Alexis' methodological points (was: Linguistic classification)

[Ralf-Stefan Georg]

----------------------------Original message----------------------------
At the moment I hope I'll be forgiven not to address the Pakawan question
(I do have Alexis' papers on it, but at present I haven't seen Campbell's
intervening reply, and I'd like to see the whole picture before).
 
The methodological claims/questions/proposals Alexis tabulates for
discussion are of course not new, nevertheless I think most of us
interested in such questions share with Alexis and others (e.g. me) that
something like a solid set of principles for linguistic relatedness, known
to everyone in the field and accepted to a degree which would allow us to
call it "established" may still be in want.
A different question is of course, whether such a thing can really exist,
but along the lines of Alexis' posting I'd venture some remarks:
 
" NO list of
differences between two languages can be an
argument against their relatedness."
 
I think there can be no serious fight over that; that argumenta e silentio
have no serious value in any scientific discussion should be clear. That
some anti-Altaicists, e.g. Gerhard Doerfer and at times A. Shcherbak have
placed too much emphasis about things which simply were'nt there (numerals
and other "basic vocabulary" stuff) instead of looking at what is there,
are palpably positions of by-gone days now (they can, however, be traced
back to Nemeth, Bang and other early researchers from the childhood-days of
comparative Altaic).
 
" I only
would like to know what people think about the
question of whether diffrences or disagreements
can ever in principle be used to argue against
a relationship."
 
And again, I cannot imagine how any meaningful answer on this question
could be in favor of counting "differences". But one tiny little thing:
 
If diachrony comes into play, i.e. if we are in the blessed situation to be
able to observe languages change over a considerable period of time, and if
the languages in question occupy a large-enough territory, so that (vaguely
speaking) notions of periphery vs. centre (with all the well known
diachronic implications) could be meaningful, the notion of "difference"
acquires a somewhat different and potentially interesting meaning:
 
*If* the languages observed show a greater deal of "difference" (and I'm
fully aware of the fact that this term should be properly defined before
thrown into the debate) in earlier stages of their attestation and greater
uniformity on different levels of their systems later (the geographical
analogon would be an observation where languages at the periphery of a
given area are "more different" than those nearer to the centre), a
potentially good hypothesis would be something along the lines of
"convergence", rather than "divergence", logically entailed by the
assumption of geneticity.
 
NOW: before Alexis slaughters me: of course, a scenario is thinkable, where
original relationship (i.e. *identity*) first led to dissolution of an
original parent language (*increasing* "differences"), and *later*
convergent processes might have happened, which have led to the observable
processes (after all, the  Balkans area is made up of languages, which are
related). But these are *two* assumptions instead of one and the principle
of Occkham's razor (which has seen so many lip-services paid to it) would
at least give the following advice: *first* investigate the *convergent*
processes and every data potentially pertinent to it, and *then* see what
is left to make a point for the relationship. My personal acquaintance with
Altaic linguistics, for a randomly chosen example ;-), is that rarely the
same people are honestly investigating both sides of the coin and rather
say, like Doerfer or Clauson, (wrongly) "the non-relationship has been
shown", or, like R.A. Miller (also wrongly) "the relationship is clear
beyond the shadow of a doubt".
 
Maybe to hope for a change in that situation would simply be too pious a
wish to be realistic. However, it would be a step forward if people would
accept antagonistic views at least as potentially interesting *working
hypotheses*. Whoever has read a little bit of the Altaic civil war, will
know that the field is far from that.
 
What is the *working hypothesis* of, say,  the Anti-Altaicists ? That those
convergence processes, which I alluded to above, are the whole story, and
that "Altaic" is a fata morgana, quite simply so.
(I may add in parentheses here that I cannot help thinking exactly that).
However, this does *not* mean that I claim to "know" that those languages
"aren't related", or, for that matter, that I'm able to "show" this here
and now (or ever) to everybody's satisfaction. I only plead for some degree
of acceptance of a congergence-model for these languages as a *working
hypothesis* for the time being. Next thing to do is to go as far as it will
be possible on this path and to see what may be explainable in terms of
areal convergence in these languages (I herewith promise that I'll do it
;-), and *then* the discussion may find itself on a new footing: for,
surprising as it might seem, an "anti-Altaic" theory simply is not on the
market. The only thing which *is* there, are occasional, sometimes
emphatic, remarks (which are all too often off the
linguistic-methodological mark) about this and that weakness in the
proposals of the Altaicists, never offering a systematically worked-out
alternative scenario about what might have happened in history . Only when
such a thing will have materialized something which might be called an
Altaic *debate* can *begin* (after all those centuries of talking in
circles).
I apologize for dwelling so long on Altaic, but I take the opportunity to
draw attention to the less-than-satisfactory climate in Altaic studies,
especially since I'm personally committed to work out (fragments of) such
an alternative theory - often promised by Doerfer et al. in an often
unpleasant voice of authority, but still lacking - and fear that at least
some people in the field might be determined to lambaster it before a
single word of it has leaked to the public.
 
So, as a *working hypothesis* the assumption of "non-relationship" clearly
has to be allowed.
 
Which brings me to Alexis'
 
"Note II: Hamp's and my position assumes that the
burden of proof is always on the advocates of
a relationship, and that therefore its opponents
can do no better than to cast doubt on the evidence
cited by the proponents--but cannot go further to
provide evidence against the relationship."
 
Yes, full agreement on the first part, but the second part deserves some
musings, I think. Casting doubt on the evidence, OK, not providing evidence
against the relationship, also OK (with the emphasis on *evidence*, which
recalls the argumenta e silentio-issue, alluded to above).
I think the problem possibly leading to misunderstandings here is of an
epistemological nature. It would be too easy to draw a caricature of this
position like "say lgs. A and B have *nothing whatsoever in common* (a
purely theoretical scenario, of course); I could say they are related after
all, but the great time-depth involved eroded even the tiniest shred
evidence". Certainly, this is not Alexis' position, nor could I imagine it
being defended seriously. But a situation where the *hypothesis* that two
(or more) given languages are quite simply unrelated is viewed as a
position which violates the rule of our trade is, I feel, somewhat less
than satisfactory. A science where only saying "yes" is regarded as
meaningful, and accordingly saying "no" is held to be an unscientific
position (thus saying "non liquet" being the only alternative to express
any doubt) gives me a mild headache at least. Every assumption of
relationship, including the craziest ones, would thus be entitled to claim
the status of a "not-yet-proven hypothesis" and by consequence make its way
into the Encyclopedia Britannica.
 
Possibly, a way out of this dilemma should envolve the proper definition of
the terms we are operating with. For a start, to say "A and B are related"
is no hypothesis at all, let alone a theory. A simple (and simplistic)
statement like this can only be the *bottomline* of a larger construct (a
"set of interdependent assumptions"). A hypothesis worthy of any
consideration consists of a set of assumptions, principles,
correspondences, "sound-laws" for that matter, etymologies and so on, the
bottomline is only a *part* of. So, a set of assumptions like
 
- the words A /paluka/ and B / firch/ are historically identical ("sprung
from some common source")
 
- so are  a lot of other words in A and B
 
- the segmental phonemes displayed by them show recurrent and largely
predictable ("regular") correspondences
 
- the assumption of a proto-system on the phonological level allows to
explain and understand some opaque forms of A and B as due to productive
morphological processes on the proto-level, obscured by later sound-changes
 
- therefore A and B are genetically related
 
*is* such a hypothesis, whereas the bottomline alone is *not*.
 
(Please note, that I don't mean this to be a set of "relatedness criteria",
let alone a full one, I'm just talking as generally as possible; different
sets of different assumptions are of course possible).
 
If, now, it can be shown that one or all of the "pre-bottomline" claims is,
well, wrong, what are we to call A and B ? Languages not-yet-proven-to-be
related ?
 
I tabulate that we should call them at least *heuristically unrelated*,
i.e. no sufficient point for the relatedness has been made, therefore the
working hypothesis that they aren't is the best one (or at least one which
is *not forbidden*).
 
*For the time being* of course, but who can claim that his/her
theories/hypotheses are here to stay for ever ?
 
The common misunderstanding involved here - one which lurks between the
lines of Alexis' posting as well, I'm afraid - is that to say "unrelated"
entails a statement of "truth", i.e. A and B "are" not related, originated
completely independently in totally different (pre-) hominid groups without
the slightest possibility of mono- or oligogenesis, and that this is an
incontrovertible "result" of my/our "scientific research" which has
"proven" this.
Any statement of such radicality could of course safely be called
unscientific, or even nonsense (if we are in that mood), but, othoh, it is
rarely formulated as such.
 
But, *given* all those methodological/epistemological precautions - and I
feel they are pretty trivial - should we not be allowed to call A and B
"unrelated" (i.e. of course *heuristically* unrelated *for the time being*)
as long as we are not convinced of the merits of any claim to the contrary
which happens to be on the market ?
 
When some non-linguist asks me what Basque might be related to, I
invariably answer "to nothing else", since that is the state of the art as
of now. I'm not claiming at the same time that this is *impossible* to
change some day. Should I really enumerate all the hypotheses re: Basque,
labelling them as "yet unproven", if asked such a question ?
 
To conclude: when I - as a facon de parler - talk about A and B being
"unrelated" I don't state that it is *impossible* that they are after all.
I'm only using "unrelated" as shorthand for the state of the art as of now.
The *possibility* remains that this statement will sometimes turn out to be
wrong. But, in the absence of conclusive evidence, it is only that, a
possibility. Furthermore, there are two kinds of possibilites: "full"
possibilities, i.e. relationships which have been claimed on the basis of
well-formulated and falsifyable sets of observations and assumptions, which
however might be still controversial for different reasons, and "empty"
possibilities, where no such claim has been made, or, for that matter, all
claims to that effect have been successfully demolished.
Given a *full possibility*, e.g. Altaic, I usually answer such questions as
the abovementioned with "possibly related, but personally I think along
different lines", given an *empty possibility*, say, Basque and Kartvelian,
I say "unrelated". Maybe we can agree on that ?
 
Of course, your "full possibility" can be my "empty" one and vice versa,
but that clearly is another (or the next step of this) discussion.
 
Which brings me to the final statement that, given the redefinition of
"unrelated" as "heuristically unrelated", Alexis'
 
"Note III: It seems to me that it follows from this
that the only way to argue against a relationship
at all would be to demonstrate a DIFFERENT relationship."
 
is not acceptable for me in this sharp formulation (although it would be
the *best* via falsificationis, of course).
 
 
Stefan Georg
Heerstrasse 7
D-53111 Bonn
FRG
+49-228-69-13-32