linguistlist.org>
Editor for this issue: Martin Jacobsen <martylinguistlist.org>
umich.edu>Karl Teeter is of course right to emphasize that we do have, and have all along had, fairly good criteria for when a proposed linguistic relationship is to be accepted and when not, although some issues remain (as in any science). He is also right that structured sets of cognates (esp. inflectional and derivational ones) are other things being equal better evidence than lists of random lexical comparisons. Nor do I think anyone would disagree with this. If I am not mistaken Krauss accepted the relatedness of Tlingit to Athabaskan-Eyak largely on the basis of (his discovery of) striking parallels in derivational morphology, for example. However, things are not always equal, and some language families just do not have enough morphology left, and even Meillet (who was perhaps the stricted critic of purely lexical comparisosn) at length admitted that sometimes you accept relationships based on lexical comparisons alone. But there is still some room for discussion here. On the other hand, it is also true, as Catherine Callaghan showed, there are (rarely but not never!) cases where totally unrelated morphological systems by chance evolve to look very much alike, so you have to be careful. Finally, we have known since the beginning of the cetury that there are (again rarely) truly mixed languages, such as East Armenian Romany (known for a century) or Mitchif (only known more recently). But their lesson is the opposite of what the self-appointed critics of comparative linguistics (such as Thomason and Kaufman and so on) seem to think. It is IMMEDIATELY obvious that Mitchif is a mix of French and Cree, for example, and every morpheme can be assigned to one source or the other. So such examples do not show that comparative linguists are helpless in the face of mixed languages. On teh contrary, we very easily deal with them in teh same way that we have all along dealt with teh question of distinguishing inherited from borrowed elements in ANY language. It is no harder to separate the French and the Cree in Mitchif than it is to determine that woman is native to English but blitzkrieg is not. The existence of mixed languages also does not mean that we can with impunity assume (as e.g. Nichols does) that the grammatical systems of the Altaic (or Nostratic) languages are related via borrowing/mixture rather than inheritance. For the point is that, knowing of the POSSIBILITY of mixed languages or borrowing, we are nevertheless obliged to demonstarte the FACT of mixture/borrowing if we wish to assert it. We may NOT assume that Altaic or Nostratic involve merely mixture/ borrowing merely because Mithcif or E Armenian Romany are mixed languages! We have to demonstrate this, and this has not been done. Which brings up an all-important point I have been avoiding making but can no longer avoid. It used to be until quite recently that discussions/debates of both methodological and substantive issues in comparative linguistics were mainly if not exclusively indulged in by people who knew the subject at issue at first hand and the form of teh discussion was to specify just what one was saying and who one was disagreeing with. The great Altaic debate until very very recently was a perfect example of this: the main combatants knew the subject and the literature and addressed specifics. Moreover, it was typically the case that everybody, no matter the side they took, was reasonably civil, and certainly wished the field of comp. ling. well (and did not pretend to 'shout down' others or prevent the discussion of their views). This has changed radically. We now have a growing number of people who openly or not so openly wish to replace comp. ling. with something else, who (often, not always!) do NOT appear to know the literature or the subject AT ALL and spread second, third or fourth hand gossip (as in Nichols' and others' statements about Altaic and almost everything that I have heard about Nostratic), who do try to silence others and conspire to misrepresent and often to suppress the public discussion of the views they disagree with and so on. Even standard reference works such as encyclopedias (esp. in the US) have articles on major topics in comp. ling. authored by people who have no record at all in the field and omit other major topics entirely, and many if not most journals studiously avoid even now even such elementary matters as reviewing the major published work on Nostratic. All this should be entirely intolerable, and yet there has been almost no reaction to this (R.A. Miller's recent book "Languages and History" does mention some of these problems). Of course, the problem of misinformation and spreading of what amounts to gossip is endemic to ALL branches of ling. "Classic" examples are often laughably wrong on simple points of fact in subareas from typology to generative syntax and phonology to phonetics, and textbooks abound in them. Moreover, the hopelessly misunderstood point Chomsky once made about the relative importance of theory vs. data is sometimes used (as I know from personal experience) to discourage or prevent the publication of corrections of such misinformation. But I think that only in the case of comparative ling (and esp. classification) that these problems have reached the point where they threaten the very survival of the field. AMRMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue
HUMnet.UCLA.EDU>Karl Teeter recently wrote: > Concerning the state of comparative linguistics, a rather >diffuse debate I have been following, it is not true that there are >no criteria for genetic relationship, for goodness sake, as suggested >by Benji when he writes, "Over time any language can change and/or >borrow anything (or everything?) from other languages under >conditions of contact." In part it seems to me that this notion >comes from the mistaken assumption that genetic relationship is a >matter of statistics with words, an American fallacy I have written >about previously. I do not dispute any of the above. The intent of my remarks was not to deny that each individual practitioner has some methodological criteria for genetic classification, but to invite comments toward a *consensus* view of what genetic relationship means and, especially, how it can be demonstrated. I am grateful to Karl for offering some comments in this direction, and I would welcome further discussion of his referred to "American fallacy" involving statistics with words -- I assume lexicostatistics. Naturally, I think of such languages as creoles for types for which lexicostatistics would be misleading for genetic assumptions about these languages as *wholes*. Similarly, the fallacy, as I guess it, would misrepresent the origin of "mixed" languages, such as Mbugu (Ma'a) and some of the Romani languages, e.g., those which have an English grammar and largely Romani vocabulary -- among various others. With regard to the latter languages in particular, I wonder if there is consensus about whether genetic relationship to one rather than more than one ancestral language applies, and, if so, what the consensus is, e.g., is it that grammar, rather than lexicon, is criterial, and therefore "Anglo-Romani" is a variety of English with a heavily Romani superstratum? Frankly, I do not see that the methodology of genetic classification anticipated such problems, or that its interests in tracing origin to a *single* ancestor have adjusted to them. Karl's posting continues: >Historical/comparative linguistics is the construciton of grammars >for protolanguages. I agree to that wholeheartedly, as long as "grammar" means *all features* of the proto-language, but not to some further assumptions embedded in Karl's elaboration of this statement. >...as Meillet knew, there is a difference between borrowing lexical >items and borrowing grammatical structure. You can hear words, but >nobody has ever heard a grammar, which is a system constructed anew >by every speaker, and this is the crucial difference which allows us >to do the history of a language and distinguish borwoings from >retentions. I don't think the distinction between words and grammatical structure*s* is as clear-cut as Karl, or Meillet, suggests. We *hear* both instances of words realised in/as particular pronunciations and grammatical patterns realised in/as arrangements of morphemes, words, etc. And, in either case, we deduce some intended "meaning" conveyed by the morpheme, word or larger pattern. Karl continues: >Words can be borrowed but not structure -- where it looks as if there >has been structural borrowing, it must be seen as a case of partial >language learning. This seems to me a matter of arbitrary definition, rather than a principled consensus view. I do not reject the suggestion out of hand, but I do not see that when we associate such English words as "phenomenon" and "alumnus" with plurals like "phenomena" and "alumni", it is any more a case of partial language learning from Greek and Latin (respectively) than when we associate "child" and "man" with the plurals "children" and "men", as far as the learner of (standard varieties of) English is concerned. Since I can appreciate that Karl intends his claim, in this example, to apply to the (erudite) bilingualism through which Latinate and Greek-ate plurals came into English, it seems to me that if his claim is accepted, then it is no different for lexical items such as English "beef", "pork", etc., which are lexical items which were introduced into English through bilingualism with (Anglo-Norman) French./FN In sum, I do not see the basis for distinguishing lexical items and grammatical structures according to the criterion of "partial language learning". (FN: I use such examples as "beef" and "pork" to emphasize that in the case of these words, they became restricted in English to the edible meat produced by the animals designated by the French names. Is that "partial language learning" of vocabulary? The general problem may be, then, that words are associated with meanings, just as grammatical structures are, and that they may be reanalysed, or not, in the process of integration into another language, just as grammatical structures may.) I assume that Karl meant something like: genetic classification is (or, should be) based on (statistical?) grammatical (morpho-syntactic?) reconstruction, rather than statistical lexical reconstruction. In the case of English, perhaps "most" of the syntactic structures of English can be traced back to Proto-Germanic, aligning it with the other Germanic languages, and more distantly, with Indo-E. That may be fine for English, and many other languages, genetically classified together. But it does not work for all languages, as creoles and mixed languages show. And it becomes quite difficult to maintain for the currently controversial language families for two reasons. 1) grammatical reconstruction is more complex and its principles are less well understood or agreed to than phonological reconstruction (of vocabulary), 2) areal studies show that grammatical features diffuse across languages differentiated and classified on the basis of lexical origin. It remains unclear, at best, and is probably false, at worst, that bundling of grammatical isoglosses necessarily circumscribes sets of languages which descend from a single common ancestor but have since been differentially injected with lexical material from other languages. The problem remains to specify what it means to say that a given language has a single "principal" ancestor, and to continue to improve methodology to distinguish grammatical convergence (even in morphology) from "original" inheritance in trying to determine a chronological order of events leading to any and all attested languages. - BenjiMail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue