LINGUIST List 5.1433

Sun 11 Dec 1994

Disc: Comparative method

Editor for this issue: <>

Directory

1. Scott DeLancey, Re: 5.1393 Using the comparative method in syntax
2. Karl Teeter, Re: 5.1409 Comparative method in syntax, continued
3. Karl Teeter, Comparative grammar, reply to Alexis Manaster-Ramer
4. , 5.1420 Comparative method in linguistics

Message 1: Re: 5.1393 Using the comparative method in syntax

Fritz Newmeyer (fjnu.washington.edu) asks for opinions on the
applicability of the comparative method to syntactic reconstruction.

I would say that *comparative* reconstruction sensu strictu
can only be of forms and paradigms; reconstruction of patterns
is impossible unless it is based on reconstruction of forms. A
fair amount of syntactic reconstruction is possible this way; for
example, once we have reconstructed the IE noun declension and verb
conjugation, we have also reconstructed a nominative-accusative language.
(The "ergative Indo-European" idea, besides being clearly wrong, is not
actually a claim about PIE, but an internal reconstruction from
PIE to an earlier putatively ergative stage). If you can
reconstruct a relative pronoun (or a set of them), you can infer
something about relative constructions in the proto-language.
If all or most of the languages of a family possess apparently cognate
ergative case forms, it may be possible to reconstruct an ergative case form
for the parent language, and thus by implication to reconstruct
ergative case marking. But in the case (not a hypothetical one;
I could adduce a number of examples of this sort from Tibeto-Burman)
where most of the languages in a family or branch are ergative, but the
ergative markers are not cognate, we cannot *reconstruct* ergative case
marking for the parent language. It is hardly unreasonable to
take such data as constituting a prima facie case for the hypothesis
that the parent language was ergative, but a hypothesis not based on
reconstructible forms is certainly weaker than an actual
reconstruction.
 I emphasized *comparative*, which is the topic that Newmeyer
was suggesting. *Internal* reconstruction of syntax is another matter.
This I think is what Martin Haspelmath (martinhafub46.zedat.fu-berlin.de)
has in mind when he refers to common patterns of grammaticalization as
useful in "undoing" grammaticalization for purposes of syntactic
reconstruction:

)> In addition to regularity of change, we need general principles of change
)> for plausible reconstruction, e.g. phonological principles that predict
)> likely changes like assimilation, lenition, segment loss, etc. In syntax,
)> similar principles of change exist as well: Spatial nouns become spatial
)> adpositions, certain general verbs become tense and aspect markers,
)> allative case markers become dative case markers, purposive verb forms
)> become infinitives, etc. All these processes (instances of
)> grammaticalization) are irreversible changes and provide safe guides for
)> linguists seeking to make sense of daughter language diversity by
)> reconstructing a proto-syntax. The massive regularities of
)> grammaticalization are generally ignored in generative studies of
)> syntactic change (indeed, Lightfoot argues that there are no genuine
)> principles of diachronic syntax), but if one takes them into account,
)> they help in the difficult task of reconstruction.

This seems to be an answer to a different question than Newmeyer was
asking, although the answer may not have been out of place given N's
question and assertion:

) word in a particular environment will be mirrored by like changes in
) other words in similar environments. But what are the syntactic analogues
) of words and phonemes? And furthermore, syntactic change can be fairly
) catyclysmic, restructuring grammars wholesale in one generation --
) unlikely or impossible with phonological systems.

I wouldn't have imagined that anyone, in 1994, could believe for a
moment in the possibility of ordinary historical processes "restructuring
grammars wholesale in one generation" (could anyone suggest an example?).
This sentence does suggest the Lightfootian notion that syntactic
reconstruction is in principle impossible by any method, which, as
Haspelmath argues, is certainly incorrect.

Scott DeLancey delanceydarkwing.uoregon.edu
Department of Linguistics
University of Oregon
Eugene, OR 97403, USA

Message 2: Re: 5.1409 Comparative method in syntax, continued

Dear Linguist listers: Once more it needs to be pointed out that the
comparative grammar is not a system of comparing units and reconstructing
lookalikes, however useful this with may be with vocabulary for a practical
first approximation of linguistic relationships. What we do with
languages when we do linguistic history is no different from what we
do when we do field work; we collect data on the language (in this case a
putative protolanguage), and write a grammar of it. If
one can include a section on syntax in a grammar, one can apply the
comparative method in syntax.

AMR says I am wrong to say "you cannot classify languages on the basis of
phonological correspondences" Perhaps the problem here is his use of the
word "classify", since it is clear that everybody's first approximation
to linguistic history begins with such classification. What I say is
just that you cannot RECONSTRUCT languages on this basis, for the simple
reason that borrowed vocabulary has just as respectable a history as that
retained from a protolanguage. For history, you need a protolanguage, not
just "shared correspondences". American linguists have largely failed to grasp
this point ever since Kroeber's paper "The Determination of Linguistic
Relationship", which appeared in Anthropos VIII (1913), pp. 389-401, and
unfortunately seems to have survived the urbane rebuttal of Antoine Meillet,
"Le probleme de la parente des langues" (sorry about accent marks),
Scientia XV (1914) virtually unscathed, going on in its checkered history
all the way to Greenberg, for whom grammar and language history become
irrelevant, all you have to do any more is statistical guessing games
involving word similarity.
 Subdivisions of AMR's message: (a) "For many language families, there
IS no other basis for classification available". It may be and often is
that enough of a language is lost that there is too little data to
seriously use as the basis for writing a grammar, as with Beothuk, and here
indeed we are seriously limited; what is lost cannot be reconstructed.
But it is the linguist's basic article of faith that languages have grammars,
and it is that fact which allows us to write language history :
words may be borrowed, structures no. Thus Meillet's sort of "deep"
structural comparisons: knowing that German has a verb "to be" with a third
singular ist and third plural sind, and that Latin has one with a third
singular est and a third plural sunt, is all by itself sufficient to
guarantee the relatedness of German and Latin. Not similarities, but shared
structures. In (b) AMR says "The danger of confusing borrowing for cognates
is always real, but it is easy to [separate them]." It is not so easy -- if
one goes back far enough it may even be impossible. That is the point.
AMR's (c) "There is nothing novel in what I am saying". Nothing new,
certainly, ever since the history beginning with Kroeber I have adumbrated
above. But novel, indeed, since long since clearly refuted (for example by
Meillet in 1914). And finally, (d) "You cannot realistically expect normal
people to spend time writing comparative grammars of languages which
have not PREVIOUSLY been shown to be related." On the contrary, my
contention (not my invention), is that the only way to establish that
languages are related is to write a grammar of the protolanguage and
show how it developed into different later grammars. Whether linguists
are "normal people" is beyond my purview. Yours, Karl (= Karl V.
Teeter, Professor of LInguistics, Emeritus, Harvard University)

In LINGUIST 5.1420 Alexis Manaster Ramer writes:

) Karl Teeter is mistaken, I think, when he says that you cannot
) classify languages on the basis of phonological correspondences
) in the lexical items....For many language families, there IS no
) other basis for classification available, because they lack the kind
) of morphological complexity so beloved of Indo-Europeanists....There
) is nothing novel in what I am saying, since it is the method
) which, for example, Edward Sapir used to establish that the Uto-
) Aztecan languages are really a family (rather than three families).

While I don't want to debate Alexis' main point here, I'm afraid I must
challenge his interpretation of Sapir's goals and methods in "Southern
Paiute and Nahuatl: A Study in Uto-Aztekan." This paper was published
in two parts (part I in in the Journal de la Societe des americanistes
de Paris 10:379-425, 1913; part II in JSAP 11:433-488, 1919, as well
as in the American Anthropologist 17:98-120, 306-328, 1915). I go into
this bibliographic detail because the publication we have is disjointed
and incomplete. What Sapir actually planned is outlined in a letter
to Kroeber:

 I am sending you...the first instalment of my paper on Uto-
 Aztekan [the section on the comparative phonology of vowels]
 ...The treatment of the consonants will follow as the second
 instalment, while a third is intended to take up the points of
 morphological similarity, many of which, indeed, are incidentally
 referred to in the course of the present instalment. (ES to
 ALK, May 30 1913).

The vowel section was published in Paris in 1913. Publication of
the remainder of the paper was delayed by the outbreak of war, and the
section on comparative grammar never made it to print. But it is clear
from the introductory paragraphs of the paper that Sapir did not intend
merely to present the comparative phonology of Uto-Aztecan, or that he would
have considered this sufficient to establish genetic relatedness:

 In his resume of the problem Kroeber summarizes in tabular form the
 lexical evidence, insofar as it affects all three Uto-Aztekan groups
 ....The rather small amount of lexical evidence that is presented by
 him, unprovided as it is with definite indications of the operation
 of phonetic laws and unsupplemented by morphological evidence, can
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 ...hardly be regarded as more than strongly suggestive....The compara-
 tive Uto-Aztekan material here presented is partly phonological, partly
 ^^^^^^
 morphological in character, the purely lexical element being taken
 ^^^^^^^^^^^^^^^^^^^^^^^^^^
 notice of merely as illustrative of these. (1913:380-3, emphases mine).
 ^^^^^

One could indeed argue that what Sapir actually published of his Uto-Aztecan
comparative work makes a very good case for the genetic unity of the family
on purely phonological grounds. But if so, it does it accidentally, not by
Sapir's design.

 --Victor Golla
 Humboldt State University
 gollavaxe.humboldt.edu