I’m always on the lookout for useful visualizations of linguistic relationships, and Nathaniel P. “Nath” Hermosa posted one back in 2013, Philippine language relations in a map. He writes:
First things first: I am not a linguist and I am not a geographer/ cartographer. I am a physicist who is in dire need of a stress reliever. Mapping this is therapeutic while in the thick of preparing a manuscript for submission. […] This map is my rendering of data used by ref [2] where they propose an algorithm whose goal is to reconstruct the ancient proto-Austronesian language. The proto-Austronesian language gave birth to the modern Austronesian languages. Comprised of 1000 to 1200 languages, the Austronesian language family is the largest language family in the world. Geographically, it is spoken from Madagascar to the Easter Islands, and as far North as Taiwan and Hawai’i, and as far South as New Zealand. The languages of the Philippines are part of this large family.
The authors of course, tested their algorithm with what the Linguists have and they say that their reconstructed relationship between languages does not differ much from what the linguists have obtained. They just have a different goal when they study the language relationship.
I specifically chose the results of the paper above because it listed 77 languages from the Philippines, the most number in recent studies about the Austronesian languages.
There are more maps and many further details, and a bunch of comments to which he responded both in the thread and in a later post, Philippine language relations: Reply to comments. We’ve discussed the Austronesian languages a number of times, e.g. in 2014 and 2016.
Very nice. It’s challenging to make a map which shows geographical and genetic relationships together in a clear way.
I notice Inabaknon, an out of place Sama-Ajaw language (in green), on Capul island in the middle of the archipelago.
Comprised of 1000 to 1200 languages, the Austronesian language family is the largest language family in the world
Ahem …
https://en.m.wikipedia.org/wiki/Atlantic%E2%80%93Congo_languages
Admittedly, I myself am not persuaded that Atlantic has been shown to be related to Volta-Congo, but I’m being rhetorical here. In any case, I think Volta-Congo alone can trump 1200 languages, but tbh it’s actually an unanswerable question for both Austronesian and Volta-Congo how many languages they contain.
Incidentally, there seems to have been a (welcome) outbreak of splitterism wrt “Niger-Congo” on WP since I last looked, with Niger-Congo now described as “hypothetical.” Progress!
IIRC a phylogenetic tree is a concept in Linguistics (or Philology as probably it was called at the time), that Darwin stole, with acknowledgment. Wikip is not unequivocally backing me up. Linguistics Tree model maybe attributed to Welshman James Howell ~1645.
The blurb on the back of John Merrill’s The Cangin Languages says “… which are of great importance in the study of the world’s largest language family, Niger-Congo.”
A very interesting book, incidentally. I approve of Merrill: he’s a proper historical linguist. As a consequence (I reckon) he doesn’t think that there is any such animal as “Atlantic” at all, and even though the Cangin speakers are culturally Serer, he thinks that their languages have no closer connexion with Serer than with any other part of Niger-Congo.
His PhD thesis, “The Historical Origin of Consonant Mutation in the Atlantic Languages”, is pretty good, too:
https://escholarship.org/uc/item/1qn4m0bh
It’s the first time I’ve really seen this done properly. In the process, he points out that the only Atlantic noun class affixes that can be plausibly related to Volta-Congo are the “human” plural and the “liquid” ones: he isn’t as negative as I am about the Atlantic-Volta-Congo relationship, but even though he accepts it, he points out that it must be very remote*, and that the actual noun class systems in the Atlantic languages must have been very heavily reworked – if not effectively restarted from scratch – subgroup by subgroup.
And it’s got Welsh data in it. Great stuff.
* Mind you, Noon fu tum-ee “you did” looks stunningly like Kusaal fʋ tʋmya “you’ve worked.” The awesome power of coincidence …
The Cangin Languages
For those who are curious, as I was, Cangin is [ˈtʃaŋ.in]
Merrill reconstructs PC *d : Noon/Lehar/Safen s : Ndut/Palor l. *d > s is an odd one.
@Y
PIE *dyḗws > PH *Dzéus > Classical Ζεύς (zděu̯s) > Modern Ζευς (ˈzefs)
Only requires z > s to get there…
@PP: But that’s conditioned by a following j. All on its lonesome, *d > s looks … unusual.
Well, Western Oti-Volta has [j] for proto-Oti-Volta *r (like Hausa with [j] for Chadic *r), the Gurma languages have [ŋ] for proto-Oti-Volta initial *s, and Welsh has [ð] for non-initial *j.
And then Schleicher stole it right back, again with acknowledgment, in an article titled “Die Darwinsche Theorie und die Sprachwissenschaft” and formatted as an open letter to Ernst Haeckel.
[d] – [ð] – [z] – [s] and [d] – [ð] – [θ] – [s] are two possibilities that come to mind immediately.
That reminds me of the sound change *s > n in Arapaho (by a series of megamergers), but it might just be [s] > [h] > 0 followed by vowel onsets acquiring [ŋ], I suppose.
Yes: it’s vowel onsets acquiring [ŋ] – actually an active process with loanwords, so from a Gurma point of view it’s not actually very mysterious at all. Non-initial *s disappears altogether, though that was probably a relatively recent thing, as it’s left epenthetic vowels behind that previously were inserted in clusters like *ls. Realisation of /s/ as [h] is a common areal thing too: Kusaal actually uses /s/ for non-initial /h/ in loans (like the name “Dasmaani” from ʿAbdu-r-Raḥman.)
In his more recent work, Merrill says that Cangin languages actually have no plain voiced stops anyway, only the corresponding implosives and prenasalised stops. That’s not too odd in an African context: Swahili is the same, for example (the orthography is misleading.)
*t > *s turns up a lot in Niger-Congo: cf Twi ɛsã “three”, soma “send” beside e.g. Kusaal atan’ “three”, tʋm “send.” It may (or may not) be connected with the late John Stewart’s idea that the protolanguage had a pervasive fortis/lenis contrast in virtually all stops.
It also turns up in a puzzling way as some sort of spontaneous variation in Oti-Volta (and “Gur” generally), as with e.g. Kusaal ta’aŋ, Yom tamɣo, but Moba sàànŋ̀ “shea tree”; Kusaal saa beside Moba tāāg̀ “rain, sky”; Moba tāānm̀ beside Yom sāmɣā “horse.” In each case there are good cognates with either /t/ or /s/ in several languages, and the division doesn’t correlate nicely with the usual language subgroups at all.
I did try at one point to make it all “regular” by setting up more palatal and alveolar series for proto-Oti-Volta, but it just doesn’t work: you end up with a new distinct phoneme for every other etymon.
@de
I suppose this is nothing at all like the sh/sk variations in English caused by NG adstrate…Are there any doublets (skirt/shirt)?
Not as far as I know. “Horse” could easily be a loanword from somewhere, but “rain” seems unlikely. Even “tree” has done this in the southern Gurma languages: Kusaal tiig, Nawdm tììb́, Gulmancema tībū but Akaselem búcīī and even Ncam bʋ́sʋ̄bʋ̄, with the root vowel umlauted by the suffix. (Southern Gurma languages have class prefixes as well as suffixes, because you can never have too many flexions.) The /t/ must be original in this word (cf Ewe àtí, Swahili mti “tree.”)
Talking of class prefixes versus class suffixes, a new (to me) thing that I recently learnt from Doris Richter (genannt Kemmermann)’s grammar of Mbembe is that the Central Jukunoid languages, along with prefixes, show traces of a full-fledged system of noun class suffixes – abundant traces, in the case of Mbembe. And, as she points out, internal evidence suggests that this is actually older than the prefix system.
This is unexpected, not to say outright paradoxical, given that some of the Southern Jukunoid languages (especially Yukuben) have, not only no trace of class suffixes, but a whole Bantu-like system of class prefixes along with the whole gender agreement package and everything, albeit a bit reduced.
I’d say that the obvious explanation for this kind of thing is phonologically driven loss of suffixes by historical sound changes, leading to a need for disambiguation with proclitic “articles”: in fact, you can all but see this happening before our very eyes in Oti-Volta: the southern Gurma languages all show considerable attrition of the final syllables of polysyllabic words (Akaselem has mostly lost them altogether), and Ditammari, the only other Oti-Volta language which has developed fused class prefixes, also shows formidable wear and tear of its endings.
One thing that strikes me is that while the situation I’ve described seems entirely natural, the reverse process is a lot harder to motivate. That hasn’t stopped the Bantu-centric theorisers from concocting implausible hypothetical scenarios, but I don’t know any examples of such things actually demonstrably having happened, the way the opposite has in Oti-Volta.
Anyway, I’m increasingly doubtful as to whether the class suffixes versus class prefixes divide in Niger-Congo is really anything fundamental at all. (There is no Gur-Adamawa-Ubangi …)
By morphic resonance, I was actually just now wondering – for real – about the change *s -> d in Moba.
This certainly did happen after short root vowels, so you get Moba bid “children”, plural of big, corresponding to Kusaal biig “child”, plural biis. Happens in loanwords too: múdĝ “cat” is from Hausa mussa.
And the Ñaña dialect of Gulmancema has /s/ after long vowels and non-root vowels where “proper” Gulmancema and Moba have /d/, e.g. píísí “uncover'”, Moba pííd́.
(This was all because I recently realised that the Moba imperfective ending d, overwhelmingly the commonest ending for three- and four-mora stem verbs, is very uncommon indeed with shorter stems, and it occurred to me that the historical explanation might be phonological rather than morphological.)
A falling tone on [g]? Or is there a vowel the spelling system isn’t telling us about?
Two vowels: it’s really [mudəgə].
I blame the French. The closely related language (or perhaps even other dialect of the same language) Bimoba, spoken on the Ghana side, writes all the vowels (though it seems to have actually lost the final schwas.) It makes the language look more comfortingly Kusaal-like.
Nawdm orthography does much the same as Moba, leading to things like begdrb “to fall over suddenly.”
The traditional Moba orthography even writes b “they” and k “and.” I suppose they’re consistent, at any rate.
Which of the schwas carries the tone? Is the other one neutral?
The whole word is HHL.
(Incidentally, my thought about *s -> d seems to be panning out quite nicely. It isn’t as implausible as you might think, given what I already know about Moba phonology, and it actually simplifies a whole lot of stuff. A great culling of epicycles … It also makes the actual conjugation system of Moba* a bit less WTF. You can at least see a bit better how it got to be that way in the first place.)
* It’s as bad as Russian. Almost every verb has an imperfective and perfective form, and there are no reliable rules to connect them. Not like Western Oti-Volta at all …
So Russian is the mother of all languages. Huh.
begdrb “to fall over suddenly.”
Just from the sound the meaning is obvious: to fall over while carrying a pile of firewood, or an armful of tubers.
In this case, the -b is merely a noun class suffix, the compilers of the dictionary having decided to list all verbs under their derived action nouns. However, proto-Oti-Volta appears to have had an actual derivational suffix *-b- meaning “do something clumsily.” I feel that all derivational systems ought to be able to make such forms at will.
“Rub a dub dub three men in a tub” – now there’s clumsy preparation for a sea journey.
Actually, Merrill’s thesis gives a fine example of the very process I foolishly claimed hadn’t been documented, viz class prefixes becoming suffixes: Fulfulde.