much more difficult to prove or even suspect more distant relationships. Earlier I cited some of the problems with some of Greenberg’s subclassifications within “Amerind”, which do not inspire confidence in the validity of the larger group.

  269. marie-lucie says:

    TR: I agree with your comment. I too have trouble understanding how to deal wirh the s’s.

    About HEART and ROOT, the alleged semantic relationship seems very dubious.

  270. @Marie-Lucie

    ” You can’t even think of reconstructing a proto-language, the ancestor of a language group, unless you have good reason to believe that the group does consist of related languages.”

    Precisely. The same, IMO, works for cognate sets: you can’t even think of reconstructing an etymon and formulating a phonetic law, unless you have a good reason to believe that the cognate set(s) that form a basis for them is/are fairly complete. But routine comparative method is not diligent with criteria for inclusion/exclusion of items in cognate sets, hence the reconstruction may get wrong. This in turn obscures long-distance relationships (of those cognate sets!).

  271. @Marie-Lucie

    “About HEART and ROOT, the alleged semantic relationship seems very dubious.”

    Not at all. The HEART set also has meanings such as MIDDLE or CORE (including their arboreal applications) and the ROOT set has somatonymic meanings such as “wart”. Also comp. “The heart of the matter” and “the root of the problem.” An etymological meaning that cuts across the two sets could easily be “to grow.”

    Forms with s- and without s- are typical in IE langages and sometimes described as the s-mobile problem. I’m not trying to hide behind it, though. On the contrary, a chronically unresolved phonetic problem can be resolved with a change in the approach.

  272. “An etymological meaning that cuts across the two sets could easily be “to grow.””

    Or even better ‘source’.

  273. Which reflexes in either set even show s-mobile? In any case, the main problem with uniting these two sets isn’t s-mobile, or even the semantics, but the fact that *wreh2d- and *k’erd are phonologically irreconcilable.

  274. By the way, the initial *w in the Greek “root” family is guaranteed by Aeolic βρίζα and Mycenaean wi-ri-za, on the assumption (not certain, however, because of the vocalism) that these and their Attic cognate ῥίζα — all meaning “root” — are related to ῥάδιξ.

  275. marie-lucie says:

    TR: Which reflexes in either set even show s-mobile?

    Indeed. I don’t see it in those examples.

    <i.By the way, the initial *w in the Greek “root” family is guaranteed by Aeolic βρίζα and Mycenaean wi-ri-za, on the assumption (not certain, however, because of the vocalism) that these and their Attic cognate ῥίζα — all meaning “root” — are related to ῥάδιξ.

    Thank you, I was wondering about that *w. Is there a regular internal alternation in Greek between ζ [dz] and δ (d)?
    ῥίζα: Now I understand “glycorrhiza” ‘liquorice’.

  276. Is there a regular internal alternation in Greek between ζ [dz] and δ (d)?

    There is a regular sound change whereby Proto-Greek *dy (and *gy) > ζ. Presumably ῥίζα is derived with the so-called devi suffix, which in Proto-Greek is *-ya (PIE *-ih2).

    Now I understand “glycorrhiza” ‘liquorice’

    I never realized that!

  277. David Marjanović says:

    Still no time to deal with everything…

    “Yet there was no discord until your arrival.”

    Discord emerged only after SF Reader’s posts. Before it was a discussion. A discussion presupposes disagreements, otherwise what’s the point of debating.

    Dude. You are the first person in the history of this blog to feel trolled by SFReader. His presence here dates back… at least a year, likely more. This thread doesn’t exist in a vacuum!

    Troll kalla mik […]

    I laughed out loud.

    it seems to me that Greek only admits “rh” as word initial, not plain “r”

    Correct. The trick is that PIE didn’t allow plain */r/ as a word onset either. Laryngeal + */r/ was allowed, but would have yielded vowel + voiced /r/ in Greek (as it did in érythros < *h₁rudh- “red ~ crimson”). So, */sr/ is an obvious option.

    But so is */wr/.

    r is indeed aspirated in the beginning but also in the middle (rrh) of Greek words as in diarrhea.

    That’s the middle of a compound word: dia- “through” + rhe- “flow”. Does rh occur in the middle of any morphemes?

    And we indeed find support for it in Gk hippos ‘horse’ (Lat equus), which I compare not just with Skrt asva but also with Skrt sapho- ‘hoof’ (Slav kopyto, with an ending just like in Gk hipotes). Not only do we have Gk h ~ Satem *s again, but also centum kw > p in Satem languages!

    So why doesn’t aśva have this p?

    Now it becomes clear why even some Anatolian languages (Luwian) show satemization signs (HLuw azuwa ‘horse’)

    The Luwian branch of Anatolian doesn’t merely show “satemization signs”, it shows the whole sound shift of palatalized velars to z in all environments that they have been found in so far (in the limited corpus). It’s a pretty obvious, “natural” thing for palatalized velars to do, don’t you think?

    I obviously have other examples to support both developments

    I’d be very interested in seeing them.

    The “old” approach leaves forms without good etymologies and generates many phonological anomalies (-h- in Skrt sapha- or s in Skrt asva)

    Well, there is no “h” in śapʰa-. Of course comes from PIE *pH, meaning *ph₂ and possibly *ph₁. If you really want to compare śapʰa- with *kopyto despite the latter’s *k… how would the laryngeal be reflected in Slavic? The y conveniently derives from a long vowel; which tone does it have?

    The ś in aśva is not at all anomalous if you derive the word from PIE *(h₁)ékʲwos, as current textbook wisdom has it. Please explain what you mean.

    The simplest explanation is that PIndAr *saskwo- dissimilated rightwards into asva and leftwards into sapha- in Indic. But I’m not sure if those were indeed /s/.

    *blink*

    …Just to make sure: you are aware that s /s/ and ś /ɕ/ are two completely separate phonemes in Sanskrit, right? Russian (in not too conservative accents) happens to use both sounds as phonemes, too: с, щ.

    In the light of Skrt sapha-, the always-problematic geminated -pp- in Gk hippos

    I thought consonant lengthening is completely regular in Greek after a stressed short vowel? There are problems in this word, but they’re the /h/ (which comes out of nowhere, but is a late innovation, absent from the name Álkippos) and the /i/ (/e/ would be expected).

    can be understood as -pp- < *-hp- as in *hihpos < *hihkwos.

    Again, what “h”? If you mean a laryngeal (any one), it would have disappeared with compensatory lengthening of the preceding vowel, and that didn’t happen.

    I would re-write Skrt sapha- as *sahpa- (perceptually likely identical) to align it with asva.

    But *-aHp- would have yielded -āp-, not -apʰ-, probably before *h₂ actually became [h]. I would certainly not call [χp] and [pχ] “perceptually likely identical”.

  278. at least a year, likely more

    Since 2012, thanks to the wonders of binary search through dates.

  279. marie-lucie says:

    Thanks David.

    GD: About “horse” and “hoof”, I don’t see why the Slavic, etc words have to be related to each other any more than the English ones. Or than “heart” and “root”. When two words are similar in one language but their potential cognates are quite different in one or more others, the obvious conclusion is that they must come from different ancestors. Especially if making them compatible with a common origin requires tweaking otherwise well-attested rules or introducing extraneous elements.

  280. Does rh occur in the middle of any morphemes?

    Only as part of a geminate, e.g. ἄρρην “male”. Breathings are a late and not wholly reliable notation, but the appearance of rough breathing over rho is completely predictable: word-initially and in the second half of a geminate (written ῤῥ). You never get a singleton non-initial ῥ.

    I thought consonant lengthening is completely regular in Greek after a stressed short vowel?

    That’s the first I’ve heard of a littera-type rule in Greek, but in any case the geminate in hippos isn’t a problem, as you say, given that it reflects two PIE segments rather than a single labiovelar.

  281. Sorry, I had to update my OS, hence picking up the thread only now.

    @TR

    “the main problem with uniting these two sets isn’t s-mobile, or even the semantics, but the fact that *wreh2d- and *k’erd are phonologically irreconcilable.”

    The truth is they are very much reconcilable – morphologically (Gk/Lat *sradiks and Slav *srudice are identical), semantically (see above) and phonetically. Slav *or- (*ord-ti ‘grow’) and Germanic *ur- (*wurtiz, with w- as a prothetic) are perfectly regular reflexes of syllabic R also found in *kRd ‘heart’. For Gk, I’d postulate the sequence k’ > s > h, which is similar to k’ > s > sh in Lithuanian or Indic. The k- vs. s- doublets are well know from satem languages (Slav *korwa vs. *serna), so the copresence of kardia and rhadiks in Greek is not extraordinary. /s/ is regularly lost in front of a sonorant in Latin. The phenomenon of s-mobile (mostly initial) is pervasive in IE languages (Slav *vendati ~ Goth swintan ‘disappear, wane’): I’m not abusing the similarity and not hiding behind this poorly understood phenomenon, but believe that we need more data to understand the contexts in which /s/ is lost. I’m not providing this additional data. Interestingly, palatalization is conventionally thought of as getting lost in kentum languages, so k’ > k already presumes as s-mobile-like phenomenon.

    @ David

    “So why doesn’t aśva have this p?”

    As we know from kentum languages, labiovelars sometimes stay as k(u), sometimes change to p.

    “I’d be very interested in seeing them.”

    Another good one is IE *esH2r-/-n (Hitt eshar, Skrt asrk, Gk ear, etc.) ‘blood’ and IE *(d)akr-/-n ‘tear’ (Hitt ishahru, Skrt asru, Gk dakruma, etc.) The two sets have already been compared to each other but only as a case of mutual contamination. The morphology is identical (heteroclitic -r-/-n-). IE *kmtom or *kntm ‘hundred’ can be compared to IE *seno- ‘old’ (Lat senatus showing the same extension -t- as centum). IE *swekuro- ‘husband’s parent’ ~ IE *sweso:r ‘sister’ (Initial sh- in Lith sesuras and Skrt svasura does not need to be interpreted as assimilation from medial -s- but a perfect pair to Gk h in hekura-.)

    I’d keep Aeolic βρίζα out of it for now, although I agree that the connection is tempting.

  282. @ David

    I continue… My comp is acting funny…

    “If you really want to compare śapʰa- with *kopyto despite the latter’s *k…”

    It’s a standard cognate set. There are a number of cases where k ~ s alternate within a “satem” or between “satem” languages. Again, there has not been a good explanation for it.

    “you are aware that s /s/ and ś /ɕ/ are two completely separate phonemes in Sanskrit, right?”

    yes, of course.

    “The ś in aśva is not at all anomalous if you derive the word from PIE *(h₁)ékʲwos, as current textbook wisdom has it. Please explain what you mean.”

    Your reconstruction doesn’t take into consideration the fact that Latin, Greek and Mycenaean attest for a labiovelar and not a cluster. We should have ak(w)a** in Sanskrit, not asva. Something else is clearly going on here.

  283. So the price of uniting the “root” and “heart” sets includes at least (a) banishing ῥίζα “root” from the former set because of its *w-; (b) divorcing ῥάδιξ from the set of English wort etc., where it fits straightforwardly, in order to give it an *s- instead; and (c) positing an irregular pre-Proto-Greek change of PIE *k’ > *s just to account for that *s-. And presumably some further fiddling to get rid of the laryngeal that shows up in the “root” words but not the “heard” words. All because you feel that “root” and “heart” are semantically close and because they share two segments. I still don’t see how s-mobile has anything to do with it, but honestly, this doesn’t inspire confidence in your new and improved methodologies for historical linguistics.

  284. @David

    “I thought consonant lengthening is completely regular in Greek after a stressed short vowel?”

    How about melo? hippos has a triple problem: h-, i- and -pp-. Pretty much the whole word. Time to rethink the fundamentals.

    “But *-aHp- would have yielded -āp-, not -apʰ-, probably before *h₂ actually became [h]. I would certainly not call [χp] and [pχ] “perceptually likely identical”.

    Yes, good thinking. Like I said, I don’t have all the answers. We need to look at a number of cognate supersets and revisit PIE phonology. -ph- in sapha- is not expected, neither is h- in Skrt hrd ‘heart’ but both seem to correspond to /s/ as seen in Slav *srudice and Skrt asva.

  285. @TR

    “divorcing ῥάδιξ from the set of English wort etc., where it fits straightforwardly,”

    Who’s divorcing it? Not me.

    “And presumably some further fiddling to get rid of the laryngeal that shows up in the “root” words but not the “heard” words.”

    You are rushing through issues. h- in Gk hippos, rhadiks is taken as a reflex of a laryngeal (H1) in a number of recent linguistic works, so we have a laryngeal in both HEART and ROOT sets.

    “positing an irregular pre-Proto-Greek change of PIE *k’ > *s”

    Why is it irregular? It’s more parsimonious (if supported by right data, of course), then postulating two /s/ – one PIE, the other one ‘satem’.

    “this doesn’t inspire confidence in your new and improved methodologies for historical linguistics.”

    Fine, at this point I would settle for your anonymous curiosity, rather than your anonymous confidence. 🙂

  286. So what is the actual form of the PIE root which you think underlies both sets, and how does it give rise both to words which seem to reflect *k’- as well as to words which seem to reflect *w-?

    h- in Gk hippos, rhadiks is taken as a reflex of a laryngeal (H1) in a number of recent linguistic works, so we have a laryngeal in both HEART and ROOT sets.

    Wait, now hippos is part of the “heart” set?! Surely not, so which “heart” word shows a laryngeal?

    Why is it irregular?

    Because PIE *k’ is otherwise not known to give Greek s, but Greek k. You’re positing an ad hoc change for the purpose of this single etymology.

    Btw, I’ve finally looked up Greek ῥάδιξ (which sounded unfamiliar) and it turns out to be a post-Classical word first attested in Nicander (2nd century BC), and very likely a borrowing from Latin.

  287. marie-lucie says:

    GD: h- in Gk hippos, rhadiks is taken as a reflex of a laryngeal (H1) in a number of recent linguistic works

    So rhadiks is no longer from *sr- ??

  288. Btw, a recent attempt by Chiara Bozzone to explain the initial aspirate of hippos can be found here. I’ve yet to fully think it through, but it looks interesting.

  289. marie-lucie says:

    TR: Thanks for looking up rhadiks! A borrowing from Latin makes the best sense.

  290. anonymous curiosity […] anonymous confidence

    Pseudonymous people, particularly lightly pseudonymous people who are using abbreviations of their real names, are by no means anonymous. A statistician who heard talk of “Gosset’s t” would be completely confused; it is always called “Student’s t“, though “Student” was a pseudonym for William Sealy Gosset (his employer was concerned that his work might reflect badly on them, perhaps because they did not understand it). The work of Nikolas Bourbaki is too well-known to discuss. And once we get out of science and mathematics, non-anonymous authors who use pseudonyms are everywhere.

    If you insist on real names, consider visiting Paleoglot.

    —John Woldemar Cowan, writing under the pseudonym “John Cowan”

  291. (Bozzone and Felisari, I should have said). The troublesome i, as mentioned in that paper, has been explained as a case of schwa epenthesis in a zero-grade form *h1k’wó-, for which there are other Greek parallels probably including our word ῥίζα; though I would want to see a good explanation of the subsequent accent retraction.

  292. @Marie-Lucie,

    “So rhadiks is no longer from *sr- ??”

    My bad. I created a confusion. We don’t need this. There’s a simpler way to counter TR’s objection: -a- in Gk rhadiks (just like -a- in Gk kardia, per the conventional reconstruction) come from syllabic R, which regularly gives ar in Gk and ur in Germanic (*wurtiz). TR assumes that Gk h corresponds to w and postulates *wreH2d- for the ROOT set. But in fact w (as I mentioned above) is just a prothesis in front of -ur-, which is a regular reflex of syllabic R. So, the only “mismatch” between germ *wurtiz and, say, Slav *srudice is the Slavic s- vs. Germanic o (zero)- (not w-).

    @TR

    “Because PIE *k’ is otherwise not known to give Greek s, but Greek k. You’re positing an ad hoc change for the purpose of this single etymology.”

    I’m giving examples where Gk /h/ corresponds to /k/. It’s not a single etymology. Of course I’m shooting for having it regular. We’re just looking at this one at the moment. A moment ago we were looking at hippos vs. sapha- (< *k'-). Stay tuned.

  293. marie-lucie says:

    GD: a- in Gk rhadiks (just like -a- in Gk kardia, per the conventional reconstruction) come from syllabic R, which regularly gives ar in Gk

    Even if rhadiks can be assumed to be a genuine Greek word rather than a Latin borrowing, your rule should give ardiks not r(h)adiks. Or hardiks, if the word cannot start with a vowel?

  294. @TR

    “Btw, a recent attempt by Chiara Bozzone to explain the initial aspirate of hippos can be found here.”

    I read it some time ago and had it in mind when I took a wrong curve and began talking about H1.

    @John Cowan

    “If you insist on real names…”

    Online pseudonyms are indeed annoying, especially when their owners make metascientific claims, passionately defend mainstream views that give them comfort but that they didn’t take part in building or launch ad hominem attacks. The use of a pseudonym is excusable when their owners are not making metascientific claims, don’t glamorize mainstream views or launch ad hominem attacks but communicate something concrete, valuable and memorable. Then they begin building a new equity behind their abbreviated name.

  295. Kindly consider languagehat, our gracious pseudonymous host, who was in fact anonymous for the first several years of this blog, to no one’s detriment.

  296. @Marie-Lucie

    “Even if rhadiks can be assumed to be a genuine Greek word rather than a Latin borrowing, your rule should give ardiks not r(h)adiks.”

    Its immediate cognates such as Slav *ord-ti (next to derived *rodu, *rasti) ‘grow’ and Germ *wurtiz have the shape that you are describing. The shape of Gk rhadiks and Lat radix is different from those but approximates Slav *srudice. We can’t divorce the two half-sets from each other by this measure.

  297. Athel Cornish-Bowden says:

    marie-lucie: That’s exactly the attitude Ruhlen tries to foster (thereby downgrading the work of linguists) in his book on the origins of language (a misleading title). He takes the reader through some (carefully chosen) material, adding here and there “linguists know that (such and such a rule applies)”…

    I realize that Ruhlen is not the most admired figure in the linguistics community (to put it mildly), but there is one point of fact I’d like to raise with you. He claims that when Greenberg first proposed his classification of African languages he was subjected to much the same sort of criticisms as later greeted his classification of American languages, but that now everyone agrees that his classification of African languages is correct. Is that in fact true, and is there a qualitative difference between African and American families that would make one correct and the other not?

  298. Athel Cornish-Bowden says:

    A statistician who heard talk of “Gosset’s t” would be completely confused; it is always called “Student’s t“, though “Student” was a pseudonym for William Sealy Gosset (his employer was concerned that his work might reflect badly on them, perhaps because they did not understand it).

    I’m surprised they would care. Did they think they would sell less Guinness if people knew that one of their employees was a distinguished statistician? How many Guinness drinkers follow the statistics literature? (I do, or at least I once did, but I don’t suppose I’m typical.)

  299. much the same sort of criticisms as later greeted his classification of American languages, but that now everyone agrees that his classification of African languages is correct

    I think there was controversy early, then a period of quiet, then doubts, and now open (though not noisy) rebellion.

    Afro-Asiatic has never been controversial, except for the shaky status of the Omotic family within it: most don’t go so far as to say Omotic is an isolate family, but some do, and certainly the quantity and quality of the evidence tying it to (the rest of) Afro-Asiatic is not very great.

    Niger-Congo is, as far as I know, fairly uncontroversial as a genetic grouping, though there is no consensus about its internal classification. There may be noise around the edges of this group as well. It’s in this family that Greenberg did most of his actual work, drawing in languages that were formerly not thought to be related to the core group whose recognition predates him.

    Nilo-Saharan is increasingly recognized as what biologists call a wastebasket taxon: a language spoken in the rough and fractal borderland between the first two families that did not look like either of them tended to get dumped into Nilo-Saharan. It should probably be broken up until proper comparative work is done. Some linguists do defend it as written, however.

    Khoi-San is typologically defined by a single property, the presence of click consonants. Nobody has ever found (or thought of, depending on how you look at it) a plausible phonological process by which non-clicks could be transformed into clicks, and so the assumption has been that it must be a primitive character. (That in itself would not be enough to demonstrate relatedness anyway, as it could have survived from the ancestral language in these groups but not others, as five-toedness survives from the ancestor of all tetrapods.) The languages are intensely diverse otherwise, and there seems to be nobody who believes any more that they can possibly all be related.

  300. I just read presentation slides by Blench drawing large conclusions from the worldwide pattern of isolates and small families. There are lots in the New World, few in Africa, concentrated in the north in Australia, and so on. But I greatly fear that what he is really investigating is the worldwide pattern of resistance among specialists to Greenberg’s lumping: very high in the New World, until recently low in Africa, mixed in Australia, and so on.

  301. I didn’t have the Student story quite right (post in haste, repent at leisure). Here’s Wikipedia (emphasis added):

    Another researcher at Guinness had previously published a paper containing trade secrets of the Guinness brewery. To prevent further disclosure of confidential information, Guinness prohibited its employees from publishing any papers regardless of the contained information. However, after pleading with the brewery and explaining that his mathematical and philosophical conclusions were of no possible practical use to competing brewers, he was allowed to publish them, but under a pseudonym (“Student”), to avoid difficulties with the rest of the staff.

  302. Online pseudonyms are indeed annoying, especially when their owners make metascientific claims, passionately defend mainstream views that give them comfort but that they didn’t take part in building or launch ad hominem attacks.

    The only person launching ad hominem attacks in this thread is you. I don’t care whether your real name is “German Dziebel” or not; I suppose if I googled it I’d find there is such a person, but how do I know you are that person? I could post as “Barack Obama”: a real name! I don’t care what name people post under (I still miss a commenter from the early days of the site who went by “a”), I only care that they not behave as jerks. Frankly, you have been behaving as a jerk far too often; I’m letting the conversation proceed because it’s interesting and everybody seems to be more amused than annoyed by your heavy-handed attempts to throw your weight around, but if SFReader or any other long-standing and valued commenter feels offended and asks me to ban you, I will gladly do so. I don’t demand politeness, but I draw the line at personal insults, which (again) you are the only one here to have indulged in.

  303. Nobody has ever found (or thought of, depending on how you look at it) a plausible phonological process by which non-clicks could be transformed into clicks, and so the assumption has been that it must be a primitive character.

    Never thought of that, but if indeed it is the case, you still have cases like the Bantu languages of South Africa, where you borrow words with the clicks in them. A couple of ten thousand years will make the borrowings unrecognizable and the language will safely possess clicks.

  304. marie-lucie says:

    Athel: Greenberg’s reputation

    John Cowan has answered your question much better than I would have about the African classification. What I have read about it was that he corrected some groupings which were based on irrelevant criteria (such as the race of the speakers, or the presence of gender) but did not really do very original work. But I am not competent to have an opinion either about his African work or the responses to it.

    JC: Initial responses among North American linguists have been either “He did a great job in Africa, so his Amerindian work must be great too”, or some time later after negative comments from specialists: “He did a lousy job in America, so his African work must not be as great as reported”. I think that both responses were inappropriate, but I know of no Amerindianist on this continent who approves of Greenberg’s classification. However, his and Ruhlen’s work seem to be very popular in France, where his comment that older linguists did not wait for PIE reconstruction to classify the languages (which is true and valid) is taken as showing that his Amerindian work was right and the gainsayers must have all sorts of axes to grind. No matter that the proposed “Amerind”, if correct, would be on the order of Greenberg’s proposed Eurasiatic (which includes Nostratic as well as yet other groups) rather than Indo-European.

  305. Athel Cornish-Bowden says:

    Thanks John and Marie-Lucie. That’s quite enlightening.

    Seeing that you’re a specialist in languages of the Pacific Northwest, Marie-Lucie, I was remembering today that Maud Menten, author of one of the most best known papers in biochemistry (though most of the people who cite it just think of her as M. L. Menten and have no idea that she was a woman, and, indeed, Canadian) was a speaker of Halkomelem, which she learned from school friends while growing up in Harrison Mills. Later in life she was also fluent in English (of course), French, German, Italian and Russian.

  306. where you borrow words with the clicks in them

    But it can’t be borrowings all the way down: somewhere there must be languages with non-borrowed clicks.

    I’m reading a paper by George van Driem with the innocuous-sounding name of A Holistic Approach to the Fine Art of Grammar Writing, which contains the following paragraphs guaranteed to bring warmth to the heart of Hat, Marie-Lucie, Etienne, and all the other pariahs from mainstream linguistics who cluster so thickly on this blog:

    Each language deserves a comprehensive grammar and extensive documentation of its lexicon. A language should not be used serendipitously as an intellectual playground to ‘test’ the linguist’s obsession of the moment or merely to serve as a source of examples for some already dated or quickly fading fashionable formalism such as ‘optimality theory’ or ‘parametrics’, e.g. Holmer (1996). Such works have a limited shelf life and are of little lasting value. Nine out of the ten primary branches of the Austronesian language family are represented exclusively by the native languages of Formosa. Fifteen Formosan languages survive, some of which are represented by several different dialects. All of these languages are threatened with imminent extinction, and none of these has been documented in the shape of a comprehensive grammar with extensive documentation of its lexicon. Many a scholar in Taiwan has dabbled in phonetic features or superficial syntactic phenomena of a particular Formosan language.Yet anything less than providing in-depth grammatical analyses and comprehensive lexical documentation of each of these fifteen Austronesian languages at this point in history will constitute a heinous form of linguistic negligence for which both future generations and the present generation of scholars will not forgive today’s linguists and research institutions.

    It would be risible if an archaeologist were to excavate a site and decide, for example, just to look for potsherds because the archaeologist in question happened to indulge an interest in the prehistory of ceramic cultures. This archaeologist would discard bones, stone implements, artefacts of early metallurgy and would neglect to use flotation techniques to recover grains of early cultivars from the site. Of course, this would strike us as ludicrous, and fortunately archaeologists do not in fact work that way. Sites are excavated thoroughly. Careful stratigraphies are done. Flotation techniques are used. All recoverable items are meticulously cleaned, preserved, catalogued and described.Dendrochronologically calibrated radiocarbon datings are ascertained. A comprehensive study and analysis of the site is conventional practice. Yet most linguists indulge in frivolous exercises very much like the wanton obsession of our imaginary archaeologist. Examples are picked out serendipitously from poorly documented languages to argue some abstruse point and buttress some formalist framework. In fact, it is fair to say that at this point in the history of the field much work conducted by professional linguists is either bogus, utterly useless or both. The principal task of linguists is to provide comprehensive grammatical descriptions of languages and extensive documentation of their lexicons.

    Preach it, brother. Tell it like it is.

  307. Wonderful, true and indicting.

  308. German, your etymology keeps jumping around with each new comment so that I can’t keep track of it; now we have a syllabic R all of a sudden, but you still haven’t answered my question: what’s the actual PIE form you’re positing? Specifically, what consonant does it begin with?

  309. It made me think of an idea I have long ago. I think that it’s important to keep the flame of historical linguistics alive and ensure that future fieldworkers all get a proper training in historical linguistics, if only for lexicons of endangered languages to be as well documented as they should.

    The current description industry privileges grammars to everything else. The model of a PhD thesis is a fairly well-written grammar with a 800-word lexicon with minimal definition, often one or two English words. This is understandable. One of the baser human motives of science for scientists is to say interesting things. Having the prospect of saying interesting things motivate people to collect data. The current grammar model works because there are often interesting things to say about morphosyntax and phonology, if you happen to land on a difficult or unusual system. On the other hand, you don’t have a universal (not UG but comparative-across-continents) way to talk about lexicon or much of morphology. There is a way to talk about lexicon, which is historical linguistics. The thrill of finding an exciting etymon should be there for languages that happen not to be Indo-European, and for fieldworkers to understand that there could be some excitement over there, the current way of teaching linguistics should be recalibrated somewhere to the pre-Saussurian, let alone pre-Chomskian, equilibrium.

  310. I for one wish more people were writing these model Ph.D. theses you’re talking about… Grammars, well-written or otherwise, constitute an almost negligible proportion of linguistics dissertations these days.

  311. I would hope those who have written the grammars for their PhD would do the reconstruction of the immediate parent group and an etymological dictionary for their Habilitationsschrift. (I would prefer the reverse order, in fact, as a really good grammar is very difficult to write.)

  312. Trond Engen says:

    That archaeology parallel is apt in another way: Conrext is everything. Shiny artifacts without context are next to nothing. To be understood for what they really are, linguistic features must be seen in the multifaceted context of overall grammar, sociolinguistics, historical phonology, cultural influences, whatever relevant background there is to collect.

  313. marie-lucie says:

    JC: I just read presentation slides by Blench drawing large conclusions from the worldwide pattern of isolates and small families. There are lots in the New World, few in Africa, concentrated in the north in Australia, and so on. But I greatly fear that what he is really investigating is the worldwide pattern of resistance among specialists to Greenberg’s lumping: very high in the New World, until recently low in Africa, mixed in Australia, and so on.

    I have been trying to see the slides, but no luck thus far, so I will just reply to the rest of the comment.

    I can’t speak for the whole New World as I am only familiar with Western North America, but I think that the number of ‘isolates and small families’ is exaggerated. I have mentioned several times that the current “mainstream” classification of North American languages is little different from what it was in 1862 when the first comprehensive survey was completed, using data compiled from missionaries, traders, and sundry other people, few of whom had much training in language studies. Sixty or so years later, the great linguist Sapir proposed a classification of these languages into six “phyla”, or groups sharing some typological characteristics and considered beyond the reach of the comparative method. Two of the “phyla” (though no longer called that) correspond to Athabaskan and Eskimo-Aleut, which are acknowledged by Greenberg as quite different from each other and from his “Amerind”. Sapir’s other phyla have been struck off the maps. One of them, “Aztec-Tanoan” still has a few – very few – adherents, as I mentioned not too long ago. The one I have been working on, “Penutian”, was defined as including about 15 language families and isolates more or less covering the Pacific coast, from California to the tip of the Alaska panhandle. It is no longer officially accepted, and is now considered to consists of a few small families and some isolates. But my own research has convinced me that most of the Penutian languages and families are indeed related to each other. Most previous scholarship has emphasized the search for “lexical sets” (potential cognate sets), which has been frustrating because of some unusual correspondences, while I have worked on morphology, especially irregular, archaic morphology, with continuing results. Also, I have been struck by many resemblances between Penutian and Uto-Aztecan, while U-Az is quite different from Tanoan. Although I found this independently, it is not an original discovery as it was anticipated by Whorf and Swadesh decades ago. So I think that the description of the New World as containing a large number of totally separate families is inaccurate. I think that further research (which might take a long time) will eventually reduce the acknowledged groups of the continent to a number closer to that of Sapir, and include some large groups and a few true isolates (such as Zuni in the US Southwest).

    As for the “very high pattern of resistance” to Greenberg’s classification (a phrase which presumes that Greenberg was right), it is easy for people outside the field to compare reactions to Greenberg’s work on different continents as if the linguistic and scholarly situations were the same. I already mentioned that “Amerind” would be of the order of Eurasiatic, not Indo-European. Also, Indo-European and some of the other families of Eurasia have long written traditions, sources of documents which do not exist in Australia or the Americas where the “window” of time in which the languages are attested is very short, and moreover these continents have seen the loss of many of the original languages.

  314. J. W. Brewer says:

    Re minus273’s point, I guess the question is how do we get there from here. Pick a random underdocumented non-Sinitic language of Taiwan. Let’s say you have someone who would be willing to write a check in the amount necessary to support a grad student with sufficiently modest standard-of-living expectations to do two or three years of fieldwork there and another year back on campus writing it all up, all with the approach minus273 suggests. (It’s not going to be a huge sum, all things considered; and it doesn’t need to be a grad student from a First World university who’s an L1 Anglophone — write up the results in Mandarin or Tagalog or Tupi-Guarani and assume someone else will be willing to fund a translation.) Do we have a supply of grad students on hand who’d be willing and able to do that sort of project if the funding were made available? If not, do we have the institutional infrastructure at hand to create such grad students if the right people decided we ought to be doing that? If not, how do we create that infrastructure?

  315. marie-lucie says:

    When I said that Eskimo-Aleut and Athabaskan are very different from “Amerind”, I didn’t mean to suggest that “Amerind” is a single, valid genetic group (which is what Greenberg said). The languages and families included in it can be very different from each other, just perhaps not quite as distinctive as the aforementioned two.

  316. Preach it, brother. Tell it like it is.

    Amen! My heart is well warmed.

  317. @JWB, I would say the problem is not the supply of grad students willing to do fieldwork so much as the supply of jobs for such grad students after they file. The institutional culture is unfortunately such that, given two job candidates of whom one has written a grammar of Shipibo and the other a thesis ironing out some kink in Stratal Optimality Theory, the vast majority of search committees are likely to prefer the latter to the former.

  318. J. W. Brewer says:

    Thanks TR. That’s helpful, if depressing. And it would require a lot more $ to endow a different-emphasis tenure-track gig for the grad student to be hired into than to support a few years of fieldwork.

  319. “very high pattern of resistance” to Greenberg’s classification (a phrase which presumes that Greenberg was right)

    Why would it presume that? I certainly meant no such thing. Resistance to error, like resistance to oppression, can be correct and even heroic. (Indeed, where error hardens into orthodoxy it inevitably becomes oppressive.) Like you, I think neither the orthodox nor the Greenberg view of the New World is necessarily right.

    I heard back from Blench, who says that he is personally convinced by the evidence for Nilo-Saharan, Central Khoi, and Pama-Nyungan, but thinks Trans-New-Guinea, Oto-Manguean, and Altaic may have insuperable problems. In my (private) reply, I urged him to look at Penutian, and took the liberty of dropping your name.

    By the way, you write “Amerindianist”. Has this displaced “Americanist”? I don’t wish to sound old-fashioned, even if I am.

  320. @TR

    “German, your etymology keeps jumping around with each new comment so that I can’t keep track of it; now we have a syllabic R all of a sudden>.”

    Nothing is jumping. You just can’t follow a simple logic, it seems. Syllabic R was there from the beginning. The HEART form is currently reconstructed as *k’Rd (where R is syllabic R). This is fully compatible with the vocalism of Gk rhadiks or Germ *wurtiz, so we don’t even need a laryngeal there (see above).

    “Specifically, what consonant does it begin with?”

    Like I said, classification precedes reconstruction, so I keep PIE reconstructions open for now because one
    doesn’t provide a new reconstruction on the basis of a single form, especially since the flaw I’m correcting is a methodological one. We can keep the traditional *k’Rd for now, which accounts for both HEART forms and ROOT forms (*Rd-). My point was that k’ > s (just like kw > p) may have been a PIE sound change, hence we see s- in Greek and Latin.

    @languagehat

    “The only person launching ad hominem attacks in this thread is you.”

    Incorrect. I haven’t launched any ad hominem attacks whatsoever (I’m only interested in the discussion of the comparative method here and the relationships between amateur and professional linguists) but countered the one launched at me. It’s bizarre that you can’t see it. And now, after several serious posts I left on your website, you’re threatening to ban me instead of banning SFReader who was digging through dirty laundry?! I wonder who taught you the difference between “good” and “bad” in life.

    Your comment is truly insulting. And I don’t care if you are the owner of the site. You have been rather unwelcoming to me. Anybody can open a site and start blogging. I guess I now understand why SFReader has been around here for a long time.

    @Athel Cornish-Bowden

    I can add to JC’s and ML’s responses that in Africa Greenberg relied on the robust continent-wide classification by the German missionary linguist Westermann. He tightened it up in places but for the most part followed in Westermann’s footsteps. These days linguists count 20 language families in Africa where Greenberg left 4. http://onlinelibrary.wiley.com/doi/10.1111/j.1749-818X.2008.00085.x/references.

    Greenberg and Ruhlen continue to hold a spell over population geneticists working on Amerindian populations. Linguists had to publish a kind of “cease and desist” article in AJHG demanding geneticists to stop using Greenberg’s classification of Amerindian languages.

    But: Edward Vajda who recently proved the reality of the Dene-Yeniseian language family acknowledged Greenberg and Ruhlen as one of the sources of his inspiration.

  321. Anybody can open a site and start blogging.

    That’s true, but not just anybody can keep it going steadily for more than a decade and attract a community of commenters who stick around for years having informative and enjoyable discussions.

  322. Anybody can open a site and start blogging.
    And you think that languagehat is no different from some random WordPress blog. What a barren taste, German Valentinovich.

  323. @Keith Ivey

    “a community of commenters who stick around for years having informative and enjoyable discussions.”

    I guess everybody, including languagehat, has bad days. I just happened to stop by during those days.

    “And you think that languagehat is no different from some random WordPress blog. What a barren taste, German Valentinovich.”

    Coming from an anonymous poster, this looks like promotional spam. That’s why you need real names. Real names make people responsible for what they post, especially when addressing other people.

  324. @Marie-Lucie,

    “Also, I have been struck by many resemblances between Penutian and Uto-Aztecan, while U-Az is quite different from Tanoan. Although I found this independently, it is not an original discovery as it was anticipated by Whorf and Swadesh decades ago. So I think that the description of the New World as containing a large number of totally separate families is inaccurate.”

    This is interesting. Do you have anything to share? As you may know, my key area of expertise is kinship terminologies (Amerindian and Indo-European especially). I know Uto-Aztecan and Penutian systems pretty well on the structural semantic level (see my Database at http://kinshipstudies.org/kinship-studies/database/) and there are indeed some interesting parrallels between the kin terminologies of the two language families. but I obviously don’t know much about the actual kinship terms from the linguistic point of view. But once I happened to be right regarding an aspect of proto-Athabascan kinship systems just by looking at their semantic structures comparatively and was happy to receive an encouraging note from Golla (http://kinshipstudies.org/category/victor-golla/).

    I’d be very curious to see if you’ve come across any lexical matches between Penutiuan and UA kinship terms.

  325. We can keep the traditional *k’Rd for now, which accounts for both HEART forms and ROOT forms (*Rd-). My point was that k’ > s (just like kw > p) may have been a PIE sound change, hence we see s- in Greek and Latin.

    But we don’t see s- in either Greek or Latin, and *k’Rd can only account for the “root” words by a lengthy series of unsupported ad hoc suppositions: an Indo-European k’-mobile which no one has ever heard of, a Greek sound change *k’ > s ditto, some kind of laryngeal loss (since there’s clear evidence of *h2 in the long ā of Latin rādix), irregular prothetic w- in Germanic, some other arbitrary change to take care of the Celtic cognates (which also point to *w). Not to mention that adopting your etymology means losing the etymology for ῥίζα. I don’t see you correcting a methodological flaw, I see a textbook example of the classic methodological fallacy of assuming two words are cognate because of vague semantic and phonological similarity and then concocting ad hoc changes to support one’s hunch.

  326. @TR

    “But we don’t see s- in either Greek or Latin.”

    What do you mean? We do! /h/ in Gk rhadiks is from *s. /s/ is regularly lost in Latin before a sonorant, so we can reconstruct *sradiks for proto-Italic.

    “(since there’s clear evidence of *h2 in the long ā of Latin rādix),”

    There’s just as much evidence for a laryngeal in Gk ke:r ‘heart’. But Gk rhadiks is lacking a:. Its a comes from syllabic R just like ur in Germ *wurtiz.

    “Not to mention that adopting your etymology means losing the etymology for ῥίζα.”

    Not losing it. Just putting it on a back burner in view of its strange vocalism (to your own point),

    “a lengthy series of unsupported ad hoc suppositions”

    They are all products of your imagination.

    “I see a textbook example of the classic methodological fallacy of assuming two words are cognate because of vague semantic and phonological similarity and then concocting ad hoc changes to support one’s hunch.”

    I disagree. You just biased in favor of conventional thinking. And since you can’t defend it, you fall into the psychological trap of attacking your opponent’s position.

  327. /h/ in Gk rhadiks is from *s.

    You keep stating this as a fact — are you just unaware that both PIE *sr and *wr- > Attic rh-? On the other hand, PIE *sr- > Lat fr- (e.g. frigus : ῥῖγος), which is yet one more reason your account is impossible.

    “Not to mention that adopting your etymology means losing the etymology for ῥίζα.”

    Not losing it. Just putting it on a back burner in view of its strange vocalism (to your own point),

    It’s not the vocalism that’s the problem (this can be explained by schwa epenthesis, as I mentioned above), it’s the fact that this word is actually attested with initial w-.

    A good rule of thumb in proposing new sound laws is that if the number of such laws is larger than the number of new etymologies they’re required to prove, something is wrong. This is one of the more extreme examples I’ve seen of that situation.

  328. @TR

    “You keep stating this as a fact ”

    Because it is. There are dozens of examples that prove it. Cases when /h/ seemingly corresponds to y or w are false positives. Scholars have started reconstructing H1 to explain Gk h in those cases, which is methodologically correct, but my idea is better because it’s more parsimonious and does not require laryngeals unattested even in Hittite. IE *w is regularly lost in Greek and cluster sw- yields s- (hekura < *swekuro- 'mother in law'), so your ad hoc postulation of sr Lat fr- (e.g. frigus : ῥῖγος)”

    Another ad hoc theory of yours. frigus : ῥῖγος is a unique isogloss, with no cognates in other languages and no other similar cases between Greek and Latin. Again, it’s likely a misanalyzed cognate set. It is interesting (potentially parallel to IE *-sr > Lat -br-, but never in the anlaut) and I thought quite a bit about it but don’t use it as proof that Gk sr- should be expected to correspond to Lat fr-. In Latin, s is regularly lost in the anlaut before any sonorant (r, l, n or m). This is the rule we should follow.

    “It’s not the vocalism that’s the problem (this can be explained by schwa epenthesis”

    Keep putting on your ad hoc show!

    “A good rule of thumb in proposing new sound laws is that if the number of such laws is larger than the number of new etymologies they’re required to prove, something is wrong. This is one of the more extreme examples I’ve seen of that situation.”

    The irony of the situation is that you don’t see that you are making all the mistakes that you think I’m making. Schwa epenthesis here, wr > hr there, no etymology for PIE *kRd heart – all the no-nos of historical reconstructions. This is because you are basing yourself in a flawed methodology where you are not even looking at the right material to deduce your conclusions.

  329. Sorry, it should read “so your ad hoc postulation of *wr- > Attic rh-” is phonetically impossible and contradicts other existing evidence.” (My computer froze on me.)

  330. @TR

    “an Indo-European k’-mobile which no one has ever heard of.”

    I treat it as a compliment. 🙂 On a more serious note, there’s certain kinship between IE s-mobile and the behavior of the “satem”s/PIE k’. The palatal feature is lost in all of kentum dialects, so there’s a loss of a palatal feature of a stop to match a palatal spirant loss, and even in satem dialects s is unstable as can be seen from my example Skrt sapha- ~ Slav *kopyto instead of expected sopyto**(and there are many more). So, both the so-called IE s and Satem s/IE k’ are mobile. Traditional s-mobile, although observed across all of IE dialects tend to pop more systematically in the “satem” area, as can be seen in, e.g., Gk nipha but Slav *sne:gu- ‘snow’, Lat nurus ~ Slav *snoxa ‘daughter-in-law.’ It’s just by this measure Germanic tends to cluster with satem languages and not with kentum languages, while Armenian with kentum languages and not with satem languages.

  331. I’m only interested in the discussion of the comparative method here and the relationships between amateur and professional linguists.

    Erunda. You also have an obsession with legal names, as if everyone had to present a passport before entering Languagehattia. I invite you once again to relocate yourself to Paleoglot, where that obsession is shared with the host and the others that can survive his fierce scrutiny.

    Coming from an anonymous poster

    What the devil do you know about whether Keith Ivey is anonymous or not? If his name is good enough for LinkedIn, it should be good enough for you.

  332. German, this is getting comical. I haven’t brought any arguments into this discussion which are not widely accepted among Indo-Europeanists. You’ve introduced some half dozen. If you want anyone to consider these seriously, your work is cut out for you — find some evidence which doesn’t rely on an infinite regress of building one unsupported theory on top of another, and show how your account is better than the unproblematic accepted etymologies. Good luck rewriting the textbooks.

  333. @TR

    ” I haven’t brought any arguments into this discussion which are not widely accepted among Indo-Europeanists.”

    Why did you bring them up? I know accepted viewed even without you. But my models work better than those “accepted ones” because “accepted” doesn’t mean “supported.” Have fun reading textbooks!

    @John Cowan

    “What the devil do you know about whether Keith Ivey is anonymous or not?”

    My comment was directed at minus273. I should have been clear about it. But you, too, should have followed the thread. What business do you have chiming in on that all the time in the first place?

  334. I know accepted viewed even without you. But my models work better than those “accepted ones” because “accepted” doesn’t mean “supported.”

    I assume most readers of this discussion will by now be able to assess the truth of both those statements for themselves, so I’m going to leave it there.

  335. @TR

    “I assume most readers of this discussion will by now be able to assess the truth of both those statements for themselves,”

    For this to happen, we need to have people in the audience who meet the following two criteria: a) know the IE material well; b) have proven experience advancing alternative models to explain facts. I doubt we have such people in attendance here, but who knows…

  336. marie-lucie says:

    JC; Thank you for the recommendation to Blench! I just found out that my 1997 IJAL paper is now on Academia. I did not put it there as it was a copyrighted article (to IJAL) and I did not think that would be legal, but there is is for all to read and even download (I think). I still agree with the article with a couple of very minor corrections which I can provide on request and which do not invalidate the major findings and conclusions.

    “Amerindianist”. Has this displaced “Americanist”

    It seems to me that “Americanist” is a more general term, including anthropology as well as linguistics. “Amerindian languages” is a common term (I think!), hence “Amerindianist”. But I could be wrong. On the negative side, “Amerind” is a back-formation from “Amerindian”, which can lead to an undesired interpretation.

  337. marie-lucie says:

    GD: Edward Vajda who recently proved the reality of the Dene-Yeniseian language family acknowledged Greenberg and Ruhlen as one of the sources of his inspiration.

    Even a very flawed work can provide “inspiration” for research, if only for checking the data. Vajda is an excellent linguist who certainly did not follow Greenberg’s advice about how to proceed. Ruhlen had earlier written an article about lexical resemblances between Ket (the last living Yeniseian language) and the Athabaskan languages, with a list of potential cognates containing many errors. Vajda would have been remiss in not consulting (and later mentioning) this article.

    Personally, I was “inspired” to start investigating Penutian after reading a few papers which were full of mistakes (of data and interpretation) about a relevant language family I knew well from firsthand acquaintance, while none of the other authors were familiar with it.

  338. there is is for all to read

    Actually it isn’t. There is a reference page giving the title, author, etc., but no one can either read or download the paper. Instead, there is a button (at least for me there is) that I can push to ask you to upload the paper as a scanned PDF. Copy shops can do this readily if you don’t have a scanner at home. I have pushed the button now. Whether you upload the paper in response is up to you.

    I interpret the “Green Open Access” section of this page to mean that the University of Chicago Press allows self-archiving by the author of articles published in IJAL in accordance with the Research Council of the UK’s policy on open access (linked from there), even though the author (you) did not receive UK funding. That policy requires social-science articles older than 24 months to be freely available to all via self-archiving. So upload away.

    I am not a lawyer, this is not legal advice; but scholarly publishers suing authors for self-archiving would be an execrable precedent and horrible public relations, and they are not going to begin with you.

  339. @German Dziebel: You probably do not care about my opinion, since I do not know Indo-European linguistics well. In fact, I am only a very amateur linguist. However, I am a professional researcher in a entirely different (and often contentious) area, so I am quite familiar with the sociological hallmarks of good arguments and poor ones. Your comments, larded with bombast, are not very convincing when they are being rebutted by a whole crowd of other commenters here—people whose demeanor and linguistic acumen I have come to respect over years of reading this site.

    And “Brett” is my real first name, by the way.

  340. marie-lucie says:

    Note about my IJAL paper

    The bulk of the paper is about morphological resemblances throughout the “phylum” (including about 35 common grammatical morphemes). That part is still strong. The paper also has an appendix with what could be called “potential cognate sets”. This part I am less happy with, as it is not organized as clearly as I would wish, and also some of potential cognate sets need to be revised in the light of my ongoing research since then.

  341. @Brett

    “Your comments, larded with bombast, are not very convincing when they are being rebutted by a whole crowd of other commenters here—people whose demeanor and linguistic acumen I have come to respect over years of reading this site.”

    There was no bombast, on my part (I just want to be crystal clear about certain things such as “methodological flaws”, so that my proposals don’t get confused with disagreements regarding individual reconstructions that are ultimately based on the same methodology), or rebuttals, on “their” part.

    90% defended, 10% taken as homework.

    BTW, there was an experiment done on human children and chimps. A treat was put in a glass box with an open door. An experiment supervisor showed both chimps and kids a convoluted path to get to the door and get a treat. (Although the door was right in front of them.) Human kids ended up repeating that convoluted path and reaching for the treat only after they’d performed all the steps showed to them by the supervisor. The chimps ignored what the supervisor was telling them and reached for the treat directly.

    So, I’m well aware of the fact that humans are social beings and they tend to follow traditional rules, seek approval from seniors and do what everyone else does. They will support trolls as long as a troll has been with them for a few years. They will check new ideas against textbooks. And they will measure a horse’s speed in “car powers.” Your response did not come as a surprise but rather as a payoff from my unintended experiment. You can follow “authority”as much as you want. I’m just gonna go directly for the treat.

    Do you have a last name? Or just Brett? 🙂

  342. marie-lucie says:

    JC: Thanks for the clarification about how to access my paper. I guess I was not aware of what “self-archiving” means. I will gladly help you access the paper. Should I email it to you? If you have changed your address, email me has mine is still the same.

  343. marie-lucie says:

    JC, sorry, I think I misinterpreted again. Can you email me?

  344. @Marie-Lucie

    “Even a very flawed work can provide “inspiration” for research, if only for checking the data.”

    Well, the inspiration came with some hard cash in the form of “at least 8 valid cognates” that Vajda confirmed with his sound correspondences (Vajda, “The Dene-Yeniseian Connection,” p. 106).

    BTW, what’s your take on Lyle Campbell’s critique of Vajda? He was the only negative outlier, as far as I know, but I hold his opinion in high regard. (He was a very helpful reader of my book, too.)

  345. marie-lucie says:

    I was present the first time that Vajda presented his hypothesis of a link between Ket and Athabaskan (in the 90’s) on the basis of morphological and morphosyntactic similarities. He then went on to do more research on Ket (including travelling there for fieldwork with some of the remaining speakers nt historical linguists had been invited (Michael Krauss, Eric Hamp, Johanna Nichols), and the 2009 SSILA summer meeting in San Francisco where Vajda presented another paper on the topic. The audience at that meeting included John Bengtson and George Starostin. The latter had a number of objections, which Vajda met very graciously. It would be surprising to me if Vajda did not acknowledge that he was encouraged in his own researches by previous work by other linguists (something that is routinely done by authors of scholarly books and papers).

    The fact that some of Ruhlen’s cognate sets were correct is not surprising (he is a linguist after all), but has no bearing on the correctness of his and Greenberg’s “Amerind” classification.

  346. marie-lucie says:

    Oops, I think an entire line disappeared itself. Here is a rewritten version of the paragraph, with missing words in italics (for now):

    I was present the first time that Vajda presented his hypothesis of a link between Ket and Athabaskan (in the 90′s) on the basis of morphological and morphosyntactic similarities. He then went on to do more research on Ket (including travelling there for fieldwork with some of the remaining speakers. I also attended the 2000 meeting in Fairbanks, to which some prominent historical linguists had been invited (Michael Krauss, Eric Hamp, Johanna Nichols), and the 2009 SSILA summer meeting in San Francisco where Vajda presented another paper on the topic. The audience at that meeting included John Bengtson and George Starostin. The latter had a number of objections, which Vajda met very graciously. It would be surprising to me if Vajda did not acknowledge that he was encouraged in his own researches by previous work by other linguists (something that is routinely done by authors of scholarly books and papers).

  347. marie-lucie says:

    sorry, the 2008 meeting, not 2000.

  348. marie-lucie says:

    GD: I have not read Lyle Campbell on Vajda, but I am aware of his work. He is very hard to please, so congratulations.

  349. Campbell on Vajda is in the 2008 proceedings published by the university.

  350. @Marie-Lucie

    “The fact that some of Ruhlen’s cognate sets were correct is not surprising (he is a linguist after all), but has no bearing on the correctness of his and Greenberg’s “Amerind” classification.”

    Absolutely agree. BTW, George Starostin later published a critique of Vajda in his Journal of Linguistic Relationship.

  351. Trond Engen says:

    I can’t seem to find my copy of The Dene-Yeniseian Connection, but as I remember it, Vajda’s attitude to Ruhlen’s work is politely dismissive.

  352. @Trond Engen

    “Vajda’s attitude to Ruhlen’s work is politely dismissive.”

    Nah…I think Marie-Lucie is right about the fact that Vajda is a genuinely nice person. He appreciates being part of a scientific community of linguists of different creeds who together solve the mysteries of prehistory. Sometimes they are right, sometimes they are wrong but their mistakes don’t belittle their essential professionalism and integrity.

  353. Trond Engen says:

    Yeah, well, OK, I might just as well have written “nice but dismissive”. But I’ll reread it if I find it.

  354. Vajda is careful to say that Ruhlen had 8 cognates right; he is also careful to say that Ruhlen proposed 35 cognates in all, for a less than stellar hit rate. (Numbers from memory, may be wrong.)

  355. P. 108: “In spring 1996 I invited Merritt Ruhlen,..to present a talk at my university. I was impressed that Ruhlen was actively working on the problem of Yeniseian genetic linguistics, while no one else seemed to have given a second thought since Trombetti…I greeted his subsequent publication of 34 putative cognate sets linking Yeniseian and Na-Dene…as an important event, though most linguists took at best only cursory notice.”

    Yes, you are right: “nice but dismissive”. Dismissive of “most linguists,” not of Ruhlen. There’s this attitude, too, you know, showed by an “excellent linguist,” as Marie-Lucie referred to Vajda.

  356. @John Cowan

    “proposed 35 cognates in all, for a less than stellar hit rate.”

    This is currently a world record. But we of course are all waiting for a linguistic Hussein Bolt who will take us out of the 8-cognate range for a long-range family.

  357. marie-lucie says:

    GD: Thanks for the link to LC’s article.

    The article is balanced between pros and cons, which is good. I am not very familiat with either of the languages (or families in the D case), but I did know (from various sources including Vajda’s early presentation and a course on Athabaskan languages) that the verbal structures in both are multipartite rather than merely bipartite as are so many languages.

    Having read several other reviews by LC of comparative proposals in various languages, I feel that LC tends to overemphasize the dangers of “onomatopeia” and “semantic latitude”, seeing the first in too many cases (what could be onomatopeic about a word for “shadow”?) and insisting on very close meaning resemblances (equivalences of lexical items) without reflecting on what apparently completely different words might be considered to belong together. Consider “hook” and “come back”: at first sight it looks like these two meanings have nothing to do with each other, but what is a hook? an object made by curving back a straight object on itself, just like the path of a person or animal may turn back to return to where the starting point.

    In conclusion, I have neither an axe to grind nor enough knowledge to really evaluate Vajda’s work (or Starostin’s or other linguists who have dealt with the case), but it does look promising.

  358. @Marie-Lucie

    Thanks! It’s very helpful. Dene-Yeniseian and even Dene-Causcasian look promising/intriguing from the kin terminological perspective as well (http://kinshipstudies.org/2015/06/26/dene-caucasian-kinship-and-dene-caucasian-kinship-terms/), but it’s still rather daunting and too good to be true. Starostin (https://www.academia.edu/2497705/Dene-Yeniseian_a_critical_assessment) does criticize the key part of Vajda’s proof, namely the multipartite verb morphology.

    ““hook” and “come back”:th

    I couldn’t agree more with you in this instance. That’s exactly one of the problems I have in mind when I mention errors in cognate set composition.

  359. @Marie-Lucie

    ” “Amerind”, if correct, would be on the order of Greenberg’s proposed Eurasiatic (which includes Nostratic as well as yet other groups) rather than Indo-European.”

    Am still thinking through what you wrote earlier….Greenberg was mislead by the impression fostered by mainstream American archaeologists that Amerindians for the most part are recently (as in 12,000 years ago) derived from a single Asian source. (And those that aren’t are even more recent!) This made Amerind look as a taxon of roughly the same age as relatively unproblematic Afroasiatic. So, Greenberg did not anticipate any problems with arriving at an essentially correct classification and leaving reconstruction to Amerindian linguists to figure out later.

    “However, his and Ruhlen’s work seem to be very popular in France, where his comment that older linguists did not wait for PIE reconstruction to classify the languages (which is true and valid).”

    Greenberg is considered “great linguist” in Russia, too. And Ruhlen a respectable one. This is because lots of mainstream Russian linguists support the Nostratic concept (and the Dene-Caucasian concept and the Khoisan concept), which is close to Greenberg’s Eurasiatic. There’s this intriguing cultural aspect to science.

  360. @Marie-Lucie

    “older linguists did not wait for PIE reconstruction to classify the languages (which is true and valid).”

    Narrowing down to this very important argument used by Greenberg, Indo-Europeanists set the wrong precedent of thinking about languages, and not cognate sets, as the minimal units of classification. Remember how often pater, pita:, pate:r and fadar ‘father’ have been celebrated as a great example of the reality of the IE family? You don’t need a reconstructed PIE to see that IE languages are related. And establishing linguistic kinship is the primacy goal of historical linguistics as a science! The fact that 1) Hittite and Balto-Slavic showed no reflexes of this presumably PIE etymon; 2) that in the Gothic corpus fadar is found only once (the regular word for father is atta, just like in Hittite and likely Balto-Slavic (but those are baby words, so you wouldn’t use them as proof of linguistic kinship); and 3) that Slavic has the form *strujus ‘father’s brother’ which is a morphological and semantic near-copycat of Skrt pitrvya, Lat patruus but with an unexpected phonetic correspondence non-Slav *p- ~ Slav *s- has never been explained or properly thought through during the past 150 years. (And Arm has yawray ‘step-father’, with a strange correspondence p > y next to the “normal” hayr ‘father’.) If you really scrutinize the IE cognate sets (as I did), you’ll see how messy they are. At some point you’ll stop seeing the critical difference between Greenberg’s one-man Amerind work and Indo-European studies. Indo-European is just a vastly younger language family, hence the sheer amount of cognates doctored out of their messy cognate sets and nicely packaged by generations of scholars to showcase their discipline leaves no doubt they are all related. But is it truly reconstruction?

    Greenberg once said that all those Amerindian linguists can’t get their collective act together to arrive at higher order groupings in the New World, so all the work falls on the shoulders of “one man” (or one man and a half, if Merritt is brought into the picture 🙂 ). In Indo-European studies, the problem is of the opposite nature: Indo-European is collectively thoroughly overcooked, and it takes “one man” (say, yours truly, for the sake of the argument) to actually think through the myriad of misnalyzed, phonetically irregular, antietymological “cognate” sets and ad hoc semantic and phonetic solutions left behind by a horde of Indo-Europeanists.

  361. marie-lucie says:

    GD, thank you.

    It makes sense that Greenberg’s Amerind concept was influenced by the then prevailing idea that the peopling of the Americas was thought to be relatively recent and resulting from a single migration. My impression (based on a variety of criteria) is that although it is plausible that 12000-odd years ago saw the first entry of humans into the New World (Bering strait etc), there must have been one or more subsequent migrations, mostly by boat. The currents in the North Pacific regularly bring Japanese flotsam and sometimes wrecked boats (and even survivors) to the North American coast. ( I don’t mean that only Japanese items are coming, but those seem to be the most identifiable).

    kinship terms

    I posted a link to some of them (mostly from Penutian languages) in the Mama-Papa thread.

  362. And establishing linguistic kinship is the primacy goal of historical linguistics as a science!

    Why so? It seems to me that historical linguistics has at least the goals of establishing kinship, genetic classification, and reconstruction. Evolutionary biology does all these things, and why should linguistics be behind?

  363. @John Cowan

    I meant it as a statement about what people (in and outside of historical linguistics) think, not what I think. I think exactly what you just described but in an even more extreme formulations: unless ALL three are done, no one of them can be considered finished.

  364. @Marie-Lucie

    “I posted a link to some of them (mostly from Penutian languages) in the Mama-Papa thread.”

    Thank you! The mama-papa terms are hard to analyze historically. They are either too short (na) or two patterned (nana, tata). They form very interesting worldwide areal clines (studied as early as mid 19th century by Buschmann in “Naturlaut”) with intriguing local reversals but they are weak as evidence for pairwise kinship between languages.

    What caught my attention regarding the potential Penutian-UA links is the so-called self-reciprocal terminology between grandparents and grandchildren and aunts/uncles and nephew/nieces. It’s a relatively rare semantic and pragmatic pattern that’s very stable in UA and it doesn’t have the problem that terms for parents have. It shows up as a complete set in some Penutian languages.

  365. marie-lucie says:

    I agree in general that kinship, classification and reconstruction all need to be done (but rarely all at once), but I disagree that comparison has to start with “cognate sets” (or rather, “sets of resemblant forms suggesting cognacy”, since cognates cannot be established just by looking at individual resemblances. Early Indo-Europeanists put great emphasis on common morphology (especially what can be shown to be archaic, now “irregular” morphology) and wrote “comparative grammars” of various families before they got really serious about reconstruction. Many current linguists minimize (or even ignore) the role of morphology, but a secufe morphological and morphophonemic foundation greatly reduces the likelihood of basic errors such as those described by Lyle Campbell in several of his works (and mentioned in his review of Vajda).

  366. marie-lucie says:

    GD: The list I linked to on the mama-papa thread does include the mother-father terms for comparison.

    the so-called self-reciprocal terminology between grandparents and grandchildren and aunts/uncles and nephew/nieces. It’s a relatively rare semantic and pragmatic pattern

    I don’t think it is restricted to just Penutian and Uto-Aztecan, but I have not looked at other languages from that point of view. I thought it must be an areal phenomenon.

  367. marie-lucie says:

    an unexpected phonetic correspondence non-Slav *p- ~ Slav *s- has never been explained or properly thought through.. and in Armenian a strange correspondence p > y

    Sorry, these cannot be right. Just because the sounds in question occur in the same place in words which have “some” similarity in shape and meaning does not mean that they constitute a valid correspondence. You have to have some phonetic plausibility. I cannot see that either p or s could derive one from the other, or that both could derive from yet another sound. Sometimes an unusual correspondence does occur, which would presume a chain of changes, but if it is valid it should be consistent over a series of word pairs or sets, and some parts of the chain would also likely be found in other languages, since the number of sounds the human vocal apparatus is capable of producing is large but not infinite.

  368. @Marie-Lucie

    “I disagree that comparison has to start with “cognate sets” (or rather, “sets of resemblant forms suggesting cognacy”, since cognates cannot be established just by looking at individual resemblances.”

    Good point.

    “I don’t think it is restricted to just Penutian and Uto-Aztecan, but I have not looked at other languages from that point of view. I thought it must be an areal phenomenon.”

    That’s what we don’t know. Everything has an areal distribution of sorts. They could be related or unrelated, or, if related, related at a deep or recent level. Self-reciprocal terminology does seem to be a (North) Amerindian phenomenon (in the Old World it doesn’t show up until such remote places as Australia and Papua New Guinea). It’s reconstructible for proto-Athabascan and proto-UA. The advantage of looking at self-reciprocal terminology is that it forms sets (sometimes very large sets). With time, a self-reciprocal set degrades, simplifies and becomes irregular. (E.g., in IE Hitt huhhas ‘grandfather’ has its self-reciprocal counterpart in OIr (h)aue ‘grandfather’ but otherwise the pattern is broken.) if a subgroup or a language has a full set of self-reciprocal terms, then it gives one a cue as to what to look for in other subgroups or languages of the family.

    “Sorry, these cannot be right. Just because the sounds in question occur in the same place in words which have “some” similarity in shape and meaning does not mean that they constitute a valid correspondence.”

    Absolutely agree. Some people tried to force IE *pH2ter and Slav *stryji (sorry I misspelled this form before) together and proposed a sequence *ptr > *ttr- > *str. (Or *ptevis for Lith tevis ‘father’). But others remained unconvinced. The shared morphology and semantics, however, are so specific that one is left without any other choice but to keep thinking about a phonetic solution. This example is not the only one. Lat nepo:s, Lith nepuotis, Gk anepsios correspond to Slav *nestera, *netiji ‘niece, nephew’.

    I don’t buy the IE *pt > Slav *tt transition but then tt > st seems very natural and well-known in Slavic. Using my methodology, I did arrive at a solution for the puzzle, which I think works better than anything proposed before but it again requires a large-scale rethinking of IE phonology. (Not a bad thing by itself, IMHO.) But recently I lost a Facebook friend (a sharp and thoughtful Indo-Europeanist) over my solution. He thought it was too radical for our friendship, so I better keep it for myself here, or SFReader and TR will call cops on me for disturbing their sleep. 🙂

  369. Erratum: “(E.g., in IE Hitt huhhas ‘grandfather’ has its self-reciprocal counterpart in OIr (h)aue ‘grandfather’…” should read: in IE Hitt huhhas ‘grandfather’ has its self-reciprocal counterpart in OIr (h)aue ‘grandson’..”

  370. David Marjanović says:

    No time tonight to catch up with the thread, so just so much for now:

    1) On the long consonant in hippos: I misremembered Donald Ringe’s blog post about this putative cognate set; probably I was thinking about the Anatolian development instead of the Greek one. In the post, Ringe never once mentions the pp as anything that would need an explanation, despite dwelling quite a bit on the h and the i.

    2)

    The ś in aśva is not at all anomalous if you derive the word from PIE *(h₁)ékʲwos, as current textbook wisdom has it. Please explain what you mean.

    Your reconstruction doesn’t take into consideration the fact that Latin, Greek and Mycenaean attest for a labiovelar and not a cluster.

    The ancestors of Latin and Greek (including Mycenaean Greek) had undergone the kentum merger: they merged [kʲ] into [k]. This turned the cluster [kʲw] into the cluster [kw]. As the next step, they both merged this cluster into [kʷ].

    “Latin, Greek and Mycenaean” don’t distinguish between the outcomes of *ḱw, *kw and *kʷ. They are useless for determining which of these three the PIE “horse” word had.

    3) So, the Germanic word for “root” has a prothetic *w? Why, then, isn’t there a *w before every *u that results from a syllabic sonorant? It’s wunbelievable that this would happen specifically with syllabic r but not with syllabic n.

    I’m off to bed reading that preprint-like object by Bozzone & Felisari.

  371. @David

    Thanks!

    I like Ringe’s post. But here are my thoughts:

    1. He doesn’t discuss gemination in depth because he doesn’t have anything to say about it. He probably ranks it lower in importance than the problem of Gk -i- and h-. But if you take the gemination -pp- in Gk and -zz- in CLuwian seriously, then it makes the cluster interpretation impossible because it would require two identical back-to-back clusters *kw turning into labiovelars *Kw (-*kwkw > *KwKw). You could argue that gemination postdated the formation of /p/ from labiovelar kw but then what about CLuw -zz-? It must have happened independently and affected the postulated original palatovelar.
    2. The cluster k’w is rare, and Ringe mentions this. So we’re already risking postulating an ad hoc development. I would consider interesting an idea of reconstructing a compact voiceless palatolabiovelar phoneme for PIE (matched by a voiced aspirated palatolabiovelar as seen in Gk the:ros ~ Slav *zve:ri ‘animal’) but the cluster is just not compelling enough. One could reconstruct a cluster in place of any labiovelar attested in Gk, Latin, Myc or Gothic. But nobody does this.
    3. A good example of a cluster would be PIE *pek’u- (Skrt pasu-, Lith pasu-, Lat pekus ‘cattle’). There’s a palatovelar, /u/ but no conversion into /q/ in Latin.
    4. It’s a minor quibble but an objection nevertheless: H1 is unattested in Hittite, so with it in the anlaut of *H1ekwo- we have an etymon with a rare cluster in the middle and an unattested consonant in the onset. This is not a convincing reconstruction.
    5. The current reconstruction *H1ek’wo- leaves the IE word for HORSE without an etymology. Ringe acknowledges it. For me, this is a big problem because if a reconstruction doesn’t elucidate the origin of a word, there must be something wrong with the reconstruction.

    Alternatively, the equation Skrt sapha-/Slav *kopyto ~ Gk hippos/Lat equus is etymologically informative. Indo Europeans after all classified the horse either as a hoofed animal or as a hornless animal. It has a number of advantages over the conventional reconstruction *H1ek’wo-: 1) it’s phonetically crisp (no soupy clusters, no attested consonants), with a necessary conclusion that palatovelars turned into /s/ and labiovelars turned into /p/ already in PIE times, or prior to the divergence of IE languages; 2) it’s semantically straightforward; 3) it’s morphologically compatible (comp. Slav *kopyto and Gl hippotes) and 4) it fills in a distributional gap: Slavic is currently thought to lack a cognate for the IE word for HORSE but it turns out that it has a form that’s potentially a phonetic and semantic archaism.

    I treat the little phonetic puzzles that the equation Skrt sapha-/Slav *kopyto ~ Gk hippos/Lat equus poses more as a bonus, than a liability. Any solution to a big problem opens the gates for some smaller problems. (Grimm’s law had issues and those issues were later successfully resolved by Werner’s Law.) The conventional etymology doe the opposite: it reflects all the busy attempts to explain every little phonetic detail but as a result it generates a completely inscrutable etymon.

  372. @David

    “So, the Germanic word for “root” has a prothetic *w? Why, then, isn’t there a *w before every *u that results from a syllabic sonorant? It’s wunbelievable that this would happen specifically with syllabic r but not with syllabic n.”

    Could you give me a couple of examples? In Germ *wurtiz < *Rd- the prothesis is motivated by the previous loss of the front consonant. Sort of like in Armenian the famous metathesis that created elbayr out of *bhreH2ter 'brother' resulted in the emergence of a prothetic vowel e- because now the word began with a resonant and this was against the rules of the ancient Armenian language.

  373. George Gibbard says:

    The current reconstruction *H1ek’wo- leaves the IE word for HORSE without an etymology. Ringe acknowledges it. For me, this is a big problem because if a reconstruction doesn’t elucidate the origin of a word, there must be something wrong with the reconstruction.

    Suppose that in the near future the Americans colonize Mars, and they bring horses, who are able to survive thanks to the brilliant technology that has been invented for adding oxygen to the atmosphere. After a few centuries or millennia, Martian English has turned into a number of mutually incomprehensible languages. At this point, the Martians rediscover historical linguistics. A perceptive scholar is able to deduce that the horse in Proto-Martian was *hors. But there are no other words in the proto-language that appear to be related to *hors (other than compounds containing *hors) and so no one can offer a clue to the etymology of this reconstructed word. This does not mean it is a false reconstruction.

    Proto-Indo-European was not in any significant sense a ‘younger language’ than modern languages: it already had a history stretching back thousands of years and so had plenty of words, no doubt originally derived, but without identifiable etymologies.

    I have read that the ancient Indian grammarians debated whether or not all nouns in Sanskrit were derived from “roots” (the answer clearly being “no”), but I don’t know if this means verbal roots or a more general concept.

  374. @Geroge Gibbard.

    It’s a nice thought experiment – science fiction as a guide for science. 🙂 However, I wouldn’t use unique cases (Eng horse, Eng dog) as a material to inform a general methodology. Especially so, if so much of Indo-European cultural reconstruction, homeland and dating hinges on its wheeled transport pulled by horses. In your thought experiment it amounts to a situation when the English word ‘rocket’ suddenly becomes etymologically opaque to future scholars.

  375. What’s also interesting about the hoof > horse transition in PIE is that the HOOF isogloss is attested in Slavic, Indo-Aryan and Germanic, while the derived HORSE isogloss in Baltic, Greek, Latin and Celtic. Indo-Aryan and Germanic have both forms. Slavic has only HOOF, while Baltic (Slavic and Baltic form a subclade with IE) has only HORSE. So there’s an east to west cline in the transition from HOOF to HORSE: western dialects are HORSE only, while eastern are both HOOF and HORSE. This can be interpreted as a lexical trace of the westward expansion of IE languages.

  376. If all knowledge of Italian had been lost, rocket would be etymologically opaque, and we would have conjectures that it was an ironic name derived from a diminutive form of rock.

  377. David Marjanović says:

    Others would, of course, point to French perroquet “parrot”, and to colorful fireworks to bridge the semantic gap.

  378. David Marjanović says:

    “So, the Germanic word for “root” has a prothetic *w? Why, then, isn’t there a *w before every *u that results from a syllabic sonorant? It’s wunbelievable that this would happen specifically with syllabic r but not with syllabic n.”

    Could you give me a couple of examples?

    I gave one: the prefix un-, from PIE */n̩/ (zero-grade of */ne/), lacks a prothetic */w/ in every Germanic language I’m aware of, and has never been reconstructed with one either.

    I can’t think of other examples off the top of my head; there can’t be examples with */r/ because */r/ was not allowed as a PIE word onset. But un- alone already makes your hypothesis rather wunbelievable.

    I’ll now try to catch up with the rest of this thread.

  379. @David

    “But un- alone already makes your hypothesis rather wunbelievable.”

    That’s an overstatement, but I’ll keep looking into this issue.

    BTW, re cluster vs. labiovelar in the HORSE set, I can accommodate a cluster-like solution, so that Gk hippos < *hipwos < *hik'w-wos, with -wo- as grammatical marker to turn the root with the meaning HOOF into an adjective "hoofed." A cluster hypothesis does not contradict the reconstruction of a labiovelar.

  380. David Marjanović says:

    German Dziebel, Nov. 4, 10:20 am:

    ANE is an Amerindian marker in the Old World by definition. Read Raghavan et al. 2013 to see that MA-1 is closer to modern Amerindian populations than it is to any modern Eurasian population.

    I have now read the whole paper again. (BTW, it has two first authors – “These authors contributed equally to this work” –, so it should be Raghavan, Skoglund et al..) Maybe you should do the same. Yes, MA-1 is indeed “closer to modern Amerindian populations than it is to any modern Eurasian population” – but the paper (p. 88) puts a direction on this:

    “However, a significant residual was observed between the empirical covariance for MA-1 and Karitiana, a Native American population, and the covariance predicted by the tree model (Supplementary Fig. 12). Consequently, gene flow between these lineages was inferred in all graphs incorporating two or more migration events (Fig. 2 and Supplementary Fig. 13). Bootstrap support for the migration edge from MA-1 to Karitiana, rather than from Karitiana to MA-1, was 99% in this analysis.”

    “This result is consistent with allele frequency-based D-statistic tests^20 on SNP arrays for 48 Native American populations of entirely First American ancestry^19, indicating that all tested populations are equally related to MA-1 and that the admixture event occurred before the population diversification of the First American gene pool (Fig. 3a, Supplementary Information, section 14.4 and Supplementary Fig. 24).”

    In other words, the authors thought of the possibility that MA-1 could be a back-migrant from America, tested it, and found it insufficient. In the next sentence (still on the same page) they even spell it out:

    “The genetic affinity between Native Americans and MA-1 could be explained by gene flow after the split between east Asians and Native Americans, either from the MA-1 lineage into Native American ancestors or from Native American ancestors to the ancestors of MA-1. However, MA-1, at approximately 24,000 cal. BP, pre-dates time estimates of the Native American–east Asian population divergence event^24,25. This presents little time for the formation of a diverged Native American gene pool that could have contributed ancestry to MA-1, suggesting gene flow from the MA-1 lineage into Native American ancestors.”

    Take a good look at figure 2, too.

    On the next page, you’ll see that the authors took the idea of a back-migration really seriously and tested it again and again, always with the same result:

    “Thus, if the gene flow direction was from Native Americans into western Eurasians it would have had to spread subsequently to European, Middle Eastern, south Asian and central Asian populations, including MA-1 before 24,000 years ago. Moreover, as Native Americans are closer to Han Chinese than to Papuans (Fig. 3c), Native American-related gene flow into the ancestors of MA-1 is expected to result in MA-1 also being closer to Han Chinese than to Papuans. However, our results suggest that this is not the case (D(Papuan, Han; Sardinian, MA-1) = −0.002 ± 0.005 (Z = −0.36)), which is compatible with all or almost all of the gene flow being into Native Americans (Supplementary Information, section 14.6). Similar results are obtained when MA-1 is replaced with most modern-day western Eurasian populations, except populations with recent admixture from east Asia (Russian, Adygei and Burusho) and Africa (Middle Eastern populations) (Fig. 3c). The most parsimonious explanation for these results is that Native Americans have mixed origins, resulting from admixture between peoples related to modern-day east Asians and western Eurasians. Admixture graphs fitted with MixMapper^27 model Karitiana as having 14–38% western Eurasian ancestry and 62–86% east Asian ancestry, but we caution that these estimates assume unadmixed ancestral populations (Supplementary Information, section 12).”

    Later on that page:

    “allele frequency-based D-statistic tests^20 show that all 48 tested modern-day populations with First American ancestry^19 are equally related to MA-1 within the resolution of our data (Supplementary Information, section 14.4)”

    I honestly wonder if perhaps you never made it past the paywall. But even if so, the paywalled “paper” is just an extended abstract. The actual paper is the so-called supplementary information, a doorstop of 12.9 MB and 110 pages that you can freely download as a PDF from the link at the bottom of this page. Section 14.4, which stresses that all Native American samples except those from Eskimo-Aleut speakers are equally close to MA-1, begins on p. 90; also check out the end of 14.5 and all of 14.6 (p. 92).

    John Cowan, Nov. 5, 8:15 pm:

    […] Indo-Pacific isn’t even a hypothesis, just a wastebasket taxon.

    I’ve certainly never heard of an attempt to discover regular sound correspondences or shared morphology for it, but it is a hypothesis. Start here.

    German Dziebel, Nov. 8, 12:39 am:

    Some authors (G & I) reconstruct a single, compact phoneme *sw- in the form for ‘six’.

    Well, that would be convienent. Nowadays, however, [sʷ] is such a rare phoneme worldwide – Wikipedia lists only Archi, Lezgian and Lao as languages that have it – that I’d require really strong evidence to reconstruct it for an unattested language.

    That’s not Science Theory As We Know It.

    This is nonsense to me. Sorry.

    So is, to me and everything I’ve heard or read about science so far, your claim that the goal of establishing “laws” as in “natural laws” that make a theory “proven.” I don’t know of anybody who’s looking for laws in all of biology! (Everything you might have read about “laws of evolution” is either flat-out wrong or grossly exaggerated, and that’s current consensus and textbook wisdom speaking, not just me.) That’s not surprising, because all laws that have been postulated elsewhere can be derived from those of physics and mathematics.

    A law is just a generalization across a lot of observations, usually expressible as a mathematical formula. It states; it doesn’t explain.

    Laws don’t prove theories. Theories explain laws.

    yes, a more relevant reference is in the bottom of the post (http://www.kunstkamera.ru/index/science/books/books/algebra_rodstva_11/).

    I have no way of accessing your paper.

    marie-lucie, Nov. 8, 1:56 pm:

    Greenberg should have used the comparative method
    The comparative method is a method of reconstruction of the common ancestor and only secondarily a method of confirming (or infirming) a proposed genetic classification.

    I mean he should have tried to find regular sound correspondences (which he only did in a very approximate way in his macrofamily work). Without that, reconstructing a common ancestor isn’t possible in the first place.

    German Dziebel, Nov. 8, 4:26 pm:

    The implications of those methodological flaws are likely considerable. Just one example: PIE *krd- ‘heart’ shows up as *srudice in Slavic and sirdis in Lithuanian. So, the velar is presumably palatovelar (PIE *k’rd-) because the so-called satem languages (named so precisely because they turn up palatals where Greek, Latin, etc. have plain velars) show s- and not k-. But then there’s another cognate set represented, e.g., by Gk rhadix and Lat radix ‘root’. The Greek form presupposes *s (< *sradiks),

    If it’s not a Latin loan, it presupposes either *s or *w; *w is present in the putative Germanic cognates. Both would have been lost in Latin.

    which is presumably not the same /s/ as the satem /s/ < PIE *k' but PIE spirant *s. But now the full morphology of this root is unmistakably the same as the full morphology of Slav *srudice and the meanings are fully compatible ('heart', 'middle', 'core', 'root'). So, if we put the two conventional sets together and treat it as one single cognate (super)set, the reconstruction will be radically (pun intended) different.

    But why should we?

    Where do Latin cord-, Greek kard-, Germanic *hert- come from? You leave them dangling.

    Are potential cognates with meanings like “middle” or “core” attested? And does a shift from “core” to “root” have any parallels elsewhere?

    At the same time, the searching for potential Eskimo-Khoisan parallels can be done by a host of amateurs.

    Only if they understand the morphology and the sound systems of the languages they’re comparing. That’s generally not the case.

    Nov. 9, 8:36:

    E.g., his DNA-Genealogy showed that there was no significant influx of immigrants from Scandinavia in the 9-10th centuries A.D. that could justify the theory of the Western (“normanist”) origin of the Russian state.

    That simply doesn’t follow. Such an immigration would not at all have needed to be large to have the effects observed in history.

    =====================

    I’m too tired to continue, I’m off to bed.

  381. @David

    “Yes, MA-1 is indeed “closer to modern Amerindian populations than it is to any modern Eurasian population” – but the paper (p. 88),”

    So it is an Amerindian marker in the Old World by definition.

    “puts a direction on this.”

    Raghavan et al’s argument on directionality is mired in assumptions (they assume that Amerindians must have diverged from East Asians before the link with West Eurasians appeared but what if East Asians and West Eurasians both split from Amerindians but didn’t admixed with each other afterwards, so they would be equidistant to Papuans). See more in the comments section here: http://anthropogenesis.kinshipstudies.org/2015/07/amerindians-are-even-more-genetically-diverse-and-older-than-we-thought/.

    “I honestly wonder if perhaps you never made it past the paywall.”

    You must be kidding. I actually read both the main text and the supplementary material. You didn’t even know the paper I initially referenced.

    “Nowadays, however, [sʷ] is such a rare phoneme worldwide – Wikipedia lists only Archi, Lezgian and Lao as languages that have it – that I’d require really strong evidence to reconstruct it for an unattested language.”

    Thanks, that’s useful.

    “If it’s not a Latin loan, it presupposes either *s or *w; *w is present in the putative Germanic cognates. Both would have been lost in Latin.”

    Why would it need to be lost in Greek or Latin? E.g., we would expect vardix** in Latin and v- would have been perfectly safe in Latin.

    “But why should we?”

    Because we need to bake an etymological solution into the cognate set composition. Instead, we’re not first building a set, reconstructing an etymon and only then trying to find a match with a different cognate set to supply an etymology. This is ahistorical and results in dead-ends as *H1ek’wos.

    “Only if they understand the morphology and the sound systems of the languages they’re comparing. That’s generally not the case.”

    But someone needs to do the work of looking at the vast amount of data in search of potential patterns. Most amateurs do have a basic knowledge of linguistics but I agree they should learn more of it.

    “Such an immigration would not at all have needed to be large to have the effects observed in history>”

    yes, that’s one of the counterarguments.

  382. David Marjanović says:

    So it is an Amerindian marker in the Old World by definition.

    *eyeroll* If you want a theory-free description, as “by definition” heavily implies, you can’t say it’s either an ANE marker in the New World or an Amerindian marker in the Old World.

    Raghavan et al’s argument on directionality is mired in assumptions (they assume that Amerindians must have diverged from East Asians before the link with West Eurasians appeared

    …No, they test this hypothesis and find it more consistent with their data than any alternative. The scenario follows straight from their data.

    See more in the comments section here: http://anthropogenesis.kinshipstudies.org/2015/07/amerindians-are-even-more-genetically-diverse-and-older-than-we-thought/.

    That’s a really telling post, you know. You present the latest papers, then you state your conclusions in the last paragraph – and how you arrived at those conclusions is only revealed when commenters pull your arguments out of your nose one by one.

    Well. There are so many of those arguments now, and there’s so much wrong with them, that I don’t have time to pick them apart now or anytime soon. So, just one thing: are you aware that all these papers contain both phenetic and phylogenetic analyses and simply expect the readers to recognize which are which?

    I honestly wonder if perhaps you never made it past the paywall.

    You must be kidding. I actually read both the main text and the supplementary material.

    Well. You cited one figure of the paper to support your conclusion, but completely ignored everything else, including the salient fact that the paper explicitly tested and rejected your very hypothesis several times. I was left with two conclusions: either you hadn’t read those parts, or you were cherry-picking the paper based on your wishful thinking. I went with the more charitable assumption.

    You didn’t even know the paper I initially referenced.

    I did; it was, after all, all over the popular media, so there was no way to overlook it. Being affiliated with an academic institution, I downloaded and read it long ago. I just didn’t remember the names of the authors.

    You see, this isn’t my field, so I’ll never need to cite the paper in a publication of mine; and that means I have no reason to pay attention to who wrote what.

    Paying attention to names, or more generally to persons I don’t interact with much, doesn’t come naturally to me. You seem to come from a completely different headspace, seeing as you have lots and lots of people as categories on your blog.

    “If it’s not a Latin loan, it presupposes either *s or *w; *w is present in the putative Germanic cognates. Both would have been lost in Latin.”

    Why would it need to be lost in Greek or Latin? E.g., we would expect vardix** in Latin and v- would have been perfectly safe in Latin.

    I didn’t say anything about Greek. About Latin, I actually don’t know what happened to syllabic */r/ there; in the “bear” word, it turned into ur – PIE *h₂ŕḱtsos, Latin ursus.

    Besides, **var- doesn’t look Latin to me. There are plenty of words with vor- < ver-, but with var- I can only think of varius and its derivatives.

    Because we need to bake an etymological solution into the cognate set composition.

    What if your assumption that the “heart” and “root” words in question need to form a cognate set is false?

    I asked a few specific questions about that that you haven’t answered.

    dead-ends as *H1ek’wos

    I’ll get to that one. 🙂

    Back to catching up.

  383. David Marjanović says:

    Oh, on phenetic vs. phylogenetic, that’s why I asked what you meant by “closer”.

  384. Lewis & Short give us vāra ‘trestle or forked pole for spreading nets on’, vargus ‘vagabond’ (said to be a Celtic borrowing), vārico ‘straddle’ (with the delightful derivative varicator ‘one who walks with his legs spread apart’), varix ‘dilated vein’, vāro ‘stupid boor’, varus ‘pimple’, and a lot of proper names. Some of these may be related to varius somehow, but I don’t see it.

  385. @David

    “either an ANE marker in the New World or an Amerindian marker in the Old World.”

    ANE is a hypothetical population. Amerindian is a real one. Remember: MA-1 is closer to Amerindians than to any Old World population. So by definition it’s an Amerindian signature found in the Old World.

    “No, they test this hypothesis and find it more consistent with their data than any alternative. The scenario follows straight from their data.”

    No, they everywhere assume that Amerindians derive from East Asians. That’s precisely why, when they see that Amerindians are closer to East Asians than to Papuans, while West Eurasians are not closer to East Asians than they are to Papuans, they conclude that MA-1 admixed into Amerindians and not the other way around. If, however, East Asians and West Eurasians are both derived from Amerindians but there was no gene flow between EA and WE since the split, we wouldn’t expect West Eurasians to be closer to East Asians.

    “You present the latest papers, then you state your conclusions in the last paragraph – and how you arrived at those conclusions is only revealed when commenters pull your arguments out of your nose one by one.”

    That’s fair. It’s a blog, you know… But you got to read previous posts, too.

    “You cited one figure of the paper to support your conclusion…”

    It’s the one that people tend to ignore… All of the authors’ tests are based on an assumption of an East Asian origin of Amerindians prior to admixture with West Eurasians. The whole thing is untenable because based in assumptions and not data. For instance, there are no West Eurasian components in Amerindian populations, only “ANE” but MA-1 is a geographically East Asian sample. But at 24,000 YBP instead of an East Asian genetic component it showed an Amerindian one.

    ” was left with two conclusions: either you hadn’t read those parts, or you were cherry-picking the paper based on your wishful thinking. I went with the more charitable assumption.”

    Thank you, I deserve it. But I won’t return you a favor. You simply didn’t know this paper, or utterly forgot about it and now you’re desperately trying to catch up. Sorry, I have to go with the facts.

    “Oh, on phenetic vs. phylogenetic, that’s why I asked what you meant by “closer”.

    That’s indeed a critical distinction. On PCAs (phenetic) MA-1 is closer to West Eurasians than to Amerindians (but still closer to Amerindians than to East Asians), while f3 stats (phylogenetic) show that MA-1 is closest to Amerindians.

    “You seem to come from a completely different headspace, seeing as you have lots and lots of people as categories on your blog.”

    People, haplogroups, everything. You gotta keep track of things.

    BTW, I’m happy to discuss both my IE etymologies and out-of-America but I’d prefer to split them into two threads. I invite you to visit my site and we can talk about out-of-America there. And we can keep the IE discussion in this thread.

    “I didn’t say anything about Greek. About Latin, I actually don’t know what happened to syllabic */r/ there; in the “bear” word, it turned into ur – PIE *h₂ŕḱtsos, Latin ursus.

    Besides, **var- doesn’t look Latin to me. There are plenty of words with vor- < ver-, but with var- I can only think of varius and its derivatives."

    Good observations, thanks. In any case, the Latin and Greek forms are hard to reconcile with the Germanic and Celtic forms, if we are really strict about phonomorphology here. And it's hard to resolve those issues having only one heavily doctored set of forms. My approach opens it up to more data and hence more chances to solve it.

    "What if your assumption that the “heart” and “root” words in question need to form a cognate set is false?"

    Sure, that's how science works. If you can provide a different set of pairs of forms that bake an etymological solution into them, I'd be happy to consider them. But the problem is the comparative method as it exists now is vulnerable at that fundamental level. I'm trying to make it less vulnerable.

  386. ə de vivre says:

    David,

    I’m fascinated by this conversation for completely non-linguistic reasons. I hope my question doesn’t offend you because it comes from a place of respect and I’m legitimately curious, but, what do you get out of your participation this thread?

  387. vargus ‘vagabond’ (said to be a Celtic borrowing)

    Or Germanic? Cf. vargr.

  388. The question of syllabic R in Latin is a red herring because the standard etymology has an e-grade: PIE *wreh2d- > Lat ra:d- (completely regular).

  389. Re: de vivre’s “I’m fascinated by this conversation for completely non-linguistic reasons. I hope my question doesn’t offend you because it comes from a place of respect and I’m legitimately curious, but, what do you get out of your participation this thread?”

    Another ad hominem post. This time from a passive-aggressive type. Confirms my hypothesis that anonymous posters are more likely to write adhominem posts.

  390. @TR

    “The question of syllabic R in Latin is a red herring because the standard etymology has an e-grade: PIE *wreh2d- > Lat ra:d- (completely regular).”

    How do we get Germ *wurtiz from *wreh2d-? If we reconstruct R (syllabic) instead, Germ *ur is a matter of course.

  391. Fair enough, the Germanic forms do point to a zero-grade. But that too would give a long a in Latin, where *CRHC > Cra:C; the Latin form is inconclusive as to the vowel grade, but the *w- etymology works unproblematically either way.

  392. ə de vivre says:

    German Dziebel,

    I love you, please never change.

  393. @de vivre

    Can you get off the German Dziebel topic? Anybody can discuss that. But do you have anything substantial to contribute? If not, just remain silent and don’t troll the thread.

  394. @TR

    “*w- etymology works unproblematically either way.”

    Just to confirm: are you now reconstructing *wRH2d-? Either way, how can it be unproblematic if it requires *w > h in Greek, which is contrary to common sense?

  395. German Dziebel:

    You are not entitled to say on this blog what is and what is not an acceptable posting. That privilege belongs exclusively to our host, who has in this (one) case delegated it to me. Please talk about languages, or indeed anything else (topic drift is always encouraged here), but not about other posters, even if they provoke you (as it seems to you) by talking about you.

    “I have strange powers, people. Don’t make me use them. Don’t even make me show them.” —The Lord Redlady of the Instrumentality of Mankind

  396. Everyone,

    I think we’ve gnawed enough on the HEART-ROOT example. Let’s switch gears a bit. Here’s a good parallel example to the HEART-ROOT situation, which will allow us to refresh our thinking on it. IE *wLkwo- ‘wolf’ is well represented across the IE dialects (Lat lupus, Gk lykos but Germ. *wulhwaz, Lith vilkas, Slav *vliku ‘wolf’, Toch B walkwe ‘wolf’, etc.), but, again, this set does not have a good etymology. Below I’m showing that it’s morphology has been misanalyzed, too.

    Here’s my analysis of the WOLF set, which parallels my analysis of the HEART-ROOT superset. I reconstruct PIE *k’we- ‘wolf, dog’ that yielded two early affixal derivatives – *k’wen- ‘dog’ (Gk kuo:n, Lat canis, Slav *senka, Arn sun, etc.) and *k’wel-kwo- ‘wolf’. Skrt svaka ‘wolf’, vrka ‘wolf’ and Avest spaka ‘dog’, verka ‘wolf’ establish the possibility of the connection between the IE sets for the dog and the wolf. Skrt svaka/Avest spaka are morphologically archaic as they don’t have either the -n-, or the -l- (> InAr -r-) suffix. Arm skund ‘young dog’ and Gk skulaks ‘same’ show that indeed affixes -n- and -l- were in alternation. OIr cuilen and Welsh colwyn ‘young dog’ (forms without the difficult s-mobile) also confirm that the l-suffix alternated with the n-suffix. (They are not product of dissimilation from *kunen, but independent l- formations from more basic *k’we-.)

    So, basically, l- in lykos and lupus corresponds to the affixal -n- is kuo:n and canis. Gk lykos (< *lhykos) is the same as -ulaks in Gk skulaks, just like Gk rhad-i-ks is the same as -ard- in kard-i-a.

    On a methodological note, if we reconstruct PIE *k'we- 'wolf, dog' we can easily compare it with Uralic *kujna 'wolf' and Eskimo-Aleut qenRa 'wolf'. Even the IE DOG form is presently compared with this "Nostratic" set but the fact that the IE term for 'wolf' can be shown to be related to the IE term for 'dog' makes this long-range equation quite intriguing. So my methodology can actually clear some obstacles for long-range comparison.

  397. @John Cowan

    “Please talk about languages, or indeed anything else (topic drift is always encouraged here), but not about other posters, even if they provoke you (as it seems to you) by talking about you.”

    Fair.

  398. Just to be clear, if I haven’t been clear yet, Germ. wulhwaz, etc. from the WOLF set show the loss of a frontal palatovelar that’s retained in the DOG set.

  399. Eli Nelson says:

    Wait, are you now disputing any Greek etymologies where PIE *w(V)r- corresponds to Greek rho? If I remember correctly, it’s a pretty widely used correspondence, regardless of what you think of it in terms of common sense. Rough breathing on word-intial rho was not contrastive, so you could also consider it as a simple loss of the *w (as happened in modern English) followed by allophonic word-initial devoicing of the rho. Unfortunately I don’t have access to a professional etymological dictionary right now, but here are some correspondence sets using it that are listed in Wiktionary: Germanic : *wurmiz “worm”, Latin vermis “worm”, Ancient Greek ῥόμος “wood-worm”; Ancient Greek ῥέζω “I do” and the “work” words; Latin ūrīnor ‎”submerge in water” and Ancient Greek ῥαίνω “sprinkle”. There are also words possibly not descended from PIE that show loss of w before r, such as ῥόδον “rose” which Wiktionary says has been connected to Arabic warda “rose, blossom” with the ultimate origin likely being Persian. Wiktionary even lists ϝρῆξῐς as an attested alternative form in Aeolic dialects of ῥῆξις “breaking.” Are all of these erroneous?

  400. simple loss of the *w (as happened in modern English)

    Despite intuition, there’s a good case to be made that /r-/ and /wr-/ merged into /wr/ rather than /r/, or at any rate that all /r-/ in English is strongly labialized, which is probably what /wr-/ was before.

  401. @Eli Nelson

    “Wait, are you now disputing any Greek etymologies where PIE *w(V)r- corresponds to Greek rho?”

    As I mentioned somewhere above, the most stable and natural correspondence is IE s ~ Gk h. Greek also shows the ubiquitous loss of IE *w. I therefore think that the handful of examples where Gk h SEEMS to correspond to IE w and y must be false positives. They are neither phonetically natural, nor systematic.

    I haven’t worked through all the cognate sets you listed but regarding Germanic : *wurmiz “worm”, Latin vermis “worm”, Ancient Greek ῥόμος “wood-worm” I compare it with IE *kwrmis ‘worm’ (Skrt krmi-, Lith kirmis, Alb krimb, etc.) I would prefer if the *kwrmis set turned up a palatovelar (to link to h < *s in Greek ῥόμος) but in any case IE w does not correspond to h in the Greek word. -w- is regularly lost in Greek.

  402. Eli Nelson says:

    Why does anything need to correspond to an h in the Greek word? As mentioned earlier, word-initial rho was regularly written with a rough breathing. This is true even for loanwords and placenames that had word-intial /r/ in the source language, where there is clearly no historically distinct /h/ segment. I think I remember reading about one exception to this rule involving a Greek goddess whose name was written with a rho with smooth breathing, but the general practice seems to show that there was no exception for words without etymological /h/.

  403. Eli Nelson says:

    My point is that the rough breathing on rho does not seem to be sufficient evidence for reconstructing an independent /h/ segment before it historically. There may have been; but I don’t see how you can say that there must have been.

  404. David Marjanović says:

    German Dziebel, Nov. 10, 9:53 am:

    This means that *s > hr in Greek and not *s > r as it happened in Latin (radix) where *s is always lost in the onset of a word before sonorants.

    You mean *sr > *hr and *sr > *r, right?

    Morpheme-initially, *sr wasn’t allowed to become r in Greek. Nothing was allowed to become r there. “Rh” was the only option.

    Perhaps an intermediate stage *hr, from earlier *sr, is the reason why “rh” was generalized to every morpheme-initial r; but the fact is that both *sr- and *wr- ended up in Greek as “rh-“, and that loans with truly initial r also did.

    In short, we cannot distinguish *sr > *hr > “rh” from *sr > *r > “rh” in Greek. We must expect the same outcome either way.

    For a voiceless r that is not descended from earlier [hr], see Welsh. It’s even spelled rh.

    But we do need to check the hypothesis against other phonetic environments. And we indeed find support for it in Gk hippos ‘horse’ (Lat equus), which I compare not just with Skrt asva but also with Skrt sapho- ‘hoof’ (Slav kopyto, with an ending just like in Gk hipotes). Not only do we have Gk h ~ Satem *s again, but also centum kw > p in Satem languages! Not only does the 19th century belief in satem languages disappears under a 21 century methodological scrutiny, but so does the 19th century belief in kentum languages! Instead, what we have is a positional conversion of velars into palatalovelars or into labiovelars on the PIE level (!) in response on some – admittedly still poorly understood – suprasegmental pressures that result in empirically attested doublets

    Wow, what a splendid return of conjectures on the investment of so little fact.

    What, if anything, makes you think that the Skt for “horse” and the Skt for “hoof” are cognate with each other and yet follow completely different sound correspondences? If at least one of them were a loan… but no…

    (comp. the mobility of aspiration in Gk thriks vs. trikhos ‘hair’).

    I thought trikhos is simply the completely regular genitive of thriks? Because if so, we’re not dealing with a doublet here.

    What we’re dealing with is Grassmann’s law: from some rather late stage onwards (*s had already turned into *h), Greek only allowed one aspiration per word. If there were two, the first was deleted. So, first, on the morphophonemic level, you get thrikhs, genitive thrikhos; on the phonemic level, gs, ks and khs merge into something where aspiration is at least not contrastive, so you get thriks at the phonemic level, and that already fulfills the new constraint; in thrikhos, you have two at the phonemic level, and Grassmann’s law deletes the first, yielding trikhos. (And ultimately Trichosurus, the brushtail possum.)

    Nov. 10, 2:22 pm:

    I’m not discarding them but the set you have in mind is a subset of a real, natural, historical set. The two sets (the HEART set and the ROOT set) were misanalyzed as two separate, unrelated sets.

    If you say so…

    Systematic morphological, semantic and phonetic resemblances between the two have been overlooked. Phonetic problems (e.g., Skrt hrd ‘heart’ that can’t be derived from PIE *k’rd but it is undoubtedly part of the HEART set) have been ignored or underestimated.

    If that form is real, I must mention that Skt h (voiced [ɦ]) can’t be derived from *s either, only from *ǵʰ.

    Nov. 10, 2:32 pm:

    classification precedes reconstruction

    It’s recursive. First you classify tentatively, then you find regular sound correspondences, then you find irregularities, then you revisit the correspondences and/or the classification, then you try again…

    I think that it’s cognate sets that are the minimal units of classification and not languages

    I don’t quite understand what you mean. After all, you haven’t claimed that different sound correspondences can apply to different sets in the same language.

    marie-lucie, Nov. 10, 2:50 pm:

    Earlier I cited some of the problems with some of Greenberg’s subclassifications within “Amerind”, which do not inspire confidence in the validity of the larger group.

    Greenberg didn’t build his “Amerind” from the bottom up, though. The subclassification is later and, by his own criteria and – IIRC – his own frank admission, less well supported than “Amerind” as a whole.

    German Dziebel, Nov. 10, 3:12 pm:

    Forms with s- and without s- are typical in IE langages and sometimes described as the s-mobile problem. I’m not trying to hide behind it, though. On the contrary, a chronically unresolved phonetic problem can be resolved with a change in the approach.

    I don’t think any approach can fully explain the German synonyms lecken (northwest?) and schlecken (southeast?) or a number of other such cases.

    Nov. 11, 12:52 am:

    The k- vs. s- doublets are well know from satem languages (Slav *korwa vs. *serna), so the copresence of kardia and rhadiks in Greek is not extraordinary.

    It absolutely is, because that would be the one and only case of s from a palatalized velar in the whole language.

    Interestingly, palatalization is conventionally thought of as getting lost in kentum languages, so k’ > k already presumes as s-mobile-like phenomenon.

    Not at all. It’s a regular merger of two (series of) phonemes, as opposed to the random presence or absence of one phoneme.

    So why doesn’t aśva have this p?

    As we know from kentum languages, labiovelars sometimes stay as k(u), sometimes change to p.

    …But that happens in different languages, not in different randomly selected words in the same language.

    OK, in Greek it does happen in the same language, but in different (and easily identifiable) phonetic environments.

    IE *(d)akr-/-n ‘tear’ (Hitt ishahru, Skrt asru, Gk dakruma, etc.)

    These are supposed to be cognate? Are you serious?

    The morphology is identical (heteroclitic -r-/-n-).

    So is that of “liver” and “water” and a bunch of others. I don’t see your point…?

    IE *kmtom or *kntm ‘hundred’ can be compared to IE *seno- ‘old’ (Lat senatus showing the same extension -t- as centum).

    Or textbook wisdom could be correct, and “100” is simply formed from “10”, *dḱm-t-ó-m, with simplification of the impressive consonant cluster by loss of the *d.

    Why is “old” *sen- and not *sem-?

    IE *swekuro- ‘husband’s parent’ ~ IE *sweso:r ‘sister’

    The latter fairly obviously contains the root *ser-, which was used to form words for women in Anatolian and also shows up in the feminine forms of “3” and “4” that only Celtic and Iranian have (*tisres < *trisres; *kʷetesres).

    TR, Nov. 11, 1:43 am:

    Btw, I’ve finally looked up Greek ῥάδιξ (which sounded unfamiliar) and it turns out to be a post-Classical word first attested in Nicander (2nd century BC), and very likely a borrowing from Latin.

    Why, though, would this Nicander borrow a Latin word for “root” as soon as the Romans show up?

    1:56 am:

    Btw, a recent attempt by Chiara Bozzone [and Clara Felisari] to explain the initial aspirate of hippos can be found here. I’ve yet to fully think it through, but it looks interesting.

    That’s fascinating!

    Section 3.1 says: “In treating the etymology of ἵππος, de Vaan (2009:203) […] explains the –i– vocalism in the root by schwa epenthesis, after Vine (1999), a phenomenon situated in Pre-Greek. The ø grade in Greek is explained as deriving from a Late PIE (=Core PIE) thematization of the original –u-stem noun (preserved in Hittite), based on the genitive singular of the inflection.”

    The last sentence says that the form *h₁éḱwos isn’t PIE at all, but belongs to the non-Anatolian branch, and was built from the actual PIE form, which was *h₁éḱus!

    And then it turns out this has been on Wiktionary all along. Just goes to show how maddeningly heterogenous Wiktionary is.

    Next, Bozzone & Felisari mention that Greek not only has ἵππος from zero-grade *h₁ḱwo-, but also a person name Ἐπειός from e-grade *h₁eḱwo-: no /h/, no /i/, and no length on the /p/.

    I conclude that the /h/ and the /i/ that still puzzled Ringe in 2009 are explained; what’s left is the /pː/, which wasn’t mentioned by either Ringe or Bozzone & Felisari, but which appears to be a development within Greek if Ἐπειός isn’t some kind of loan.

    John Cowan, Nov. 11, 2:08 am:

    If you insist on real names, consider visiting Paleoglot.

    Heh. That would be a vicious kookfight… but a very short one, alas.

    TR, Nov. 11, 2:09 am:

    though I would want to see a good explanation of the subsequent accent retraction.

    Some kind of nominalization after a non-trivial amount of derivation?

    John Cowan, Nov. 11, 9:42 am:

    Nilo-Saharan is increasingly recognized as what biologists call a wastebasket taxon: a language spoken in the rough and fractal borderland between the first two families that did not look like either of them tended to get dumped into Nilo-Saharan. It should probably be broken up until proper comparative work is done. Some linguists do defend it as written, however.

    Have there been any new developments since this thread (where we both commented)?

    Khoi-San is typologically defined by a single property, the presence of click consonants. Nobody has ever found (or thought of, depending on how you look at it) a plausible phonological process by which non-clicks could be transformed into clicks, and so the assumption has been that it must be a primitive character. […] The languages are intensely diverse otherwise, and there seems to be nobody who believes any more that they can possibly all be related.

    I’m surprised you haven’t encountered the idea that clicks come from consonant clusters. Start from clusters; proceed to coarticulated consonants like this kind of situation; then get the timing subtly wrong, and suddenly there are clicks all over the place.

    “Nobody” is a strong word; this page links to PDFs of five papers on the subject by G. Starostin. I haven’t read them, though, except probably the one on Hadza long ago.

    marie-lucie, Nov. 11, 12:21 pm:

    Greenberg’s proposed Eurasiatic (which includes Nostratic as well as yet other groups)

    Not quite. Greenberg’s Eurasiatic included Nivkh and, puzzlingly*, Ainu; Nostratic excludes Ainu and at least nowadays also Nivkh, but… used to include Afro-Asiatic, though now the Moscow School has changed its terminology and excludes Afro-Asiatic while maintaining that it’s the closest relative of “Nostratic”. Clearly, they could have called that “Eurasiatic”, but didn’t because they don’t want to be seen as accepting Greenberg’s work.

    * Typologically, the other language families in the group range from lightly agglutinating with suffixes and postposed clitics (like Japanese) to fusional with suffixes (IE); Ainu uses long chains of prefixes instead. It’s probably Austric, but that’s another story.

    John Cowan, Nov. 11, 12:54 pm:

    Fifteen Formosan languages survive, some of which are represented by several different dialects. All of these languages are threatened with imminent extinction, and none of these has been documented in the shape of a comprehensive grammar with extensive documentation of its lexicon.

    None of them?!? That’s surprising. And horrifying.

    German Dziebel, Nov. 11, 5:38 pm:

    That’s why you need real names. Real names make people responsible for what they post, especially when addressing other people.

    People who use the same pseudonym consistently for years and years on different sites are hardly anonymous for this purpose. They build up a reputation, for starters.

    You really haven’t thought this through. I’ll have to compare you to Zuckerberg again.

    6:23 pm:

    But we don’t see s- in either Greek or Latin.

    What do you mean? We do! /h/ in Gk rhadiks is from *s.

    No, that’s not how it works. “Rh” is just the word-initial allophone of /r/. The aspiration/devoicing is automatic and inevitable. They did it to Rome, for crying out loud.

    It so happens that PIE didn’t allow /r/ at the beginnings of words; so, if this word is inherited from PIE (rather than borrowed from Latin), it must have begun with either */sr/ or */wr/.

    You assume the former possibility and ignore the latter for no discernible reason… other than your desire to connect the word to the Balto-Slavic “heart” word, that is.

    There’s just as much evidence for a laryngeal in Gk ke:r ‘heart’.

    Nope, that length comes rather transparently from Szemerényi’s law.

    But Gk rhadiks is lacking a:.

    How do you know? If it doesn’t appear in poetry somewhere, you don’t know. Length on /a/ wasn’t written.

    “Not to mention that adopting your etymology means losing the etymology for ῥίζα.”

    Not losing it. Just putting it on a back burner in view of its strange vocalism (to your own point),

    Can’t the /i/ be explained as an epenthetic vowel inserted into zero-grade */wr̩d/- as evidently happened in ἵππος?

    Nov. 13, 12:34 am:

    Starostin (https://www.academia.edu/2497705/Dene-Yeniseian_a_critical_assessment) does criticize the key part of Vajda’s proof, namely the multipartite verb morphology.

    …Everyone should read this paper and Vajda’s reply which is included in the same file. Thanks a lot for the link!

    And so, to bed, it’s 1:42 am over here. As a more famous Austrian has said a couple of times: I’ll be back.

  405. David Marjanović says:

    Long comment currently in moderation.

    David,

    I’m fascinated by this conversation for completely non-linguistic reasons. […] what do you get out of your participation this thread?

    http://xkcd.com/386/

    On top of that, I’m learning interesting things here. Particular highlights so far are the explanation of the /h/ and the /i/ in hippos, and G. Starostin’s critique of Dene-Yeniseian – with Vajda’s reply attached.

  406. As Eli says, there is no /h/ in the Greek words that are transcribed rh-. It’s just a voiceless R. In Greek all initial R’s are voiceless. Since Attic Greek lost /w/, initial /wr-/ became /r-/, and /r-/ was pronounced [r̥]. There’s nothing either controversial or phonetically unnatural there, and it involves no change *w- > h-.

  407. @David

    ” *sr- and *wr- ended up in Greek as “rh-”, and that loans with truly initial r also did.”

    You want to have the cake and eat it too: on the one hand, r is always rh in Greek regardless of its etymological source; on the other hand, rh comes from *sr or from *wr. I don’t care if all loans with r- turned into rh-, I care that all rh that are not in loan words come from *sr and IE wr ~ Gk hr is a false positive (in reality, s > h and w lost).

    “Wow, what a splendid return of conjectures on the investment of so little fact.”

    This is the opposite from truth. I offer facts and logic, you offer only tradition and belief. I’ve noticed that this blog is notorious for its ad verecundiam thinking. Even trolls with a tenure becomes authorities here.

    “What, if anything, makes you think that the Skt for “horse” and the Skt for “hoof” are cognate with each other and yet follow completely different sound correspondences?”

    There’s work to be done here, indeed. That’s what reconstruction is all about. It doesn’t just land on your lap. But you need to have the right material to even pose the right questions. *H1ek’wos, on the contrary, is a dead-end.

    “What we’re dealing with is Grassmann’s law.”

    Yes. That’s why it’s interesting. Something of similar suprasegmental nature may have caused the disparity between asva and sapha, or svaka and vrka.

    “If that form is real, I must mention that Skt h (voiced [ɦ]) can’t be derived from *s either, only from *ǵʰ.”

    Indeed, Skrt hrd- contradicts the traditional reconstruction of *k’rd. Hence, we need more material and better thinking. I tentatively note similarity between -ph- in sapha and h- in hrd.

    “E *(d)akr-/-n ‘tear’ (Hitt ishahru, Skrt asru, Gk dakruma, etc.)

    These are supposed to be cognate? Are you serious?”

    Yes. You asked me for more examples and now you are complaining. Scholars have already noticed similarity between the two, so why not interpret it as “identity by descent” and not contamination if it’s logical, semantically valid (tear and blood are both bodily liquids) and supported by other examples???

    “Why is “old” *sen- and not *sem-?”

    Fait. *sem- looks better in some respects. Both would work formally at the moment. We need to better understand how the whole numeral system of IE evolved to see which one is better.

    “The latter fairly obviously contains the root *ser-, which was used to form words for women in Anatolian and also shows up in the feminine forms of “3″ and “4″ that only Celtic and Iranian have (*tisres w? *s > h and *w > 0, on the other hand, are natural and well attested. Discernible enough?

    “Can’t the /i/ be explained as an epenthetic vowel inserted into zero-grade */wr̩d/- as evidently happened in ἵππος?”

    I didn’t buy that epenthesis argument for hippos, so it’s hard for me to use it as a model. But -i- from *ye would make sense if the front palatovelar in PIE *k’ek’wo- ‘hoof, horse’ was reanalyzed with as a cluster *kye- and then became -i in Greek.

    “No, that’s not how it works. “Rh” is just the word-initial allophone of /r/. The aspiration/devoicing is automatic and inevitable. They did it to Rome, for crying out loud.

    It so happens that PIE didn’t allow /r/ at the beginnings of words; so, if this word is inherited from PIE (rather than borrowed from Latin), it must have begun with either */sr/ or */wr/.”

    So, I’m only interested in the cases when it’s inherited from PIE. *wr > hr doesn’t work, so one is left with *sr > hr and *w > 0. I hope you’re not denying multiple uncontroversial cases when rh- in Gk corresponds to IE *sr.

    “Nope, that length comes rather transparently from Szemerényi’s law.”

    It’s so funny. You tend to say something with utmost confidence but then it quickly turns out to be wrong. Your sources says literally this: “The PIE reconstruction for “heart” is the single instance where *d is deleted after *r, with compensatory lengthening of the preceding vowel. It is not clear whether this is an isolated example, or a part of a broader process such as Szemerényi’s law.”

    Why is it that what’s transparent to you is unclear to others and the other way around? 🙂

  408. @David

    ““The latter fairly obviously contains the root *ser-, which was used to form words for women in Anatolian and also shows up in the feminine forms of “3″ and “4″ that only Celtic and Iranian have (*tisres w? *s > h and *w > 0?”

    Something happened here. My response should have read the following: “This is one of the worst IE myths, the one fueled by generations of thinking using a flawed methodology. There’s not a single IE language in which *ser/*sor is attested as a term for ‘woman’. The evidence is circumstantial (HLuw nanasrai-, hassusara, Celt *tisres are all compounds) and very gappy. How can it be used to explain such a pervasive IE form as *sweso:r?! Indo-Europeanists don’t realize how kooky their etymologies sometimes are. Sister = ‘one’s own woman’. What an intellectual breakthrough!

  409. @David

    “People who use the same pseudonym consistently for years and years on different sites are hardly anonymous for this purpose. They build up a reputation, for starters.”

    And then they lose it in one day! 🙂

    “You really haven’t thought this through. I’ll have to compare you to Zuckerberg again.”

    In my personal experience, over the past 10 years, the most meaningful contributions were made by people with real names or with easily traceable identities. I don’t want to oversimplify the situation, but I just follow the facts as they come along. On this blog: SFReader, minus209676, joie vivre are all ad hominem. Marie-Lucie, David Marjanovich are not. TR in between. I don’t why, though. Maybe anonymous people don’t like strong opinions. they want everything, not just names, to be understated. 🙂

  410. Ainu uses long chains of prefixes instead. It’s probably Austric, but that’s another story.

    Not to derail from the horses and heartroots, but… so far the best arguments I’ve seen for Ainu as part of “Greater Austric” are a bunch of loose wordlists where, for example, proto-Ainu initial /k/ corresponds to /k/, /x/, /s/, 0, etc, “ear” and “dirt” are supposed to be cognates, etc. Overall the evidence for this “Greater Austric” including Ainu hasn’t seemed to me any more convincing than the evidence for “Greater Altaic” including Korean and Japanese (or GAlt including Ainu, for that matter). What makes you come down on the “probably” side?

  411. And then they lose it in one day! 🙂

    The only one who has lost it here is you. The others you mention are all longtime contributors who have always had sensible and interesting things to say; you show up parading your (alleged) real name and (alleged) credentials and produce endless strings of (to be generous) inadequately supported ideas, and when called on the obvious failings say something like “There’s work to be done here.” I have been staying away from the thread because I find you so irritating, but since others seem to enjoy interacting with you despite your obnoxious ways I have let the thread take its course, asking John Cowan (who I know is not easily offended) to keep an eye on it and let me know if you were crossing the line. This is just a note to urge you in the strongest terms to stick to discussing the linguistic issues and refrain from any personal comments whatever about your interlocutors, however strongly you feel the urge to call them trolls or ignorant or the like. Treat everyone here as though they were respected equals; if you can’t bring yourself to do that, then just go away. Thanks!

  412. @languagehat

    You just proved my point. Get a real name, real creds and learn how to counter an argument with an argument and not with an ad hominem attack. I haven’t attacked anybody on this blog and only defended myself a few times against the personal attacks by others. “when called on the obvious failings say something like “There’s work to be done here.” and “inadequately supported ideas” are not rational arguments.

    “I have been staying away from the thread because I find you so irritating.” Sure, a carefully argued and well-supported position from a self-critical commenter is always irritating to someone like you.

    I don’t need to blog here. And don’t try to bully me with your rants. If you block me, you will leave me with no other option than to have you join my Kunstkamera collection of dishonorable online individuals at http://www.anthropogenesis.kinshipstudies.org.

  413. Otherwise, as Hat is too polite to say but I am not, the Strange Powers come into play: third time is the charm.

  414. marie-lucie says:

    Back to semantics:

    horse/hoof

    Even though horses have big hooves, they are far from being the only animal that has them (see the name “Ungulates”), so I don’t see the rationale for naming the one after the other (in any order).

    tears/blood

    If the common meaning is “liquid”, there is at least one other liquid rejected by the body: urine. The appearance, smell and feel of these various liquids are quite different, as well as the conditions under which they leave the body. I would be extremely surprised to learn of a language in which their names (in twos or threes) derive from a common ancestor.

  415. A sense of infallibility is a wonderful thing to possess.

  416. Get a real name

    Your powers of observation apparently do not extend to casting your eye to the right, where it clearly says “My name is Steve Dodson.” I have been languagehat all over the internet since 2002; it is a nice, memorable moniker and has served me well, but my real name is not a mystery. You could of course respond “How do I know that’s your real name?” — to which I would respond “a-HA-a!” You will of course have contempt for my responding with a video clip of a joke instead of a reasoned argument, but that’s OK, because you’re out of here.

    I don’t need to blog here. And don’t try to bully me with your rants. If you block me, you will leave me with no other option than to have you join my Kunstkamera collection of dishonorable online individuals at http://www.anthropogenesis.kinshipstudies.org.

    You sure don’t, I’m not bullying but banning you, and I am thrilled to join your collection of dishonorable online individuals; believe me, you’re far from the first internet kook/jerk who’s become annoyed with me. We’ll just each have to pronounce anathema on each other and accept our respective fates. At any rate, any further comments from you will be deleted, but I urge everyone else to carry on discussing whatever interesting topics have arisen.

  417. ə de vivre says:

    David:

    Fair enough. It’s often easy to forget that not everyone has the same temperament as I do.

    L-hat:

    I apologize if I’ve needlessly aggravated the situation. I saw behaviour I didn’t understand and so I asked about it, but I probably could have phrased my question better for the sake of the overall tone of this thread.

  418. I apologize if I’ve needlessly aggravated the situation. I saw behaviour I didn’t understand and so I asked about it, but I probably could have phrased my question better for the sake of the overall tone of this thread.

    Nonsense, I’ll have no self-flagellation among my valued commenters! It was perfectly in order to ask, and there’s no requirement to be ostentationsly polite; the only requirement is not to stoop to open insults, and the only one who has done that is “German Dziebel” (if that is his real name). Well, and maybe me, but I don’t respond well to jerks, which is why I was staying out of the thread.

  419. @David:

    Why, though, would this Nicander borrow a Latin word for “root” as soon as the Romans show up?

    Well, I don’t know, but it’s not like there aren’t Latin loanwords in Greek (though of course not nearly as many as the other way around). If it’s a native Greek word, it stayed in hiding for a very long time. Not that it matters for the ultimate etymology, which is the same either way.

    [hippos accent retraction] Some kind of nominalization after a non-trivial amount of derivation?

    Yes, presumably that’s what you’d have to say, but it still strikes me as needing specific parallels. It’s not like there’d be anything wrong with a Greek noun ἱππός.

    [ῥίζα] Can’t the /i/ be explained as an epenthetic vowel inserted into zero-grade */wr̩d/- as evidently happened in ἵππος?

    That’s more or less how it has been explained, by Brent Vine: http://www.pies.ucla.edu/IESV/1/BV_rhiza.pdf

  420. The Strange Powers have done their work, and the Bone-Hammer (< OE ban-hamor, but widely misunderstood on the Internet) has struck a blow for peace, order, and good government.

  421. marie-lucie says:

    David, TR: Why, though, would this Nicander borrow a Latin word for “root” as soon as the Romans show up?

    It makes a big difference whether the “root” in question is the part of a tree that is below ground, for which a word would hardly need to be borrowed, or an edible root such as a carrot, radish, parsnip, etc. Unlike green vegetables, many edible roots can be stored for quite a long time and transported with a minimum of care, making them possible items of long distance commerce witness the occasional appearance of a strange root with a Japanese or other foreign name in the vegetable section of a supermarket). One would have to know the semantic contrast between the referent of the borrowed word in question and other “roots” the writer was familiar with (but did not call by the normal “root” word).

  422. Oh dear, I haven’t been here for just a few weeks and look what happened in the meantime. Hundreds of posts to catch up with and a brawl I seem to have missed.

  423. Well, welcome back in any case! The kerfuffle was confined to this thread, and hardly worth it from a linguistic point of view.

  424. [ῥίζα] Can’t the /i/ be explained as an epenthetic vowel inserted into zero-grade */wr̩d/- as evidently happened in ἵππος?

    That’s more or less how it has been explained, by Brent Vine

    To be a bit clearer: Vine’s idea isn’t purely phonological, since *wr̩h2d- would regularly give Greek (w)ra:d-. He assumes the later analogical creation of an oblique zero-grade stem wrd-y-, and it’s this stem which would have undergone schwa epenthesis into wridy- and later been generalized.

  425. David Marjanović says:

    *phew* Well, catching up just became a lot easier, even though I won’t stay up much longer tonight. The most urgent things:

    Not to derail from the horses and heartroots, but… so far the best arguments I’ve seen for Ainu as part of “Greater Austric” are a bunch of loose wordlists where, for example, proto-Ainu initial /k/ corresponds to /k/, /x/, /s/, 0, etc, “ear” and “dirt” are supposed to be cognates, etc. Overall the evidence for this “Greater Austric” including Ainu hasn’t seemed to me any more convincing than the evidence for “Greater Altaic” including Korean and Japanese (or GAlt including Ainu, for that matter). What makes you come down on the “probably” side?

    Well, first of all I have a lower threshold for saying “probably” than you do. 🙂 There’s evidence (however little) for Ainu being Austric, there’s no evidence for it being anything else, so it’s probably Austric. I’d rather have a weakly supported hypothesis than none at all.

    Also, I find some of the evidence for Altaic – including Korean and Japanese; they don’t fit in worse than, say, Mongolic does – convincing enough. But of course plenty of work remains to be done; in particular, loans between Turkic and Tungusic that bypass Mongolic are not in fact impossible, and the Proto-Japonic reconstruction that was used in the 2003 reconstruction of Proto-Altaic was outdated (at least the vowel system was).

    The paper I’ve seen I only have on paper, and not here. Maybe I’ll find it later. There was a word list, but I can’t remember “ear” and “dirt” being declared cognate; “blood” and “blood” were. Concerning the sound correspondences you mention, that’s exactly what you can see within Indo-European today…

    It’s so funny. You tend to say something with utmost confidence but then it quickly turns out to be wrong. Your sources says literally this: “The PIE reconstruction for “heart” is the single instance where *d is deleted after *r, with compensatory lengthening of the preceding vowel. It is not clear whether this is an isolated example, or a part of a broader process such as Szemerényi’s law.”

    Oh, sorry. I expected a nominative *-s, which would disappear with compensatory lengthening by Sz.’s law, but – being neuter/inanimate – it wouldn’t actually get this ending. The long vowel in Greek, which is projected straight back to PIE in the Wikipedia article for unstated reasons, could of course be analogical to those produced by Sz.’s law in other words, but that says nothing about the *-d that’s missing from the nominative.

    Would *ḱérd actually be a permissible word in PIE, though? It certainly wouldn’t be in Greek, which only allowed very few consonants at the end of a word, and /d/ isn’t one of them. Latin seems to have disliked consonant clusters at the ends of nouns (the nominative of lact- is lac). In Germanic the *d is saved by turning the word into an n-stem, so no evidence is left either way. The article mentions Hittite ker, apparently without evidence for vowel length; whether the *d was genuinely missing or just not written because there’s no clear way to do that in cuneiform is perhaps another question.

    Well, welcome back in any case! The kerfuffle was confined to this thread, and hardly worth it from a linguistic point of view.

    A few interesting problems were brought up. It’s just the proposed solutions that wander off the deep end.

  426. David Marjanović says:

    Vine’s paper, incidentally, confirms long /aː/ (and long /iː/) for rhadix… which means “branch, frond” rather than “root”, but otherwise really does look like a very early Latin loan.

  427. It makes a big difference whether the “root” in question is the part of a tree that is below ground, for which a word would hardly need to be borrowed, or an edible root such as a carrot, radish, parsnip, etc.

    Well, the same Latin word is where English gets radish, so…

  428. @David: what’s left is the /pː/, which wasn’t mentioned by either Ringe or Bozzone & Felisari, but which appears to be a development within Greek if Ἐπειός isn’t some kind of loan.

    I’m not sure why you see a problem with the standard account, by which PIE *ḱw gave Greek /p:/ (basically, just like a single labiovelar *kw gives /p/, but preserving the phonological weight of the cluster). I don’t know if there are supporting parallels, since this cluster was rare, but it seems natural enough. Of course if this is so, Ἐπειός can’t be related, but there’s no particular reason it needs to be.

  429. (Off topic, can someone tell me why the HTML sup and sub tags I’ve been using for kw and h2 respectively are getting ignored?)

  430. I’m afraid sup and sub tags only work in posts, not in comments. It’s a WordPress thing.

  431. Got it. What I should have asked was, how do savvier commenters get their subscripts and superscripts to display correctly?

  432. By using Unicode characters in plain text rather than HTML makeup. Here are some useful samples: h₁ h₂ h₃ kʷ pʰ. Anything that’s in IPA can be created using the Weston Ruter virtual keyboard, which provides a clickable IPA chart and puts what you type into a box, where it can be copied or cut and then pasted elsewhere. You can, for ease of use, type ordinary letters to the page as well.

    If you use Windows and either the US or the UK physical keyboard, see the page for my Moby Latin keyboard (you can google it), which allows typing almost 1000 characters including a useful subset of IPA, mostly for transcribing English.

  433. Arrgh. HTML markup, of course.

  434. Thanks, JC!

  435. I’m afraid sup and sub tags only work in posts, not in comments. It’s a WordPress thing.

    So you’re saying that the violence is inherent in the system?

    David: Thanks!

    There’s evidence (however little) for Ainu being Austric, there’s no evidence for it being anything else, so it’s probably Austric. I’d rather have a weakly supported hypothesis than none at all.

    But “not discernably related to any other known language” is a hypothesis too, admittedly supported only by negative evidence (necessarily). I can certainly see good reasons for encouraging the exploration of non-null hypotheses instead of just waiting for some genius to come along and convincingly demonstrate a relationship in a single blow, but that “shouldn’t be ruled out” seems very far from “probably”, even allowing for different thresholds (we want a baseline of “more likely than not,” surely).

    Concerning the sound correspondences you mention, that’s exactly what you can see within Indo-European today…

    Oh, absolutely! But the situation there is not higgledy-piggledy; the data has been combed through and convincing arguments made that in environment X, you get /x/, in Y, you get /y/, and so on. Not having seen any of that for Ainu, the only reason I can see to rank the Austric hypothesis above, say, the Semitic one (for which I have seen wordlists too) is because of geography.

  436. marie-lucie says:

    TR: Latin radix, Greek rhadix: the same Latin word is where English gets radish, so…

    Actually the English word is not borrowed directly from Latin but from French le radis. The plain French word is for those cute little red (and white) raw titbits, as well as for le radis noir ‘horseradish’, which is a much larger root, dark outside and white inside.

    David: rhadix… which means “branch, frond” rather than “root”

    That would make sense if the PIE root has wr- and is related to English wort, which always refers to a green plant, not its root. Actually the young fronds of red/white radishes are edible (preferably cooked, as in soup).

  437. David Marjanović says:

    German must have borrowed radix more directly: Rettich. Just the long consonant doesn’t make sense.

    English wort, which always refers to a green plant, not its root

    German Wurzel means “root”. German does have unextended -wurz in a bunch of plant names, but there are dialects which use this form for “root”.

    …Well, no, they extend it with -n, but analogical meaningless -n on feminines is not surprising at all.

  438. David Marjanović says:

    Rettich means “radish” of course.

    Concerning special characters, I copy & paste them from the Windows character map. It’s intricately hidden in Start > All Programs > Accessories > System Programs* > Character Map, but after you’ve used it once it stays in the start menu.

    * Because where else other than in Accessories would you ever put System Programs. *facepalm*

  439. Ксёнѕ Фаўст says:

    @Piotr Gąsiorowski

    Lat. pastor : PSl. *pastyrь ‘shepherd’ (note that Slavic uses *-tel-, not *-tor- in occupational terms).

    Is the apparent raising of ō to ū in this word the same thing as the one seen in, among others, *kamy, *mati (and *netopyrь??).

    (How did Latin acquire the -tor : -tōris alternation anyway, some kind of open syllable lengthening?)

  440. @Ксёнѕ Фаўст

    My suspicion is that it’s the same process (yes, in the ‘bat’ word too, and also in ‘four’)

    The vowel quality and length in Latin were levelled out to match the nom.sg. Then long vowels were shortened in final syllables before any consonant other than /s/, which paradoxically made the nom.sg. different again from all the case-forms it had influenced.

  441. David: Sounds like you’re a candidate for my (projected) QWERTZ version of the Moby Latin keyboard driver.

  442. David Marjanović says:

    Perhaps, but the character map works fine for my purposes. I don’t need to write whole texts…

  443. David Marjanović says:

    A few loose ends…

    German Dziebel, Nov. 13, 11:42:

    5. The current reconstruction *H1ek’wo- leaves the IE word for HORSE without an etymology. Ringe acknowledges it. For me, this is a big problem because if a reconstruction doesn’t elucidate the origin of a word, there must be something wrong with the reconstruction.

    lolwut?

    1) What if the word is simply inherited, turtles-all-the-way-down? Perhaps it goes back to the origin of language, and its first ancestor meant “zebra” or “eland” or who knows what?

    Seriously, what did he want, circular derivation of all PIE roots from each other, like Tangut characters?!?

    That would be as wrong-headed as the Tangut script itself, now that I’m thinking of it.

    2) The Moscow School has proposed that */(h₁)ekʲwos/ is a loan. That’s because they reconstruct the Proto-(North-)Caucasian “horse” word as */ɦint͡ʃwi/ (IIRC). Imagine a language closely related to PNC that had lost */n/ in this environment, apparently a typologically common phenomenon. Arguably, */kʲ/ was the closest thing PIE had to [t͡ʃ]…

    3) An IE etymology for */(h₁)ekʲus/ has of course been proposed: as a noun derived from a root */(h₁)ekʲ/- which would underlie the word for “swift”. More on this here.

    Nov. 14, 9:41 am:

    so much of Indo-European cultural reconstruction, homeland and dating hinges on its wheeled transport pulled by horses. In your thought experiment it amounts to a situation when the English word ‘rocket’ suddenly becomes etymologically opaque to future scholars.

    I can’t see any reason to think that important words would be etymologically more transparent than others. The domestication of the horse may have been recent to PIE speakers, but the species certainly wasn’t!

    Nov. 15, 11:21 am:

    Remember: MA-1 is closer to Amerindians than to any Old World population. So by definition it’s an Amerindian signature found in the Old World.

    I want to quote this in Comic Sans.

    No, they everywhere assume that Amerindians derive from East Asians. That’s precisely why, when they see that Amerindians are closer to East Asians than to Papuans, while West Eurasians are not closer to East Asians than they are to Papuans, they conclude that MA-1 admixed into Amerindians and not the other way around.

    That’s not how their analyses work!

    You simply didn’t know this paper, or utterly forgot about it and now you’re desperately trying to catch up. Sorry, I have to go with the facts.

    Projection from someone who cares more about persons than about facts.

    Nov. 15, 6:16 pm

    Germ. *wulhwaz

    What is this, some kind of internal reconstruction? Or forward projection from PIE? From Germanic data you can only get */f/, not */xʷ/.

    (And yes, that requires an explanation – personally, I suspect that assimilation to the other labial consonant in the same root is regular, but that’s not much more than a guess.)

    I reconstruct PIE *k’we- ‘wolf, dog’

    A PIE root that doesn’t end in a consonant?

    *k’wel-kwo-

    Yay, unexplained suffixes (*-l-kʷo-) out of nowhere (just like *-n-), plus */kʲ/ mobile again.

    Arm skund ‘young dog’ and Gk skulaks ‘same’ show that indeed affixes -n- and -l- were in alternation.

    Also, */s/ mobile on top of */kʲ/ mobile. Panta rhei.

    So, basically, l- in lykos and lupus corresponds to the affixal -n- is kuo:n and canis.

    Suddenly an unexplained suffix can become an unexplained prefix in PIE?

    On a methodological note, if we reconstruct PIE *k’we- ‘wolf, dog’ we can easily compare it with Uralic *kujna ‘wolf’ and Eskimo-Aleut qenRa ‘wolf’. Even the IE DOG form is presently compared with this “Nostratic” set but the fact that the IE term for ‘wolf’ can be shown to be related to the IE term for ‘dog’ makes this long-range equation quite intriguing. So my methodology can actually clear some obstacles for long-range comparison.

    This is really, really funny. First he desperately tries to explain away the *-n- of the PIE “dog” root, even at the cost of assuming a PIE root that ended in a vowel, and then he doesn’t blink when this same -n- shows up in Uralic and Eskimo-Aleut?

    Nostraticists do indeed compare these three words – -n- included. The Proto-Nostratic reconstruction, last time I looked, was *küjna, with a vowel that would nicely explain the strange PIE cluster; I have no idea if the PEA *-ʁa is a suffix (EA wasn’t considered at all in the first round of Nostratic research).

    Me, Nov. 15, 8:43 pm, way down in a long comment:

    marie-lucie, Nov. 11, 12:21 pm:

    Greenberg’s proposed Eurasiatic (which includes Nostratic as well as yet other groups)

    Not quite. Greenberg’s Eurasiatic included Nivkh and, puzzlingly*, Ainu; Nostratic excludes Ainu and at least nowadays also Nivkh, but… used to include Afro-Asiatic, though now the Moscow School has changed its terminology and excludes Afro-Asiatic while maintaining that it’s the closest relative of “Nostratic”. Clearly, they could have called that “Eurasiatic”, but didn’t because they don’t want to be seen as accepting Greenberg’s work.

    Turns out my memory was pretty bad on this. Greenberg’s E-A excluded Kartvelian and Dravidian, which are very much part of Moscow School Nostratic (and also of Bomhard’s); and G. Starostin does use the name E-A (while very much excluding Ainu).

    Finally, German Dziebel, Nov. 16, 1:20 am:

    You want to have the cake and eat it too: on the one hand, r is always rh in Greek regardless of its etymological source; on the other hand, rh comes from *sr or from *wr. I don’t care if all loans with r- turned into rh-, I care that all rh that are not in loan words come from *sr and IE wr ~ Gk hr is a false positive (in reality, s > h and w lost).

    …Perhaps there was a period when unattested early forms of some Greek dialects had both [hr]- from *sr- and *r- from *wr-, and then the aspiration/devoicing was generalized. Wouldn’t be the first time: every word that would be expected to begin with υ in Greek got a /h/ in front of it, whether etymologically expected or not. The only ὐ that was left in the entire language (not counting οὐ-) was the one in the name of the letter, “naked y” (y psilon), called “naked” precisely because it lacked the spurious [h]. Probably */hu/- was so much more common than */u/- that the latter was abandoned in favor of the former by simple analogy; perhaps the same happened with */hr/- vs. */r/-.

    The alternative is obvious: first step: */sr/- > */hr/-; second step: */hr/-, */wr/- > *r-; third step: *r- is devoiced like in Welsh, with no /h/ anywhere in sight (because it was lost in the second step).

    I tentatively note similarity between -ph- in sapha and h- in hrd.

    But the similarity between the voiceless [pʰ] and the breathy-voiced [ɦ] does have its limits.

  444. David Marjanović says:

    both [hr]- from *sr- and *r- from *wr-

    Ah, editing. Both *[hr]- from *sr- and *[r]- from *wr-.

  445. David Marjanović says:

    not counting οὐ-

    …or εὐ- or αὐ-, obviously…

  446. I enjoy all discussions about etymologies (as that is my main forte) thus I noted the one on hound / wolf.

    The trolling side-snipes I prefer to eschew as much as possible, especially as at the Ancient, and always at the *Proto level we all have to involve ‘educated guesses’!

    Before getting apoplectic about the minutae of conjectures perhaps some of your contributors may wish to review assumptions that pass for foundations. Thus if any wish to see what Proto-Indo root-words look like when purged of all the false ‘ghost laryngeals’ please consult the following 2 books, bound as one.

    ● Kaledon Naddair–”Proto-INDO-EUROPEAN √Root-Word DICTIONARY–Thematically Sorted
    & -“Proto-INDO-EUROPEAN√Root-Word DICTIONARY-Alphabetically Sorted whilst they can be bought seperately – I have bound them together as one volume, 2014 InPrint
    hence £20 (+£7 Gt.B) / £20+£14 (Eur.), / £20(+£24 P&P USA), number of pages (c.350 A4 size in total);
    for more details on this and many other Ancient Language Dictionaries consult my website
    http://keltia-publications.com

  447. Thus if any wish to see what Proto-Indo root-words look like when purged of all the false ‘ghost laryngeals’…

    Oh, dear… (shuts the door).

  448. This post is magical.

  449. Having taken the time to read all the entries in this thread I might be able to contribute a few items of worth, especially as regards Keltic materials. Thus I beg to differ with German Dzeibel in his attempted linkage of KRED-heart & radix-root. This is mainly due to having a different etymology for heart, I presume he was following Matasović in the notion that it was the ‘middle’ or ‘central organ. I differ thus

    *KRED-, (Pr.Kelt.) > CRIDE, (O.Ir.) `heart’ ie. the Pulsing, Throbbing, quivering, trembling,warm pumping thing’. [n.] this concept is surely the most striking feature of this organ, for when it stops pulsing, pumping and quivering death quickly ensues. The Anceint Kelts (& other Indo-Europeans) knew this not only from feeling humans’ throbbing chests, but also observing the gore of battle, and seeing hearts of animals as they killed and dismembered them after a successful hunt for food.
    KERDa > KERDe, KERD, KERT, KRID = heart (the quivering, trembling, throbbing, pulsing organ) (Pr.I-Eur. CRITHEACH, CRITHEANN, (Sc.G.) for the quivering-leafed Aspen tree(whose leaves shake in the slightest breeze), and the Ir. & Sc.G. Critheach for a shaking Bog (once found in some peat-bog parts of Eire, Alba, Cymru & Kernow – one could stand on the somewhat wobbly surface, but if one ‘shoogled’ too much one was in danger of sinking into the liquid mud underneath, farmers lost some cattle & sheep drowned in such spots before they did extensive drainage in the last 100 years; & criathar (O.Ir.) = a sieve that one shakes. Also one might wish to involve the quivering-stringed harp (CRUT > CRUTH O.Ir.,Pict.) whose very shape (cruth, cruith) reminded the Gaels of a heart, and was employed both in bardachd & artworks displaying that very notion, although the Bards & Fili did enjoy multi-level puns note!
    *KRIDos, *CRIDos // CRIDIon (Goid.) n.c.neut. = heart > CRIDE[ion]; CRIDE = Heart (O.Ir.)
    CROÍ, CREID, CRITH (Sc.G.) = heart (the quivering, trembling, pulsing organ) ; CRADIon > CRAEDIon > CREDIon // CRIDIon (Goid.) n.c.neut. heart (mostly figuratively);
    CRADIA // CRIDIO = Heart (O.Ir.) ; CRAIDD = Heart (Kym.)
    CRI / CRIDHE = the Heart, Heart (Sc.G.)
    *CRIDICos = Hearty or Courageous (Goid.) > CRIDHEACH = Hearty or Courageous (Sc.G.)

    NOTE : Thus I contend that other notions about (X1) ‘center / middle’, for the heart is not in the middle of the body, rather the softer organs around the navel (omphalos) are; or (X2) creed, belief, faith etc are but homonyms from originally slightly different roots, and are not a source concept related to the heart!
    (X1) CRAIDD ’center’ (M.Kumb.) ; (X1) CREIS `center’ (M.Bret.) ; (X1) CREYS `center’ (Kor.)
    (X2) CRAT = Faith (Skt.); (X2) CREDA = Creed (A-Sax); (X2) CREDO = I Believe (Lat.)
    (X2) CREID = to believe (with one’s whole Heart) (Sc.G.); (X2) CREIDSINN = Belief (Sc.G.)
    (X2) CREIDEASACH = Credible, Worthy of Belief (Sc.G.)
    (X2) CREIDHEAMH, CREIDEAS = Faith, Credence (Sc.G.)

    (to avoid being misled by puns into false etymologies one should use a dependable source for Ancient Celtic materials- by far the best is : ● J.MONARD – “ANCIENT CELTIC DICTIONARY”, 2000 & reworked as ● J. MONARD – ANCIENT CELTIC DICTIONARY – arranged as a THEMATIC LEXICON- (extensive Anc.Celtic > Français & English) Keltia Publications 2013 £35 (+ £8 P.&P. in Gt.B.) / £35 (+ £16 P.&P. to Europe), / £40 (+ £25 P.&P. to USA,etc), a huge tome – 386 A4 size pages

  450. David Marjanović says:

    all the false ‘ghost laryngeals’

    And Hittite is what, chopped liver?

  451. David Marjanović says:

    Answering another question of mine:

    Both points require more precise comments. First of all, it must be made clear that in a lot of situations it is hard to make a clear distinction between the two types of scoring. “Historically studied” is not an absolute definition: no two language groups in the world have received a completely equal amount of study, and our knowledge of the regularity of correspondences is always relative rather than absolute. Even Indo-European is prone to cases where it may be reasonable to sacrifice regularity and resort to scoring on the grounds of phonetic similarity instead.

    More in the next comment if this one gets through.

  452. David Marjanović says:

    Well, I still have no idea where in the next sentence the problem lies.

    Case in point: do we judge Old Indian hr̥d ‘heart’ as cognate to Germanic *xirt­, Slavic *sьrdь-ce, Greek κρ, etc. ← IE *ḱr̥d­, or do we score it differently, since it violates the regularity principle (the Old Indian form should reflect IE *ǵhr̥d­)? In Pokorny’s dictionary, an authoritative but by no means dictatorial source, the Indo-Iranian root is judged to represent a separate “Reimwort” [Pokorny 1958: 580], not to be related to *ḱr̥d­. Intuitively, however, it is extremely hard to think of the two variants as having nothing to do with each other — apart from complete regularity in every other respect, there is also the important issue of representativity: the two variants are in complementary distribution throughout Indo-European, and no non-conjectural evidence can be found as to their co-existence in at least one branch of the family. Hence, probably, the “compromise” solution of *ǵhr̥d- as a “rhyme word”, adopted by Pokorny — a solution that achieves nothing, since nothing is explained about the mysterious origins of this “rhyme word” (did it exist in Proto-IE? was it an original concoction on Indo-Iranian grounds? how did it originate? are its origins related to the existence of *ḱr̥d- or is it just a fortunate coincidence? etc.), but at least spares the author from the painful Neogrammarian duty of declaring the phonetic similarity between the two variants as the result of pure coincidence.

    The representativity criterion — which, in this case, merely represents a particular application of Occam’s razor — would strongly speak in favor of judging the Old Indian form as cognate with the rest of Indo-European. The exact reason that underlies the irregularity remains unknown, with several ad hoc explanations possible (idiosyncratic development of some old non-trivial cluster, perhaps with a laryngeal; assimilatory influence of two ensuing voiced segments; analogy/contamination with some other word; taboo, etc.) but none of them supported by strong independent arguments. But the assumption of a lexical replacement in this case would reduce the Neogrammarian model to absurdity, and, more importantly, leave us with a far larger number of unanswered questions (see above) than the assumption of an unexplainable irregularity.

    From about the middle of this page. A bit farther down, there’s an example of a nontrivial regular sound correspondence among clicks between North Khoisan and South Khoisan (disclaimer: I have no idea if those groupings are themselves uncontroversial).

  453. If you guys are taking requests, I’d like a pan-IE cognate family for “shoogle” next, please.

  454. Vladimir Diakoff says:

    @ David Marjanovic

    “2) The Moscow School has proposed that */(h₁)ekʲwos/ is a loan.”

    Well, this indirectly supports Dziebel’s views, for, if it’s a flawed Indo-European reconstruction, as he claims, then obviously some folks may end up deriving it from a non-Indo-European source.

    “Remember: MA-1 is closer to Amerindians than to any Old World population. So by definition it’s an Amerindian signature found in the Old World.

    I want to quote this in Comic Sans.”

    I’m not very familiar with the topic, but what exactly is so comical about it?

    @marie-lucie

    “Even though horses have big hooves, they are far from being the only animal that has them (see the name “Ungulates”), so I don’t see the rationale for naming the one after the other (in any order).”

    Indo-Europeans apparently saw this rationale, as they had no problem calling bear (in Germanic languages) using an earlier word meaning ‘brown’, although there are many more animals that are brown besides the bear.

  455. Well, “that brown guy” is obviously apotropaic. Surely there’d be no need to fear the arrival of horses, as long as they were one’s own.

  456. And not, for example, those belonging to the mistranslated Man-Afraid-Of-His-Horses, recte ‘They even fear his horses’.

    The bear is a taboo animal, so ‘brown one’ and ‘honey eater’ are suitable replacements, whereas beaver needs no such replacement (though it got one in French).

  457. David Marjanović says:

    Well, this indirectly supports Dziebel’s views, for, if it’s a flawed Indo-European reconstruction, as he claims, then obviously some folks may end up deriving it from a non-Indo-European source.

    His claim strongly implied that all PIE words must have an internal etymology.

    I’m not very familiar with the topic, but what exactly is so comical about it?

    Comic Sans is a font best used for extreme mockery. “MA-1 is closer to Amerindians than to any Old World population. So by definition it’s an Amerindian signature found in the Old World” is a really embarrassing thing to say if you consider basic logic: it means either that MA-1 had ancestry in America – or that (some of!) the ancestry of Amerindians comes from a source that was very close to MA-1!

    bear (in Germanic languages) using an earlier word meaning ‘brown’

    That’s one hypothesis. The other hypothesis is that the bear is The Beast, PIE *gʷʰēr (from which Greek θήρ “wild beast” is also derived, for example).

  458. beaver needs no such replacement (though it got one in French)

    Looks to me like French castor was taken pretty much as is from Greek κάστορας.

  459. Well, via Latin castor. I don’t know the order in which castor (Greek) and beber (supposedly Celtic) moved in on the territory of native fiber.

  460. marie-lucie says:

    The old French word for ‘beaver’ was bièvre. While the replacements for ‘bear’ (whatever the original was) were due to taboo avoidance, the French replacement with castor was probably due to the widespread use of the beaver glands for pharmaceutical purposes, while the animal itself was getting scarcer in the country because of the destruction of its habitat.

  461. That’s clearly descended from beber rather than fiber, which as far as I know was lost entirely.

  462. Vladimir Diakoff says:

    @Rodger C

    “Surely there’d be no need to fear the arrival of horses, as long as they were one’s own.”

    I think there’s a difference between a semantic pattern and cultural content. Whether the Germanic words for ‘bear’ originated from a taboo because early Germanic peoples were afraid of bears is beyond the point. It’s cultural history. Semantically, there’s a pattern whereby one member of a class appropriates the name for the whole class. The hoof-to-horse fits this semantic pattern. BTW, I read around and Russian (my native language) has a word for ‘male horse’ which is kon’. Other Slavic languages have it, too. Mallory & Adams (Oxford Introduction…) derive it from *komni and compare with other Indo-European words that mean ‘hornless” (Eng hind). So the Slavic form for horse appropriated the name for the whole class of hornless animals.

    @David Marjanovic

    “That’s one hypothesis. The other hypothesis is that the bear is The Beast, PIE *gʷʰēr (from which Greek θήρ “wild beast” is also derived, for example).”

    That’s interesting. But even if bear is from ‘beast’ there are many more beasts out there but only one of them claimed the name of the whole class.

    BTW, Vasmer’s Etymological Dictionary connects the Germanic word for ‘bear’ with the IE word for ‘beaver’ (as a reduplicated ‘brown’ stem). This would mean that Balto-Slavic (Lithuanian bebras, etc.), just like Germanic, turns up a labial outcome for a labiovelar.

    “is a really embarrassing thing to say if you consider basic logic: it means either that MA-1 had ancestry in America – or that (some of!) the ancestry of Amerindians comes from a source that was very close to MA-1!”

    I’m sorry, I must be very far away from this topic – I don’t see where Dziebel’s statement contains a logical mistake. Seems to be pretty matter-of-course.

  463. Dziebel assumed that MA-1’s genetic pattern was a North American innovation, but it could equally well be a relic that has been outcompeted in the Old World in the last 24ky.

  464. @Vladimir

    The hypothesis supported by Vasmer is that both Germanic ‘bear’ and PIE ‘beaver’ derive from the colour adjective ‘brown’. However, there is no other reason to connect bears with beavers, and the comparison is a mere root “equation” (there is or might be a *bʰ(e)r- element in all these words). What exactly was the form of the adjective, and how do you derive both animal names from it? Without answering this question, i.e. without a morphological analysis, no IE etymology is worth the ink it’s written with (or the typing effort).

  465. marie-lucie says:

    bear

    I don’t know about the merits of PIE bear/beaver, but in several Penutian languages (North American West coast) there is a definite relationship between words for bear and badger. Since these animals do not look much like each other, the naming connection might have to do with their habits of hibernating in deep burrows.

    On the other hand, words referring to smaller furbearers (no more than beaver size) often apply to different species in different languages. This may have to do with the preferred or locally most common animal sought for its pelt.

  466. Vladimir Diakoff says:

    David Marjanovic: “This is really, really funny. First he desperately tries to explain away the *-n- of the PIE “dog” root, even at the cost of assuming a PIE root that ended in a vowel, and then he doesn’t blink when this same -n- shows up in Uralic and Eskimo-Aleut?”

    Sorry, I’m still catching up…. This is all fair and true but I know that Afroasiatic has a very similar root *k(w)alp- ‘wolf, dog’ (Bomhard & Kerns, no. 319) that shows -l- where Uralic and Eskimo-Aleut have -n-. Bomhard and Dolgopolsky still believe that Afroasiatic is part of Nostratic. In Uralic, Eskimo-Aleut and Afroasiatic languages the same root means both ‘dog’ and ‘wolf’. The Indo-Europeans forms with the meaning ‘wolf’ (*wlkwko-) and ‘dog’ (*k’wo:n) are both of Proto-Indo-European age, so it would be unusual if they were unrelated. I don’t particularly like long distance comparisons, as the picture gets grainier and grainier, but taking this material at face value, Dziebel’s idea of connecting PIE *wlkwko- and *k’wo:n phonetically may mean that Indo-European preserved the traces of ancient Nostratic/Eurasiatic morphological variation, whereas Afroasiatic generalized -l- and Uralic and Eskimo-Aleut generalized -n-.

    BTW, Gamkrelidze & Ivanov reconstructed roots of the shape CV- for the early stages of Proto-ndo-European.

  467. in several Penutian languages (North American West coast) there is a definite relationship between words for bear and badger.

    In Japanese the word for bear is kuma/くま/熊 and the word for badger anaguma/あなぐま/穴熊 (ana=hole, burrow + kuma).

  468. Vladimir Diakoff says:

    @Piotr

    ” What exactly was the form of the adjective, and how do you derive both animal names from it?”

    Good question. AFAIK, there are only two reduplicated roots of PIE age: *bhebhro- ‘beaver’ (< *gwher- 'beast' or *bher- 'brown') and *kwekwlo- 'wheel' (< *kwel- 'go around'). Gamkrelidze & Ivanov (p. 225) note that in both Kartvelian and Indo-European languages the words for 'bear' (or 'beaver') and for 'wheel' are reduplicated. (The actual forms are unrelated but the morphological approach is the same.) Ironically (considering that Marie-Lucie is an Amerindianist) they also cite a reduplicated Squamish (Salishan) form for 'bear' as a typological parallel.

    I understand that Nostratic, Glottalic, etc. are not most Western linguists' cup of tea, but we Russians love our linguists…

  469. Vladimir Diakoff says:

    John Cowan: “Dziebel assumed that MA-1′s genetic pattern was a North American innovation, but it could equally well be a relic that has been outcompeted in the Old World in the last 24ky.”

    Sounds like an equally possible but different interpretation, not a logical mistake.

  470. marie-lucie says:

    Dziebel has written a book arguing that humanity originated in America, not in Africa.

  471. Vladimir Diakoff says:

    Marie-Lucie: “Dziebel has written a book arguing that humanity originated in America, not in Africa.”

    Yes, I know. I even bought it for 140 US dollars. It’s mostly on kinship systems, which is a topic that is hard to follow. There’re only a few specialists on kinship left (at least in Russia) and Dziebel came out of one of the academic traditions of kinship studies in Russia. One of his arguments is that there are many more language families in America and Papua New Guinee than in Africa, so diversity means greater age, etc. As an Amerinidianist, what are your thoughts on that?

  472. Good question. AFAIK, there are only two reduplicated roots of PIE age: *bhebhro- ‘beaver’ (< *gwher- 'beast' or *bher- 'brown') and *kwekwlo- 'wheel' (< *kwel- 'go around').

    There were probably quite a few of them. “Wheel’ and ‘beaver’ (and possibly ‘meat’) are simply the most widespread ones. Paul S. Cohen has been working on them for several years. He has collected a number of plausible candidates and has made substantial progress towards identifying the morphological processes that produced them. I have discussed some of those etymologies with him and even suggested one or two, but since Paul has not published it all yet, I’d rather not divulge any details.

    Gamkrelidze & Ivanov (p. 225) note that in both Kartvelian and Indo-European languages the words for ‘bear’ (or ‘beaver’) and for ‘wheel’ are reduplicated. (The actual forms are unrelated but the morphological approach is the same.) Ironically (considering that Marie-Lucie is an Amerindianist) they also cite a reduplicated Squamish (Salishan) form for ‘bear’ as a typological parallel.

    (1) The IE word for ‘bear’ is not reduplicated (nor is the Germanic one). (2) Klimov’s (1998) reconstruction of the Kartvelian ‘bear’ word is *datw-, which doesn’t look reduplicated to me. According to Klimov, Svan -št- is a reflex of *t in this position (resulting from some kind of preaspiration), not evidence of reduplication. (3) Georgian borbal is isolated; *br- ‘whirl, twirl’ is Proto-Georgian/Zan, not Proto-Kartvelian, and the connection between the two is uncertain.

    I understand that Nostratic, Glottalic, etc. are not most Western linguists’ cup of tea, but we Russians love our linguists…

    I wonder if I count as “Western”, being Polish and living in Poland. It wouldn’t occur to me to classify historical linguists into those I love and those I do not love depending on where they come from.

  473. marie-lucie says:

    Vladimir: they also cite a reduplicated Squamish (Salishan) form for ‘bear’ as a typological parallel

    Since reduplication occurs in widely separated families, I don’t think that this parallel really counts as particularly relevant. Salishan make use of some quite complex forms of reduplication.

    [Dziebel’s] arguments is that there are many more language families in America and Papua New Guinee than in Africa, so diversity means greater age, etc.

    I understand that Papua New Guinea tops the rest of the world in terms of language diversity, most likely related not only to the great age of its settlement but also to the fact that its geography makes internal communication quite difficult.

    About the Americas, as I have written before I think that the number of “families” is exaggerated by the fact that many of the acknowledged small families and isolates probably belong together, in the way that Romance, Germanic, Slavic etc (which are internally quite recognizable even by untrained persons) are not isolated but belong together as members of the larger Indo-European group. I don’t think that “Amerind” will prove valid, but there are probably several groups on the order of IE.

  474. It wouldn’t occur to me to classify historical linguists into those I love and those I do not love depending on where they come from.

    Hear, hear! Nationalism has no place in science (or rational thinking in general).

  475. Vladimir Diakoff says:

    @Piotr

    “I wonder if I count as “Western”, being Polish and living in Poland. It wouldn’t occur to me to classify historical linguists into those I love and those I do not love depending on where they come from.”

    Russian linguistics has always been somewhat isolated from the West due to world politics, so it spawned some regional peculiarities of different degree of extremism: first, Marrism, then Nostratics, then Glottalics. I know that Russian Indo-Europeanists and Nostraticists tend to reject Laryngeal Theory (apart from what’s actually attested in Anatolian) but Western linguists mostly support it. (Szemerenyi and his followers in Finland and elsewhere also reject it and Szemerenyi was Hungarian.) At the same time, Russian historical linguistics are largely supportive of the Nostratic grouping, while Western linguists are very critical of it. I don’t want to push the geopolitics of science too far but there may be some rationale behind it.

    I suppose Poles are lucky because they have options to choose from. 🙂

  476. One of his arguments is that there are many more language families in America and Papua New Guinee than in Africa, so diversity means greater age, etc.

    This is an artefact of classification. If you follow Greenberg, there are four families in Africa and three in the New World, so Afica wins. I believe there are more splitters in American Indian studies than among Africanists, which doesn’t necessarily mean there are more New World language families.

    If humans evolved in America, what did they evolve from, being the only catarrhines in those parts? Bigfoot?

  477. Vladimir Diakoff says:

    Marie-Lucie: “About the Americas, as I have written before I think that the number of “families” is exaggerated by the fact that many of the acknowledged small families and isolates probably belong together.”

    Thank you for your insight. But I wonder if this is balanced off by the number of extinct small families and isolates that were exterminated during the first 300 years of the European conquest of the New World (before linguistic data began to be collected). Africa did not experience this level of language extinction in the past 500 years.

    @ Piotr

    “If humans evolved in America, what did they evolve from, being the only catarrhines in those parts?”

    What I gathered from Dziebel’s website is that he believes that modern humans originated from an East Eurasian archaic hominin (so, yes, a catarrhine), migrated to the New World, speciated into modern humans (I guess isolation is good for speciation) and then migrated back to replace all the archaics in the Old World. That’s why it’s the New World and Papua New Guinee that have more language families than elsewhere. Because they were first to be peopled by modern humans armed with language and culture.

    I also read on Dziebel’s website that there’s a ethnomusicologist Jordania who argued that the reason there are more language families in the New World vs. Africa (but Africa has more diverse musical traditions) is because humans transitioned from a “musical” language to a real language earlier in Asia and later in Africa. Jordania is a multiregionalist, though.

  478. Africa did not experience this level of language extinction in the past 500 years.

    European colonisation has not been the only cause of language death. Agricultural revolutions, in particular, can be quite destructive. Certainly lots of tiny language families in sub-Saharan Africa went extinct as a result of the Bantu expansion, and the same happened in North Africa as the Afro-Asiatic languages swept across it (the Sahara was still green and well-populated).

  479. Vladimir Diakoff says:

    @Piotr

    “Agricultural revolutions, in particular, can be quite destructive.”

    Certainly, and the New World had a fair share of their own agricultural expansions (Uto-Aztecan, Oto-Mangan, etc.). What’s noteworthy is that the New World also knew recent large-scale forager expansions such as the Athabascan spread. Considering that there are more language isolates in the New World than elsewhere, they could have easily been an easy prey for more successful foragers. So chances of pre-Columbian language extinction seem to be comparable to those in Africa, if not higher.

  480. Considering that there are more language isolates in the New World than elsewhere

    How do you know?

  481. Vladimir Diakoff says:
  482. diversity means greater age

    That is only a heuristic, anyway. Etienne pointed out that Corsica is the most diverse portion of the Romance-speaking area, yet we know that the Romance languages did not originate on Corsica.

  483. Linguistic diversity is hard to measure, since “language families” are not a natural level of grouping. Instead, they reflect the current limit of our reconstructions. In some parts of the world families tend to be bigger just because the languages in question are better studied and have some historical documentation. In other places we have “documentation” in the form of short word lists, family classification is to some extent a matter of faith, and nobody can tell with much confidence how many families (including isolates) could be distinguished if we were to go about classifying them with any rigour. So many currently recognised groupings should be taken with a grain of salt. Africa has more languages than any other continent (and I suppose some twenty-odd bona fide families), but there seems to be more fragmentation into small families and isolates in the Americas. But Atlantic-Congo (including Bantu but probably excluding some outliers) and Afro-Asiatic (there are doubts about “Omotic” and some languages tentatively grouped with Cushitic) are huge families, which have spread over most of the continent, leaving only residual areas occupied by a large number of small families (I don’t think the genetic unity of “Nilo-Saharan” and “Khoisan” is defensible). Nobody knows how old the family divisions are, either in absolute terms or in comparison to the New World.

    Anyway, where are the bones of our American ancestors? May I see some archaic humans or pre-humans from America, please? There were anatomically modern humans in the Indo-Pacific region well before 40 kya, still earlier in the Middle East, and much earlier in Africa, so the American ancestors should be still older. DNA markers don’t suggest a New World human homeland unless you focus on carefully selected haplotypes and ignore all other evidence. Dziebel sees himself as a latterday Florentino Ameghino, but Ameghino died more than 100 years ago and speculation that could be entertained during his lifetime is ruled out by what we have learned since.

  484. Vladimir Diakoff says:

    @Piotr

    “Linguistic diversity is hard to measure, since “language families” are not a natural level of grouping.”

    I wholeheartedly agree. I think only with Nichols’s Linguistic Diversity in Space and Time linguists have started to seriously apply global and regional linguistic patterns to understanding population prehistory. A lot still needs to be figured out. I found the following discussion interesting:
    https://www.academia.edu/13246715/Wichmann_S%C3%B8ren_Andr%C3%A9_M%C3%BCller_and_Viveka_Velupillai._2010._Homelands_of_the_world_s_language_families_A_quantitative_approach._Diachronica_27.2_247-276

    This said, at least one well studied case seems to support the argument that greater diversity means greater age. It’s the dispersal of Austronesians from Taiwan. The Formosan branch is by far the most diverse of all and this seems to be in good agreement with both archaeology and genetics.

    “Dziebel sees himself as a latterday Florentino Ameghino,”

    He does mention Ameghino on his website, but I think he believes as strongly as you that Ameghino has been disproven.

    “Anyway, where are the bones of out American ancestors? May I see some archaic humans or pre-humans from America, please?”

    I don’t know what his answer to this would be. At one point, he made a comparison between Louis Leakey’s approval of some super-old sites in the America as an example of how America is both understudied linguistically and archaeologically but to a different effect: more comparative work will necessarily reduce the number of linguistic families in the Americas, while more archaeological work will generate proof of greater human antiquity in the New World. Also, Russians are very proud of Denisovan man (and Derevianko is almost a national hero there) and what’s left of that hominin? Just a couple of teeth and fingers. Who knows how many of those are hidden in caves in America.

  485. David Marjanović says:

    This is all fair and true but I know that Afroasiatic has a very similar root *k(w)alp- ‘wolf, dog’ (Bomhard & Kerns, no. 319)

    That’s a pretty long root.

    AA historical linguistics hasn’t advanced very far, because hundreds of AA languages are underresearched. What evidence did Bomhard & Kerns use to arrive at *k(w)alp-, and why aren’t they sure about the *w? I don’t have their book.

    that shows -l- where Uralic and Eskimo-Aleut have -n-. Bomhard and Dolgopolsky still believe that Afroasiatic is part of Nostratic. In Uralic, Eskimo-Aleut and Afroasiatic languages the same root means both ‘dog’ and ‘wolf’. The Indo-Europeans forms with the meaning ‘wolf’ (*wlkwko-) and ‘dog’ (*k’wo:n) are both of Proto-Indo-European age, so it would be unusual if they were unrelated. I don’t particularly like long distance comparisons, as the picture gets grainier and grainier, but taking this material at face value, Dziebel’s idea of connecting PIE *wlkwko- and *k’wo:n phonetically may mean that Indo-European preserved the traces of ancient Nostratic/Eurasiatic morphological variation, whereas Afroasiatic generalized -l- and Uralic and Eskimo-Aleut generalized -n-.

    All this is possible. But without a reconstruction of Proto-Nostratic morphology that would explain what this *l/n suffix is, it’s just a stab in the dark, I’m afraid.

    BTW, Gamkrelidze & Ivanov reconstructed roots of the shape CV- for the early stages of Proto-ndo-European.

    That’s internal reconstruction, which has the disadvantage of being unconstrained…

    I understand that Nostratic, Glottalic, etc. are not most Western linguists’ cup of tea

    The “Glottalic theory” isn’t Eastern or Western in origin. It also used to have a pretty large following in the West.

    The Moscow School Nostraticists do not accept it – but Allan Bomhard does!

    Sounds like an equally possible but different interpretation, not a logical mistake.

    There are two (a priori) equally possible interpretations. German Dziebel pretended there was only one; he even added “by definition”. That is a logical mistake.

    I understand that Papua New Guinea tops the rest of the world in terms of language diversity, most likely related not only to the great age of its settlement but also to the fact that its geography makes internal communication quite difficult.

    It also makes research quite difficult; not as much has been done as in most other places. Consequently, it is easily possible that large language families still await discovery there.

    because humans transitioned from a “musical” language to a real language earlier in Asia and later in Africa.

    …Right.

    Jordania is a multiregionalist, though.

    So, a crackpot.

    I don’t think the genetic unity of “Nilo-Saharan” and “Khoisan” is defensible

    Khoisan, including Sandawe but not necessarily Hadza/Hatsa, looks fairly promising. So do parts of Nilo-Saharan, but probably not the whole thing.

  486. David Marjanović says:

    A long comment with three links is in moderation; it will appear above this one.

    This said, at least one well studied case seems to support the argument that greater diversity means greater age. It’s the dispersal of Austronesians from Taiwan. The Formosan branch is by far the most diverse of all and this seems to be in good agreement with both archaeology and genetics.

    There are linguists now who wonder if PAN was instead spoken in eastern Borneo… and the Formosan languages are apparently all woefully underresearched.

    Who knows how many of those are hidden in caves in America.

    Native Americans don’t carry a greater proportion of Denisovan DNA than other people, though.

  487. The orthodox view is that Taiwan is an AN dispersion zone, but it’s also possible that it’s a convergence zone, in which case the Formosan languages descend from random Austronesian languages that have strongly affected each other. I linked a paper by Blench to this effect.

    As for the archaeological evidence, Taiwan was certainly settled long before the time-depth of AN.

  488. The Indo-Europeans forms with the meaning ‘wolf’ (*wlkwko-) and ‘dog’ (*k’wo:n) are both of Proto-Indo-European age, so it would be unusual if they were unrelated.

    Why would it be “unusual”? It’s perfectly normal for words to be unrelated. And assuming two words are related because they have a vague partial similarity is something historical linguistics quite rightly got away from well over a century ago.

  489. @David Marjanović: The Denisovan issue confuses me. The commonly stated position, in line with the paper by Reich et al., is that Denisovan DNA is largely absent outside of Australia, Melanesia and eastern Indonesia – but then there are things like the Genographic Project, which has told legions of people of West Eurasian ancestry that they have significant shares of it. For example, my mother’s results were 3.1% Denisovan and only 2.0% Neanderthal, even though her ancestry is entirely European and southwest Asian.

  490. This said, at least one well studied case seems to support the argument that greater diversity means greater age. It’s the dispersal of Austronesians from Taiwan. The Formosan branch is by far the most diverse of all and this seems to be in good agreement with both archaeology and genetics.

    But in this case we are talking of a securely established family. We know that the Formosan branches are related to each other and to the rest of AN, and we have a fairly good idea how they are related (we know the structure of the family tree pretty well). If we didn’t know of the “basal” status of the Formosan languages, there would be no way of knowing that they are “more diverse” than other AN-speaking areas. This can’t be compared to counting known families and isolates in Africa and the Americas.

    Will you say that, since South America is more “diverse” than North America in terms of family count, it must be the cradle of Homo sapiens? It’s an extraordinary claim that requires extraordinary evidence. The burden of proof is on the shoulders of the proponents.

  491. David Marjanović says:

    the Genographic Project, which has told legions of people of West Eurasian ancestry that they have significant shares of it

    That’s indeed puzzling, but if nothing has come of it since January 2013 – I’m not aware of any published papers on this subject; bizarrely, there’s no trace of a comment function on the blog –, then probably there was something wrong with the algorithm.

  492. David Marjanović says:

    LH thread about Roger Blench’s hypothesis that Proto-Austronesian wasn’t spoken on Taiwan after all.

  493. Vladimir Diakoff says:

    @David Marjanovic

    “Native Americans don’t carry a greater proportion of Denisovan DNA than other people, though.”

    Actually, Dziebel claims (citing Qin and Stoneking 2015) that Amerindians have more Denisovan than East Asians, West Eurasians and Africans (http://anthropogenesis.kinshipstudies.org/2015/07/amerindians-are-even-more-genetically-diverse-and-older-than-we-thought/). I assume they have less Denisovan than Papuans and Australians but still more than everyone else. It’s interesting that archaic admixture is higher in the areas where linguistic diversity is the highest, too.

  494. Vladimir Diakoff says:

    @ Piotr

    “Will you say that, since South America is more “diverse” than North America in terms of family count, it must be the cradle of Homo sapiens?”

    I’d say that the expectation from a theory like that is that North America should be more diverse but then I don’t know what’s the geographical unit (continent or subcontinent) for which the diversity-age principle works and how to factor in subsequent replacements.

  495. Vladimir Diakoff says:

    @David Marjanovic

    “What evidence did Bomhard & Kerns use to arrive at *k(w)alp-, and why aren’t they sure about the *w?”

    I checked again and it was my mistake: they are affirmative about the labiovelar in Proto-Afroasiatic (*kw(h)alp). But the evidence they cite is restricted to Semitic (from Hebrew to Tigrinya), hence for Semitic they write *k(h)alp-.

    Russian Nostraticists cite a bunch of forms, too, with more variable onset and affixation: http://starling.rinet.ru/cgi-bin/etymology.cgi?single=1&basename=%2Fdata%2Fsemham%2Fafaset&text_number=2599&root=config.

  496. Vladimir Diakoff says:

    @languagehat

    “Why would it be “unusual”? It’s perfectly normal for words to be unrelated.”

    I know but if we take Uralic, Eskimo-Aleut in the north and then Afroasiatic in the south and they all link dog and wolf under one form, then Indo-European that’s geographically in-between looks “unusual.”

  497. Vladimir Diakoff says:

    @David Marjanovic

    “So, a crackpot.”

    I don’t carry this word in my personal vocabulary. Jordania is a solid specialist in ethnomusicology, which is an arcane subject. He just began thinking about human origins before out of Africa emerged as an alternative to Multiregionalism.

  498. I think we should also mention here a theory proposed by certain Russian “paleoantropologists” that monkeys of South America are actually degenerate forms of ancient humans.

    “Degeneration theory” doesn’t limit itself to New World monkeys, but also claims that all apes and most monkeys of the Old World are also descendants of ancient humans.

    And of course, currently prevailing mainstream theory of human origin is wrong – humans already existed 10 millions years ago and since then never stopped degenerating…

  499. Vladimir Diakoff says:

    @David Marjanovic

    “LH thread about Roger Blench’s hypothesis that Proto-Austronesian wasn’t spoken on Taiwan after all.”

    Thanks for the link. I like Blench’s unorthodox views. I’m afraid though that he’s going to be as unsuccessful with upsetting the out-of-Taiwan orthodoxy as Dziebel is going to be with upsetting the Clovis-First model.

  500. Oops, I posted this in the wrong thread. Let me try again:

    Re: Denisovans

    I’m a little worried by the fact that so much is being made of just three Denisovan genomes (two of them only fragmentary).

  501. Vladimir Diakoff says:

    @ Piotr

    “I’m a little worried by the fact that so much is being made of just three Denisovan genomes (two of them only fragmentary).”

    It’s always prudent to be skeptical about the shortage of reliable data. One thing that goes for Denisova DNA is that it seems to over and over again cluster with Sima de los Huesos, which is a good cross-check because it has more solid cranial material to go with the DNA and it’s geographically remote from South Siberia.

  502. marie-lucie says:

    Vladimir: the number of extinct small families and isolates that were exterminated during the first 300 years of the European conquest of the New World (before linguistic data began to be collected)

    Actually, many people did collect linguistic data, starting from the early days of the conquest. Traders and missionaries, among others, needed to learn local languages. Chapter 2 of Campbell’s American Indian Languages: The historical linguistics of Native America contains a lot of information on the data collectors and early linguists. Small families and isolates, and their speakers, tend to persist in regions which are difficult to get to. On the other hand, the Algonkian expansion in the central plains after the adoption of the horse may have wiped out some languages (not necesssarily through massacres but also by language shift on the part of non-Algonkian speakers).

    the New World had a fair share of their own agricultural expansions (Uto-Aztecan, Oto-Mangan, etc.). What’s noteworthy is that the New World also knew recent large-scale forager expansions such as the Athabascan spread.

    Some of the U-Az spread may be related to agriculture, but the largest territory of U-Az speakers is in the Great Basin, a desertified area where “forager expansion” is more likely to have happened.

  503. I know that Afroasiatic has a very similar root *k(w)alp- ‘wolf, dog’ (Bomhard & Kerns, no. 319)

    Looks similar to Arabic Kalb and Hebrew כלב k-l-v. Wolf in Hebrew is זאב ze’ev.

  504. David Marjanović says:

    Actually, Dziebel claims (citing Qin and Stoneking 2015) that Amerindians have more Denisovan than East Asians, West Eurasians and Africans

    Oh, sorry, I forgot about this. Yes, the evidence for an Australian-like migration to America is interesting.

    I don’t carry this word in my personal vocabulary.

    I smiled when I read that, and then I giggled out loud. 🙂 There really is such a thing as a crackpot. Denying this may sound noble and all, but it’s… denial. 🙂

    Jordania is a solid specialist in ethnomusicology, which is an arcane subject.

    So? I’m not talking about his workd in ethnomusicology.

    He just began thinking about human origins before out of Africa emerged as an alternative to Multiregionalism.

    Multiregionalism has been completely out of the question for at least 20 years. Before that, it couldn’t perhaps be completely ruled out by what little was known of the evidence, but it never made sense: why would populations all over the world evolve in the exact same direction? What selection pressure could cause that?

    I like Blench’s unorthodox views. I’m afraid though that he’s going to be as unsuccessful with upsetting the out-of-Taiwan orthodoxy as Dziebel is going to be with upsetting the Clovis-First model.

    …Clovis First is long gone. Several sites that are older than Clovis have withstood all scrutiny, Dziebel or no Dziebel.

  505. Vladimir Diakoff says:

    @David Marjanonic

    “There really is such a thing as a crackpot.”

    Not at that level, not at the level of Jordania. There are plenty of crazy people in the web world, but as you go up (or down) an academic path crackpots are unlikely. All scholars look at humans origins out of their respective disciplines and they may see something other disciplines don’t. In the recent years, out of Africa got falsified twice: first, when it was showed that there was archaic admixture outside of Africa (a step back toward multiregionalism), second when it turned out that there was no serial founder effect as population colonized the globe. Should we now call all the people who developed the out of Africa theory crackpots?

  506. I can’t agree with you there. Plenty of academics are crackpots when out of their subjects: creationist physicists and chemists, e.g.

  507. David Marjanović says:

    There are plenty of crazy people in the web world, but as you go up (or down) an academic path crackpots are unlikely.

    Unlikely, yes. But the extreme case is Linus Pauling, who got the Nobel Prize in chemistry and the one in physics – and then went on to invent vitamin C woo, the idea that eating huge amounts of vitamin C is somehow good for you. (In fact, as one person found out the hard way, it kills you if your kidneys can’t get rid of it quickly enough.) Outside his original fields, Linus Pauling was a crackpot, even though he justly deserved both of his Nobel Prizes.

    In the recent years, out of Africa got falsified twice: first, when it was showed that there was archaic admixture outside of Africa (a step back toward multiregionalism),

    …OK, but that’s a very small step. The combined Neandertal and Denisovan contributions to extant human genomes hardly ever reach 10%; that’s very much unlike what multiregionalists thought.

    second when it turned out that there was no serial founder effect as population colonized the globe.

    That depends on how big the founding populations were! And to some extent, a founder effect has been discovered: open-access paper.

  508. Vladimir Diakoff says:

    David Marjanovic: “the extreme case is Linus Pauling.”

    Chemistry and Peace, not Chemistry and Physics. Yes, it is an interesting case of a genius who lost control of his own mind. “Crackpot” is too simplistic though to describe Pauling’s phenomenon. There must have been some kind of molecular change that caused it. 🙂

    BTW, if not “musical language,” what heuristic regarding the origin of language do you maintain?

  509. …got the Nobel Prize in chemistry and the one in physics

    That was Mme Curie (but she remained sane).

  510. Vladimir Diakoff says:

    Marie-Lucie: “Some of the U-Az spread may be related to agriculture, but the largest territory of U-Az speakers is in the Great Basin, a desertified area where “forager expansion” is more likely to have happened.”

    I read somewhere that Uto-Aztecans actually abandoned agriculture in the course of expanding into the Great Basin. If so, this is pretty amazing and… funny: an agricultural expansion gone backwards.

    Paul Ogden: “Looks similar to Arabic Kalb and Hebrew כלב k-l-v.”

    Yes, they do list those forms. In Jibbali, Harsusi, Mehri, etc. the respective cognates mean both dog and wolf, which Illich-Svitych (Opyt sravnenija…, no. 238), citing Thomsen, called “archaic semantics.”

  511. this is pretty common thing actually

    Eurasian pastoral nomadism actually arose among settled agriculturalists who abandoned farming and expanded deep into great steppe.

    Even European settlers in Americas or South Africa often did the same – all these trekboers, gauchos, vaqueros, cowboys, etc.

  512. Vladimir Diakoff says:

    SFReader: “this is pretty common thing actually.”

    Your examples include pastoralists, which is indeed common. A shift from agriculture (back) to hunting and gathering is not. As far as I know…

  513. Many early Euro-Appalachian settlers practiced agriculture only to supplement hunting and foraging. Crevecoeur, long ago, explained the economic rationality of this shift.

  514. The Moriori come to mind: the Chatham Islands are too cold and bleak to support agriculture, though they were settled by Māori agriculturalists.

  515. Vladimir Diakoff says:

    @John Cowan

    “Plenty of academics are crackpots when out of their subjects: creationist physicists and chemists, e.g.”

    It’s an interesting subject for a philosophy of science study and I don’t know all the cases of course. But it’s possible that quackery is more of an absolute category in the exact and natural sciences and a relative category in the historical and social sciences because disciplinary boundaries and methodologies are generally fuzzier in the latter. Cultural biases (again, more likely to occur in the historical and social sciences) can also create an illusion of an evidence-based consensus, on the one hand, as well as false crackpots (if there are pseudo-scientists there could be pseudo-crackpots), on the other. Creationism is an example of a religious subculture bias that cuts across natural and social sciences but there could be others, less pronounced, often mainstream.

  516. The poster baby for crackpots who were proved right in the end is, of course, Alfred Wegener who campaigned for the embryonic theory of plate tectonics 50 years before evidence was available to finally prove him right.

    With a degree in astronomy and a career in meteorology he was indeed an outsider in geophysics, but he wasn’t as alone with his views as the ‘crackpot’ label implies — he built on earlier works, and a minority of researchers inside the specialty shared his opinions already then.

    But when the ‘CONTINENTS MOVE!’ headlines hit the Sunday supplements (early 70’s, I guess, since I vaguely remember it), Wegener did get the vindicated crackpot writeup.

  517. David Marjanović says:

    what heuristic regarding the origin of language do you maintain?

    That’s a hard question! I haven’t thought about it much. Judging from what other primates do, and what they don’t do in the wild but are capable of, I imagine a very gradual origin. There are multispecies monkey communities in West Africa where each species has a set of different alarm calls for different types of danger…

    Music is certainly old, and draws on some of the same capabilities as language, but I’m not aware of a reason to think it was directly involved in the origin of language.

    early 70′s, I guess

    Yes! It was a short revolution – 1968 to 1973 or so; then it was all over, and no “fixists” were left.

    However, Wegener wasn’t vindicated across the board. He had thought that his equivalent of continental crust lies on top of his equivalent of oceanic crust, and that the latter is a single solid piece all around the Earth. In part for this reason, the only idea he had to offer for what moves the continents was inertia from the Earth’s rotation, which turned out to be completely wrong.

  518. Vladimir Diakoff says:

    @ David Marjanovic

    “but I’m not aware of a reason to think it was directly involved in the origin of language.”

    It’s all speculative but with a good pedigree (Darwin was the first one to propose). Scholars like Fitch (“The Biology and Evolution of Music: A Comparative Perspective.”) observe that musical vocalizing is found outside of human species, while language is unique to humans. If they are related and music preceded language, then it’s likely that language evolved from music. Then there are at least two areas in which music and specific language forms seem to intersect (phonetically it’s tones and syntactically it’s polysynthetic languages). Apart from music, I don’t know of any other cultural system that could potentially provide clues to language evolution and I don’t expect archaeology or genetics to cast any significant light on this any time soon.

  519. Also (if it hasn’t been exploded yet), the association between a tonal L1 and absolute pitch.

  520. I would say music is almost pure syntax without (compositional) semantics. Human communication is multimodal. We have articulated speech (and of course its written representation), gestures (body language), and non-linguistic or paralinguistic vocalisation. The first is unique to humans, as far as we know, the other two are shared with many other animals. To me, music and language are parallel, co-evolving types of code. They both require complex syntactic structures, but there’s little reason to believe that language is historically derived from music.

  521. Vladimir Diakoff says:

    Piotr: “I would say music is almost pure syntax without (compositional) semantics.”

    Yes, that’s exactly how the supporters of a “musical protolanguage” express the similarity between language and music.

    “gestures (body language)”

    The gestural origin of language is a competitor of the musical protolanguage theory.

    “there’s little reason to believe that language is historically derived from music.”

    Skepticism is indeed a tempting position in this issue because it’s hard to offer anything but speculations.

  522. This is a fascinating subject which I know nothing about, but the musical protolanguage idea would be a lot more convincing if other primates had anything similar to human music, which (AFAIK) they don’t. There’s also the fact that every neurally normal human is a competent language user, but very many people literally couldn’t carry a tune to save their lives, which seems a bit problematic if you think musical ability was subject to strong selection pressure.

  523. To me, the musical protolanguage idea is one of the many, many things that fall into the category of “interesting but unresolvable,” and I try not to waste too much mental effort on such things. (See also “What if we’re all just brains in a vat, part of some huge AI experiment?”)

  524. ə de vivre says:

    Appeals to ‘musical’ proto-languages always seem really cagey about what they mean by ‘musical’. Primates make use of things like pitch modulation and rhythm in their calls, and many (most? all?) human languages use f0 changes for some kind of linguistic purpose, whether it’s at the syllabic/moraic level in tone languages or a more global phenomenon like sentence polarity in English, but is that music? It seems plausible, but as you say unprovable, that before the current system of contrasts in place/manner of articulation, f0, and formant frequency, only f0 movement had the kind of discrete contrastiveness that appears to be unique to human speech, but it seems fundamentally misleading to call that ‘musical’. Or are we really suggesting that at some stage of hominid evolution a rising perfect 5th had discrete, contrastive meaning compared to a rising 4th? Or that some pre-linguistic groups spoke in makam while others spoke in equal-tempered scales in the same way that languages today can differ in dividing stops between voiced-unvoiced or aspirated-unaspirated? Everything that makes music music seems like a formalized elaboration of more rudimentary capabilities found in naturally acquired speech rather than vice-versa.

  525. David Marjanović says:

    if other primates had anything similar to human music, which (AFAIK) they don’t

    Gibbons spend a lot of time singing together to reinforce their pair bonds. Other than that, there are the howling monkeys, and that’s pretty much it.

  526. Vladimir Diakoff says:

    “if other primates had anything similar to human music, which (AFAIK) they don’t”

    Highly synchronized, antiphonic duetting (the closest parallel to Jordania’s polyphony, which he believes is ancestral in humans) sporadically occurs in lemurs as well as in monogamous, pairbonded, alloparenting primates (especially New World monkeys such as titi, but not in all platyrrhine species). Duetting is not known among any other mammals, which is interesting, but it’s more complexly developed among birds than among primates. Big apes don’t have duetting, which fits with the general observation that apes are socially very unlike humans. They are not paibonded, monogamous or alloparenting.

    What’s also remarkable is that the same paibonded, monogamous, alloparenting, arboreal creatures developed rather complex language-like communications because they couldn’t rely on either smell or sight to communicate with each other. Too mobile and active to recognize each other by smell, too covered with leaves to be able to see each other.

    https://www.academia.edu/10923561/Duetting_in_the_Titi_Monkey_Callicebus_cupreus_Structure_Pair_Specifity_and_Development_of_Duets

  527. David Marjanović says:

    the general observation that apes are socially very unlike humans. They are not paibonded, monogamous

    Eh… humans aren’t that pairbonded or monogamous either, on average. We seem to span the whole range from gibbons to bonobos and back.

  528. Vladimir Diakoff says:

    @David Marjanovic.

    ” humans aren’t that pairbonded or monogamous either, on average. We seem to span the whole range from gibbons to bonobos and back.”

    True. But we don’t know if polygyny was ancestral to humans or re-invented later. Typically, polygyny is most frequent in large, agricultural and pastoral societies, which emerged very late in human history. Small bands are more monogamous, but also more prescriptive in the choice of the partner. Also, low sexual dimorphism observed from the earliest Homo on is more compatible with monogamy, as a socially more egalitarian mating pattern, than polygamy.

  529. David Marjanović says:

    Oh, monogamy and polygyny aren’t the only two choices. Have a look at chimpanzees and bonobos. Even gorillas aren’t as classically polygynous as people used to think.

  530. Vladimir Diakoff says:

    @David

    Yes, that’s a good point. It’s possible that the big apes retained more of the ancestral (assuming that New World primate attestations are phylogenetically significant) primate paibonded, monogamous and alloparenting behavior than we think, but hominins and humans have retained them in a stronger form. This would fit with growing body of evidence that apes, although the closest to us genetically, are not necessarily the best proxies for our MRCA.

  531. David Marjanović says:

    Follow-up on Khoisan looking promising, including Sandawe and even Hadza, with much discussion of problems (mostly: first, lack of data; second, lack of adequately transcribed data) and some speculation that is clearly marked as such. Long.

  532. David Marjanović says:

    Whoa. Even more recent paper by the same author finding a pretty large number of parallels between Hadza, Omotic and Cushitic in the basic lexicon and wondering if Hadza is basically an Afro-Asiatic language with a Khoisan substrate. Shorter than the previous one, but in Russian, so I’ve only read it rather superficially.

  533. Vladimir Diakoff says:

    Thanks David. I didn’t see them. Now, I read them both and you pretty much got all the gist out of them. Pretty intriguing! Hadza’s population structure is very different from that of San. They are far less heterozygous than most Sub-Saharan Africans, especially San, and approximate North African heterozygosities.

  534. David Marjanović says:

    That tells us little about their language. The Basques hardly differ genetically from their neighbors, having just the amount of Yamnaya admixture that you’d expect for their location; and the people with the largest percentage of Yamnaya ancestry today are the Estonians, whose language isn’t Indo-European but Uralic.

  535. Vladimir Diakoff says:

    @David. Not directly of course, but it does add to the overall picture of the population/language/society, and who knows maybe subsequent research will show that those variables are more in sync than we currently can see. Lower heterozygosities mean that, while a language shift is still possible, it’s less likely to have happened as genetic admixture between an Afroasiatic- and a Khoisan-population clearly didn’t happen because an admixed population would have been more heterozygous.

    One thing that Starostin doesn’t consider is that Afroasiatic and Khoisan may form a megaphylum, with Hadza as its linchpin.

  536. David Marjanović says:

    Well, Afroasiatic already forms a megaphylum with Eurasiatic… it would be a bit strange if Khoisan were the closest relative of Nostratic.

  537. Vladimir Diakoff says:

    I didn’t even think about it! But who knows…

  538. marie-lucie says:

    The Basques hardly differ genetically from their neighbors, having just the amount of Yamnaya admixture that you’d expect for their location;

    Does this admixture have to do with the higher proportion of Rh-negative blood among the Basques? (my mother, of Languedoc ancestry, not that far from earlier Basque territory, had Rh-negative blood).

  539. David Marjanović says:

    No idea. Rh-negative is indeed very common in the Basque country and in the Caucasus, thought to be an effect of isolation.

    I should have mentioned that the Basques have only slightly more Early European Farmer ancestry than their northern neighbors, and even slightly less than their southern neighbors; in spite of this, there are reasons to think that their language (and a few odd words in Sardinian) are all that’s left linguistically of the EEF people.

  540. Trond Engen says:

    thought to be an effect of isolation

    Caucasus isolated? It’s a showroom displaying samples from just about every migration that ever swept across the Eurasian continent.

  541. Trond Engen says:

    A mosaic made out of shards from bottlenecks?

  542. It’s not really PIE until there are potsherds involved, as *méh₂tēr used to say.

  543. David Marjanović says:

    A mosaic made out of shards from bottlenecks?

    Heh, pretty much. The Caucasus as a whole isn’t all that isolated, but isolated mountain valleys are isolated!

  544. I don’t think it’s reasonable to refer to Athabaskan as having undergone a long-range expansion in the sense of Indo-European, Uto-Aztecan, or Afroasiatic, just because there are some outlier Athabaskan languages in the U.S. Southwest and one on the Plains. Surely nobody supposes that the Algic outliers show that Algic was once spoken from coast to coast and that the historic distribution is just a remnant of it, any more than we think that because English is spoken not only in the British Isles but also in certain parts of Western and Southern Africa, it was once spoken all along the Atlantic coasts of Europe and Africa.

    Do we in fact have any notion of how the Western Algic outliers did get there? If, as seems to be the standard view, they aren’t closely related, it suggests something very odd happening.

    ObHat: Yurok, one of the two known outliers, is remarkable for its unusually large number of nouns that are suppletive for number: ko:ra’ ~ ko’r ‘1 person’, ni’iyel ‘2 persons’, nahkseyt ‘3 persons’, though what the general plural is I don’t know. English has only person/people plus the native titles mister, missis, which are pluralized (if at all) with French forms.

  545. David Marjanović says:

    Do we in fact have any notion of how the Western Algic outliers did get there? If, as seems to be the standard view, they aren’t closely related, it suggests something very odd happening.

    Algonquian comes from its current western end or thereabouts, and expanded eastwards. Wiyot and Yurok may have moved south instead.

    Siouan, conversely, apparently got started in a more central or eastern location and expanded northwestwards.

  546. marie-lucie says:

    JC: Surely nobody supposes that the Algic outliers show that Algic was once spoken from coast to coast and that the historic distribution is just a remnant of it, …. Do we in fact have any notion of how the Western Algic outliers did get there? If, as seems to be the standard view, they aren’t closely related, it suggests something very odd happening.

    For a summary of opinions on this topic, see Campbell: American Indian languages: the historical linguistics of Native America (Oxford UP 1997), pp. 152-154. Several authors have argued that Algic started rather than ended up on the West Coast, and although there is apparently no official consensus, Campbell seems to favour this opinion. Although I know little about those languages, I understand that Wiyot and Yurok (neighbours on the Northern California coast) are considered to be linguistically about as distant from each other as either is from the very large Algonkian language group found in a wide swath of East and Central North America. This suggests to me independently that the Algonkian differentiation is relatively recent and that Wiyot and Yurok must be the result of separate migrations (perhaps at very different times). In addition, they must also reflect a certain amount of mixture of the common incoming ancestor with two very different language groups already spoken in the area.

    California languages, like others on most of the Pacific Coast of North America, are hemmed in by high, rugged mountains and other difficult terrain, while several language groups of the North American interior extend over vast territories with few obstacles to long distance travel. Thus it should not totally surprising that Wiyot and Yurok should have remained confined to California while a third, related language group should have expanded its territory toward the East and eventually spawned a very large language family. Nomadic speakers of this third element would not have had to remain in the intervening territory but could have traversed deserts and other unsuitable areas before finding places where to settle, gathering members of other groups and assimilating them in the process.

  547. marie-lucie says:

    JC: Yurok nouns which are suppletive for numbers … English has only person/people plus the native titles mister, missis, which are pluralized (if at all) with French forms.

    I am puzzled by “pluralized with French forms”. All the English nouns you cite are adaptations of (old) French forms, and only people is plural. I can’t think of what other French plural forms you mean.

  548. David Marjanović says:

    I’ve often seen Messrs. in written English, clearly Messieurs; you can’t, after all, pluralize Mr. to Mrs. which already means something else.

  549. marie-lucie says:

    I have seen Messrs too, apparently limited to business correspondence, instead of Sirs (I don’t think either can followed by names). . I don’t think there is a plural for Mrs..

  550. The masculine plural form Messrs /ˈmɛsərz/ must always be followed by one or more surnames; if not, Sirs is used. The feminine plural is Mesdames /meɪˈdɑm(z)/ or more recently Madames /məˈdɑmz/. But the plural forms are thought of as much less nativized than the singular forms, though you’re of course right that all of them are from French originally.

  551. Mssrs. is pronounced /ˈmɛsərz/, and it does precede surnames – e.g., “Mssrs. Smith and Jones”. The female equivalent is Mmes., pronounced /meɪˈdɑm/ or /meɪˈdæm/ .

  552. I don’t think I have ever seen “Messrs.” in American English. Although I’m sure it exists to some extent in America, I very much associate it with British (or Commonwealth) English. I suppose this goes along with my remark, from several weeks ago, that “M.” for “Monsieur” is seldom used in America, for Frenchmen or anyone else.

    In spoken American English, “Misters” before multiple surnames seems unremarkable to me, although it’s not particularly common. I don’t think it can be abbreviated though.

  553. Yeah, I’ve never actually encountered those plurals in the wild.

  554. Messrs.

    Is (or was) widely used in Canada in business correspondence.

  555. I just remembered where I first encountered “Messrs.” It was on a two-record set of Tom Baker reading The Strange Case of Doctor Jekyll and Mister Hyde. (It was directly abridged from the original, including the using the full title, but the script didn’t bother trying to hide that Jekyll and Hyde were manifestations of the same person; still, even knowing the reveal, it was pretty effective dramatically.) Jekyll addresses a letter to a chemist shop to “Messrs, Moore” (so that may have been part of the firm’s official name), and Baker pronounces the word as “MESS-uhs,” not sounding French at all. The meaning was obvious, but the word seemed very odd to my young-ish American ears.

  556. Re “Messrs.”: That reminds me of my boyhood, reading Karl May (a German adventure story writer, many of whose books take place in the Wild West). He renders the pronunciation “Mesh’shurs” . Is that something he made up (wouldn’t be the only thing) or is a pronunciation like that attested?

  557. David Marjanović says:

    How old is the pronunciation of /sr/ as something more like [ʃɹ] in… well, whatever English May may actually have heard? For groceries it even seems to be the only option left.

  558. What, what, what? I say /groʊsriz/, with no shibilation whatever.

  559. marie-lucie says:

    From what I have heard, grosheries seems to be spoken in the same geographical areas as warsh and rinshe, somewhere in the middle of US landmass.

  560. David Marjanović says:

    What, what, what? I say /groʊsriz/, with no shibilation whatever.

    Oh, good. I thought “great, another case where English spelling is a plain lie”.

    I’ve never been to the middle of the US landmass, though, and I don’t think the people I’ve talked with much have spent much time there either.

  561. How old is the pronunciation of /sr/ as something more like [ʃɹ] in… well, whatever English May may actually have heard?

    /s/ may undergo place assimilation and become distinctly retracted especially if the /r/ has an apical or retroflex (rather than mid-tongue) articulation. The result, however, is an allophone modified towards [ʂ] (apical) rather than [ʃ] (domed). The same may happen to /zr/, as in Israel or newsreel.

  562. I hear grosheries all the time here in NorCal.

  563. According to Bert Vaux’s 2003 survey, 52% use /s/ and 45% use /ʃ/; I think this places it on a much stronger footing than “warsh”-style /r/-insertion. Interestingly, the same survey finds an 88%-to-11% preference for /s/ in “nursery” and a 94%-to-6% preference for /s/ in “anniversary”, showing that at least in most cases, “groshery” isn’t the result of a regular process.

    And because why not, here’s a simple map I just made to give you a feeling for the regional distribution of “groshery”. (Vaux’s data presentation is a bit lacking in that regard.) The most notable things seem to be a particular concentration in Minnesota, Iowa and South Dakota and a relative absence from the urban Northeast, with most states falling in the middle ranges. Consistent with my state, I use /s/.

  564. I would expect that as assimilation to “bunched” (palatal) /r/.

  565. George Gibbard says:

    I, from Michigan, say gro[ʃ]ries, and I remember going to nur[ʃ]ry school (before kindergarten). The [ʃ] is not apical. I didn’t know the term “domed”, but I think it applies to both /ʃ/ and /ɹ/ for me: per Wikipedia “Palato-alveolar sounds are normally described as having a convex (bunched-up or domed) tongue, i.e. the front, central part of the tongue is somewhat raised compared to the tip, back and sides, which gives it weak palatalization.” No /ɹ/ in “wash” (which I mentally associate with hillbillies in West Virginia, where my dad grew up) or /ʃ/ in “rinse” (which I don’t recall hearing).

  566. David Marjanović says:

    The same may happen to /zr/, as in Israel

    I think I’ve heard that.

  567. @George Gibbard: According to Vaux’s data your “grocery” is typical for your state, but your “nursery” isn’t. Also I enjoyed making that map, so if anyone would like another one, just ask.

  568. David Marjanović says:

    The map has disappeared.

  569. Silly imgur. This one should work.

  570. Sounds like an equally possible but different interpretation, not a logical mistake.

    My point is that a statement like “Fact A, therefore conclusion B” is undermined if fact A is subject to a second interpretation that does not entail conclusion B.

  571. Vladimir Diakoff says:

    @John Cowan

    “My point is that a statement like “Fact A, therefore conclusion B” is undermined if fact A is subject to a second interpretation that does not entail conclusion B.”

    I guess when Dziebel writes that “ANE is an Amerindian marker in the Old World by definition” he means that, if a genetic marker is ever given a geographic label, this label should come from the continent/area in which the said marker is found at highest frequencies. If this pattern is interpreted in the simplest of historical terms, this means ANE is an Amerindian population that came to the Old World. There could be other interpretations (ANE is an ancient Old World marker that declined in frequency in the Old World but increased in frequency in the New World, or that ANE is a New World marker that declined in frequency in the New World but increased in frequency in the Old World – still without reaching the New World maximums, etc.), but those are more special ones that require special proof. The problem, as I understand it now, is that mainstream science has adopted a “special” interpretation (which is rare for mainstream thinking), while Dziebel who is a heterodox thinker calls for a more basic, simplest interpretation of the data.

  572. Yurok […] is remarkable for its unusually large number of nouns that are suppletive for number.
    Yurok has a large number of dependent-specific numerals. Most nouns are not marked for number. So koːraʔ ʔoːɬ ‘one person’, niʔiɬ ʔoːɬ ‘two people’, naʔap’ puːwiʃ ‘two bags’, etc.

  573. marie-lucie says:

    Y: Yurok is not the only one, several other Pacific Coast languages also has different sets of numerals depending on noun classes (eg humans, animals or their skins, roundish objects, longish objects, etc). This is true on both sides of the Pacific.

  574. David Marjanović says:

    while Dziebel who is a heterodox thinker calls for a more basic, simplest interpretation of the data

    No. One marker cannot tell us much about its history; that it’s found on both sides of the Pacific doesn’t tell us which one it came from. Currently it’s known earlier in Asia (24,000 years ago) than in America, which is evidence that it came from Asia, while it has a stronger presence in America than in Eurasia in today’s populations, which is evidence that it came from America – that’s a stalemate, and Dziebel only looks at one side of it.

  575. different sets of numerals depending on noun classes

    Sounds like what happens when in a classifier system like Chinese or Japanese the numerals merge with the classifiers in idiosyncratic ways. In Dungan, something different again has happened: all the Mandarin classifiers have been lost except the most common (and default) classifier 個 ge, which has merged with the preceding numeral or demonstrative as a suffix, so that one counts yigɨ, lyongɨ, sangɨ, sigɨ, vŭgɨ, ljugɨ, čigɨ, bagɨ, jyugɨ, šigɨ.

  576. David Marjanović says:

    ge

    Simplified: 个.

  577. (I meant head-specific, not dependent-specific.)

    That indeed seems to be the case. Mithun’s The Languages of Native North America calls these classificatory numerals, and gives examples in Carrier, Wiyot, and Nuuchahnulth (Nootka). In Nuuchahnulth the number-classifier compound can be used alone, in contrast to Yurok.

  578. I meant head-specific, not dependent-specific

    Yes, I wondered if you were saying that in two boots the head is two in these languages. Which I suppose is conceivable.

  579. Vladimir Diakoff says:

    David Marjanovic: “that’s a stalemate, and Dziebel only looks at one side of it.”

    So looks like both Dziebel and mainstream science are committing a logical mistake by favoring one interpretation over the other. But it does seem that the Old World attestation of an Amerindian component at 24,000 years is evidence not for the directionality of migration but for it’s timing. It must have happened prior to 24,000 years before present. As Dziebel point out in one of his posts, the fact that this Amerindian component in most frequent in the New World combined with the fact that typical West Eurasian components (WHG, EHG, etc.) are not attested in the Americas makes it likely that migration went from America to the Old World and not the other way around. The fact that it’s not yet attested in America earlier than 24,000 years is a matter of archaeology, ancient DNA recovery, etc., but not so much evidence for the directionality of the migration from the Old World to the New World.

  580. David Marjanović says:

    What about the principle of parsimony…?

    Mal’ta is far east of anywhere where WHG or EHG have been found. East Asian ancestry, in contrast, does occur in America.

  581. Vladimir Diakoff says:

    David Marjanovic: “Mal’ta is far east of anywhere where WHG or EHG have been found. East Asian ancestry, in contrast, does occur in America.”

    Mal’ta/Amerindian ancestry (distinct from East Asian/Amerindian ancestry) is found all over West Eurasia (any West Eurasian population is closer to Amerindians than it is to East Asians). Mal’ta is also known to contain no East Asian/Amerindian ancestry, although it’s geographically in East Asia. This leaves Amerindians as the common denominator between (or source of) West Eurasians and East Asians.

    Note also that Oase is geographically a West Eurasian sample is just as close to Amerindians (if not closer) and East Asians than it is to modern West Eurasians. So, it’s a mirror image of the Mal’ta situation. West Eurasians and East Asians come and go but Amerindians are always around when it comes to the ancestry detected in ancient samples.

    So, the principle of the economy of interpretations (parsimony) seems to favor a migration out of the Americas into the Old World, with subsequent divergence into East Asians and West Eurasians.

  582. Trond Engen says:

    John Cowan: Vajda is careful to say that Ruhlen had 8 cognates right; he is also careful to say that Ruhlen proposed 35 cognates in all, for a less than stellar hit rate. (Numbers from memory, may be wrong.)

    Here’s me quoting Vajda five years ago.

  583. David Marjanović says:

    Mal’ta is also known to contain no East Asian/Amerindian ancestry, although it’s geographically in East Asia. This leaves Amerindians as the common denominator between (or source of) West Eurasians and East Asians.