LINGUIST List 9.431

Sat Mar 21 1998

Disc: State of Comparative Linguistics

Editor for this issue: Martin Jacobsen <>


  1. 00hfstahlke, Re: 9.424, Disc: State of Comparative Linguistics
  2. Mohamed Diriye Abdullahi, Re: 9.397, Disc: State of Comparative Linguistics
  3. Peter T. Daniels, Re: 9.424, Disc: State of Comparative Linguistics
  4. bwald, Re: 9.424, Disc: State of Comparative Linguistics

Message 1: Re: 9.424, Disc: State of Comparative Linguistics

Date: Fri, 20 Mar 1998 16:02:16 -0500 (EST)
From: 00hfstahlke <>
Subject: Re: 9.424, Disc: State of Comparative Linguistics

I am puzzled by the following statements by Peter Daniels: 

>Alexis Manaster Ramer's most recent contribution is unclear to me:
>since one of Greenberg's aims is the proper *sub*grouping of languages
>within a phylum, the following position doesn't seem reasonable:

>> What Peter Daniels says may be correct but please please note that
>> it does NOT in any contradict the thesis that the Niger-Kordofanian
>> language family IS a valid (and I think uncotroversially valid)
>> language family. The question of branching he alludes to has to do
>> with the interal structure of this family, much as
>> Indo-Europeanists keep debating teh branching of IE without ANYONE
>> taking this to mean that Indo-European itself is invalid. As for
>> Peter's second comment, I must object to the rhetoric here: I
>> myself pointed out that parts of Greenberg's African langauge work
>> are not acceptable, even though Niger-Kordofanian (and other parts)
>> ARE.
> The question of branching -- the internal structure -- is *exactly*
>the point of Greenberg's classification, and that is what has been
>challenged and reworked by the specialists; and it is especially
>perverse to continue to use the name "Niger-Kordofanian" for a
>phylum, Niger-Congo, in which the Kordofanian group is simply another
>constituent among many, branching off at a level coordinate with

I've read much of Greenberg's work on language classification,
including his earlier work on the African classification problem. I
don't find him saying that "The question of branching -- the internal
structure -- is *exactly* the point of Greenberg's classification."
The point of his African work was to put order to conflicting
classifications that had been based on everything from bad ethnology
to equally strange theology. As he looked at the African data he
found arguments of linguistic relationship based on whether or not two
languages were spoken by nomadic herders or by sedentary agricultural
groups. Often lists of words were cited that contained obvious
borrowed and cultural vocabulary, and typological features were
introduced, such as the presence or absence of lexical tone or of
monosyllabic stems. It was this work, from the late '40s to the early
'60s that led to his African language classification, his claim that
the languages of Africa belong to four large families. The criticisms
raised against this conclusion in the '50s and '60s, which were
collected into a unpublished volume by Charles Kraft in the mid '60s
were largely of a variety that neither Campbell nor Greenberg would
consider relevant.

Greenberg viewed the sub-groupings within the major families as less
definite, not as his primary point. He says in his essay "The problem
of linguistic subgrouping" (Essays in Linguistics, University of
Chicago Press, 1957, p.49), in a comment on Semitic and Afro-Asiatic,

"We may presume that the groups of languages which at an earlier stage
were recognized as independent families are valid branches of the
larger family, since differences among the branches are here so great
that each was recognized as a separate entity before the family as a
whole came to be accepted. Howver, grouping errors may arise even
here, and the whole problem should be re-examined after the
establishment of the larger family."
Greenberg's goal was the identification of larger families. Up to his
time people didn't question the identity of Bantu, Gur, or Kwa. His
contribution was to show that these, and other group including some
not yet identified at the time, together formed a larger family,
Niger-Congo. Generally, he made use of existing groupings unless
those, as in the case of Nilo-Hamitic, were not based on linguistic
evidence and were not valid.

>The question of branching -- the internal structure -- is *exactly*
>the point of Greenberg's classification, and that is what has been
>challenged and reworked by the specialists; and it is especially
>perverse to continue to use the name "Niger-Kordofanian" for a
>phylum, Niger-Congo, in which the Kordofanian group is simply another
>constituent among many, branching off at a level coordinate with

The initial claim in this bit, "that internal structure is *exactly*
the point of Greenberg's classification" is simply false. In fact,
Greenberg has himself pointed out that his method is less effective on
smaller subgroupings than on the larger groupings. In _The Languages
of Africa_ (IJAL series 1963), Greenberg lays out the four major
families, Niger-Congo, Afro-Asiatic, Khoi-san, and Chari-Nile. (I'm
referring to the 1963 version because I unwisely lent out my 1966 and
haven't seen it since.) He then adds chapters on Nilo-Saharan and
Niger-Kordofanian. These latter chapters--and families--are clearly
more tentative. He was the first to suggest that there was a link,
for example, between Songhay and Chari-Nile. This link has always
been considered doubtful, and there question has been raised often
about his Nilo-Saharan; not, though, about his Chari-Nile, which is an
important innovation in itself. His recognition that the Kordofanian
languages were related to Niger-Congo is another contribution. A
review of his data suggests two conclusions. First, Kordofanian is
itself a valid grouping. Second, the nature of the relationship
between Kordofanian and Niger-Congo is not clear, whether it is a
sister group or simply a highly divergent subgroup. That still is not

At the end of the 1963 work, Greenberg makes the following tentative
proposal (p. 153):

"In summary, the present investigation leads to the classification of
African languages into four families: Congo-Kordofanian, Nilo-Saharan,
Afroasiateic, and Khoisan. The extinct language of Meroe is left
unclassified for lack of evidence."

In a later paper, in Current Trends in Linguistics, Vol 7: Sub-Saharan
Africa, Greenberg adds Meroitic to Nilo-Saharan, confirming the
judgment of other scholars whom he cites in the article.

Niger-Congo has undergone considerable revision since 1963, as shown
especially in the Bendor-Samuel volume where Eastern Kwa is split of
from Western Kwa and made into several branches of Benue-Congo. There
remains doubt over the status of Songhay, Hatsa, Sandawe and a few
other languages, but what is remarkable is that in the last 30 years
there has been so little major revision. No one has proposed fewer
groupings. Bender has separated Western Cushitic off from Cushitic as
another branch of Afroasiatic that he calls Omotic.

Greenberg's work is, contrary to what Peter Daniels says, not about
subgrouping but about large families.

Herb Stahlke
Ball State University
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 9.397, Disc: State of Comparative Linguistics

Date: Fri, 20 Mar 1998 11:58:42 -0500
From: Mohamed Diriye Abdullahi <diriyeamMAGELLAN.UMontreal.CA>
Subject: Re: 9.397, Disc: State of Comparative Linguistics

At 10:59 PM 18/03/1998 +0000, Manaster wrote:

> My learned friend(and occasional collaborator) Peter Daniels'
>comments on the state of Semitic and Afro-Asiatic linguistics, while
>phrased to disagree with some of what I said, state things I
>completely agree with (namely, that there is a (small and I think
>shrinking) number of excellent Semitic comparative LINGUISTS who
>recognize the value of linguistic reconstruction as well as the
>validity of the Afro- Asiatic language family, of which Semitic is a
>small part). If one confines oneself to talking to or reading the
>works of these distinguished scholars, then indeed there is no need
>for alarm. However, what does alarm me is that many (maybe even
>most) departments that teach Semitic comparative linguistics (and
>this usually means depts. of Near Eastern Studies or the like, not of
>linguistics) are very largely hostile to both reconstruction and
>classification of languages, and that students of the subject are
>discouraged from pursuing such topics or acepting the validity of
>such obvious constructs as Proto-Semitic or the Afro-Asiatic language
>family. Or even if these concepts are technically accepted, they are
>paid no more than lip service. Even within linguistics itself, we
>have the spectacle of Gerhard Doerfer, a distinguished student of
>Turkic and Mongolic languages and the leading critic of the Altaic
>theory, repeatedly dismissing the Afro_Asiatic language
>family--without the linguistic community at large rising up in arms
>at this. Yet Doerfer's explicit position is a major threat to
>classificatory linguistics and perhaps to comparative linguistics as
>a whole: namely, he claims that related languages must have cognate
>numerals between 2 and 5 and a set of cognate basic body part terms.
>Since he realizes that the different branches of Afro-Asiatic do not
>in fact meet this criterion of relatedness (invented as a quick and
>painless way to dispose of Altaic), he then has to reject
>Afro-Asiatic (and also incidentally Uralic), because otherwise he
>would lose what has become his favorite argument against Altaic.

As far as I know a basic set of numerals and a basic set of body parts
can be established for the so-called 'Afro-asiatic' (I will explain
below I call it always the 'so-called Afro-asiatic.') Two of the most
consistent words for body parts in the so-called Afro-asiatic are the
words for mouth and eye. Skinner (Skinner,=20 Neil. 1977. "'Fly'(Noun)
and 'Mouth'" in Afroasiatic." Afroasiatic Linguistics 4,1:51=8462.)
made a list of the word for mouth. As for eye, from the top of my
head, the cognate set exists in both Semitic Arabic and Cushitic
Somali as, respectively, Eyn and Il.

As for the a set of numbers between 2 and 5, the cognate set exists in
Semitic Arabic and Cushitic Somali (approximate transcription):

Arabic 			Somali
Salasa/Talada 		Sadeh/sadi
Arbaca				Afar
Khams				Shan

The reason why I do not agree with the label Afro-asiatic is that: (a)
it portrays a false equality of geograhical distribution;

(b) it obscurs the origin of the whole group;

All the languages execpt Semitic are spoken in Africa. Semitic
languages are a minority numerically in the group.

Semites emigrated from the African side of the Red sea anyway and
exported an African language to Asia where they met Asians such as the
Persians, Assyrians and others. This is the only logical way to
account for the presence of an African branch in a place where it is
surrounded by unrelated languages. And those who know the shortness of
the crossing between Africa and Asia especially at Bab-el-mandab would
probably agree with me. Moreover, the most conservative Semitic
dialect or dialects (Jibali) with the so-called true DAD or hard DAD
are situated in southern Arabia opposite historical Ethiopia (today's
Ethiopia, Eriteria, Djibouti, and Somalia). 

It would make sense to change the name of the group to something like
Ethiopic or Ethio-Chadic.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Re: 9.424, Disc: State of Comparative Linguistics

Date: Fri, 20 Mar 1998 18:44:31 -0500
From: Peter T. Daniels <>
Subject: Re: 9.424, Disc: State of Comparative Linguistics

Alexis Manaster Ramer wrote:

> "Mr." Doerfer is Professor Gerhard Doerfer, only the senior
> Turkologist and Mongolist around, and THE critic of Altaic. 

You'll recall, Alexis, the reverse snobbery at the University of
Chicago where it was quite infra dig to call a professor anything
other than "Mr." ...

> His rejection of Uralic AND Afro-Asiatic first appeared, I believe,
> in: Doerfer, Gerhard. 1974. "Ist das Japanische mit den altaischen
> Sprachen verwandt?" Zeitschrift der Deutschen Morgenlandischen
> Gesellschaft 124: 103-142.

Dare I confess that, if I had come across that title in a table of
contents, it wouldn't have occurred to me to look in it for a comment
on Semitic or Afro-Asiatic?

> In a later publication, viz.:
> Doerfer, Gerhard. 1993. "Nostratismus, Illic-Svityc und die
> Folgen. Ural- altaische Jahrbu"cher, N. F. 12: 17-3

I did look in this annual once: for Robert Austerlitz's investigation
of Old World vs. New World "Genetic Units" (1980, a concept
subsequently used by Johanna Nichols).

[Everything else Alexis says is quite reasonable; the
Trubetzkoy/Kotwicz theory that Alexis keeps bringing to everyone's
attention really does deserve to be left alone.

Note also that I titled a chapter in *The World's Writing Systems*
"Scripts of the Altaic Languages," and the author, Gyorg Kara, made no

- Peter T. Daniels
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 4: Re: 9.424, Disc: State of Comparative Linguistics

Date: Fri, 20 Mar 1998 21:52:23 -0800 (PST)
From: bwald <bwaldHUMnet.UCLA.EDU>
Subject: Re: 9.424, Disc: State of Comparative Linguistics

Peter Daniels wrote recently:

>The question of branching -- the internal structure -- is *exactly*
>the point of Greenberg's classification, and that is what has been
>challenged and reworked by the specialists

This is only partially true in historical context. The first
important contribution of Greemberg's classification scheme was that
the given groups were related into a larger family. Before this, the
dominant opinion was that they were not -- and most notably that the
Bantu languages were an independent family unrelated to any of the
others. A transitional group in what is now univerally accepted as
Benue-Kwa, which includes Bantu and most of the Cameroons and Nigeria,
and more, was called "Semi-Bantu" at that time and assumed to have
borrowed much vocabulary from Bantu, while its separate "genetic"
origin was seen to be betrayed by its "un-Bantu" grammar. Since that
time the phonological causes of the loss of Bantu features for this
group of "Bantoid" languages in the Cameroons have become clear, and
can be accounted for with the rigor required of "genetic"

Greenberg's subclassification was suggestive. However, subgrouping of
branches in any family is known to be notoriously difficult according
to the tree concept -- for ALL families. Such difficulty does not
constitute a demonstration that there is anything wrong with the
genetic classification of the family, only with the tree concept of
diversification (as opposed, say, to areal diffusion from different
centers of innovation leading to crossed isoglosses of great
historical importance, which lead to ambiguities for the bifurcating
tree concept). The problem of whether Germanic is closer to Slavic
than to Italic, etc., remains without anyone doubting the notion that
all these branches of IE are genetically related. With regard to
Niger-Congo ( a name preferred by specialists to "Niger-Kordofanian"),
a large number of branches are quite continuous and difficult to
subgroup unambiguously, some are relatively discrete, and some have
been questioned for bona fide membership by some specialists, esp
Mande, North Atlantic (and perhaps Kordofanian). Much remains to be
worked out. Subgrouping problems will remain as long as historical
linguists remains wedded to the tree concept of linguistic evolution.
I hope that it will not be long. Where it works, fine. But in some
cases it will never work, because that's not how the linguistic
diversification came about.

Finally, AMR reported a bizarre episode in which one linguist, Doerfer
I think, laid down as a criterion that languages were not to be
considered genetically related if they did not maintain cognate
numbers for 1-5. As a heuristic it may have some (not much) merit,
but as a definition of genetic relatedness it is arbitrary and
unjustifiable. It remains to be seen what criteria can be agreed to
for genetic relationship, since over time any language can change and
/ or borrow anything (or everything?) from other languages under
conditions of contact. Presumably, tracing the history of any and all
languages is more important than classifying them by one arbitrary
method or another. In my opinion, a classification is useful but not
secure until a complete account of the evolution of the classified
languages is given. This is necessary simply to understand what the
classification (genetic or otherwise) means, and what parts of a
relevant language it applies to. -- Benji


Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue