Joel of Far Outliers has posted an excerpt from A Labyrinth of Kingdoms: 10,000 Miles through Islamic Africa, by Steve Kemper, about the history of Agadez; I’m going to quote the part of linguistic interest:

Like most port towns, Agadez had a mongrel population that reflected all the peoples who passed through it, beginning with the Berber tribes that had founded it. There were Tuaregs, Hausas, Fulanis, Tebus, Kanuris, and Arabs. And also, Barth was puzzled to find, Songhais, a black ethnic group based 600 harsh miles to the west. All this diversity made Agadez a polyglot town where interpreters did good business.

But Agadez also had its own unique language, Emgedesi, spoken nowhere else in the region. To a linguist such as Barth, this was a mystery to pursue. He detected the influences of Hausa, Tamasheq, and Songhai in Emgedesi, but remained puzzled about the dialect’s origins and exclusivity to Agadez. Then came the clue that connected the dots: several Tuaregs who had been to Timbuktu told him that Emgedesi was also spoken there, 800 miles west. Barth was surprised, then thrilled as he realized the implications.

Songhai had been the most extensive empire in Central Africa’s history, greater than Mali or Ghana. It had covered portions of present-day Mali, Burkina Faso, Guinea, Senegal, and Niger. Songhai had conquered Timbuktu, another Sahelian port city of Tuaregs and Arabs. The language of the conquerors mixed with Timbuktu’s other tongues, creating a distinctive language unique to the town.

Then early in the sixteenth century, Askia, Songhai’s king, decided to extend his realm to the east, into central Sudan and Hausaland, and to curb the pesky Tuaregs to the north. He conquered Agadez in 1515 and left an occupying force there before proceeding on a haj through Egypt to Mecca, scattering legendary amounts of gold in his wake.

By the end of the sixteenth century the empire of Songhai had disintegrated. But in Agadez the descendants of the occupying army had melded with the local population. So had their language, and the resulting hybrid dialect evolved along similar linguistic lines as the hybrid language of Timbuktu, like related bird species on separate islands. This link, wrote Barth, “throws a new light over the history and ethnography of this part of the world,” and is “of the highest importance for the whole ethnography of North Africa.” It also gave him his first whiff of the fabled city of Timbuktu, a place he never expected to see.

“Emgedesi” should be Emgedeši, or in Anglicized form Emgedeshi [per Lameen’s comment below, the accepted English spelling is actually Emghedesie]; Maarten Kossmann, in “On relative clauses in Northern Songhay: Tuareg and Songhay components,” lists it as one of the five varieties of Northern Songhay, calling it “Emgedeshi, the now-extinct language of the city of Agadez (Niger), which seems to have been very similar to Tasawaq.” I’m guessing the name is somehow derived from the city name Agadez, but I’d love to know how. As always, I enjoy learning about obscure and forgotten forms of language.


  1. David Eddyshaw says

    the resulting hybrid dialect evolved along similar linguistic lines as the hybrid language of Timbuktu

    No, not really.
    I feel (appropriately) diffident in commenting on this when Lameen is around as an actual bona fide five-star expert on this very matter, but:

    The comments about Timbuktu Songhay are highly misleading. The only full-dress account I’ve read of this language is the Mouton Grammar Library grammar by Jeffrey Heath; he recounts in the introduction that he expected it to be basically a creolised version of Gao-type Songhay, and there are indeed a good number of reasons to think along those lines, like the loss of tone and the adoption of SVO word order. However, he discovered that there are evidently old isoglosses separating the western Songhay languages from the eastern, and that the eastern forms are not consistently more conservative by any means. He ended up feeling that the data could not be accounted for by a once-off “big bang” creolisation but must reflect a much more complex history of long term interaction between different Songhay languages along with Fulfulde, Tuareg etc.

    Timbuktu Songhay is at least not in any meaningful sense a “hybrid” language, and is not at all like the extraordinary northern Songhay languages, which seem to have been put on this earth specifically to refute all simple theories of creolisation whatsoever.

    Robert Nicolai, Grand Old Man of Songhay linguistics, thinks the entire family derives from some sort of Tamasheq-lexifier creole, but the idea doesn’t seem to have acquired much traction. The less peculiar Songhay languages do have an awful lot in common structurally with Mande, though.


  2. David Eddyshaw says

    Incidentally, the Songhay Empire led to some Songhay loanwords travelling far afield in West Africa: the most notable probably being the word for “noble person”, which (via Mooré) ended up as the first component in the name “Burkina Faso”, and percolated even as far as Yoruba bọrọkinni “gentleman.”

  3. Wow, that’s fascinating!

    *sends up Lameensignal*

  4. While we are on the subject, where do the forms like “Songhay”, “Songhai”, and even “Sonrhaï” come from? The native forms in mainstream Songhay seem to be [soŋaj] or [soŋoj], so is there a reason that the familiar spellings seem to indicate the presence of some back fricative or approximant?

  5. I’ve often wondered about that.

  6. John Cowan says

    I have always automatically segmented it Song-hai, a position in which /h/ would be easily dropped in English. But it seems that it was originally soŋay ‘the Imperial ruling class’ in some Songhay language. That was borrowed into Arabic, where [ŋ] exists only as an allophone of /n/ before /k/. Since there was plainly no stop in the word, Arabs must have heard it as /nɣ/, and this accounts for the /nr/ in languages were /r/ is [ʁ]. Languages without voiced back fricatives probably brought it in from Arabic as /ŋg/, simplified back to /ŋ/ in English and German.

  7. David Marjanović says

    rh is a newfangled French representation of Arabic /ʁ/.

    German does not drop foreign h in unstressed syllable-initial positions, preferring instead to not reduce the e in Sahel. I was quite surprised by Stephen Colbert dropping it, consistently, in Cohen (which more or less becomes cone), Graham and Ingraham (two syllables: [ˈɪŋgɹəm]) – no ham for you!

  8. David Eddyshaw says

    Arabic does indeed haveسونغاي , and French spellings like Sonr(h)aï are out there too.
    I had forgotten also to say that Heath discovered north-south isoglosses cutting across the west-east division between Timbuktu-like and Gao-like Songhay, which is more pertinent to the point that the matter can’t be reduced to a simple notion of creolisation.

  9. Jen in Edinburgh says

    I’ve never heard ‘Graham’ pronounced with an h, although it does have two syllables (unless you’re Very English).

    But I’ve never seen ‘Ingraham’ at all, only Ingram.

  10. John Cowan says

    Laura Ingraham is a reactionary American TV talk show host. I neither know nor wish to know how her name is pronounced; the Wikipedia article was more than enough for me. The etymology is at least well established: from the personal name Ingelram < Ængle ‘Angle’+ hramn ‘raven’. It got into French as Enguerrand, borne by several of the mediaeval French nobility, notably including many of the Lords of Coucy, who were very proud of not having any higher title: “Je suis ni roi, ni prince aussi / Je suis le seigneur de Coucy!” (modernized).

    The etymology of Graham is more obscure: it may be greg ‘gray’ + hama ‘cloak’, as in Greyhame, Gandalf’s nickname in Rohan, or it may be a by-form of Grantham.

  11. David Marjanović says

    Ah, so the h in Ingraham is hypercorrect in the first place. Shades of yellowhammer.

  12. Since when is Songhai in Central Africa? Maybe if your point of reference is Morocco?

  13. David Eddyshaw says

    I think he means “Subsaharan Africa” (itself a pretty silly term, but at least people know what you mean by it.)

    On the name soŋay/soŋoy, Heath says:

    … it is part of a set of terms for patrilineal clans or castes, each of which was associated with particular occupations, rituals, and customs […] Soŋoy was associated chiefly with the original Songhay-speaking group which founded Gao and Hombori, and lost a crucial battle with an invading Moroccan army in the late Middle Ages which spelled the end of the Songhay Empire and remains the subject of popular legends. Currently soŋoy in this limited sense is associated with the patronymic meyga (Gallicized as Maiga), and more particularly with those Maiga who continue practicing sorcery and other traditional practices frowned on by Islam. Currently, under the influence of French […] soŋoy is increasingly used in the French sense as a general language name and ethnic label.
    The descendants of the Moroccan soldiers […] are called arma, and associated with the French patronymic Touré.

  14. David Eddyshaw says

    There isn’t any velar fricative in the soŋay/soŋoy etymon in any of the Songhay languages I’ve seen descriptions of (Timbuktu, Djenné, Hombori, Gao, Kikara.) It might have been simply an attempt to write ŋ in Arabic, I guess; or maybe the sound really was ŋɣ in the relevant Songhay language at the time the Arabic transcription was first made. Lameen will know if anybody does, I imagine.

  15. Thanks for the answers on the origin of the “h” in “Songhay”. John Cowan’s explanation makes sense. I did suspect some sort of Arabic involvement, especially as I know French speakers who know Arabic will often pronounce غ gh as the French r [ʁ]. It’s good to see the dots connected.

    By the way, I have to check pronunciations of topical names for a side gig with the National Institute of the Korean Language to help them standardize Korean spellings of foreign names, and I did find out that Laura Ingraham pronounces her name [ˈɪŋ.ɡrəm]. But we ended up spelling it 잉그러햄 Inggeureohaem based on the pronunciation [ˈɪŋ.ɡrə.hæm] on the logic that the former could be a reduced version of the latter and also because Bahamian politician Hubert Ingraham is already spelled 잉그러햄.

    Then again, some people are inconsistent—Lindsay Lohan has used both [ˈloʊ.ən] and [ˈloʊ.hæn].

  16. Emghedesie is fun – glad to see it getting more attention! Here are some references: Northern Songhay: Emghedesie. And here’s Barth’s original account – the Emghedesie phrases are on pp. 26-28, 41-61.

    Timbuktu Songhay is, as David notes, nowhere near as heavily Berber-influenced as Emghedesie (which is itself the least Berberized of the Northern Songhay languages). However, surprisingly and counter-intuitively given the geography, they do seem to belong to the same branch of Songhay, with a number of shared innovations: at least, that’s what I try to show in Souag 2012.

    Barth elsewhere writes the word “Songhay” as Soṅɣai, so he presumably heard it as such. A Songhay-internal explanation is conceivable: some varieties show a change of g > ɣ / _{a|o}. But I’m not aware of any recent attestation of a pronunciation /soŋɣay/ within Songhay, so an explanation in terms of Arabic remains plausible.

    For Central Africa, read Inland Africa – at the time, the coasts were rather well-known to Europeans, while the interior was still essentially inaccessible to them, so that was the big divide.

  17. David Eddyshaw says

    I’ve been overthinking it: would of course be the obvious way to write /ŋ/ in any form of Arabic which lacks [g]. I know nothing to speak of about spoken Arabic in West Africa, but certainly the literary sort used by Islamic scholars has the [d͡ʒ] rendering of ج (or whatever the nearest equivalent is in the local vernacular.)

  18. As for the name’s derivation; checking Ritter’s monumental dictionary of Tuareg, it seems to be an Arabic nisba with -ī or -iyyah formed from the Tuareg gentilic Emăggădăz “inhabitant of Agadez”.

  19. Thanks, Lameen!

  20. David Eddyshaw says

    Thanks indeed for linking to your 2012 paper, Lameen. Very interesting. The idea of Timbuktu/Djenné Songhay being heavily decreolised is fascinating, and makes a lot of sense.

    Seems that Kemper may have been less wrong than I thought, too (although still pretty wrong …)

    Very interested (naturally) in the Gur connections. I’m not sure that I believe even in “Central Gur” as a thing. Still, the materials for reconstructing Proto-Oti-Volta at least are vastly more copious and reliable nowadays than when Manessy was pursuing these issues. It’s all there just waiting for somebody with the right skills and background and energy to take it on.

  21. David Eddyshaw says

    While Askia Muhammad may very well have been liberal with his gold when on the Hajj, “legendary” makes me wonder whether Kemper is getting him confused with Mansa Musa.

  22. George Grady says

    Like most port towns, Agadez had a mongrel population that reflected all the peoples who passed through it, beginning with the Berber tribes that had founded it.

    After reading this, I was mildly surprised when I followed the Wikipedia link to see that Agadez is, in fact, very much not a port town. I realized that’s because, when reading it, I had mentally stressed the “most” in “like most port towns”, rather than the “port”. I hadn’t ever noticed before that this sort of construction was ambiguous in writing.

  23. No, you read it correctly the first time; here’s the first paragraph of the passage in Joel’s post:

    In the fourteenth century the restless Moroccan traveler Ibn Battuta called Agadez “the largest, handsomest, and strongest of all the cities in Negroland.” In Battuta’s day 30,000 people lived there. It flourished as a caravan crossroads, where the Sahara met the Sahel, a band of semiarid land 300 to 600 miles wide that stretches for 2,600 miles along the Sahara’s southern edge and buffers the desert from green Africa. “Sahel” came from an Arabic word for shore or coastline. The sea was the Sahara. When travelers from the north reached the Sahel after crossing the desert, they felt the relief of stepping ashore after a long sea passage. Travelers heading north from the Sahel felt that they were casting off. Agadez, like Timbuktu, was a desert port town.

  24. David Eddyshaw says

    Ship of the desert …

    Astonishly, camels seem to be American megafauna. The world is truly a strange and wonderful place.

    “Sahel” came from an Arabic word for shore or coastline

    As in “Swahili” (from the plural.)

  25. David Eddyshaw says

    The West African words for “camel” are hardly less interesting: most (including Kusaal yʋgʋm and Hausa raƙumi) seem to be ultimately from Berber, and the Berber itself may be connected (via metathesis) with the Semitic words for “camel”, and thus with the very word “camel.”

    (Here I am deliberately Lameen-baiting.)

  26. George Grady says

    languagehat: Ah, thanks.

  27. David Marjanović says

    Astonishly, camels seem to be American megafauna. The world is truly a strange and wonderful place.

    Rhinos, too, are fundamentally North American. It’s not just horses!

  28. John Cowan says

    very much not a port town

    Well, in the U.S. at least the inland marine is very much a thing, and it doesn’t refer to lacustrine or riverine traffic, either.

  29. not just horses!

    Your Fido ultimately harks back to America, too:


  30. The puzzling thing is that, while both Manding and Hausa use Berber loans for “camel”, the Songhay term, *yoo, has no plausible Berber source.

    The idea of metathesis to link alghem with camel has been proposed, but not convincingly demonstrated. One would expect a loan for the camel, given its history, but it could just as well derive from a repurposed native word.

  31. David Eddyshaw says

    The Western Oti-Volta “camel” words seem pretty clearly to go back to Berber too; the more southerly ones have adopted the Hausa word (e.g. Dagbani laakumi) whereas the northern languages have forms like Kusaal yʋgʋm, where y- goes back to Proto-Oti-Volta *ʎ- (cf also Koromfe logomde “camel.”)

    (This Proto-Oti-Volta *ʎ is my own invention/discovery, and not part of Manessy’s reconstructions, but the evidence for it seems pretty solid: I really must get round to writing it all up properly.)

  32. I really must get round to writing it all up properly.

    The combined voice of the Hattery commands: get to work, man!

  33. This past summer I ate at a strangely exotic little ramshackle restaurant near Takadanobaba station in Tokyo that specialized in wild game meats. A Japanese linguistic fieldworker friend took me there. They had camel on the menu, but we chose horse, deer, and goat for our sashimi course and kangaroo, emu, and crocodile for our grilled course. (The crocodile tasted like chicken, but the emu didn’t.)

    On one menu describing the game animals, ‘kangaroo’ was written in kanji (ateji) as 長尾驢 ‘long-tail-ass’, which we at first confused with ‘long-tail-camel’, but the last kanji is the ro of 驢馬 roba ‘ass-equine’ and was probably chosen for its similarity to the -roo of kangaroo as much as for the kangaroo’s similarity to donkeys.

    The Japanese term for camel, rakuda, is borrowed from Chinese 駱駝 luòtuó, which seems to be descriptive in origin: 駱 ‘white horse’ (in Japanese) + 駝 ‘hunchback, humpback’ or ‘load (on top of animal)’. There’s another luó ‘mule’ with different tone and hanzi. So ‘camel’ in the Sinographic world seems to derive from something like ‘humpbacked equine’.

  34. January First-of-May says

    The Russian/Slavic word for “camel”, верблюд, is somewhat infamously said to come from Latin elephantos “elephant” via Gothic ulbandus. (The connection is clearer in Polish wielbłąd.)

    …Or, at least, that is the traditional theory; apparently some modern linguists suspect that the Gothic word might have actually been a borrowing from an Anatolian (!!!) language (with the original meaning “hump-backed”, which is apparently attested in Hittite, but not in reference to camels).

  35. David Eddyshaw says
  36. The Russian pronunciation is quite good, but the grammar leaves something to be desired: “караван верблюд” (intended to mean ‘caravan of camels’) should be караван верблюдов (genitive plural).

  37. January First-of-May says

    should be караван верблюдов (genitive plural)

    In addition, as far as I can tell, near the end he’s saying жалко ему (which would literally mean “he’s sorry”) where I would have said жалко его (“feel sorry for him”).
    Though I’m not entirely sure what the intended reading actually is in this particular case, if any (and I might well have misheard).

    But yes, караван верблюдов (second syllable stress in the second word, as in the nominative) is correct.

    Of course, the version of this apparently popular and varied folk song that I recall actually has (incorrect) final stress in this word:

    Шёл один верблюд,
    Шёл второй верблюд,
    Шёл целый караван верблюдОв…

    (This does not appear to be on YouTube – or if it is, I’m using the wrong keywords.)

  38. This one just has целый караван, with no following camels.

  39. January First-of-May says

    …On further thought, I’m not entirely sure if it was второй верблюд or другой верблюд in the second line. I think the latter would actually make more sense?

    I’m fairly confident of the third line, though (complete with the incorrect final stress).

  40. Chinese 駱駝 luòtuó is one of the few words borrowed by Chinese from the language of Asian Huns (Xiongnu).

    So is Japanese rakuda, obviously.

  41. David Eddyshaw says

    Donald Swann’s parents (despite his being a veritable incarnation of Englishness) were a Russian and a Turkmen (his middle name was Ibrahim); I don’t know if he grew up speaking Russian at home (I suspect not), but he did study it at Oxford.

  42. This page has lyrics; I note they’ve corrected верблюд to верблюдов in one case but not the other. Also, that’s quite astonishing about Swann’s background.

  43. Also, they have “Óчень жа́лко э́то” for what is clearly “Очень жалко его.”

  44. Sadly, I can’t find any online video/audio of the Russian song, which was apparently very popular in the 1960s.

  45. Swann’s parents (despite his being a veritable incarnation of Englishness) were a Russian and a Turkmen

    This is odd and also common. Think of Peter Ustinov and Johnson the current PM (0,25 Turkish), also a Hungarian I was at school with and a Polish friend of my mother’s who liked to go grouse shooting in the rain at weekends, but no one here knows them.

  46. David Marjanović says

    In the rain! 😮

  47. January First-of-May says

    Sadly, I can’t find any online video/audio of the Russian song

    I was able to find it, but only on very shady sites. It really does not appear to be on YouTube.

    (Again, it probably is there and I’m just using the wrong keywords.)

    Also, they have “Óчень жа́лко э́то” for what is clearly “Очень жалко его.”

    I don’t hear either of those two – I hear “Очень жалко ему” (which doesn’t make as much sense in context).
    That said, the last word is not very clear; I would not be especially surprised if it’s actually none of those three but something completely different entirely.

  48. I don’t think he preferred the rain; being England, it just usually was (I was taken along as a beater, only once and that was enough).

  49. That’s quite a review, and delightfully written (“He fills out on screen as the audience thins out”).

  50. Yes, he’s good! I thought so too.

  51. PlasticPaddy says

    Re ajp crown link:
    “cobbled together from survivor’s accounts” is overly charitable (and the author is unlikely to sue for libel, since he committed suicide after being accused of plagiarism, as I think is discussed on another thread).

  52. Starlings are perfectly happy to murder their recognized conspecifics, with no misleading paint job necessary.

  53. PlasticPaddy says

    @AJP Crown
    I just realised the author (I thought he would have been too old to be your friend) might have been your mother’s friend. If so, I was overly uncharitable and am genuinely sorry.

  54. I’ve only learned about Jerzy Kosiński on this thread.

    But it turns out, I was only four degrees of separation from him.

    Me -> My grandfather -> marshal Budenny -> Svetlana Allilueva -> Jerzy Kosinski

    It also puts me in five degrees from Tate murders victims and in six degrees from Manson family.

  55. You’re very kind, Plastic. No, he wasn’t my friend nor my mother’s; his last name was Romer and he had a wife called Mrs Romer.

  56. On that Proto-Oti-Volta *ʎ, by the way: that’s actually very interesting. I would normally expect a Gur form to be borrowed from Mande. But the Mande forms reflect initial palatal n, which is presumably derived from a Berber variety which, like Zenaga, had palatalized the l; Proto-Oti-Volta would thus actually be more conservative than can be reconstructed for Mande in this regard. (Cf. Kossmann’s “Berber Loanwords in Hausa”, p. 51.)

  57. David Eddyshaw says

    As I say, this *ʎ is not part of the standard inventory implied by Manessy’s work, but I think it’s justifiable. The only Oti-Volta language in which it remains as a completely distinct phoneme is Nawdm, which has r throughout (except that rr -> d, as an active synchronic rule.) However, its merger with various /r/ /l/ and /j/ segments in the daughter languages must be a fairly late development in most cases, as the outcomes differ even in quite closely related languages.

    In Western Oti-Volta it became *y [j] everywhere except except when geminated or between a long vowel or a consonant and *i, where it became *r; this *r is itself preserved as a distinct phoneme only in Mooré and Agolle Kusaal (not even in Toende Kusaal); it has different outcomes even in Mampruli (r) and Dagbani (l), despite those languages being so close as to be partly mutually intelligible.

    My use of the ʎ symbol is of course just a guess at the putative sound in the protolanguage; it might have been some sort of palatalised rhotic. It certainly wasn’t nasalised, at any rate.

    Leaving aside such speculative stuff, assuming that Kusaal yʋgʋm, Mooré yʋgemde, Farefare yʋgnɛ (pl yʋgma) really are from the Berber etymon in question at all, they definitely don’t show any trace of original nasalisation (all three languages would have nasalised the first vowel if the original consonant had been *ɲ.)

    I had wondered about the Mande forms: thanks for the explanation!

    I put a bit of justification of Proto-Oti-Volta *ʎ in previous versions of my Kusaal grammar (it has some bearing on the synchronic variation in form of CV/CVV roots) but eventually decided that although it was all doubtless very interesting it was not contributing much descriptively and needed to be treated more rigorously anyway. I really will put up a proper treatment of the issue at some stage, honest.

  58. David Eddyshaw says

    The l/r divergence between Dagbani and Mampruli must be later than the adoption of the Songhay “freeborn” word, incidentally: Dagbani bilichina, Mampruli birikyina.

    I should clarify that Mampruli has merged Proto-Western-Oti-Volta *r with *d; the actual reflex of both is [ɾ] (written r) word-internally after vowels, and [d] elsewhere.

  59. David Eddyshaw says

    There’s actually something odd about Western-Oti-Volta *l itself; although /l/ corresponds straightforwardly in all the languages (except where Dagbani l is from *r), there is a puzzling rule ld -> nd, which applies with great regularity in flexion of both nouns and verbs throughout the group. I have wondered whether this might mean that *l was originally nasalised, but contrastively nasalised l-sounds seem to be rare to non-existent cross-linguistically. There do seem to be some cognates with /n/ in other Oti-Volta languages (as with the words for “egg” and “horn”) but I haven’t looked at it properly.

    If the original Proto-Oti-Volta contrast really were of plain vs nasalised l, it would be easy to imagine it being unstable enough to drive all sorts of divergent developments in the daughter languages, at any rate.

Speak Your Mind