LINGUIST List 5.1448

Wed 14 Dec 1994

Disc: Comparative Method

Editor for this issue: <>


  1. Sally Thomason, Re: 5.1433 Comparative method
  2. Karl Teeter, Comparative method
  3. Helge Dyvik, Re: 5.1433 Comparative method
  4. , Comparative Linguistics, Response to Teeter and Golla

Message 1: Re: 5.1433 Comparative method

Date: Sun, 11 Dec 1994 20:52:19 Re: 5.1433 Comparative method
From: Sally Thomason <>
Subject: Re: 5.1433 Comparative method

 I've been enjoying reading Karl Teeter's messages on the
 comparative method, since I agree with his arguments, but one
 item in his 12/10 comments seems misleading (maybe I misunderstand
 it, but some clarification might still be useful): he says that `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 a third plural sind, and that Latin has one
 with a third singular est and third plural sunt, is all by itself
 sufficient to guarantee the relatedness of German and Latin.'

 Of course this is normally quite true, but there are exceptions.
 Structures do get borrowed, sometimes. So, for instance, there is
 general agreement that the Tanzanian language Ma'a (also called
 Mbugu) was not originally a Bantu language; in his African
 classification Greenberg grouped it with Cushitic languages because
 there are cognates in the basic vocabulary with (other) Cushitic
 languages, and Ehret used Ma'a data in reconstructing Proto-Southern-
 Cushitic phonology in a 1980 book. But the language has long been
 famous for its dramatically mixed structure: it has few structural
 features that are clearly of Cushitic origin, and it has an entire
 inflectional morphology (as well as other features) adopted wholesale
 from Bantu lgs., mainly KiPare and KiShambala. One of these features
 is the irregular negative + 1sg prefix, which (as in some Bantu languages)
 contrasts with other members of the negative paradigm, which have
 separate negative and person/number prefixes. This is just the same
 type of feature that Teeter cites as obvious evidence of the
 relationship between (say) German & Latin. But nobody claims that
 Ma'a, at least as it was in the 1930s through about the 1960s, is
 a Bantu language (though today its speakers have apparently shifted to
 speaking mainly Bantu, with Ma'a reduced to a kind of jargon);
 the few Cushitic structural features that still remained thirty years
 ago seem to be gone now, to judge from recent reports by Matthias Brenzinger
 and, most recently, Maarten Mous (whose detailed lexical analysis even
 casts doubt on the previously accepted Cushitic origin of the basic

 There are other cases of mixtures as spectacular as Ma'a, and
 many more cases of structural borrowings that are less disruptive of
 inherited structure. Some structural borrowing involves borrowed
 morphemes as well; in other structural interference only the
 structure is transferred -- the new features are expressed with
 native morphemes.

 But Teeter's larger picture is still correct, because one cannot
 use Ma'a as a whole in reconstructing Proto-Bantu or Proto-Cushitic
 or proto-anything else, since there is a serious discrepancy between
 its basic vocabulary and its morphology (among other things).
 That's what Boas predicted about language mixture, many years ago:
 mismatches in different subsystems of the language. A case like
 Ma'a is a nice example of the problem with relying on vocabulary
 comparison alone to demonstrate genetic relationships among
 languages: you are liable to get a wrong answer. Teeter is also
 correct, I believe, in saying that `if one goes back far enough
 it may even be impossible' to distinguish shared lexicon due to
 borrowing from shared lexicon due to inheritance; and again, this
 has been known since Boas (at least).

 Sally Thomason
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Comparative method

Date: Mon, 12 Dec 1994 12:14:44 Comparative method
From: Karl Teeter <>
Subject: Comparative method

 [BOn and on we go! This is a response to the message of Alexis
 Manaster-Ramer concerning mine on the methodology of comparative grammar:
 (1) AMR: "Karl Teeter refers to Meillet's celebrated statement about
 being able to relate German and Latin on the basis of the irregular
 allomorphy of the verb to be (ist : est, sind : sunt). This is very true,
 of course, but it is also true that one can immediately tell that German
 and Latin are related by looking at a small part of the vocabulary, e.g.
 Vater : pater, Mutter : mater, etc." I am very sorry, this is not true,
 and shows that, unfortunately, the point I was trying to make was not clear!
 First, the "celebrated statement" referred to is not, to the best of my
 knowledge, Meillet's, but my own, in the style of Meillet. Second, it has
 nothing to do with irregular allomorphy, but is a demonstration of a real
 example of GRAMMATICAL lookalikes, guaranteed to prove genetic
 relationship because grammar (short of learning a language) is exempt
 from borrowing. The vocabulary lookalikes Alexis shows may guarantee
 relationship, as he says, but not genetic relationship. This is the
 problem Meillet, and I following him, finds with vocabulary comparisons.
 My example (following Meillet) is about genetic relationship; that of Alexis
 is about similarities which may as well be the result of borrowing as of
 genetic relationship.
 (2) AMR: "Karl says that it is only by writing a comparative grammar
 that we can show a group of languages to be related. I simply do not
 understand where this comes from." Sorry, Alexis. Once again it comes
 from the same place, from me following Meillet. Since vocabulary
 comparisons cannot in principle be probative of genetic relationship, the
 question is, what is? Meillet's frequent discussion of this point makes
 it clear that what is needed is grammatical rules (not necessarily
 paradigms, irregular or not) which can be ascribed to a protolanguage and
 traced to the descendant languages. Summing his many assertions of this
 principle leads me to assert as his principle that what is probative for
 genetic relationship is, as I have stated previously, a demonstration of the
 possibility of writing a grammar of the protolanguage (not a
 "comparative grammar").
 (3) AMR: "Karl does not respond to my most important point,
 that it is possible to write a spurious comparative grammar for
 unrelated languages." Actually, writing a grammar is not that easy,
 as I can fervently testify from my experience when my teachers sent me
 into the field to interview the last living speaker of Wiyot with the
 assignment "bring back a grammar". A real grammar is hard enough to
 write, a spurious one even more difficult. And in fact, this is an
 area where we in 1994 have a major advantage over Meillet. That
 advantage results from the work of Noam Chomsky and many others who
 have studied language typology within his framework and also following
 Roman Jakobson, such as Morris Halle, Ken Hale, and John Whitman. A
 grammar of a human language, thanks to this work, is coming to be a
 recognizable object. Many years ago Mo"ller wrote a phonology of
 Proto-Semitic Indo-European which is now easily recognizable as
 spurious because basically all it does is add the units from
 Proto-Semitic to those of Proto-Indo-European, to come up with a
 phonology which is impossible for a natural language. AMR: "I could
 write a French-English comparative grammar, for example." You could
 not. Actually, maybe you could. What you could not write would be a
 plausible non-spurious grammar of Proto-French-English, the example I
 (4) AMF: "There is to this day no comparative grammar of
 Uto-Aztecan in the relevant sense, yet no one would dispute their
 relatedness". It has been amply demonstrated by a number of scholars
 that is possible to write a grammar of proto-Uto-Aztecan, the
 way I put the claim, and a number of scholars have performed a portion of this
 task. Were it not for the tragic death last summer of my friend and fellow
 student Wick Miller, an excellent grammar of proto-Uto-Aztecan would be well
 on the way.
 (5) Finally, AMR: "Karl, and many others; appear to have come up
 with the idea that you CANNOT show languages to be related EXCEPT by
 finding such paradigmatic oddities as those which Meillet pointed to".
 Sorry, Alexis, this is simply not true; it bears no resemblance to
 anything that I have stated or anything that I believe. I agree that an
 understanding of the methodology of comparative grammar is important in
 the search for wider genetic relationships, and when I see a grammar of
 proto-Nostratic, it will do a great deal for my skepticism. Karl (=Karl
 V. Teeter, Professor of Linguistics, Emeritus, Harvard University)
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Re: 5.1433 Comparative method

Date: Mon, 12 Dec 1994 11:52:05 Re: 5.1433 Comparative method
From: Helge Dyvik <>
Subject: Re: 5.1433 Comparative method

 Karl Teeter writes:

)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.

 With respect: can this be read as anything but a dismissal of the whole
 methodological discussion as meaningless?

 Helge Dyvik

 Helge Dyvik
 Department of Linguistics and Phonetics
 University of Bergen Phone: +47 55 212261
 Sydnesplass 9 Fax: +47 55 231897
 N-5007 Bergen, Norway E-mail:
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 4: Comparative Linguistics, Response to Teeter and Golla

Date: Mon, 12 Dec 94 09:35:00 ESComparative Linguistics, Response to Teeter and Golla
From: <>
Subject: Comparative Linguistics, Response to Teeter and Golla

 Let me ask this: What are people so afraid of that, instead of
 looking at the substantive issues surrounding the Nostratic
 hypothesis (looking at, I say, not necessarily accepting the
 theory), they are so busy inventing all kinds of unprecedented
 and unmotivated a priori methodological strictures?

 Now, as to specifics, Victor Golla gives an accurate account
 of the story of Sapir's paper on Uto-Aztecan, but the points
 I am making are (a) that, whether he intended to publish a
 morphology or not, he never did (nor is there a ms., I don;t
 think), and YET both he and everybody else accepts that these
 languages are related, and (b) given the information avaialable
 to Sapir and given the way Uto-Aztecan languages actually are,
 I don't think that any morphology he might have written at
 the time COULD have contained the kind of paradigms or morphological
 oddities that delight the Indo-Europeanist and that Karl Teeter
 seems to insist on before he will recognize the relatedness of
 a group of languages.

 As for Karl's response, he now says that he is merely talking about
 reconstruction not about classification. But surely he does not
 mean that a comparative grammar is a prerequisite to a reconstruction;
 it is PART of a reconstruction. And since the whole debate started
 with the question of classification (e.g., are the Nostratic languages
 really related?), I will reiterate that we canNOT demand a detailed
 morphological reconstruction UNTIL the languages are acceptyed as
 FIRST and then ONLY if the languages in question HAVE morphology
 to speak of.

 If we accepted Karl's reasoning, we would not I think be able to
 recognize that Vietnamese is related to Khasi, for example, or Sinhalese
 to Afrikaans, or certainly the various Uto-Aztecan languages to
 each other (or Mandarin to Cantonese, I suspect).

 I should add, too, that I have NEVER accepted the validity of
 Nostratic but rather have been exploring its various implications
 rejecting many of Illich-Svitych's proposals but also finding
 significant evidence that he was right (see, e.g., my paper in
 the latest Diachronica). It is this kind of substantive work
 which will one day decide the correctness of any theory, not
 the methodological innovations which are being proposed as a way
 of justifying not ever having to look at the substantive issues.
 I should also say that, since no one in this discussion has
 so far been interested in the substantive issues, we have NOT
 established that there is NO reconstruction of Nostratic
 morphology comparable to those for some well-established families.

 What if there IS? The real question will then be whether
 the reconstruction is valid, for as I have said earlier I can
 write a comparative grammar of Proto-Wilhelmian, the ancestral
 language of modern English and modern French which we will
 assume was spoken by William the Conqueror. Even though such
 a grammar would be absurd. The point, pace Karl, is not whether
 we can write a comparative grammar; the point is whether the
 actual, substantive proposals regarding phonology, lexicon,
 and as much grammar as might be available (very little in
 some language families) are valid or not.

 Alexis MR
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue