Nelson Goering posted on Facebook as follows (I’ve added links and italics):
Roan eats a lot of banana these days, and as is inevitable in such circumstances, we got to talking about the word “banana”. English Wiktionary claims it goes back either to Wolof banaana, or to a similar word in a related language, but goes no further. Le Trésor de la langue française informatisé says it’s probably from “le bantou de Guinée” (the Bantu of Guinea): since Guinea doesn’t, as far as I know, have any native Bantu languages, I guess this either means Equatorial Guinea, or is using “bantou” in a very broad sense to refer to the larger Atlantic-Congo family.
I’m curious if anyone I know here has any further light to shed, both whether the Wolof/Atlantic-Congo origin is regarded as likely, and if so, if there’s anything to say about the history of the word *within* Atlantic-Congo.
I responded “There are a number of Africanists at LH, so I’ll post this and see what they have to say,” and I am now so posting it. Thoughts? (Forget the OED — its entry is from 1885.)
The Ghanaian Oti-Volta languages all borrow the Twi kwadu, not surprisingly, as does Moba, and even the relatively distant Mbelime.
Mooré, on the other hand, has bànándè, probably a back-formation (sound familiar?) from the plural banana; Gulimancema has (probably) borrowed the first element of its báanáantìbu “banana tree” from Mooré (the SIL dictionary does not give a word for “banana” as such.) Nawdm has àyàbà, which is obviously from Hausa in the first instance; I think Hausa got it from Yoruba. (I’m away from my books at present.)
Niggli says “emprunt français” for Mooré bànándè, but he’s often wrong about the source of loans. Dyula has bàràndá; a lot of Wanderwörter got to Mooré via Dyula, but I must say that in this case French does look more likely.
Assuming that the Wolof word exists and is not itself from Portuguese, Wolof looks as likely a source as any for “banana” to me.
To describe “banana” as “from an Atlantic-Congo language” is like saying “weekend” is “from an Indo-European language.” (Or “from Nostratic”, even.) It means nothing (even in the event that it might be technically true.)
References to “the Bantu of Guinea” are helpful inasmuch as they show that the writer has no idea what they’re talking about and are just bloviating.
Didn’t we discuss this before? (Generally a safe bet at LH.)
Another possibility is that French “Guinee” much like English “Guinea” previously described a much larger if vaguely defined region of coastal Africa adjoining the shores of le https://fr.wikipedia.org/wiki/Golfe_de_Guin%C3%A9e. There’s some Bantu toward the southern end of that coast. Although Wolof is at the other end …
https://en.wikipedia.org/wiki/Canaan_Banana
—
Banana was a controversial figure, especially after his criminal conviction. As president, he did not always command respect (a law was passed in 1982, banning Zimbabweans from joking about his surname).
—
Apologies if this has come up before.
Didn’t we discuss this before? (Generally a safe bet at LH.)
[2025] Banana is supposed to have come to Portuguese on loan from a West African language …
[2020] The word banana is said to have come to Spanish and Portuguese from Wolof. How did it get there?
[2012] Actually, Trond, there is a more northerly dialect, generally known as Banana-nana Fofana.
Assuming that Wolof banaana is not itself a loan, tracing its origin would be hard. Contrary to the overoptimistic statements of many of those who have attempted comparative work on the “Atlantic” languages, Wolof is not particularly close to other supposedly “Senegambian” languages. One of the few who have applied genuine rigorous comparative methods to languages in this grouping is John Merrill; in his 2018 thesis, “The Historical Origin of Consonant Mutation in the Atlantic Languages” he says
(I’m fairly sure that I’ve quoted this before, too, in service of casting shade on the as-yet-still-mainstream idea that Atlantic-Congo is a demonstrated real thing.)
Didn’t we discuss this before? (Generally a safe bet at LH.)
I had the same feeling, but couldn’t find the discussions; a tip of the Languagehat hat to Stu for doing so!
@JWB:
“Guinea” did indeed originally mean the part of West Africa south of the savanna zone (“West and Central Sudan”), and certainly, if “banana” really is originally from a West African language, it must be from one spoken in this region, as bananas are no more native to the savanna than they are to Europe. (And echt Bantu languages do indeed manage to creep in to the Guinea Zone at its Eastern margin.)
But, unhelpfully for the hypothesis, most of the Guinea-zone languages have words for “banana” which are nothing like “banana” at all (or even much like each other, mostly.) Wolof seems to be about the only exception (which looks a bit suspicious in itself.)
I see that Bété has batɛ, to add to the parade of the unbananalike Twi kwadu, Ewe akɔɖu*, Yoruba ọ̀gẹ̀dẹ̀ etc etc.
I must admit that there is a bit of a danger of circular argument here, as any real bananalike word that actually turns up could be falsely accused of being a mere loan from French or English or Portuguese. But ah hae ma doots; ah hae ma doots.
* A much more likely source of the Nawdm and Mbelime words than Twi, on reflection.
Rather belately, it’s occurred to me that bananas are not actually native to Africa anyway: which surely increases the likelihood that this Wolof banaana is actually a loan. Though that also raises the question of where all the nonbanana West African words came from. Repurposed words for “plantain”, presumably …
https://en.wikipedia.org/wiki/Musa_(genus)
But I’ve no idea how long banana-bananas have been in West Africa. Calling ROGER BLENCH (again …) Unfortunately, the paper I previously linked to by him seems to have disappeared along with his website.
His very long academia.edu page is still up.
We did the banana and the plantains, their cousin the enset, Wolof, Blench and all, back in 2020/21 (Direct link to the first banana in the thread linked by Stu above).
As my wife remembers vividly, the itinerant banana-sellers who used to come to our compound in Ghana would call out
Da’ kɔdʋ! Da’ kɔdʋ!
I was never able to parse this. Da’ is “buy” in Kusaal, and kɔdʋ is “banana”; but if it means “Buy banana(s)!”, you’d expect the imperative: Da’am kɔdʋ!
Unfortunately, it never occurred to me at the time to interrogate a banana-seller on this point.
On reflection, they were probably just being polite, and saying
M sɔsi ya ye ya da’ kɔdʋ.
I ask you-pl that you-pl buy banana
“I entreat you to buy bananas.”
with ellipsis of the main clause and the non-salient bits of the subordinate clause.
After all, it would be rude to command potential customers to buy your bananas.
Obviously there is further research to be done on the language of Kusaal street cries.
Wiki claims Magellan’s expedition encountered bananas in the Philippines (or maybe Marianas or maybe both) but the chronicler didn’t have a word for them so he described them in Portuguese as weirdly long figs. But the Portuguese were traveling to all sorts of places at similar times and presumably veterans of the different expeditions sometimes compared notes and culinary anecdotes and newly-acquired lexemes back in Lisbon …
The Malay banana-word “pisang” was brought to Europe in Dutch (where it now allegedly sounds “dated” but maybe “piesang” is still current in Afrikaans) and also supposedly found its way into Hobson-Jobson and various Scandinavian languages. The word in many Filipino languages is “saging” or something that looks similar to that, but I’m not sure if that lexeme got exported. Japanese has its own name for the so-called Japanese banana (Musa basjoo, fruit is inedible), which has long been cultivated in the Ryukyus, but seems to just use katakanified “banana” for sort you eat.
Oh and here’s an interesting banana-related cryptolect from Taiwan: https://en.wikipedia.org/wiki/Banana_colloquial_speech
G. Banane as defined in J. C. A. Heyse’s Allgemeines verdeutschendes und erklärendes Fremdwörterbuch, 1807, quoted in the 9th edition of Hermann Paul’s Deutsches Wörterbuch (1992):
— so it was still compared to a fig in Germany at the beginning of the 19th century.
The first attestation of banana everyone still cites is from Garcia de Orta (1563) Colóquios dos Simples e Drogas e Coisas Medicinais da Índia, available here (Colóquio 22; middle of the page in image file no. 210 at that site). Garcia de Orta describes bananas as a type of fig:
Garcia de Orta was resident in Goa for 30 years and had made the voyage down the Atlantic coast of Africa in 1534. He seems like an upright, serious kind of guy who wouldn’t just make stuff up or fob off to his reader any old story he considered dubious. For Garcia de Orta’s quelli, cf. Konkani केळें keḷẽ, Marathi केळे keḷe, Gujarati કેળું keḷũ, Hindustani کیلا केला kelā, etc.; for palã, Malayalam പഴം paḻaṁ; for piçã, Malay ڤيسڠ pisang; for musa, Arabic موزة mauzah (singulative of موز mauz). The question remains, what did de Orta mean by Guiné ?
For decades I have been charmed by the idea that banana is the singulative بنانة banānah of بنان banān ‘fingertips, fingers’ (cf. the expression a hand of bananas), picked up by the Portuguese somewhere and spread around the world. But where? The Arabic of northwestern Africa during Portugal’s various incursions into Morocco? The Arabic of the Indian Ocean traders on the eastern coast of Africa? But I could never find any evidence that would back any of this up. Garcia de Orta says that he found the bananas at Sofala in Mozambique to be especially delicious: e de muito bom sabor os de Çoffala ja os prouei, sã muyto gabados, eu os achei de bõ sabor (that would have been in 1534).
The other early attestation (1591) of banana that everyone cites is here, page 41 of the text (page 59 of the pdf), from Filippo Pigafetta, relaying information from a Portuguese trader, Duarte Lopes (a somewhat obscure figure), who had been in kingdom of Kongo for 12 years. (Searching on the word banana in the pdf will find it.)
@DE: I’ll just put this bit of Danish banana rock here.
https://www.youtube.com/watch?v=_cNRu80wCck
Banana cobs.
That is a lovely idea, Xerîb, and it would parallel the history of musa. When and how did the latter enter European languages?
The name Pigafetta always makes me think of “Mrs. Pigafetta Swims Well.”
“To describe “banana” as “from an Atlantic-Congo language” is like saying “weekend” is “from an Indo-European language.” (Or “from Nostratic”, even.) It means nothing (even in the event that it might be technically true.)”
Those are rather different cases! To say a loan is from an Indo-European language (let’s leave aside weekend, which is a late compound) can certainly be meaningful, and even helpful. If you’re a bit more cautious about attributing everything in Eurasia to Tocharian, it’s perfectly reasonable to suggest that Finnish mesi is a loan from an Indo-European language, and quite useful too for pursuing the further etymology of the word.
From your later comments, it sounds like we’re closer to Nostratic territory with Atlantic-Congo, which is interesting to learn. Though disappointing, since it sounds like modern Wolof might indeed be the end of the road. I’d naively assumed that one could at least back-project a modern Wolof word into a hypothetical earlier form, just as one could do with, say end, which if inherited, would phonologically have to go back to something like *andī̆- (comparative data specifies the exact stem class more precisely, of course).
Wolof really is related to at least some other “Atlantic” languages, and TBH I know much less about this than I do about Volta-Congo. But what I have read on attempted reconstruction at the level of the supposed top-level branches of Atlantic is notable for pretty gruesome methodological problems. Lower-level stuff (as you might expect) is often a lot better, and I find Merrill particularly good. The Atlanticists are learning that you need to learn to walk before you can run.
Though, as with “Khoisan”, actual specialists in the languages involved have all along tended to be a lot more sceptical about the high-level relationships than Greenbergian uberlumpers are. Way back in the eighties, W A A Wilson reckoned (using the voodoo science of lexicostatistics, but even so) that the branches of “Atlantic” were as remote from one another and as isolated within “Niger-Congo” as Ijoid (nowadays often regarded as an isolate.) The fact that the grouping-previously-known-as-Atlantic is, at the very least, not actually a single branch within “Atlantic-Congo” now seems to be generally accepted.
My guess is that at least parts of Atlantic really are related to Volta-Congo, but at a level where convincing demonstration by proper comparative methods is never going to be possible. But I think the tide is turning on all this lumpery: people seem often to express scepticism about Kru being part of “Niger-Congo” now, for example, and frankly there are better reasons for including Kru than any part of Atlantic.
But Greenberg’s zombie groupings are hard to kill off. “Ubangian”, on the other side of the supposed phylum, is another grouping which is long past its sell-by date. Far from being a genuine subgroup of “Niger-Congo”, it’s a Frankenstein’s Monster of some Volta-Congo languages along with several quite unrelated groups.
A lot of traditional subgrouping even within pukka Volta-Congo is really just based on vibes, too. There is no “Gur” …(at least, not as Manessy understood it, and as Greenberg did.)
Of course, all that is part of what makes comparative work on “Niger-Congo” so interesting. There is still everything left to play for (and vastly better and more copious data to play with than in Greenberg’s day.)
And once you shake off the illusion that Greenberg’s classification was in any way definitive, then you can get properly started.
Thanks for all that — it’s always interesting to learn where things are with languages/families/non-families I know (much) less about!
Pisang vel sim. was indeed a thing in Scandinavia, but it was before living memory. SAOB has the latest attestation in 1815 (apart from a dictionary entry in 1911), ODS in 1874. (And as a synonym in a lexicon of colonial goods in 1929).
We say banan now.
people seem often to express scepticism about Kru being part of “Niger-Congo” now, for example, and frankly there are better reasons for including Kru than any part of Atlantic.
Via a post yesterday by Mair at LLog: in a story at Sachan Journal on How Minnesotans are preserving six rare languages (five immigrant languages, one indigenous), one of them is a Kru speaker, who now teaches the language via Zoom to other adults and children. The journalist says Kru “is part of the Niger-Congo language family”, but I don’t mind that — at least she looked something up. I mind a bit that she calls it “the Kru dialect”. Also that the OED’s entry for Kru, revised December 2025, defines it as “A Niger–Congo linguistic group” without mentioning any dispute.
(The story has video clips of each of the six interviewees speaking their language with English subtitles.)
I wouldn’t have expected the OED to have caught up yet (if it ever does …) They don’t seem to currently have the advice of anybody with much of an interest in African languages, even, let alone specialised knowledge of the state of the art in comparative work.
While I’m being all splittery: though I long since realised that Greenberg’s “Ubangian” is Not A Thing, I only recently realised that his “Adamawa” is at least as bad, if not worse. Parts of it are undoubtedly Volta-Congo (like Tula-Waja and Leko); others show really no sign of any such relationship at all. (I started looking into this in more detail after reading the late Stefan Elders’ nice grammar of Mundang, one of the really-not-Volta-Congo kind. If it’s related at all to Volta-Congo, it’s at least as remotely as any part of “Atlantic.” There’s no question of it being part of a single real branch of Volta-Congo along with Tula-Waja or Leko.)
Greenberg’s whole Adamawa-Ubangi thing is a real unholy mess. A standing indictment of the flakiness of his language-classification methods. Right up there with his “Amerind.”
I only recently realised that his “Adamawa” is at least as bad, if not worse.
I think you might find some things to dislike in Pozdniakov and Vydrin’s recent Mande and Bantu in the Niger-Congo context. Besides that, what do you think of their central thesis, linking Mande and Bantu?
I think you might find some things to dislike in Pozdniakov and Vydrin’s recent Mande and Bantu in the Niger-Congo context.
Good grief, yes. An Awful Warning of the perils of trying to do comparative work without any grasp of the basic principles at all.
Frankly, comment is superfluous. (Though I will say that Bangime, which even Roger Blench thinks is an isolate, has nɛ̀ɛⁿ “four.”)
Pozdniakov has form: he’s published extensively on comparative Atlantic using similarly valueless methodology. For an antidote, I’d recommend John Merrill, who, unlike Pozdniakov, actually understands the comparative method and has done proper comparative work in Atlantic.
There’s no evidence worth a damn that Mande is related to Bantu.
Bantu is, demonstrably, related to Oti-Volta*. Any non-crank comparison of Mande with Bantu should instead work via comparison of proto-Mande with the common ancestor of Bantu and Oti-Volta. This is elementary stuff. Pozdniakov, of course, ought to be working to compare Mande with the common ancestor of “Atlantic” and Volta-Congo.
I was, incidentally, just looking at Dame Ndao’s nice account of Pepel, a Bak language of Guinea-Bissau. As usual with “Atlantic”, some striking typological similarities with Volta-Congo but precious little else to suggest an actual genetic relationship. But I noticed, alongside dɛ “eat”, dʊm “bite”, which is one of Pozdniakov’s pet words for “proving” relationships. It’s certainly pan-Volta-Congo, but I was reflecting just the other day that like the Ubiquitous Turtle, it’s odd that it should be so highly conserved when e.g. “child” or “woman” aren’t. Phonaesthetic?
* I discovered quite early that it’s fatally easy to get lookalikes between Bantu and Oti-Volta by cherry-picking from the vast array of proto-Bantu variants. Unlike P & V, I regard this as a bug rather than a feature.
This paper in a nutshell: Greenberg in, Greenberg out.
Forms like Kusaal (a)naasi “four” are indeed extremely widespread in West Africa and in Bantu. But when you turn up similarities like this in languages which show no other evidence of close (or, sometimes, any) genetic relatedness, the proper response is not “it’s statistically impossible that this is chance: this proves that the languages are genetically related after all!”
Statistically impossible it may be (though many such statements turn out to be mathematically mistaken.) But even if it is not coincidence, the correct response should be “why does this same form keep turning up in languages which show so little other evidence of being related? Should I be reconsidering my assumptions about the borrowability of words for small numbers? Has this been shown to occur elsewhere?”
I learned of it from Lameen’s Bluesky feed, and from Pystynen’s Txitter feed (@JLingPystynen, perusable through Nitter: pick a server from here, and substitute it for that other one). Lameen is cautiously encouraging, Pystynen more heartily so. Y’all play nice, now.
There’s so much wrong with this …
It’s a kind of Gish gallop …
https://en.wikipedia.org/wiki/Gish_gallop
Just picking one of so many: “bone.” They reconstruct this as PB *kʊdɪ (“regional”) and equate it with PM *xɔlɛ.
Now, as it happens, “bone” is reconstructable to proto-Volta-Congo. The actual PB form is *-kúpà (cf Swahili mfupa.) The POV form was *kṍb-; forms of the shape *kup turn up in several languages in between OV and Bantu. There’s an issue with the vowel correspondences, perhaps related to the fact the the vowel was nasalised (something lost in PB), but there’s little doubt that this is the form that should be compared with PM *xɔlɛ, not this cherry-picked “regional” form.
The lengthy list at the end is pure Mass Comparison stuff.
I wonder why Miyobe is called Miyove?
On crankery: how comapring PM with PB is different from comparing OV with PB? Otherwise, I don’t understand what people are arguing about. Either the results are sufficient to work on them or they aren’t.
It ain’t what you compare, it’s how you compare it.
In principle, there is no sin in comparing Oti-Volta and Athabaskan. But if anyone claims that their comparison has resulted in a demonstration of genetic affinity, a certain scepticism about the soundness of their methodology would be called for.
I am by no means hostile a priori to the idea that Mande might be genetically related to Volta-Congo. That would be way cool. (I hope that it really is, though I’m pretty sure that the question does not admit of a convincing answer either way through rigorous comparative methods.)
Nor do I think there’s anything at all wrong about investigating the question. What I object to is bad methodology, especially from eminent and learned academics. They should know better.
[In my day job, bad science kills and maims. The stakes in comparative linguistics are a bit less. Perhaps I should lighten up a bit …]
That frankly sounds like Proto-Bantu hasn’t actually been reconstructed at all, just a blur of forms that are old within or around Bantu, but no attempt to figure out how old exactly.
The IEists have noticed (slide 22–31) that PIE is actually in the same situation, and they still haven’t done a lot about it, though of course they have it easier than the Bantuists.
As the biologists found out around, very roughly, the same time: “nothing makes sense in evolution without a phylogeny”.
I should perhaps mention that proto-Mande is generally thought (including by Vydrin) to be of a time depth at least as great as PIE, and possibly a good bit more; both POV and PB are probably quite a bit younger than PIE, but their common ancestor must surely be at least as old as PIE. That’s before you start trying to factor in “Atlantic”, which is as internally diverse as it is, overall, different from Volta-Congo. So this Mande-Bantu thing is Altaic level stuff, or at least Nostratic level. By common agreement, there is zero morphological support for it. Zip. It’s all based on supposed lexical correspondences (along with a steadfast refusal to consider that a region currently notable for widespread extensive borrowing across multiple languages might perhaps have been notable for extensive borrowing in the past too.)
The default response to this kind of proposal should be O RLY?
Of course, P & V may wish to imply that West African languages change much more slowly than languages elsewhere …
Or they may wish to maintain that Mande and Bantu form a subgroup within Niger-Congo, thus accounting for their remarkable resemblance …
Feh.
@DM:
Very interesting slideshow (as you’d expect from the author.) Thanks.
The Bantuists are pretty clued up on issues like these, at least on a theoretical level; Guthrie was already careful to distinguish between the way comparisons of existing languages are notated, and the putative spoken last common ancestor of all the Bantu languages itself.
Current issues (among many others) are whether the weirdo Northwest Bantu languages can be made to spring from the same protolanguage at all as the “classic Bantu” languages that Guthrie and Meinhof etc mainly relied on; and the related question of just where the boundary between Bantu and not-Bantu even is, up there in Cameroon. Or if there is one …