consensus view

[bwald]

----------------------------Original message----------------------------
It occurred to me that among our various misunderstandings, there is a
misunderstanding about
what I mean by "consensus view" of establishing language families.  I
cannot speak for
Campbell or Goodard, or anyone but myself, but I thought it was clear from
what I said.  However,
it might be helpful to look at it in terms of the commonplace logical
distinction between SUFFICIENT
and NECESSARY (to establish a family).  I put "sufficient" first, because
my construal of the
consensus view is that it is about what is SUFFICIENT to establish a
language family (beyond "reasonable" doubt).  What is NECESSARY is no doubt
another matter, one for which there
is obvious disagreement, and for which I do not have a fixed opinion, but
which must somehow be
empirically grounded in such a way that we can go from what we all agree on
to how we can enlarge agreeement about the things we currently (and for a
long time) have disagreed about.  All that I can envisage is that what is
necessary must be less than what is sufficient, and that there is a
relationship (subset?) between what is sufficient and what is necessary,
though just what that relationship is has to be a matter for continued
discussion, if we are going to "establish" (or even get widespread
"acceptance") of deeper family relationships where evidence is (presumably)
lacking to the current consensus view of sufficiency (cf.  "there's
INSUFFICIENT data to establish the family according to the consensus view,
the consensus methods of demonstration.")

So I think AMR misunderstood my use of  "consensus view" to be about what
is NECESSARY to
establish a family relationship, when, in fact, it is about what is
SUFFICIENT.  We are still
trying to figure out what is necessary.  At least I am -- for purposes of
this discussion.

I'm gonna try to keep this message short, so I can't give all my thoughts
on the subject of what is necessary (and I don't know the answer anyway).
Foremost in my mind is the basic concept of language family.  Previously I
characterised it as the idea that all the members descend from a SINGLE
ancestor.  I think that is what people usually have in mind, but in the
final analysis it's a big problem.  All languages have historical lexical
strata with respect to source.  Only the "oldest" stratum seems to count as
criterial to "single ancestor".  And for languages for which there seem to
be insufficient data that's a big problem:  how to identify the oldest
stratum.  That's where such issues as "basic vocabulary", presumably most
resistant to "borrowing", comes in -- and there has been a lot of
contention about all aspects of this problem, when it comes to the
practical details (I haven't mentioned "grammatical" morphemes yet but
that's also an aspect of the "layer" problem). In fact, it seems to me that
"oldest" stratum is only a relative term.  If we consider families that are
recognised by the consensus view/methods, if they have deeper relatives, as
we infer by the uniformitarian principle of going from what we know to what
we don't know, then what appears to be the oldest layer is actually the
COMPRESSION of a bunch of older historically arranged layers.  How to
separate them out and arrange them chronologically, in order to extend the
"oldest" stratum notion?

Moving on to grammar, I have never been able to keep straight in my head
why some scholars have proposed that Mbugu demonstrates the "borrowing" of
(Bantu) noun classes, etc., rather than the "borrowing" of non-Bantu
vocabulary.  Given that Mbugu speakers are all fluent in a Bantu language
(either Shambaa or Pare), the direction of "borrowing" seems moot to me.
Thus, we seem to have the possibility of historically simultaneous layers
from different sources -- assuming that we're not gonna dismiss Mbugu as
some kind of (in-group) "slang" (which does not seem to me to be grounds
for dismissal, and which nobody has suggested anyway).  So far, it seems
that the larger world of historical linguistics chooses to dismiss such
phenomena as "rare" (uniformitarian principle?), which seems a bit shaky to
me when we are dealing with distant historical situations which represent
social situations that we don't know about.  How to reconstruct
bilingualism and "simultaneous layers" (two or more different "layers" NOT
chronologically distinguished in time)?

Not so removed from this is the (former) issue of whether Haitian (Creole)
is a Romance language.  Until relatively recent times when pidgin-creole
studies started to be taken seriously by historical linguists in general
(or have they? AMR, to be sure, seems to take them seriously),  lexicon
alone (i.e, through regular sound correspondences) seems to have been
criterial of inclusion in "language family" (for the "mainstream").
Grammar did not count for much, just as it did not count for much in
synchronic linguistic description.  Of course, the whole insight into the
"family" hypothesis came from the fact that Sanskrit conjugates and
declines pretty much like Classical Greek and Latin, but as historical
linguistics progressed it was "sound laws" (based on sound correspondences)
that seized the attention of theorists.  There was palpably less interest
in grammatical change, and work on it was often motivated by how it
interfered with sound correspondences and inferred processes of sound
change.  I must add that it was not usually a problem for demontsrating
families because families were demonstrated *sufficiently* at a level where
grammatical similarities could be taken for granted -- but such issues as
Haitian brought up DEFINITIONAL problems of language family.  How to
classify when there is historical discontinuity in the grammar?

(NB:  I could say more about "political" motivations for different sides on
whether or not Haitian should be [have been] included in Romance, but I am
saying enough that can be misinterpreted without getting into that fruitful
topic.  Currently there is still disagreement about how to account for
the grammar of Haitian and various other "creoles", e.g., whether it comes
from an innate "bioprogram" or is relexified Fon (a Kwa Niger-Congo
language), etc etc.  There are also issues involving whether many so-called
creoles actually descend from earlier pidgins, e.g., Berbice Dutch,
sometimes called Berbice "Creole" Dutch, which seems to be a "mixture" of
Dutch and Kalabari (the latter a variety of Ijoid, a branch of Niger-Congo
that has contentious aspects for (sub)classification), or various
non-European varieties of Portuguese that seem to have been "restructured",
but not necessarily descended "whole cloth" from pidgins, etc etc.  Does
"(radical) restructuring" exclude them from their lexical source "family"
affiliation?  These seem to be matters of DEFINITION of "family", not
whatever the historical "facts" may be)

I won't take AMR to task for saying that matters of sub-classification or
questions about whether or not a particular language (group) belong to an
otherwise "accepted" family are UNINTERESTING, because I assume he was
fixing on some narrowly focussed issue which takes the notion of "family"
for granted, but I disagree for several reasons.  First, the case of
Kadugli (whether or not it is Niger-Congo) turned out to be very important
in whether or not a genetic link could be established between Niger-Congo
and Nilo-Saharan.  Was it an otherwise "missing link"?   Do deep
reconstructions to demonstrate deep relations depend on proposing "missing
links" -- what are the criteria, and what are the assumptions behind this
(tacit?) notion?  Apart from that, I continue to wonder what the point of
deep family reconstuction is, apart from its original motivation to
reconstruct historical human socio-cultural trends (or some rationalised
form of "ancestor worship"), if historical details, which also reveal
historical socio-cultural trends, are "uninteresting".  It seems like
tunnel-vision to me.

(NB.  Similarly, if proposals to relate Niger-Congo to some other family
were to rest heavily on
Mande and/or the northern Atlantic languages, and they are still under
contention, then those proposals would remain shaky. Also, if membership in
a family is uninteresting, what's left?  Let Turkic be Altaic, why is it
interesting whether Mongolian and Tungusic, not to mention Korean and
Japanese or whatever, are also Altaic,  but not whether Mande and Atlantic
and Kadugli and whatever are NC?  Which group doesn't matter?  Why?)

Agh.  This is getting long -- one last comment.  AMR made the attractive
point that the actual reconstruction takes more work than the demonstration
of relationship.  Indeed, Greenberg quipped that of all the languages that
have changed in the last century, Proto-Indo-European has changed the most.
I hasten to add, no guilt by association is intended (though I, for one,
wish I had thought of the quip first, without wishing to be associated with
everything Greenberg has said).  So maybe we can abstract (sound)
correspondences (subject to lexical and probabilistic controls) as
NECESSARY, from reconstruction, which may be SUFFICIENT .  When I first
read some author who insisted that reconstructed sounds were actually
"formulae" for sound correspondences, I thought the point was overly fussy,
but clearly this is an area where we may be able to distinguish NECESSARY
(sound correspondences -- does anyone doubt it, not even disregarded by the
"crackpot" literature, for the most part) from what is SUFFICIENT
(phonological reconstruction to a plausible sound system, "plausible" also
being a continuing problem, with plausible sound changes, cf.  common
criticisms of many deep reconstructions is "too many phonemes", i.e., too
many sound correspondences, cf. the emperor to Mozart "too many notes", to
which, unlike Mozart with his retort "just as many as NECESSARY", the
response is usually to agree and hope to reduce the inventories by
discovering conditioning factors, as per the shining example of Verner's
Law).  In fact, to return to Greenberg's quip and what I take to be the
implications of AMR's observation, little has changed in Proto-IE in terms
of sound correspondences over the past century, glottalic theory and the
various proposals for the substances of the laryngeals seem to rest on the
same old sound correspondences, so there is clearly a point to draw the
line between sound correspondence and the details of a plausible
reconstruction.  But in view of the deeper "too many notes -- sound
correspondences" the question is -- where?

[NB  At some point revisions of reconstructions become interminable, just
as are revisions/new versions of synchronic grammatical analyses of
commonplace grammatical structures]

I'd like to see a systematic OUTLINE of these and further relevant issues,
esp. grammatical reconstruction -- but enough for now.   -- Benji