Lameen at Jabal al-Lughat has a post on the Songhay languages, which he’s studying for his dissertation; I hadn’t realized that the group’s membership in the Nilo-Saharan family was so shaky. (Lameen says “if it were spoken in the Americas, it would undoubtedly be classed as having no relatives whatsoever.”) He links to a number of resources he’s finding useful, including Jeffrey Heath’s webpage, a treasure trove of material about not only Songhay but Dogon. Dogon, it turns out, is not a single language (despite the claims of French colonialists and the Malian government) but a group of related languages; as Heath says in his 730-page Jamsay Grammar (pdf, to be withdrawn once the book is published):
Using the test of mutual unintelligibility as diagnostic, on the other hand, there are clearly many distinct Dogon languages in Mali. We do not yet have Dogon-wide data in a form that would permit accurate identification of language boundaries and of genetic subgrouping. However, having surveyed the varieties spoken in the northern and northeastern parts of Dogon country, I can report the following as distinct languages…: 1. Jamsay (aka Diamsay)…; 2. Beni-Walo, spoken in three separated microzones…; 3. Nanga (naŋa)…; 4. Tabi-Sarinyere, spoken by the people sometimes called Tandam…; 5. Najamba (= Bondu)…
The grammar has a substantial bibliography of material on the Dogon (not just language-related). My thanks to Lameen for pointing me to this great site!
As always, the obvious Dravidian origin is ignored.
I’m tired of peaceful methods. The only language the linguistics profession understands is force.
Call up the Basque Legions!
Sir, Private bulbul reporting, Sir!
The only solution I see is to nuke language isolates from orbit. Then, and only then, will the eldritch influence of the dread Cthulhu be cleansed from the Earth.
Dogon-the-language was still present in the 14th (2000) edition of the Ethnologue, but in the current edition (and a fortiori in ISO 639-3) it’s been split into ten languages:
Bangeri Me Dogon [dba]
Dogul Dom Dogon [dbg]
Bondum Dom Dogon [dbu]
Donno So Dogon [dds]
Jamsay Dogon [djm]
Kolum So Dogon [dkl]
Tene Kan Dogon [dtk]
Tomo Kan Dogon [dtm]
Toro So Dogon [dts]
Toro Tegu Dogon [dtt]
Uh? Pardon me, the Vogon?..
>Bangeri Me Dogon [dba]
>Dogul Dom Dogon [dbg]
>Bondum Dom Dogon [dbu]
>Donno So Dogon [dds]
>Jamsay Dogon [djm]
>Kolum So Dogon [dkl]
>Tene Kan Dogon [dtk]
>Tomo Kan Dogon [dtm]
>Toro So Dogon [dts]
>Toro Tegu Dogon [dtt]
…and the dozen of Sami languages, too. American and British English must be unrelated, at the least.
“if it were spoken in the Americas, it would undoubtedly be classed as having no relatives whatsoever.” – this was meant to reflect not only the unusually relaxed attitude to high-level families in African linguistics but also the hyper-skeptical attitude to such families that seems characteristic of American linguistics. The one produces a lot of false positives (heterogeneous “families” like Ural-Altaic); the other a lot of false negatives (isolates that aren’t really isolates). Which you prefer is to some extent a matter of taste.
That said, the evidence for Songhay being Nilo-Saharan really is fairly weak. As Nicolai has shown, many of the best etymologies proposed fall down when you look at them in more detail (although the alternative proposal he puts forward is, if anything, even more easily demolishable.) The pronouns are suggestive, but a lot more work needs to be done – like reconstructing Proto-Songhay, for a start – before its ancestry can be more confidently identified.
Incidentally, “Dogon” is not only not a single language, but quite likely not even a family; one Dogon language, Bangi-me, is completely different from the rest. See homepage.ntlworld.com/roger_blench/Language%20data/Bangime%20wordlist%20paper.pdf
“the hyper-skeptical attitude to such families that seems characteristic of American linguistics”
How about Greenberg? Is he the exception?
“if it were spoken in the Americas, it would undoubtedly be classed as having no relatives whatsoever.” – this was meant to reflect not only the unusually relaxed attitude to high-level families in African linguistics but also the hyper-skeptical attitude to such families that seems characteristic of American linguistics. The one produces a lot of false positives (heterogeneous “families” like Ural-Altaic); the other a lot of false negatives (isolates that aren’t really isolates). Which you prefer is to some extent a matter of taste.
Lameen is quite correct in implying that the two attitudes, to African and American language classification, are opposite extremes, but it can’t just be a matter of “take your pick of which one you prefer”, as if there was no possible intermediate position – and no hope of finding one – between “false positives” and “false negatives”. American language classification shows both extremes: Greenberg/Ruhlen’s laxness (not to say sloppiness) at one end (as in Africa), “mainstream historical linguists’ ” insistence on a rigid interpretation of historical and classificatory methodology at the other.
The truth is still elusive in many cases, but the truth must lie somewhere in between. Unfortunately, many large-scale proposals are based on methodological shortcuts which may work in straightforward cases but are grossly insufficient in more complex situations, and on the other hand, at least in the Americas some “mainstreamers” are so rigid (and sometimes narrow-minded) that they stifle research.
My impression of Nilo-Saharan (based on what I have read) is that it is not a genetic grouping in any sense but a kind of catchall group including a range of very diverse languages. It is therefore very misleading to place it on the same level as Afro-Asiatic or Niger-Congo as one of 4 “families” in Africa.
I think that part of the problem in trying to explain issues in language classification is the vagueness of the word “family”. The author of a popular textbook in historical linguistics argues that any group of languages which have been demonstrated to be related constitutes a “family”. But it is one thing to say, for instance, that French and Spanish belong to the same language family (Romance), and another to say that French and Persian belong to the same family (Indo-European), and it is something else again to propose that Spanish and let’s say for the sake of argument, Etruscan, belong to the same family (perhaps Nostratic, assuming such a group is justified). In the first case the resemblances in both structure and vocabulary are obvious to the most casual observer, and if one knows one of the languages it is not very difficult to learn the other. I propose to call such a group a “1st order family”. In the second case, the resemblances are far from obvious to the average person, who knowing only one of the languages has a hard time learning the other, and it has taken much time and effort on the part of dedicated specialists to show the relationship, by studying not just the two languages in question but any other languages demonstrably related to those, as well as (where possible) evidence of their history, and by finding resemblances in places where the average person would not think of looking (e.g. verb suffixes, etc), before reconstructing what can be salvaged of the presumed common ancestor. I call this type of group a “2nd order family”, that is, a family of obvious families. As for the third case, that of a potential “3rd order family” (and even higher orders of relationship), it involves languages which are even more diverse than in the second, and in spite of various attempts (still vigorously going on) no one has so far succeeded in proving the existence of even one such group (e.g. Nostratic – which exists in more than one version, Eurasiatic – which partially overlaps with Nostratic, Dene-Caucasian, etc) to everyone’s satisfaction.
Whether Nilo-Saharan is a 2nd, 3rd or even higher order family, or simply a group of unrelated 1st or 2nd order families and/or isolates, is not something I am qualified to comment on, but as a group it does not seem to be of the same nature as Afro-Asiatic or Niger-Congo, which appear to be 2nd order families. Similarly, in the Americas, Greenberg’s “Amerind” proposal lumping together all the (1st order) families except the very well-defined Eskimo-Aleut and Dene families, into a third, all-encompassing and very heterogeneous “family” gives the impression that this third grouping (whatever its merits, which are few if any) is of the same order as the first two, so that scholars in that area must have been singularly incompetent not to recognize its validity. One might as well say that there are only two “families” in Europe: Basque and Eurasiatic (both, of course, descended from Dravidian!). The truth, as usual, is far from being so simple, and it is likely that there are in the Americas a number of intermediate groups awaiting definition and general recognition, as well as a number of irreducible isolates.
About the word “etymology”:
I commented on the apparently new meaning of this term a few days ago in another post. It now seems to mean “a collection of words with the same or very similar meanings and some phonetic resemblances, therefore presumed to be related”, rather than “words proven to be related to each other by descent from a common ancestor, according to a theory of the relationship between the languages they belong to”. For instance, I quoted “Eng head: OE heafod, Latin caput, German Haupt“, the kind of listing one might find in a comprehensive dictionary of English. One can assemble many such groups of words, listing these groups according to alphabetical order in a reference language, usually that of the researcher (eg. according to English words and meanings as in arm, arrow, axe, bee, boat, boy, etc). With closely related languages, this is not much of a problem, since many related words will have the same meaning and not too many differences in sound, but on the other hand, after a few hundred years the meaning as well as some of the sounds may have changed: as someone pointed out, German Haupt does not mean “head” as a part of the body, but “head” as in a compound like “headmaster”, so that a list of current German parts of the body compared word-for-word with the English ones will not include the word Haupt. On the other hand, starting with a Latin word-list, one might be tempted to match caput with German Kopf, which does mean “head” and also starts with a k sound as does the Latin word. Thus unless one is very careful to also notice recurrent correspondences of sounds, as in Latin k = Germanic h (not k), one risks making the wrong matches between one language and another.
In this case, the early pioneers, such as Grimm, could not fail to notice correspondences which occur in a large number of words, as with Latin k = Germanic h, or similarly English t = German tz, but a correspondence which only occurs infrequently, or which seems to be phonetically unlikely, can be totally missed, especially when one of the words (or both) has changed meaning. (Re-)organizing the word collection around sound correspondences, rather than exact meanings, is more likely to yield the true matches, especially unsuspected ones (as with head : Haupt but also the semantically less likely timber “wood used in building” : Zimmer (pronounced tzimmer) “room in a house”), and eliminate erroneous matches (as with caput : Kopf), no matter how exact the semantic match seems to be. But this is not always done, or even doable on a large scale, depending on how well the languages are documented. Also, collecting words in this manner is likely to gather “false cognates” if the languages are spoken in the same geographical area and there has been extensive borrowing: for instance, comparison of English and French vocabularies will show up large numbers of similar words, with regular sound correspondences (as between the diphthongs oi in English and wa, written oi, in French), so that one could conclude from the vocabulary alone that the two languages must be closely related, whereas the truth is that English, a Germanic language according to its basic morphology, has borrowed very extensively from French at one time in its history, far enough in time for several sound changes to have taken place since then. The two languages are related nevertheless, because they are members of separate branches of the 2nd order family known as Indo-European. In “family” terms, they are distant cousins, not siblings.
Extending this example to the discussion above, this is why “etymologies fall down”, because many of the matches (and language classifications) which have been prematurely proposed on the basis of word-lists, before the sifting of correspondences has been done, turn out to be “false cognates” – just as two persons can resemble each other without having any family connections, while two brothers or sisters can look quite different from each other despite being born to the same set of parents.
A classification which relies primarily on “etymologies” as in the second definition is also likely to ignore, or at best, treat as secondary, morphological clues which are often the best evidence for relationship, such as, for English and German, the formation of “strong” verbs, where changes of tense are indicated by internal vowel change, as in sing – sang – sung : singen – sang – gesungen. Just listing vocabulary items such as sing : singen, while ignoring the shared morphological alternations, would miss a very valuable clue to the actual relationship of the languages.
In this instance as in many others, it is easy to conclude that the relationship between English and German (which are both members of the same 1st order family) is “obvious”, but at the next higher level, even in the Indo-European “family” which some people now consider obvious, the membership of some languages or families has not always been evident: for instance, after Hittite was deciphered about 100 years ago, the linguist who first suggested that it could be Indo-European was ridiculed, even though he had been well-respected until then. If this could happen in a “family” which had already been thoroughly studied as such for a hundred years, by large numbers of scholars who had extensive knowledge of several of the languages, past and present, it is not surprising that denial of validity (and its converse, blind acceptance) would occur even more with languages which are not as well documented and are not widely studied outside of very small scholarly circles, so that very few people are in a position to make an informed judgment on the classification of those languages.
I am personally optimistic that many more large-scale or higher-order families will be found than are currently accepted, and conversely some proposals for higher-order groupings will be abandoned or at least reduced, but any conclusions will have to be based on comprehensive comparison of the languages involved, not just on words taken in isolation and a few other items. In the meantime, “constructive skepticism” (= an open mind, but neither acceptance nor denial without good cause) seems to be the best attitude unless one is personally in a position to evaluate the proposals, or, even better, participate in them.
An excellent summary, and “constructive skepticism” is my preferred attitude as well.
: it can’t just be a matter of “take your pick of which one you prefer”, as if there was no possible intermediate position – and no hope of finding one – between “false positives” and “false negatives”… The truth is still elusive in many cases, but the truth must lie somewhere in between.
The truth is out there, no doubt about it… but we won’t know the whole truth for many more decades of work. In the meantime, methods that yields less than the whole truth in significantly less time are far from worthless. If some crude, superficial comparison lets you generates a mostly correct tree with a substantial number of errors in a reasonably short time (which is essentially what geneticists do!), the result may well be unreliable in detail, but will still be a great guide for future researchers, suggesting useful avenues to explore. Conversely, an ultra-rigorous approach that refuses to acknowledge even such manifestly obvious resemblances as are found between, say, Semitic and Berber until someone puts together a full list of sound changes has the merit that anything that does pass the test can be relied upon absolutely. A physicist may insist on four digits of precision in astronomical calculations, but in fluid mechanics they’re happy to accept results with a 30% error; likewise, the amount of false positives or false negatives a linguist is willing to accept can reasonably depend on what they’re trying to do.
Morphological similarities and shared irregularities are great, of course (and, incidentally, Greenberg used them frequently.) But they only get you so far; a lot of languages just don’t have very much morphology, or haven’t been described in sufficient detail. Moreover, morphology too is susceptible to borrowing, as Songhay illustrates nicely (Songhay and Mande, though not discernibly related, appear to have borrowed from one another 3rd person pronouns, a locative postposition, and a suffix forming participles and ordinals, among other things; and northern Songhay has borrowed productive plural and feminine circumfixes from Berber.)
In your proposed “order” terminology, Songhay would probably be a 1st order family with no 2nd order relatives. My general impression, from looking through Greenberg, Ehret, and some of their critics, is that Nilo-Saharan is a very promising (ie, probably broadly valid) 3rd order family with some questionable members thrown in – which, come to think of it, is also my general impression of Eurasiatic…
“Paleo-Siberian” is another catchall family. It seems to mean every language in NE Eurasia which is not Turkic, Finnic, or Dene.
The IE scholars were lucky to have so much morphology. From what I’ve casually seen, the non-Sino-Tibetan, non-IE, non-Dravidian languages of E and SE Asia have all proven hard to classify, partly because of lack of morphology, but also because of the “areal effects” which sound pretty deadly to the genetic approach — e.g., non-tonal languages picking up tone from a neighbor.
Seemingly “isolates” are just single languages surrounded by families of very different languages (i.e., contrastive to well-placed languages). Until SE Asia, the Americas, Australia, and Africa are well-mapped, we won’t really know how many isolates there are in those places. (But from what Lameen says, it seems that Africa is close to being well-mapped, with only one catchall group).
Lameen,
I don’t think my position is not so far from yours! but perhaps we are not talking/thinking about quite the same things.
The truth is out there, no doubt about it… but we won’t know the whole truth for many more decades of work. In the meantime, methods that yields less than the whole truth in significantly less time are far from worthless.
Of course! but it depends whether a method is even likely to yield accurate classification, as opposed to wishful thinking.
If some crude, superficial comparison lets you generates a mostly correct tree with a substantial number of errors in a reasonably short time (which is essentially what geneticists do!), the result may well be unreliable in detail, but will still be a great guide for future researchers, suggesting useful avenues to explore.
I agree in general. This is what happened in Indo-European, a 2nd order group. There were many errors to start with, but the linguists kept working at it. But they did not start just from vocabulary lists: the crucial test was the amount of shared morphology in Latin, Greek and Sanskrit to begin with. I am not sure what criteria geneticists use, but they are probably not superficial ones such as the amount of curl in one’s hair.
For North American languages, in the 1920’s Sapir proposed to regroup the 60-odd acknowledged (1st order) language families into 6 “phyla”, mostly on typological grounds. His use of the term “phylum” (a much higher order than “family” in biological classification) reflected the fact that he did not think that genetic relationship within those groups could ever be proven by traditional methods. Some of these have been accepted, others rejected, and one of the “phyla” is officially disputed. I myself have been working with that group for more than 10 years and have more and more documented reasons to think that most of the languages in it are indeed related (and that this is not the end of this particular phylum).
However, “crude superficial comparison” (which usually means vocabulary) in higher order groups needs to be supplemented considerably before they can even be considered probable. In Greenberg’s “Amerind” proposal, the comparison is extremely crude (lists of words which are unanalyzed, or analyzed by him to fit his purpose), riddled with errors of data, and the conclusion is presented as fact rather than hypothesis.
Here is an example of crude comparison, by the lumper of all lumpers, Shevoroshkin: after an American magazine had published an article on the controversies in language classification, he wrote a letter to the editor in support of his ideas, quoting two words for “woman”: “Greek gyne, Tsimshian hana-k’ ” as obvious proof that the two languages are related. I happen to have worked extensively with the language family from which the second example is taken, which is why I remember this example very clearly. The word for ‘woman’ is hanáq’, which is indeed analyzable, but not as Shevoroshkin assumes: it is really ha- ‘having, characterized by’ and naq’ ‘dress’ (originally probably ‘apron’, a woman’s garment hanging from the waist). In two of the languages of this family, the plural shows this analysis very clearly: ha-ná:naq’, where ná:naq’ ‘dresses’ is the plural of náq’. There is absolutely no independent reason for analyzing it as Shevoroshkin does, except that it seemed to fit his ideas. Greenberg’s book is full of similar errors as well as other factual errors of many kinds.
Conversely, an ultra-rigorous approach that refuses to acknowledge even such manifestly obvious resemblances as are found between, say, Semitic and Berber until someone puts together a full list of sound changes has the merit that anything that does pass the test can be relied upon absolutely.
Perhaps here there is a confusion between classification, subgrouping, and proto-language reconstruction, which are different even if they are related. In the early stages it is easier to show that languages are related than that individual words are related (see my comment on “etymologies” above). I don’t know much about Afro-Asiatic, but I got the impression that Berber is considered part of it. Perhaps the dispute is about how close Berber is to Semitic, rather than whether they are related at all? Unravelling the sound changes is something that takes a long time, and it is a waste of time to try to do it until one is pretty sure that the languages are indeed related. Some linguists think that you cannot consider languages to be related until there is a full reconstruction of the proto-language: in my opinion, for purposes of classification, reconstruction is frosting on the cake, not the cake itself.
A physicist may insist on four digits of precision in astronomical calculations, but in fluid mechanics they’re happy to accept results with a 30% error; likewise, the amount of false positives or false negatives a linguist is willing to accept can reasonably depend on what they’re trying to do.
Not being familiar with statistics, I am wary of percentages. There must be a very good physical reason why there is such a difference in accuracy in astronomical and fluid mechanics measurements, and in physics the degree of accuracy which is achievable under various circumstances can be calculated. But how does this apply to linguistics? With a rough classification such as Greenberg’s, how are linguists supposed to measure the percentage of possible (not factual) errors in the classification?
Morphological similarities and shared irregularities are great, of course (and, incidentally, Greenberg used them frequently.)
Perhaps he did in his African work? in his American book, he mentions a few affixes and pronouns (about which he grossly overgeneralizes) , but that is all. “Shared irregularities” can only be determined if one knows what the regularities are, therefore if one has done or at least consulted a competent morphological analysis, which unfortunately is not always available – more work for us.
But they only get you so far; a lot of languages just don’t have very much morphology, or haven’t been described in sufficient detail.
That is true, but where there is a lot of morphology, it would be counterproductive to ignore it: witness Indo-European. In English there is not much productive inflectional morphology left from PIE, but there is still enough of it to place the language in the Germanic family rather than any other.
Moreover, morphology too is susceptible to borrowing, as Songhay illustrates nicely (Songhay and Mande, though not discernibly related, appear to have borrowed from one another 3rd person pronouns, a locative postposition, and a suffix forming participles and ordinals, among other things; and northern Songhay has borrowed productive plural and feminine circumfixes from Berber).
Yes, that point is often made, but notice what part of morphology is borrowed: besides independent pronouns, all the examples I have seen are of affixes, not for instance vowel alternations as in sing, sang, sung, song or the “interweaving” of vowels and consonants and other complex modifications of the consonantal root as in Semitic. These are the core of the morphology in the relevant language families (whether they are still productive or not), and they are very unlikely to be borrowed.
In your proposed “order” terminology, Songhay would probably be a 1st order family with no 2nd order relatives. My general impression, from looking through Greenberg, Ehret, and some of their critics, is that Nilo-Saharan is a very promising (ie, probably broadly valid) 3rd order family with some questionable members thrown in – which, come to think of it, is also my general impression of Eurasiatic…
I am sure that the existence of 3rd order groupings will eventually be demonstrated. How far they will agree with any of the present proposals is another matter. In the meantime, our work is cut out for those of us working in that area.
What a great discussion!
he wrote a letter to the editor in support of his ideas, quoting two words for “woman”: “Greek gyne, Tsimshian hana-k’ ” as obvious proof that the two languages are related
And people wonder why I don’t take Shevoroshkin seriously…
“I” and “me” in Etruscan are mi and mini… I’m just saying. 🙂
It does, but this meaning is nowadays restricted to poetic language. (Incidentally, it’s neutral, like caput.) My point was that if we compare unwritten languages that lack a separate poetic register, we have to allow for some semantic leeway, and — initially at least — for a lot of leeway in vowel correspondences (English ea can not only be pronounced as just about any English vowel, it can also correspond to any German vowel phoneme except /ʊ uː ʏ yː/).
Incidentally, the normal word for “head” in German, Kopf, is a loan from Latin (cuppa “some kind of bowl”, > English cup). So much for occasional claims that any part of the vocabulary is never borrowed.
I don’t know if anyone ever thought it was real. The uses I’ve seen were all explicitely purely geographical wastebaskets.
Could you tell us which one?
What? Shevoroshkin? Are you sure he wasn’t joking? Because… I’ve seen quite different stuff from Shevoroshkin. Sure, a relationship between North Caucasian and Salishan-Wakashan sounds outlandish, but many of the cognates are amazing. Keep in mind that we are talking about language families with huge, and similar, consonant inventories. I can send a list as a pdf on request.
Well, genes don’t work like that. Geneticists tend to pick as many genes as possible that they believe evolve with about the right speed for their problem — on some there is so little natural selection that any signal gets swamped by noise in a short time; on others there is so strong selection that only a couple of small changes have occurred in animals and plants together — basically count the shared derived innovations, and apply Occam’s razor. That’s called cladistics.
Cladistics with morphological data works about the same way. Traditionally (before cladistics was systematically used), people tried to pick “important” characters (and got more and more pessimistic the closer they came to the conclusion that everything can evolve convergently); in the last 20 years, however, it has become evident that as much evidence as possible should be used. The signal will add up, the noise will cancel itself out. Analyses with 80 species and 430 characters exist, and the trend to ever larger analyses continues. One analysis has over 2500 characters.
BTW, a molecular analysis has shown that in a group of South American birds beak shape is much more volatile than feather color… be careful with what you say about the amount of curl in one’s hair! 🙂
Cladistics has been used way too little in linguistics. Its potential is enormous.
David –
“I” and “me” in Etruscan are mi and mini…
I agree that some version of Nostratic is probably right, but my point is the degree of relationship between languages or families.
German Haupt does not mean “head” as a part of the body
– It does, but this meaning is nowadays restricted to poetic language.
I was going by what you said earlier about the meaning of this word. Surely a word like Hauptmann is not particularly poetic.
My point was that if we compare unwritten languages that lack a separate poetic register, we have to allow for some semantic leeway, and — initially at least — for a lot of leeway in vowel correspondences (English ea can not only be pronounced as just about any English vowel, it can also correspond to any German vowel phoneme …).
I think that it is not just a question of separate registers. Semantic change as well as word replacement, for whatever reason (e.g. taboos that we moderns might be totally unaware of) are going to occur. As for vowels, not being “anchored” to specific articulators as are consonants, they are more likely to change and therefore are less useful than consonants in the initial stages. The case you mention about English ea and most German vowels is almost exactly paralleled in a similar case in the Tsimshianic family of Western Canada.
Incidentally, the normal word for “head” in German, Kopf, is a loan from Latin (cuppa “some kind of bowl”, > English cup). So much for occasional claims that any part of the vocabulary is never borrowed.
“Never say never” should be a useful maxim to remember in comparative linguistics.
I did not know that Kopf was from cuppa: similarly, Italian testa and French tête are from Latin testa “clay pot”. In all these cases the metaphorical link seems to be “cranium” (there is documented archeological evidence that drinking bowls were made from enemies’ skulls in some ancient European cultures).
one of the “phyla” is officially disputed. I myself have been working with that group for more than 10 years
– Could you tell us which one?
I mentioned it in the Burushaski thread: it is Penutian (in Sapir’s sense, not Greenberg’s). My wording “officially disputed” is perhaps not strong enough: the Smithsonian Institution’s volume on Languages in the massive collection “Handbook of North American Indians” does not recognize it at all (it uses the word for only a small part of Sapir’s original grouping). There are a few of us chipping away at the problem of demonstrating its reality: I was totally unconvinced at first but changed my mind after I started to do serious work in the subject (at first with the intent of demolishing the hypothesis!).
… Shevoroshkin … quoting two words for “woman”: “Greek gyne, Tsimshian hana-k’ ” as obvious proof that the two languages are related.
– What? Shevoroshkin? Are you sure he wasn’t joking?Because… I’ve seen quite different stuff from Shevoroshkin.
Why should he joke in support of his own work in a letter to the editor, after an article which appeared skeptical of his claims (and of others’)? I have not seen that much of his work, but what I have seen does not inspire me to invest in his books. I am willing to be convinced, when I see evidence that appears convincing (see above).
Sure, a relationship between North Caucasian and Salishan-Wakashan sounds outlandish, but many of the cognates are amazing.
Instead of cognates I would use potential cognates or better resemblant forms. Cognates are words which have been demonstrated to be related, not those which simply suggest relationship. To go back to earlier examples, English headis cognate with German Haupt and Latin caput although it does not much resemble the latter form, but the relationship can be demonstrated through detailed comparison of a number of changes, occurring not only in these words but in many others where the same corresponding sounds occur, in the corresponding places in the words. Demonstrating cognacy is done at the level of comparing individual words, and is another process from demonstrating language relatedness (although some people would disagree with these priorities).
Cladistics has been used way too little in linguistics. Its potential is enormous.
Thank you for the explanation of genetic work and cladistics. I had heard of it more superficially and agree that linguistics could do with more of it.
And you were right. Sorry for not being clearer: the meaning “head” (and the usage as a noun) is poetic nowadays.
I only brought that up because such registers (and of course written records) are likely to conserve evidence of shifts in meaning.
The usual explanation is that it was a dysphemism, like the other “bowl” words (and indeed the rare use of “bowl”, Schüssel, for “head” in German today), and like the Latin word for “eat” (edere) was replaced by “munch” (manducare “to chew powerfully”) in the ancestry of most Romance languages (French manger, Italian mangiare… French has acquired a new dysphemism, bouffer). This direct assault on the core vocabulary seems to be common in pre-teen lects.
Sorry! Must… not… comment… blog posts… after… 1 at night.
I like “comparanda”. 🙂
So Shevoroshkin was the author of the… slightly cranky paper I found somewhere online years ago that compared Proto-Tsimshian and Proto-Indo-European, assumed a number of consonant mergers that looked a bit too convenient to be real, allowed for a hardly plausible amount of semantic leeway, and concluded that Tsimshian and IE were close relatives? I forgot.
Marie-Lucie, should I send you the word list? Unfortunately it’s only the appendix to a paper which I don’t have, and it’s a large file because it was scanned in.
I did not know that Kopf was from cuppa: similarly, Italian testa and French tête are from Latin testa “clay pot”. In all these cases the metaphorical link seems to be “cranium” (there is documented archeological evidence that drinking bowls were made from enemies’ skulls in some ancient European cultures).
🙂
I love this stuff. “Would you like a Kopf of coffee?”
David,
You remember accurately, but no, that cranky paper (still on the web) is not by Shevoroshkin. I don’t want to give the author extra publicity by broadcasting his name, but I know him and heard him present the paper at a conference. The stuff is worthless, but I don’t want to rail about it in public, although I am willing to do so privately. Is the word list you are referring to, from that paper? Put it in the shredder.
*phew*
No, of course not.
Please be more precise! Do you mean Salishan-Wakashan-Caucasian etc ?? Sure, I will look at anything (if it is not so big that it will clog up my email).
Yes. I was only sent the file myself; I’ll ask for the reference.
Back to Bangime: it seems that the Bangande formerly lived on the opposite side of the (normal) Dogon-speaking area: not only are the oldest layers of loanwords from eastern rather than adjacent western Dogon varieties, but the very word Bangime, which means nothing to its speakers or their neighbors, is interpretable as a medio-passive verb ‘it is hidden’ in eastern Dogon, although alternatively -me may be a native suffix meaning ‘language spoken by’. In any case the root bang- is certainly not Bangime, and means ‘secret, hidden, furtive’. In the alternate form Bangerime, the -ri is the Dogon valence-increasing suffix.
“Bangime: Secret Language, Language Isolate, or Language Island?”.
Also, on the original topic: as of 2016 even Official Notorious Lumper George Starostin now thinks that Songhay shows no clear evidence of relationship to anything within “Nilo-Saharan” (and per him, neither do the Saharan languages, which would basically cut the family back down to just “Macro-Sudanic”). Gerrit Dimmendaal seems to remain slightly more optimistic, but he also agrees that at least Songhay and Koman-Gumuz should be removed from the family.
Interesting — thanks for the update!
Starostin’s paper, in English.
I never did. I’ll search for it later this weekend.
Thanks, David. I was thinking of doing some look-around for a published version, but apparently not (yet?).
From Dimmendaal, see e.g. his 2011 book “Historical Linguistics and the Comparative Study of African Languages” or Stable and unstable features in Nilo-Saharan (2016).
Turns out “eat” in Turkana is -/ɲam/.
There’s quite a number of these odd Pan-Africanisms. Not so odd, I guess: people have been around (and talking) in Africa for long enough that it shouldn’t be surprising that widespread borrowing of some words goes back a very long way. Hausa nama “meat” looks pretty Niger-Congo-ish (cf Kusaal nim “meat”), for example. Mind you, Hausa is a magpie language, like English, forever stealing other people’s words. Even the Hausa word for “two”, biyu, seems not to be Chadic, but borrowed from some Niger-Congo language.
I mean, it’s so beautifully sound-symbolic.
Ah. True enough. I thought you might be alluding to (for example) Wolof ñam “taste”, which is commonly (and I would say, very implausibly) said to be the origin of English “yum.” Sound symbolism sounds a lot more likely.
“Yam”, on the other hand, really does seem to go back to the Wolof or to a cognate word in some other West African language, via the perfectly plausible intermediary of Portuguese or English-based creoles.
As a noun, Wolof ñam is “food”, and seems quite likely to be connected with Kusaal nim and numerous cognates elsewhere in Niger-Congo meaning primarily “meat”; Guthrie has Proto-Bantu yàmà. One man’s meat is another man’s yam, after all.
Data point: German mjam, Lolcat OM NOM NOM. Versions with the actual palatal nasal are all over Spanish and French comicbooks.
“Nyam” is obviously just Human (and lolcat) for “eat.”
Actually, sound symbolism shows up in unexpected places: we had a discussion about very-similar-yet-plainly-not-cognate words for “butterfly” a while back.
I was puzzling about verbs of the form *LAB- with the highly specific meaning “hide behind something to eavesdrop”, found in Kusaal, Hausa and various Songhay languages. The obvious explanation that the words are loans runs up against problems with how well integrated the words are with the internal derivational systems in the various languages, and I did seriously wonder whether the form is just sound-symbolic. Somehow. For West Africans, anyway.
Nom in LOLcat is actually a borrowing from Muppet.
Is it a learned borrowing from Classical Muppet or transmitted via Lolcat/Low Muppet bilinguals?
nam-nam is Danish talking-to-babies for “this tastes good, cross my heart, don’t spit it on the wall this time!”
Oh, that explains a few things.
the Hausa word for “two”, biyu, seems not to be Chadic
Though English two is native, the corresponding ordinal second is borrowed (via French) from Latin secundus, displacing the original term other. Bizarrely, secundus in this use (its basic meaning was ‘following’, ‘according to’) displaced the original Latin word alter, whose basic meaning is also ‘other’!
@David Eddyshaw: In spite of the seeming absurdity of that question, it turns out to have a documented answer, that it was not a learned borrowing. According to Know Your Meme, om nom was already in vernacular use as early as 2004, when it was first added to Urban Dictionary; however, it did not appear in a LOLcat caption until 2007.
By the way, here is my favorite LOLsnail nom image.
@Brett:
I am deeply grateful for your taking the trouble to research the matter. And how much groundbreaking research has been due to freethinking investigators daring to investigate “absurd” questions?!
(Probably not much, to be honest. But one can dream …)
Songhay shows no clear evidence of relationship to anything within “Nilo-Saharan” (and per him, neither do the Saharan languages, which would basically cut the family back down to just “Macro-Sudanic”). Gerrit Dimmendaal seems to remain slightly more optimistic, but he also agrees that at least Songhay and Koman-Gumuz should be removed from the family.
On the genetic side of the things it turns out that the Gumuz (until now, the only studied “Nilo-Saharan” group) have no relationship with the Lugbara of Upper Nile in Kenya (a more acceptable Nilo-Saharan group). The Lugbara in turn are the most genetically distinct people within the wider region. The Gumuz-like ancestry seems to have contributed to the Afro-Asiatics but the Lugbara seems to have been quite insulated from the ancestors of other regional peoples in the past.
https://www.sciencedirect.com/science/article/pii/S0002929720302378?via%3Dihub
For some time, Zande people living in the north of present Democratic Republic of Congo and in South Sudan were known to Europeans by the name given to them by their Dinka enemies – Niam Niam.
The “Niam Niam” name was a rather childish insult implying that they were cannibals.
According to many 19th century explorers they indeed were, but politically correct anthropologists now proved that it was all a misunderstanding.
Bangime has turned out to be an isolate (and now has a Mouton Grammar Library grammar all its own, courtesy of Jeffrey Heath and Abbie Hantgan.)
The speakers (the Bangande) think that they are Dogon, but nobody else agrees with them.
>I love this stuff. “Would you like a Kopf of coffee?”
I remember sniggering with a barista long ago at a college kid, generically Asian American in my mind, who would regularly order a coff of cuppi, seeming very amused with himself. What a stupid not-even-pun to try to pass off.
It was only years later that I read a list of coffee words and realized he must have been from a Pacific Island background where kopi was his genuine word for it. He wasn’t amused with himself. The American phrase must have inherently sounded ridiculous to him.
Can’t resist latest TikTok joke.
Starbucks coffee shop in Moscow.
Barista: Americano?
Non-Russian customer (apparently first time in Starbucks): No, I am from Daghestan…
What is the current stereotype of Caucasus folks in Moscow or generally in Russia? (I am fully aware that like all ethnic stereotypes, it’s at best unflattering, at worst awful, and that the jokers aren’t always joking.) Back in the day, what I heard was that Chukchis were the stereotypical wild and uncivilized people (and there was some joke about how you weren’t supposed to say that Konstantin Chernenko looked Chukchi.) Have Daghestanis taken their place? Or is their stereotype that of religious fanatics?
>>Or is their stereotype that of religious fanatics?
I think it’s more like they are seen as show-offs, body-building, aggressive, organized crime connected…