AMR on Trask on Dixon on...

[Larry Trask]

----------------------------Original message----------------------------
Alexis writes:
 
[By the way, Alexis -- something went wrong with the wrap in your
posting, and it arrived scrambled.  I had to edit it before I could
read it.]
 
> Larry Trask says about the language families of Africa (posited by
> Greenberg and for the most part widely or even universally accepted
> by competent observers): "...  it appears, the families are set up
> on the basis of a few recurring grammatical characteristics,
> characteristics which involve actual morphological material in the
> AA case but only typological features in the NC case."
 
Actually, I was quoting Dixon.
 
> I am not an "Africanist" and I certainly cannot pretend to have a
> cure for whatever troubles Dixon, whose book I have not read, but I
> really think that Larry Trask would do well not to lend credibility
> to the statements he is repeating about African language
> classification by repeating them without any criticism.
 
Mea culpa, I suppose, but I have grown somewhat weary of hearing
assertions that such and such a family is "universally accepted", only
to find later that it has been severely called into question by
someone with specialist knowledge of the languages involved.  No doubt
I should read more widely, but the British week sadly has only 168
hours in it, and that just isn't enough.
 
> It is certainly true that some of Greenberg's proposed African
> language families have been questioned by competent scholars and in
> some cases are indeed poorly supported (and may well be
> wrong). Khoisan is the clearest example where Greenberg's arguments
> are inadequate to establish the family.  However, what Larry says
> Dixon says about Niger-Kordofanian and Afro-African is just not
> right.
 
> The topic of the validity of the Niger-Kordofanian language family
> is one I know something about and which was discussed in Baxter and
> Manaster Ramer (1996), following on the article by Schadenberg in
> the very useful 1981 compendium Die Sprachen Afrikas (sorry, I am
> not up to giving a fuller reference).  It is true that as of that
> time anyway there was very very little lexical evidence for this
> family, but it is completely erroneous to say that the evidence was
> typological.  Rather, the nominal class system of the major
> subfamilies agree in detail as to the markers for the different
> classes.
 
I am glad to hear this, but this is precisely the sort of evidence
that Dixon was apparently hoping to hear about, but did not hear about
from the specialists he consulted.
 
> As a result, Baxter and MR use Niger-Kordofanian as an example of a
> well-established language family established purely on morphological
> grounds and hence contradicting the claims of Donald Ringe (and his
> many admirers in the linguistic community) that tests on Swadesh
> lists are sole way to determine linguistic kinship.
 
I do not know if Ringe has ever made such a bald statement.  In any
case, I do not endorse such a view.
 
*However*, if it is indeed the case that the Niger-Congo languages
exhibit shared noun-class markers but few or no shared lexical
cognates, then I at once become suspicious.  Why should noun-class
markers be so amazingly resistant to obliteration or replacement when
nothing else is so resistant?  Should we not at once be wondering
whether noun-class markers might have diffused across language
boundaries, thereby producing a spurious family?
 
Don't think that's possible?  Well, consider Basque.  Basque
historically has no trace of grammatical gender in any guise.  But, in
this century, Basque has started to acquire a bit of grammatical
gender from Spanish.  And not just gender, and not even just the
Spanish masculine/feminine contrast.  The Spanish gender-markers /-o/
and /-a/ have diffused into Basque, and not only in loan words: *even
a couple of native Basque nouns have acquired masculine /-o/ and
feminine /-a/.*  Of course, this is on a trifling scale now, but what
might happen if Basque endures another thousand years next to Spanish?
Might it not acquire a Spanish-type gender system, complete with the
markers?  And might a linguist of the future not look at Basque and
Spanish and conclude "Hmmm...shared gender-suffixes.  Good evidence
for a genetic relationship."?
 
Don't get me wrong: shared morphological markers surely constitute
good evidence for a relationship.  But, in such a seemingly odd case
as Niger-Congo, I am not eager to jump to the conclusion that no other
explanation is possible.
 
> As for Afro-Asiatic, it is also a very well-established language
> family, albeit there is some dispute about one small subgroup
> (Omotic), but there are no competent scholars who have disputed in
> modern times that Semitic, Egyptian, Berber, Chadic, and Cushitic
> are all related (although some people without the competence in the
> field, such as Gerhard Doerfer and a few Semitic philologists, HAVE
> questioned AA).  The relatedness of these languages is based on both
> morphological and lexical connections too numerous to list here, but
> it is certainly not true that only morphology is involved.
 
OK; I stand corrected, but I wasn't questioning AA anyway.
 
> Finally, I find it a nice irony that the extremist critics of
> progress in linguistic classification (and even some more sober
> minds, like at times Meillet) have historically often insisted
> precisely that it is ONLY morphology which can serve as the basis of
> linguistic classification.  The reason for this is obviously that
> many language families which they did not like were set up on
> lexical evidence alone.  Here we find the opposite situation, where
> the morphology is precisely the basis for classification, and now
> THIS is not good.  All this seems to indicate that the
> self-appointed critics are united by no coherent intellectual
> position other than rejectionism.
 
Unfair, I think.  I certainly wouldn't want to claim that "only"
*anything* can be accepted as evidence.  Evidence is evidence,
wherever we can find it -- though not necessarily good evidence, and
not necessarily compelling evidence.
 
Anyway, it is not my policy to reject anything just because I like
rejecting things, and I don't know many other people who are different
in this respect.  We all have our soft spots and our prickly bits, and
what bothers me will likely not bother the next person.
 
We also all have our tastes and our priorities.  You clearly have a
passion for finding new families.  I don't.  I'm happy if somebody can
find new families, but my interests lie elsewhere, and I don't lose
any sleep if a couple of years goes by with no new families.
Moreover, I am naturally cautious, and my caution has been accentuated
over the years by experience.  So I don't rush to embrace any proposed
new families, and I try not to assume that a posited family is valid
just because this has been repeatedly asserted.
 
> What we need in comparative linguistics is a recognition that some
> proposed language families are valid (e.g., Uto-Aztecan, Altaic,
> Afro-Asiatic, Niger- Kordofanian, Pakawan, etc.),
 
This is what we need?  Alexis, I have talked to several specialists in
Altaic languages, and they keep telling me they don't buy Altaic.  I
know this drives you up the wall, but I really think I have to attach
some weight to their views.  I simply am not prepared to believe that
every specialist I run into is by nature a bloody-minded ferninster.
 
> others possible but still unproven (e.g., Nostratic, Khoisan,
> Coahuiltecan, all theories regarding Tonkawa, etc.), others still
> impossible (e.g., Hungarian-Turkic), and that even with respect to
> families that are valid it may be that some of the work proposing
> them is itself invalid (usually because it is premature, e.g., the
> pre-Ramstedt work on Altaic or the pre-Hubschmann work on Armenian
> as an Indo-European language or Sapir's work on Coahuiltecan).
> There is it seems to me increasingly clearly a dichotomy between (a)
> linguists who see this, whether they agree about particular
> proposals or not, and (b) linguists who do not see this and who,
> usually without understanding the particular factual issues, either
> reject/accept any language family that they can get away with
> rejecting/accepting.
 
Here I must agree with you, though I myself am more annoyed by
enthusiastic acceptance than by curt dismissals.
 
> There is in my book no intellectual difference
> between these latter two kinds of extremists--and all the difference
> in the world between both kinds of extremism vs. real work in
> linguistic classification which has over the course of this century
> solved major problems esp. In African, East Asian, Australian, and
> North American linguistics (e.g., Uto-Aztecan, Algic, Sino-Tibetan,
> Anatolian as part of Indo-European, Vietnamese as part of Mon-Khmer,
> Afro-Asiatic, Niger-Kordofanian, etc. etc.) and which is now
> producing new results (e.g., Vovin's Ainu-Austroasiatic, my Pakawan,
> all the work, e.g., Hayes's, on Austric, etc. etc.).  Can we not
> stop all the madness?
 
Doubt it.  Every proposal is going to be criticized, and it must be
inevitable that evidence that looks persuasive to one linguist will
look anything but persuasive to another.
 
Anyway, I reject the term "madness".  One of the things I *am*
interested in is models of linguistic descent, and I happen to believe
that the recent work in this area is potentially of profound
importance.  I was brought up to believe that languages just turn
into daughter languages, and that's the end of it.  But the message
coming through a good deal of recent work is "Look, folks -- things
just ain't that simple."  The study of diffusion and convergence
phenomena, and also the increasing weight attached to social factors
in language change, have begun to persuade me that we need to tread
more carefully than we once thought in setting up genetic links.
 
Accordingly, I am much exercised by the idea that we need to think
long and hard about what exactly we can count as evidence, and I am
being persuaded that we need a lot more empirical data from the study
of language change.  The whole enterprise of identifying genetic links
rests upon an understanding of what constitutes evidence, and I am
beginning to suspect that our understanding of evidence is more naive
than I formerly thought.  Ten years ago, I would have said "Shared
noun-class markers? That settles it."  But today I am not so sure.
 
Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
England
 
larryt at cogs.susx.ac.uk