Spruik, kayayei, obroni wawu.

This ABC News (Australia) piece by Linton Besser in Ghana is excellent (and infuriating — stop buying too many clothes, wearing them twice, and discarding them, people!), and it has several passages of decided LH interest. First comes “Clothes are spruiked by song and are quickly discounted by day’s end.” “Spruiked” looked so weird I thought it might be a misprint, but no, it’s a normal word Down Under; the OED (updated March 2019) has a thorough entry s.v. spruik:

Pronunciation: Brit. /spruːk/, U.S. /spruk/, Australian English /spruːk/, New Zealand English /spruːk/

Etymology: Either < spruik n. (although first attested earlier), or directly < its etymon German Sprüche (plural noun) patter, sales pitch, spiel (see spruik n.). Slightly earlier currency is probably implied by spruiker n.

Australian and New Zealand slang.

1. intransitive. To speak in public on a particular topic, to ‘hold forth’; spec. to attract custom to a show, shop, etc., by speaking outside the premises; to act as a spruiker.
1894 Clipper (Hobart, Tasmania) 15 Dec. 6/4 When the lamplighter will dislocate his jaw spruking.
1902 Truth (Sydney) 14 Sept. 5/6 ‘Lockie the Spruiker’ that ‘spruiked’ for years at the Gaiety door, Has gone out of the ‘spruiking’ business, and never will ‘spruik’ any more.
[…]
2001 Sunday Mail (Brisbane) 3 June 74/2 Kay McGrath, the honorary compere, gamely spruiked over the table cackle.

2. transitive. To discourse on (a subject) in a public forum; to promote or publicize (something).
1901 Sydney Sportsman 16 Jan. 1/5 How did they ‘sprook’ compliments to each other.
1907 Truth (Sydney) 26 May 1/3 Some of the women spruiking politics and posing as patriots are paid pimps of the Liberal League, and householders should shoo them off the premises.
[…]
2013 Smith Jrnl. Winter 141/1 A clue..can be found on the signs outside Mickey Bourke’s Pub, near the chalkboard spruiking Guinness on tap.

Then comes this:

But transporting the 55kg bales around the teeming bazaar, with its narrow passageways and thousands of customers, is impossible by mechanical means. So the job falls to Accra’s ranks of head porters, or kayayei, “the women who carry the burden”.

Kayayei has its own Wikipedia article, from which we learn:

The term kayayei (singular, kaya yoo) is a compound formed from two languages spoken in Ghana. Kaya means “load, luggage, goods or burden” in the Hausa language, and yei means “women or females” in the Ga language. People in Kumasi refer to the porters as paa o paa.

I’m not sure why kayayei is written as one word and kaya yoo as two, but never mind, it’s what I needed to know, and I can confirm from my Hausa dictionary that kāyā (pl. kāyàyakkī) means ‘load; goods; stuff; property; clothes.’ And finally, we get:

Wander around Accra and every spare inch of pavement seems occupied by a hawker, a new batch of old clothing folded and hung among their wares. They call them “obroni wawu” — dead white man’s clothes.

Again Wikipedia comes to the rescue, s.v. Oburoni:

Oborɔnyi is the Akan (or more specifically, the Fante) word for foreigner, literally meaning “those who come from over the horizon.” It is often colloquially translated into “white person.” […]

The word oborɔnyi derives from the word bor (Fante), which means “from beyond the horizon,” and nyi, which is a suffix that means “person”. The plural form of oborɔnyi is aborɔfo (fo is the plural form of nyi), which is often used to refer to the English language or English people.[citation needed]

There is another theory that oborɔnyi is derived from the similarly sounding phrase aburo foɔ, which means “trickster”, “one who frustrates” or “one who cannot be trusted.”[citation needed]

And speaking of “one who cannot be trusted,” some of that seems a bit murky (a monosyllable meaning “from beyond the horizon”?). Perhaps our resident West Africanist can help out.

Comments

  1. David Eddyshaw says

    The word ɔbŭróní rapidly becomes very familiar to any European wandering around in the south of Ghana, as small boys feel impelled to shout it out at you in case you haven’t noticed that you were European before.

    Unfortunately about the only thing I can say in Twi is “Do you speak Twi?”, which, given that I don’t, is not particularly useful in practice …

    FWIW, I seem to recall that ɔbŭróní originally meant “bush person” (i.e., anybody not like us civilised Akan.)

    Christaller’s dictionary sheds no light on its origin, and has no word like bor “beyond the horizon.”

    Abra is “deceit.” The derivation from that looks like folk etymology to me. People are very ready to offer folk etymologies when it comes to Twi/Fante, which unfortunately suggest themselves all too readily given the simple phonological structure of the language(s).

    The word you hear shouted after you by small boys in the north is Nasaara, which is of known provenance; it comes (via the local Hausa dialect, though not Kano Hausa) ultimately from the Arabic for “Christians.”

    Talking of opaque words for “European”, I’ve often wondered about the Wolof word

    https://en.wikipedia.org/wiki/Toubab

    Looks (again) like nobody really knows. Lameen would, if anybody …

  2. Cuconnacht says

    If obroni means white man, how does obroni wawu mean “dead white man’s clothes”? Does wawu mean clothes that belonged to someone now dead?

  3. Maybe they’ve all been reading too much Heinlein and decided that the far horizons draw no nearer.

  4. David Eddyshaw says

    @Cuconnacht

    Yes. The wu bit is (mirabile dictu) cognate with both Kusaal kpi and Swahili -fa “die” (isn’t comparative linguistics fun?)

  5. The wu bit is (mirabile dictu) cognate with both Kusaal kpi and Swahili -fa “die”

    God, that’s wonderful. That’s even better than Mandarin wu = Cantonese ng.

  6. Wait, does that mean you think Akan, Kusaal, and Swahili are genetically related to the point of reconstructing stuff?

  7. I wonder if Russian inostran-ets/-ka] “other lander-m/f” would be translated as “white man” if we were green.

  8. Another favorite pair is French nous /nu/, English us /ʌs/.

  9. Proto-Bantu *-kúa ‘to die’: “From Proto-Niger-Congo. Compare Yoruba (from Proto-Yoruboid *kpú), Tiv ɑgbe, Proto-Jukunoid *kwu-, Proto-Edoid *ghu.”

  10. Proto-Edoid *ghu

    And now we know the source of Ghu.

  11. For the curious, Delafosse‘s article on the origin of toubab can be read at the following link, beginning on page 205:

    https://gallica.bnf.fr/ark:/12148/bpt6k1227325/f213.item

    (It’s not the greatest scan.)

    I had always assumed it was somehow Arabic طَبِیب ṭabīb “doctor”, but West Africa is very far from my area of expertise. Delafosse deals with this proposal.

  12. David Eddyshaw says

    Wait, does that mean you think Akan, Kusaal, and Swahili are genetically related to the point of reconstructing stuff?

    Oh yes. Not as much stuff as I would like, but yes.
    The Akan reflex comes from the work of John Stewart, who unfortunately died before he had progressed even further. The Bantu reflexes are well-established; the Swahili is just what you’d expect from *ku, bizarre as it looks (the -a part is actually a flexion.)

    In citing Kusaal kpi I’ve concealed a bit of a problem; but cf kum “death.” There’s something odd going on with CV-stem verbs in Oti-Volta, which I think (though I’m a long way from sorting it all out) is because the Oti-Volta protolanguage formed the verb imperfect aspect with the suffix *-u, a formation which subsequently became non-viable in all the branches except Eastern Oti-Volta because phonological changes caused suffixed -u to fall together with -i (though Nawdm shows extensive traces of -u in the form of rounding of the root vowel in imperfectives.) Different branches then adopted different strategies to repair verb morphology, all still retaining the basic perfective/imperfective opposition By Whatever Means Necessary.

    Still, there are a good few things that can be reconstructed back to Proto-Volta-Congo, including pronouns, noun class affixes, the verbs “eat”, “drink”, “die”, “send”, “bite”, the nouns “tree”, “mouth”, “bone”, “child” …

    Serious work at that level is really just getting started though. And (although John Stewart explicitly disagreed) my own feeling is that nothing will be achieved securely without first reconstructing the intermediate protolanguages (like my own pet project of Proto-Oti-Volta.)

  13. For those who don’t want to struggle through the whole bad scan, Delafosse‘s conclusion is that toubab is from ثَوَّاب‎ ṯawwāb “one who sells garments” (derived from ثَوْب ṯawb ‘dress, garment’) — very appropriate in this thread!

  14. David Eddyshaw says

    Like Delafosse, I wondered about “doctor”, bur rapidly rejected it as semantically implausible, to say the least.

    I must say I’m not that much more convinced by “clothes merchant”, though. Better than “trousers-wearer”, though, as D rightly concludes; as he says, there are plenty of West African traditional trouser-wearers around (including the Kusaasi, in fact.)

  15. So, David, refresh my memory. What parts of the sprawling Niger-Congo do you think are not supported as part of the same phylum?

  16. David Eddyshaw says

    There is no real evidence that Mande is related; there’s nothing in the morphology, the syntax is very different from most of Volta-Congo, and lists of putative cognates are (a) very short and (b) full of onomatopoeic and mama/papa type words. To a lesser degree, the same applies to Dogon. (Like Mande and Fulfulde, Dogon has one or two number words clearly related to Volta-Congo; but despite the view from Indo-European and Semitic, number words are actually very borrowable.)

    Atlantic has three branches which are difficult to prove related to one another, let alone to anything else. I stumbled across a remark of Mark Liberman’s relating to when he’d been supervising a postgrad working on some of these languages and had come away thinking that their mutual relationship was on a level with Altaic, which seems fair. What has led to the fairly confident declaration of the relationship with Volta-Congo is basically morphology, notably exuberant noun class systems and verbal derivation by suffixes. However, striking as the typological similarities may be, when you actually start looking for correspondences in detail things start to melt away. And indeed the very similarities are, in a paradoxical way, suspicious: if the languages are really related, morphology of this kind has had many millennia to get eroded away, and it would be very odd if it had all remained so pristine over a period where the loss of any cognate vocabulary has been near-total. Having said that, I just recently got hold of Segerer’s Bijogo grammar (thanks in fact to you!) and recognised several familiar etyma which I hadn’t seen in the other two Atlantic branches, so it may well be that that part of Atlantic does indeed belong with Volta-Congo. And I would not be astonished if Fulfulde ultimately turned out to be related too, but not for the reasons commonly adduced, which are really just typological.

    John Stewart actually did think that Atlantic was provably related to Volta-Congo, and indeed that he had found regular phonological relationships. I don’t think his papers substantiate this, myself.

    Khordofanian is rather similar. It seems to be internally very diverse (including some parts which don’t look like plausible candidates for “Niger-Congo” membership to anybody), and again the noun class systems were what made Greenberg put them with his Niger-Congo, even he admitting that there was no real lexical evidence at all. Again, when you look at the noun class systems, although the general MO is similar, with paired sg/pl prefixes, the actual segments match no more often than chance. And as with Atlantic, how come all this morphology has survived the millennia when essentially no vocabulary has?

    Volta-Congo is undoubtedly real, but its actual boundaries are open to dispute. Within Greenberg’s “Kwa” there are groups like Ijaw which are in reality isolates. Greenberg’s “Adamawa” is a real grab-bag; part of it, like “Ubangian” is not really Volta-Congo; part is close enough to his “Gur” that it has rightly led to the conclusion that “Gur” itself is not a real node (though it’s all surely Volta-Congo); this is why I go on about Oti-Volta rather than “Gur.”

    That’s my take; but I’d say the current consensus is that although Mande doesn’t belong, pretty much everything else in Niger-Congo does. But anybody who’s actually looked at the evidence would concede that you’re not really talking about something like Indo-European or even Afro-Asiatic there; it’s more like the more sensible takes on Nostratic.

  17. Xerîb: Thanks for the link! Not terribly convinced by Delafosse’s proposal though; the only way I can think of for ww to become b in Wolof would be if it passed through Zenaga first, and as far as I know Zenaga doesn’t have it.

  18. David Marjanović says

    the more sensibel takes in Nostratic

    Speaking of which: nontrivial sound correspondences between Proto-Indo-European and Proto-Uralic, only one of which (supported by only two presented examples) looks bizarre. (Slides of a conference presentation in Russian.)

    Taken together, they make Proto-Indo-Uralic look noticeably similar to Altaic, phonologically speaking… that’s not in the presentation, I’ll try to elaborate on it tomorrow.

  19. “Spruiked” looked so weird I thought it might be a misprint,

    NZ’er here. I can confirm that’s a common enough word. I haven’t seen it written often, but I’d say no-one’s very sure how it’s spelled — yes that spelling looks weird. So I’ve seen “spruked/spruke”, also “sprewk”.

  20. Annette Pickles says

    As a joke… Assuming a shift of i to u in the first syllable due to the following labial, the form of the Wolof word tubaab looks like an Arabic plural fiʕāl, one of the plurals of the form faʕīl. There is an Arabic تبيب tabīb (root tbb, “to perish, be destroyed, suffer loss”) defined by Kazimirski as “faiblesse, impuissance, néant, nullité”. Can we propose a plural tibāb for this? Wolof could have then borrowed this plural as tubaab yi *“les nuls” > “foreigners” (y- noun class), with backformed singular tubaab bi “foreigner”. (A bit like Hi ða sende to Angle ⁊ heton heom sendan mare fultum ⁊ heom seggan Brytwalana nahtnesse ⁊ ðæs landes cysta “They then sent to the Angles, and desired them to send more assistance. They described the nothingness of the Britons, and the richness of the land”, but in reverse.) 😀

    I was actually trying to get this to plural trick, either with the plural pattern fiʕāl or the plural pattern fuʕʕāl, to work for ṭabīb طبيب “doctor” as the etymon… But there are too many objections in all aspects.

  21. David Eddyshaw says

    I wondered about the Arabic ta:ba twb “repent” (whence Hausa tuba), but anything to do with repentance as an ethnonym bestowed by Africans on Europeans seems a bit … anachronistic.

    Arabic ṭbb has turned up in Kusaal as tip “healer” presumably by extension of the meaning of ṭibb “medicine” (*bb -> p is a regular intra-Kusaal development), and, I think, also influenced the verb ti’eb “get ready” into acquiring the additional sense “heal.”

    Toubib is French argot for “doctor”, of course, but seems after all to have nothing to do with toubab (although, like me, one can of course be both.)

  22. There’s also a noun – spruiker.

    A spruiker is a person with a microphone who stands outside a shop and tells you all about the specials and great prices inside.

  23. David Eddyshaw says

    A booster (as in “boosterism.”)

  24. Rob Solheim says

    That’s a coincidence. I came across this word for the first time in an article in today’s Guardian about Covid anti-vaxers in NSW. I thought it was a misprint (you know, it’s The Grauniad), but looked it up.

    For reference, the article is here: https://www.theguardian.com/australia-news/2021/aug/14/when-covid-came-to-the-anti-vax-capital-of-australia?CMP=Share_AndroidApp_Other

  25. Another favorite pair is French nous /nu/, English us /ʌs/.

    … with Hittite anzas (German uns) sitting in the middle of the fork.

  26. PlasticPaddy says

    One definition for the word spreuk in Dutch is
    3. bordje of kartonnetje met een tekst er op
    So a spreuker could be a sandwich-board man or talking notice board, although the word does not exist in Dutch.

  27. Stu Clayton says

    Toubib is French argot for “doctor”

    JTODIL that from the French woman who lives in the house behind. I had never encountered the word, so my initial reaction was to consign it to Lyon dialect (she comes from Lyon, and often enthuses about the dialect).

    The very next day I ran across the word in Simenon (b. Liège) or Revel (b. Marseille), I forget which.

    Larousse sez

    (arabe maghrébin ṭbīb, de l’arabeclassique ṭabīb, médecin)

  28. (Fixed your Arabic transcription so it wasn’t so confusing.)

  29. Stu Clayton says

    Thanx. Blâme attaches to Larousse, I merely cut and paste. As if *I* would notice such a thing !

  30. David Marjanović says

    Another favorite pair is French nous /nu/, English us /ʌs/.

    That’s another ablaut grade, though, right?

    “les nuls”

    If that’s not true, it had to be invented!

  31. Trond Engen says

    Anette Pickles:

    There is an Arabic تبيب tabīb (root tbb, “to perish, be destroyed, suffer loss”) defined by Kazimirski as “faiblesse, impuissance, néant, nullité”. Can we propose a plural tibāb for this? Wolof could have then borrowed this plural as tubaab yi *“les nuls” > “foreigners” (y- noun class), with backformed singular tubaab bi “foreigner”.

    ON skrælingar “poor or helpless people; native Greenlanders and Americans” is a semantic parallel.

  32. J.W. Brewer says

    If “spruik” really comes from Sprüche, why isn’t it “shpruik,” or at least pronounced with a ʃp (like “spiel” for many AmEng speakers) even if the spelling doesn’t make that explicit? Maybe speakers down under have less contact with German-or-Yiddish-speakers than many Americans, so that’s more likely to be blocked by their phonotactics? I am equally puzzled by why the alternative spelling “sprook” did not become the dominant one. Do AustEng and NZEng have other lexemes ending in -uik so that it’s part of a larger pattern?

  33. “I wondered about the Arabic ta:ba twb “repent””

    The word originally meant “return” (same Semitic root as teshuvah, via Aramaic). So, assuming early contact between Wolof and Aramaic, we can take tubaab as meaning “those who should go back where they came from”.

    (Sadly, no of course we can’t, but it is tempting:)

  34. David Marjanović says

    one of which (supported by only two presented examples) looks bizarre

    Oops, seven examples. Some of them look fishy, though.

    More tomorrow!

  35. John Cowan says

    And as with Atlantic, how come all this morphology has survived the millennia when essentially no vocabulary has?

    That question makes me think of Afroasiatic, where solid cognates across all five branches (setting Omotic aside) are rare; there are two dictionaries of PAA which agree on essentially nothing, Ehret’s and the other one. Yet the triliteral roots and interdigitating morphology can’t possibly be just chance resemblance or even a Sprachbund effect.

  36. David L. Gold says

    @ Lameen. “assuming early contact between Wolof and Aramaic.”

    Since the area of maximum extent of Wolof and that of Aramaic are far apart, is there even weak evidence for contact of any kind between speakers of the two languages?

    Or is “Aramaic” a slip of the pen for “Arabic”?

    Might Arabic > Songhay > Wolof be possible?

  37. David Eddyshaw says

    triliteral roots and interdigitating morphology can’t possibly be just chance resemblance or even a Sprachbund effect

    Triliteral roots are more of a Semitic thing; in fact Semitic shows clear signs of having adopted various strategies to make biliteral roots conform to the triliteral pattern.

    Still, a good point: Hausa, for example, has pronouns which boldly proclaim their evident cognacy with Semitic after untold millennia.

    With the weird Afroasiatic ablaut-on-steroids thing it’s not so much that there are parallels in detail (the common factors really boil down to “insert -a- somewhere to make plurals and sometimes durative verb forms”) as the typological oddity that has been conserved so widely.

    However, I suppose you could indeed draw a parallel with the phonologically arbitrary sg/pl affix pairs that characterise so much of Greenbergly Niger-Congo: again, the actual details fail to match once you get outside the Volta-Congo core, but Greenberg himself seems to have felt that this strategy for noun inflexion was so exceptional that its mere presence more or less proved cognacy. It’s not a stupid proposition, though I don’t find it persuasive myself: at any rate it seems to me to be something more likely to have spread by diffusion than non-concatenative morphology*, particularly as the noun-affix ordering clearly can’t have been fixed in the protolanguage, and in many modern languages the fusion of class affixes with noun stems is even now very loose, so that for any Proto-Niger-Congo you’re really talking about a sort of classifier system that used different sg and pl classifiers rather than already-agglutinated morphology.

    * To my disquiet, I’m finding signs of what looks horribly like a/i ablaut in Oti-Volta, despite my best efforts to explain all cases away as “umlaut” (which certainly is a thing in several branches of the family.) It’s disquieting because there seems to be precious little evidence of such a thing elsewhere in Volta-Congo; also because it seems rather liable to open the door to the sort of etymologies in which consonants count for very little and vowels for nothing at all …

  38. David Eddyshaw says

    @David Gold:

    I am going to take it on myself to say that Lameen’s proposal was not in earnest …

    There are plenty of Arabic loans in Wolof, though. Arabic words get pretty much everywhere in the West African savanna and Sahel; even Kusaal has a fair number, though the Kusaasi are overwhelmingly not Muslim. (Come to that, there are a fair number of Aramaic loans in Arabic. But let’s not go there …)

  39. @JWB If “spruik” really comes from Sprüche, why isn’t it “shpruik,” or at least pronounced with a ʃp (like “spiel” for many AmEng speakers) even if the spelling doesn’t make that explicit? Maybe speakers down under have less contact with German-or-Yiddish-speakers …? I am equally puzzled by why the alternative spelling “sprook” did not become the dominant one. Do AustEng and NZEng have other lexemes ending in -uik so that it’s part of a larger pattern?

    There’s much more Scots influence in Aus/NZ than German-Yiddish-Dutch.

    I propose Penicuik as the model.

    (Hat gives us licence for speculative linguistics, I believe?)

    That’d jive with the general fakery that comes with the territory.

  40. Stu Clayton says

    it seems rather liable to open the door to the sort of etymologies in which consonants count for very little and vowels for nothing at all …

    As in German. Traditional etymology can severely hinder intelligibility. Confronted with an unfamiliar German dialect/accent, I simply dismiss the vowels and let the consonants run free-range (before that forum known as the mind’s eye). Pretty soon I figure out what’s going on, more or less.

  41. I thought the smiley face at the end would be sufficient indication that I was not entirely in earnest in supposing Wolof-Aramaic contact…

    There is one probable Hebrew loan into nearby Hassaniya Arabic, though (presumably via Zenaga given the prefix): amālāẓ “translator” < mēlîṣ.

    (Edit) This, of course, must stem from medieval contact with multilingual Jewish merchants, not from direct contact with people who still spoke Hebrew as a first language.

  42. David Marjanović says

    That question makes me think of Afroasiatic, where solid cognates across all five branches (setting Omotic aside) are rare

    I don’t think anybody knows how common they really are. The first reconstruction of Proto-Central-Chadic came out a few years ago, Proto-Chadic is a work in progress, I’m not aware of anyone working on Proto-Cushitic… before we have those, we can’t compare them in much detail.

  43. John Cowan says

    We discussed this before, but the trick is to pretend that some suitably conservative-looking Chadic or Cushitic language actually is the protolanguage, as was done in the early days of PIE study. Etienne also has an argument that it is methodologically unsound to pile protolanguages on protolanguages, or Pelion on Ossa for that matter.

  44. “This, of course, must stem from medieval contact with multilingual Jewish merchants, not from direct contact with people who still spoke Hebrew as a first language.”

    If the argument for a recent (medieval rather than from antiquity) loan is that “interpreter” is less stable than [onion-learn-…], one can always consider t-r-g-m-(n)🙂

  45. David Marjanović says

    the trick is to pretend that some suitably conservative-looking Chadic or Cushitic language actually is the protolanguage, as was done in the early days of PIE study.

    That assumes you can find one. (…Sure, Teeter’s law will give you one, but that doesn’t make a consensus unless everyone has been studying the same language as you.) There are reasons it isn’t done much in PIE study anymore, and less and less in Proto-Uralic study among others (most people in that field used to just project Modern Standard Finnish backwards, with varying amounts of internal reconstruction perhaps, and that has increasingly turned out to be misleading).

    Also, there doesn’t seem to be complete agreement on which languages exactly are Cushitic and which perhaps aren’t. It’s a very large and diverse group in any case

    it is methodologically unsound to pile protolanguages on protolanguages

    It piles error margins on error margins. You just need to keep in mind how large the error margins really are. As long as they aren’t larger than the data themselves, keep going.

    It’s not like we do anything different in biology. (It’s just less obvious there.)

  46. John Cowan says

    I simply dismiss the vowels and let the consonants run free-range

    And that’s a pretty good strategy not just for German but for written Germanic as a whole, if you can wrap your head around the High German sound shift and remember that Zeit and tide wait for no man. But elsewhere things are otherwise.

    When dealing with Polynesian, the vowels are the classic Spanish five, long and short, and they are the same everywhere — but the consonants, ah, the consonants, they play musical chairs as you move from west to east and the total number shrinks. Only Tongan retains Proto-Polynesian /h/, and only a few more retain the PPn glottal stop, though the languages east of Samoa have changed /s/ to a new /h/, and the Marquesan languages have /h/ from the merger of PPn /r/ and /l/. Here’s a chart of sound correspondences.

    Finally, when studying a new Tai language (Thai and Lao are the best known), your first step is to work out the extremely regular tone correspondences with known varieties, and everything else just flows from there.

  47. “the trick is to pretend that some suitably conservative-looking Chadic or Cushitic language actually is the protolanguage”

    It feels like a throwback to the 19th century, but the more I look at those two families the more convinced I become that the northernmost members of each are the most conservative, at least morphologically. The prefix conjugation, a clear AA inheritance in Cushitic, is only still productive in Beja, Saho, and Afar; internal plurals for nouns are essentially lost in Chadic, apart from occasional irregular archaisms, everywhere south of Lake Chad…

    But blithely reconstructing the lexicon on the basis of such “conservative” languages would be very risky, because those morphologically conservative languages are also the most strongly influenced by Semitic and Berber in the lexicon, significantly increasing the risk of spurious cognates.

  48. John Cowan says

    Sure. You will probably end up with something like Schleicher’s sheep-and-horses story. In 1868 he wrote “Avis, jasmin varnā na ā ast, dadarka akvams, tam, vāgham garum vaghantam, tam, bhāram magham, tam, manum āku bharantam”, whereas Byrd’s 2013 version has “h₂áu̯ei̯ h₁i̯osméi̯ h₂u̯l̥h₁náh₂ né h₁ést, só h₁éḱu̯oms derḱt. só gʷr̥hₓúm u̯óǵʰom u̯eǵʰed; só méǵh₂m̥ bʰórom.” Truly, PIE changes faster than any living language! (WP has eight other versions.)

    Some of the difference is just spelling convention: u̯ for w and i̯ for j aggravates me, because they mean the same thing and you are supposed to use the simpler symbol when possible, but perhaps the first at least helps Germans and Poles. But clearly Schleicher doesn’t get e/o ablaut, or thinks it is secondary, and he puts far too much weight on Sanskrit.

    But back to AA: what else can you do with “five ‘undred fousand bleedin’ ole [languages] all as alike as a regiment o’ bleedin’ dragoons”? Not to mention that Hausa eats one or two Chadic languages a week, especially the ones spoken by about three people each (quoth David E), thus putting priceless data for reconstruction to the torch? The main thing is that Schleicher’s generation founded a research tradition that burns (with varying levels of brightness) to this day, and without that, the final (so far!) version could never have been written.

  49. David Marjanović says

    When dealing with Polynesian, the vowels are the classic Spanish five, long and short, and they are the same everywhere —

    The given examples actually show a few a-e correspondences and an a-o one; they look like umlaut. But yeah, Germanic this is not.

    But clearly Schleicher doesn’t get e/o ablaut, or thinks it is secondary, and he puts far too much weight on Sanskrit.

    Yep, well into the next decade it was believed that the “European” a/e/o diversity was due to random unconditioned splits, and the Sanskrit state of affairs, where e & o (always long) are transparently and synchronically from *ai *au, was original, mostly because Sanskrit was the oldest known documented IE language.

  50. David Eddyshaw says

    it is methodologically unsound to pile protolanguages on protolanguages

    Then we are all doomed, because it seems to me that there are considerable problems with not doing it that way:

    (a) You could match pretty much any arbitrary Swadesh-200-type lexeme from any language with a plausible lookalike from some Volta-Congo language. There’s a lot of choice. It’s only by looking at protolanguages that you can be sure of avoiding this sort of thing (which is a big part of the reason why the two Afroasiatic comparative dictionaries are so bad.)

    (b) What about morphology? It really is no use at all comparing the noun class systems of Swahili and Kusaal directly, for example. We know from comparative work on Bantu and Oti-Volta that both have in fact simplified their inherited systems quite a bit, and moreover simplified them differently; moreover, Swahili has at least one common class [ki/vi] with no analogue at all in Oti-Volta, but virtually universal in Bantu (and well represented in some branches closely related to Bantu, but not at all in languages farther west.) There’s no earthly way you could know this without looking at intermediate protolanguages. And noun morphology is the easy bit in Volta-Congo; good luck with making any sensible deductions about verb morphology at that level without trying to reconstruct what’s happened in the individual branches first. Gabriel Manessy (peace be upon him) actually gave up altogether when it came to reconstructing proto-Gur verb flexion, and I can see why, but I think there is still a prospect of some progress. It may well be that it is no longer possible to reconstruct anything about Proto-Volta-Congo verb flexion, but if it is possible, it certainly won’t be done by leaping straight from the modern languages to Proto-Volta-Congo. What happens when you try to do that is that you end up projecting what is in fact the Proto-Bantu system back to Proto-Volta-Congo, as indeed is all too often done. That’s a methodological error, if you like …

    (c) Come to that, what about phonology? Two noun classes actually do match pretty well between Kusaal and Swahili (and indeed across pretty much all of Volta-Congo where the classes are preserved well enough to tell), but the only reason I can be confident that the second of these is not a case of coincidental resemblance is that the reconstructed phonemes match in Proto-Oti-Volta and Proto-Bantu. Neither Kusaal nor Swahili preserves them in anything like their pristine glory.

  51. I just recently got hold of Segerer’s Bijogo grammar […] it may well be that that part of Atlantic does indeed belong with Volta-Congo

    By “that part” do you mean the whole of Senegambian? which really appears to be the bulk of Atlantic, to the extent that it even seems to be today again called just “Atlantic” by some people who excise the southern parts (either as total or as within-NC isolates).

    they make Proto-Indo-Uralic look noticeably similar to Altaic, phonologically speaking

    Given all the uncertainties that remain in Nostratic subclassification, it’s still entirely possible that Proto-Indo-Uralic (= the last common IE / U ancestor) is in fact Proto-Altaic (≈ the last common Turkic / Mongolic ancestor).

  52. David Eddyshaw says

    do you mean the whole of Senegambian?

    No. I think it very much remains to be proven that the Bak languages are related to the Fula/Wolof languages.

  53. John Cowan says

    Etienne also has an argument that it is methodologically unsound to pile protolanguages on protolanguages

    Well, I should rightly have attributed that to Piotr. As usual, I found this out today by sheer serendipity, which is why the “Random Link” is there.

  54. David Eddyshaw says

    Piotr says there:

    Proto-branch languages are reconstructed first and foremost for the sake of quality control. The nodes of the family tree are where the most conservative features of the whole branch roughly coalesce, and where it is convenient to check the consistency of the reconstruction. Proto-Germanic is not reconstructed just by comparing English with German, Dutch, Icelandic, Gothic, etc. The reconstruction is informed by the rest of the family tree as well.

    I’ve no quarrel with that at all (and indeed it’s more or less what I was struggling to express above.) “Quality control” is a happy expression for what I meant. Piotr says it better, as you’d expect (it’s worth looking at the whole thing, naturally.)

    Mind you, nothing in what he says means that reconstruction of the intermediate protolanguages isn’t still important. Quality control works both ways in this. There needs to be feedback in both directions.

    I think it also makes a difference to the dynamic that reconstruction of Proto-Volta-Congo, even, and much more Proto-Niger-Congo, is very much less advanced than that of PIE, and likely to remain so; there are major obstacles in the way of ever achieving that level of rigour, and this limits the potential for top-down quality checking. On the other hand, if proposals for Proto-Niger-Congo shed no light at all on developments in individual branches, Proto-Niger-Congo itself would become an explanation that didn’t actually explain anything, and an answer to questions nobody had asked.

  55. You don’t treat intermediate reconstructions “as data” in the way you treat attestations as data, of course, but this does not mean that making reference to Proto-Finnic *kala ‘fish’ or Proto-Mordvinic *kuz ‘spruce’, or comparing these with further relatives, would be somehow necessarily fundamentally more unreliable than making reference to standard literary Finnish kala or Erzya куз /kuz/. If anything, in these particular kind of cases it's really more so, since attaching an asterisk tells that we also have enough reliable cognates in other languages of the branch. IMO it's distribution that is maybe the most important type of argument you get from intermediate reconstructions. Once that’s in hand, 95% of the time you might as well go right back to just operating with one relatively archaic descendant. This will superficially look like, but not actually operate as the same as, comparing a handful of these to begin with. (E.g. directly comparing just Northern Sami and Finnish runs into a lot of trouble due to hundreds of loanwords between the two, while comparing NS and Fi representatives of etyma known to reconstruct to PS and PF would do much better already in weeding these out.)

    Besides the well known attestation benefit in IE, I suspect a lot of the motivation behind this stance probably also comes down to how a reasonably accurate reconstruction of Proto-Germanic or Proto-Hellenic or Proto-Iranian etc. is pretty much impossible just from modern descendants. Well known issue here in Uralistics, too — ask ten people and you’ll get thirteen different reconstructions of Proto-Permic. But then in the end that’s no major problem for the formal side of wider-ranging comparison as long as everyone agrees on what the underlying sound correspondences are.

  56. David Marjanović says

    Taken together, they make Proto-Indo-Uralic look noticeably similar to Altaic, phonologically speaking… that’s not in the presentation, I’ll try to elaborate on it tomorrow.

    […]

    More tomorrow!

    One and a half more mornings later, here goes.

    The second slide of the presentation shows the PU and the PIE sound inventories. As the third slide points out, PU has a simple consonant inventory and a fairly complex vowel inventory; PIE has a complex consonant inventory and a very simple vowel inventory. If we just start with that, we’ll reconstruct a PIU consonant inventory that is basically the PIE one, and the PU one would be derived from that by a series of megamergers, making it too easy to find possible cognates. (See also: Tocharian.) Rather, like Illich-Svitych (the founder of Moscow School Nostratics), we should look for correspondences between PU vowel distinctions and PIE consonant distinctions – but not in the same way as Illich-Svitych, it turns out.

    (Not mentioned, because presupposed to be known by the audience, is that the eight PU vowel phonemes were only distinguished in the first syllable. The other syllables appear to have had just a two-way contrast, *a/*ä vs. *ə, with the choice of *a vs. *ä dependent on whether the vowel in the first syllable is back or front. Zhivlov follows the Finnish-based tradition in writing *ə as *i; but in any case keep in mind that *i, in the first syllable as well, is neutral with respect to vowel harmony.)

    Then follows a slide on PU *-l- (between the two vowels of the root) corresponding to PIE *-h₂j- and *-h₃j-* (at the end of the root, after its only vowel), then the six examples of PU *-č- (probably a retroflex [ʈ͡ʂ]) plus one of PU *-nč- corresponding to PIE *-r-, then two of PU *š- (probably retroflex again) corresponding to PIE *h₂-, then three of PU *l next to *e corresponding to PIE *j.

    The 8th slide shows the famous cooccurrence restrictions of plosives in PIE roots: the types on the left are allowed/attested, the three on the right are not. Like the glottalists, Zhivlov infers from this on the next slide that voicelessness vs. voiced aspiration was a feature of the root [or affix] as a whole, not of individual segments – a prosodic feature. “What could that correspond to in Uralic?”

    Vowel harmony. Vowel frontness was a feature of the whole word, not of individual segments.

    Next slide: PU *a corresponds to voiceless PIE consonants (4 examples).
    Next slide: PU *o corresponds to voiceless PIE consonants (3 examples).
    Next slide: PU *u corresponds to voiceless PIE consonants (6 examples).

    Next slide: PU *i sometimes corresponds to voiceless PIE consonants, too. Perhaps, then, it’s a merger of pre-PU *i and a pre-PU *ɯ. One of the the two examples actually gets a suffix *-ma rather than *-mä in Finnic, but that’s hidden in a parenthesis.

    Next slide: PU *ä corresponds to voiced aspirated PIE consonants (4 examples).
    Next slide: PU *e corresponds to voiced aspirated PIE consonants (5 examples).
    Next slide: PU *i corresponds to voiced aspirated PIE consonants (3 examples), except see above.
    Next slide: PU *ü corresponds to voiced aspirated PIE consonants (2 examples).

    Next slide: PU *ɤ corresponds to voiceless PIE consonants (3 examples). That’s interesting, because *ɤ is supposed to be a back vowel. Zhivlov points out, though, that it became *ü in Proto-Mari, where all other front & back vowels stayed on their sides.

    There’s a typological parallel, as the next slide points out: Adjarian’s law in a number of Armenian dialects, where voiced consonants front the vowels next to them (creating a Turkish-style vowel inventory). Bert Vaux (1992) explained this as voicedness causing advanced tongue root (ATR), which then caused frontness. Voiced consonants have also caused ATR in a number of Austroasiatic languages, and then the Mongolic languages are cautiously mentioned.

    The next slide asks if Adjarian’s law can run backwards: frontness causing ATR, ATR causing voicedness. Zhivlov postulates that this happened between PIU and PIE, adding that “at the moment of this transition” the PIE “plain voiced” consonants must have been distinguished not by voice, but by something else, “cf. various versions of Glottalic Theory”.

    ~~~~~~~~~~~~~~~~~~~~~~~

    Yes, Adjarian’s law can run backwards. It has in Oghuz Turkic, which developed voiced plosives – absent from Proto-Turkic at least in most environments – in front of front vowels. That’s textbook wisdom (assuming there are textbooks of Turkology). J. Pystynen has called it “Nayraca’s law” by spelling Adjarian backwards in Turkish.

    Now for something not totally different at all: internal reconstruction of the PU vowel system. As discussed at length in the “Comparative reconstruction” part of this presentation, it has a few oddities, for example *ü but no *ö, *ɤ but no *ɯ, evidence that *o was at the same height as *ä rather than *e, evidence for a rounded back vowel between *u and *o. One of the proposals in the presentation is that there was a chain shift which turned pre-PU **u into PU *ü, *o to *u, and so on.

    ~~~~~~~~~~~~~~~~~~~

    On the Altaic side, recent work mostly listed here (I recommend starting with the overview at the very bottom, the “draft”) has argued quite convincingly that the Mongolic, Tungusic and Middle Korean vowel harmony systems involve presence/absence of retracted tongue root (RTR) – an areal feature that covers most or all of northeast Asia and extends deep into North America, and apparently into Turkic as well. (Even the very restricted harmony systems of e.g. Old Japanese seem to fit RTR better than anything else.) Clearly, if we’re going to reconstruct any Proto-Altaic, we should reconstruct it with RTR-based harmony. Martine Robbeets does that. And if we’re not going to, we should probably reconstruct all the eastern branches of Nostratic with RTR-based harmony anyway.

    RTR-based vowel systems have a few interesting differences from ATR-based ones (which have long been known from west & east Africa, where they’re also areal). RTR-based systems have a big gap in the vowel space: there are, conservatively, no open front vowels. (This gap can be filled by umlaut phenomena, as common in Mongolic; by coalescence phenomena, e.g. /ɑ̙j/ > [æ̙] as in Khalkha and widespread in Tungusic; and/or by drift phenomena, e.g. /ə/ > [ɛ] and /o/ > [ɵ] in Khalkha.) This is not found in ATR-based systems. Also, when the harmony is partially lost, the first contrast to be lost in RTR-based systems is that of /i/ and /ɪ̙/, followed by /u/ and /ʊ̙/, while ATR-based systems lose the contrast beween /ɜ̘/ and /a/ first and have a tendency to turn /ɪ ʊ/ into [e o], preventing the mergers that are most common in RTR-based systems.

    If Khalkha-style drift goes far enough, you end up with a tongue-root-free frontness-based harmony system, as is found in (most of?) Turkic – and in those Uralic branches that show harmony. (The geographically central branches lack it, perhaps because they’ve lost too many unstressed vowels to support it.)

    So perhaps the PU *ü comes specifically from a -RTR **u, and the PU *u from +RTR **ʊ̙; and the unruly PU *ɤ, which becomes a front vowel in Mari and corresponds to voiced aspirates in PIE, could come from a -RTR *ə. Likewise, the PU *i, which corresponds to both voiceless plosives and to voiced aspirated ones in PIE without a known conditioning factor, and which is harmony-neutral despite being phonetically the frontest front vowel of all, could be a merger product of -RTR **i and +RTR **ɪ̙, the first merger to be expected.

    Finally, an extra PU vowel between *u and *o ( < **ʊ̙, **ɔ̙) would be particularly easily accommodated ( < -RTR **o). Crucially, this entails the prediction that this vowel corresponds to voiced aspirates in PIE. If its existence can be substantiated for PU, but it is found to correspond to voiceless PIE plosives, the whole hypothesis is most likely wrong. I’ve piled speculations upon speculations, all derived from different sources, and ended up with a testable prediction. Let the fun begin.

    I can go on: the PU *ä and *e should be coalescence products, perhaps like in Korean from *ɑ̙j and *əj respectively. But *ä corresponds to voiced aspirates in PIE…?

    One more thing: by deriving the voiceless and the voiced aspirated PIE plosives from a single series, but leaving the “plain voiced” ones alone, we get two series that must have merged in Uralic. Robbeets-style Proto-Altaic also has two series. But of course that’s very common worldwide, notably being restored in most of IE today.

  57. David Marjanović says

    Given all the uncertainties that remain in Nostratic subclassification, it’s still entirely possible that Proto-Indo-Uralic (= the last common IE / U ancestor) is in fact Proto-Altaic (≈ the last common Turkic / Mongolic ancestor).

    Yes. The distribution of RTR, if the optimism of these page proofs is justified, is actually an argument for projecting RTR straight back to the protolanguage of some sort of northern Nostratic.

    a reasonably accurate reconstruction of Proto-Germanic or Proto-Hellenic or Proto-Iranian etc. is pretty much impossible just from modern descendants.

    Well, in the first two cases that’s trivial, because East Germanic and Ancient Thessalian are dead.

  58. David Marjanović says

    Odd. I posted a long comment with 4 links. It survived two edits (in which I didn’t add any links). But posting the next comment made it disappear.

  59. I have freed it from durance vile!

  60. @DM … (4 examples) … (3 examples) … (6 examples) …

    How do I interpret the statistical significance of these counts? Presuming the language has thousands of lexemes, a handful of examples would seem well within the bounds of chance coincidence.

    Or is it that we have hardly any extant texts? Is it that all occurrences of PU *a corresponds to voiceless PIE consonants in the corpus?

  61. You can probably find 2–3 examples for any Uralic ~ IE correspondence you desire, if not applying any further quality controls. Most of the significance I think would depend on how solid the reconstructed PIE and PU items being relied on are exactly.

    an extra PU vowel between *u and *o ( < **ʊ̙, **ɔ̙) would be particularly easily accommodated (…) this entails the prediction that this vowel corresponds to voiced aspirates in PIE

    Sure enough, PU *porə- (> west *purə-) ‘to bite (into)’ ~ PIE *bʰerH- ‘to pierce’; but then also e.g. *kokə- ‘to check traps’ (> Samoyedic *ko-) ~ PIE *h₃ekʷ- ‘to see’ (though arguably the weaker of the two, since a lot of things in PIE can be reasonably compared with PU *k).

    in the first two cases that’s trivial

    If those were unattested, we certainly would be calling what is today known as Northwest Germanic as just “Germanic”, likewise Modern Greek + Tsakonian would be the whole of “Hellenic”.

  62. John Cowan says

    Yes: after all, we can only reconstruct on the basis of the languages we know at any given time. If a Proto-Oti-Volta equivalent of Mycenaean were to be discovered, everything would change.

  63. David Marjanović says

    How do I interpret the statistical significance of these counts?

    Well, you don’t – you wait for the paper. 😐 This is just a conference presentation after all (and then just the slides of it – I have no idea what Zhivlov said when he presented them). On the one hand, if you put more than 7 or 8 examples on a slide, your audience can’t read any of them (…if they’re projected, as opposed to the conference being online…). We’ll see if Zhivlov has more, and just chose to present the most impressive ones or a random sample. On the other, maybe this is all he had when he made the presentation; maybe he has found more since, maybe he has abandoned some of these hypotheses now. We’ll see.

    Sure enough, PU *porə- (> west *purə-) ‘to bite (into)’ ~ PIE *bʰerH- ‘to pierce’

    Oh, perfect.

  64. David Marjanović says

    Even more so because if *porə- ~ *purə- had this extra vowel, that’s a good argument against the obvious alternative that it’s a borrowing from IE: IE is not likely to have had an actual [o] in *bʰerH- and its descendants long before Northwest Germanic times (bore as in boring mollusks; modern German bohren “drill/bore”). In particular, borrowings from Pre-Indo-Iranian, Balto-Slavic and Germanic are definitely excluded that way.

  65. David Marjanović: bore as in boring mollusks

    Nice example of using the word in the phrase without actually distinguishing the meaning.

    The two relevant senses of bore may have come from the same root, but about the “thing which bores or causes ennui; an annoyance, a nuisance,”-related senses, the OED says they:

    … arose after 1750; etymology unknown
    (Usually supposed to be < bore v.2, which is then regarded as a figurative use of bore v.1, with the notion of ‘persistent annoyance’ (compare German drillen). But it seems impossible in this way to account for sense 1, which is apparently the source of the other senses, and of the verb itself. If related at all to bore v.1 or bore n.1, the connection must be much more indirect; possibly there is an allusion to some now forgotten anecdote. The phrase ‘French bore’ naturally suggests that the word is of French origin; bourre padding, hence (in 18th cent.) triviality, bourrer to stuff, to satiate, might be thought of; but without assuming some intermediate link these words do not quite yield the required sense.

    The sense 1 mentioned is:

    1. a. The malady of ennui, supposed to be specifically ‘French’, as ‘the spleen’ was supposed to be English; a fit of ennui or sulks; a dull time.
    b. One who suffers from ‘bore’ or ennui, or affects lack of interest in anything.

    The entry may have been sitting untouched since 1887; there are not attestations for either 1. a. or 1. b. except from the years 1766–1767.

  66. David Marjanović says

    Nice example of using the word in the phrase without actually distinguishing the meaning.

    I was trying to quote the title of a conference poster, which started with BORING MOLLUSKS, exhibited at the University of Vienna. I had to think twice about what it meant… I’m not sure if I had encountered that sense of bore before, actually.

    compare German drillen

    I’m not aware of any meaning like “annoy” of that word, and neither are the DWDS or Wiktionary in German or in English.

  67. @David Marjanović: Yeah, the mention of drillen looked funny to me, but I am certainly not acquainted with the full range of colloquialisms in German. As I said, the entry looks like it may be very old, so the reference to drillen may to an obscure and extinct nineteenth-century usage, or it may simply be an uncorrected error.

  68. https://www.dwds.de/wb/dwb/drillen#drillen,

    7 quälen, plagen, belästigen, foppen, necken, ebenso im engl. schwed. und dän.

  69. David Marjanović says

    Ah. I’m not surprised the ¹DWB is where that’s documented.

  70. John Cowan says

    “The borer on our peach-trees bores that she may deposit an egg; but the borer into theories and institutions and books, bores that he may bore.” —Ralph Waldo Emerson

  71. Lars Mathiesen says

    Drille is indeed the common term for ‘to tease’ in Danish, but we don’t use it for boring.

  72. Lars Mathiesen says

    It was remiss of me to omit the fact that E bore is in fact Da bore, but the other one is kede.

Speak Your Mind

*