LINGUIST List 5.1488

Tue 20 Dec 1994

Disc: Comparative method

Editor for this issue: <>

Directory

1. , Comparative Method
2. gor05, "SYNTAX AND THE COMPARATIVE METHOD
3. Karl Teeter, Re: 5.1468 Comparative method

Just as Morris Halle once pointed out that ANY phonological
system could be analyzed as having two phonemes, so too it
seems clear that any two languages, related or not, could
have a reconstruction done for them, involving BOTH a set
of correspondences among the words or morphemes AND a comparative
grammar. And just as in fact a phonology with only two
phonemes would be just a formal excersise and not a real
phonology, so too the "reconstruction" I am referring to
would have no validity UNLESS the languages really are
related (or in the case of mixed languages, unless the
relevant parts were related). But the formal possibility
is important to the following extent: it shows that we cannot
assume that the ability to write a comparative grammar is
either necessary or sufficient for showing that the languages
in question are related.

It is the ability to write a CONVINCING comparative gramm`r
that is significant here, but by the same token language
relatedness can be sjown by establishing a CONVINCING
set of correspodences between SUITABLE sets of words or
morphemes of the languages being compared. I will even conceded
that it is harder, in general, to be convincing in the case
of a comparative grammar than it is in the case of a set
of correspondences (since the former would seem to include
the latter PLUS a lot more). But what no one has yet shown
here or anywhere else is that the ONLY way to establish
language relatedness is by writing a complete comparative
grammar, and even Meillet conceded by 1925 that you cannot
even demand a fragment of a grammar because there are
perfectly well-established families where this either has not
been or indeed cannot be done.

It is another matter 9and ultimately a more interesting one)
that significant amount of grammatical comparison (though
not a whole grammar) HAS been offered for Nostratic (a theory
I find likely though perhaps not yet established) as well
as for Amerind (a theory which I find basically as unsupported
as do Karl and Victor, for example).

And it is perhaps not uninteresting that Illich-Svitych
proposed to write a Nostratic comparative grammar but never
did, much as Sapirpromised to write one for Uto-Aztecan but
failed to (though we should in all fairness note that
Sapir lived for another two decades, whereas Illich-Svitych
died without even finishing the lexical and morphemic comparisons
which are the bulk of the evidence that we have in the public
domain for Nostratic). This is no criticism of Sapir, of course,
IF we agree that there was NO NEED to write a comparatve grammar
of Uto-Aztecan in order for everybody to be able to see that
the Uto-Aztecan languages are related. Just as there is
no need for this in the case of Nostratic 9and indeed given
the age of Proto-Nostratic, if there was such a language,
it seems unlkikely that we will ever have more than a very
partial grammar

Concerning syntax and the comparative method:

If I understand it correctly, amrares.cs.wayne.edu (NO NAME)
apparently believes that under certain circumstances
one can indeed use syntax as part of the comparative method and cites among
other things, the case of a neuter plural in Classical Greek and Old
Iranian used with a verb in the singular and its
possible implications for a similar structure in the proto-
language. While I do in theory believe that a particular construction can
indeed survive the ravages of time and give us clues to the syntax of
earlier levels, this must ALWAYS be the last criterium we apply and can
almost never be used as a sort of proof - unless of course, (I believe
Scott De Lancey mentioned it, but I'm not sure) we can show that, for
example, all ergative case endings (I believe the example was Tibeto-
Burman) can be traced back to a common form and the MORPHOLOGY, when
reconstructed from the 'daughter' languages, conclusively points, in this
case, to an ergative 'mother' language.

Otherwise, John Cowan seems to have
summed it up best, citing Robbin Burling's book: Man's Many Voices...
Just as every language at one point or another borrows lexemes from another
langauge, there are literally thousands of examples of langauges borrowing
syntactic constructions from other languages, whether substrats or not -
even English has a good share of these borrowings. We cannot forget that
an extremely large portion of the world's population is by necessity bi- or
tri-lingual, many of whom cannot even read their own native language, not
to mention the other(s) or have any kind of linguistic training or sensiti-
vity and would therefore pay much attention as to what language a
particular construction comes from .

There are also the famous 'Sprachbunds', such as the Balkan area and to
a certain extent South Asia, where through everyday necessity, a large part
of the population speaks at least two langauges, as was shown in the case
of the village along the Maharastra-Mysore boundary, (mentioned by John
Cowan above) where one can
translate word-for-word from one language into another, although the
languages are not - strictly speaking - related. In this case, though, it
doesn't even seem to make sense to speak of language families. In my
opinion, the only possible connection is that of language contact or
Sprachbund. Maybe that is part of the problem - what are 'related'
languages, anyway?

We must also remember that this process has undoubtedly been
going on ever since people could speak. It wouldn't be difficult to propose
, going back to our starting point, a theoretical language contact area for
Greek and Iranian. For example, Georgian, at least the modern language,
only uses the plural verb for humans. Although objects can appear in
the plural, the verb appears in the singular. As there have always been
Greek 'colonies' throughout the region, one could just as well assume a
kind of sprachbund for the whole region at that time, which would be even
easier as we know so little about the pre-historic era in that region,
making it difficult to disprove.
Even today, there are certain 'areal' characteristics such as evidential
perfects, and many languages are, or have been at one point, split ergative
in the region stretching from Nepal to the Caucusus, etc. Seen in this
light, then, we can't really speak of the 'oddity' of the construction with
a neutral plural taking a verb in the singular, which, by the way,
in my opinion at least, does have a certain semantic justification and is
by no means 'odd'.

Concerning Indo-Aryan, the comparative method applied to
syntax would certainly give us the (wrong) conclusion that Vedic Sanskrit
was a split-ergative language, as all modern Indo-Aryan languages except
Standard Oriya and Begali (and Assamese?) are either ergative or split-
ergative languages, and many dialects of these three langauges are also
split-ergative languages. As there are several hundred IA 'langauges',
wouldn't we be forced to assume that this goes back to the parent language?
But there is also a problem here with the traditional method. We have (to a
large extent) cognate endings for the ergative case. So why isn't Vedic a
split-ergative language? Because we have gone back too far. Split
ergativity IS part of the 'parent' language, but this is presumably to be
found in Middle Indic. For a language family about which we know next to
nothing, then, this would certainly lead to tremendous errors.

So, it seems to me that, although there ARE indeed useful
insights to be gained by looking at the syntax, it should, as I stated
earlier, always be the LAST criterium, if used at all, in arguing for a
kind of genetic relationship. When we have no written history of the
peoples of the area, very little in the way of documented sound changes,
etc., we simply cannot say anything yet about the syntactic structure of
the proto-language. That will undoubtedly come about in the course of time,
but it is just one of many features, none of which alone can 'prove'
anything, except perhaps contact. If there are enough of the 'unlikely'
correspondences (i.e. 'ist/sind' and 'est/sunt'), then one can speak with
relative security of a genetic relationship, whatever that exactly means.

John Peterson, Kiel, Germany

Message 3: Re: 5.1468 Comparative method

In response to Lloyd Anderson's looooong and interesting disquisition in
5.1468 I have the impulse, which I hope the editors of the List will
allow, to introduce a light touch (something I once tried to do at the
Haas Festival Conference when discussing my field work, and was
excoriated for). In reading from Lloyd how Greenberg had this "single
flexible flap in his notebook bearing the language names, and sheets
which he matched up against that to enter data for particular lexical
items. This is a mechanism subject to errors..." This brings to mind an
anecdote I have heard several times, never daring to ask Chas Hockett if
it were true, in whole or part: When Bloomfield was doing Menomini, in
the 1920s when paper was expensive and linguists were paid even more
poorly than now, Bloomfield compensated by using each 3x5 slip four
times, thus getting maximum entries with minimal use of paper. When his
notes were handed on to Hockett, in cardboard file boxes of slips all
carefully alphabetized, it is said that the movers dropped one box and
its slips fell out...the rest is best left to imagination, which may be
the source of the whole story, for all I know. But I wanted to pass it on
in the hope of relieving the tedium I feel from time to time. Sorry,
folks. Yours, Karl