Dravidians and Sergent

[Lars Martin Fosse]

Michael Witzel wrote:

> 
> At 12:22 +0200 7/3/99, Lars Martin Fosse wrote:
> 
> > When we start comparing reconstructions of proto-languages with other
> >reconstructions of >proto-languages, the error potentiality becomes
> >enormous
> 
> That really still is up for grabs, as so few people have tried
> *systematically*,  and even less with the strict methods of traditional
> comparative linguistics (sound laws and comparison of grammatical
> elements). Until recently, we just got one-to-one comparisons of the kind
> mentioned by Lars : >Dyen 's  criticism of the Austronesian-Indo-European
> Hypothesis<

In principle, Michael is of course right: the rigorous use of the comparative method may in principle take us further back than we have come so far. My scepticism is due to the fact that this very rigor seems to get frayed at the edges when we move beyond the single families to create superfamilies. At least this is the impression I get when I see Nostraticism in practice. And I am not alone in my sceptical attitude: Dixon in his "Rise and Fall of Languages" expresses a similar scepticism not to mention a bunch of Indo-Europeanists. I suppose I shall simply have to live with the fact that Ruhlen disagrees with me. The point in my opinion is this: as languages move apart through time, the number of ancient elements that can be compared becomes smaller and smaller. I suspect that at some critical point it sinks below the limit for randomness, in other words: similarities may equally well be explained by chance as by historical relatedness (or as Michael says: any 50 words in any language can be compared with any 50 other in any other language.) This is when comparisons cease to be meaningful, and I think it is a matter of importance to scholars to learn how to decide when they are about to go off on a wild goose chase where good, well-documented results simply will not be forthcoming. Research is a precious thing, but some research simply seems to be a waste of time and money. 
    Fixing a limit for randomness, in other words, deciding when to quit a research project because it will not produce meaningful results, is of course difficult. There is no absolutely foolproof formula. But I think that responsible scholars will have to think about this because it concerns us all and the status of our science. 

> >Admittedly, some languages contain extremely conservative elements -
> >Lithuanian "sunus"  <<Latin sunus, Rgveda suunus> >
> 
> Correct, but that is not what I said or meant. Some sounds in many
> languages are more stable than others.  Which leads to results such as
> sunus (and Engl. sun), or German Jahr / Avestan yaar-... as I said:
> Resonants and sibilants often are  more stable. Coincidence, certainly,
> Avest & Germ.  developed in the same direction (aa), but  y- and -r are
> stable (unlike in Scandinavian, of course, where y- has long gone the way
> of all sounds, > zero).

I was not really thinking about what Michael said regarding sounds, I just wanted to give a simple illustration of a linguistic element that has survived for an incredibly long time. But Michael are of course right that some sounds are more stable than others. HOWEVER: "more stable" is a relative concept, and the stability is not rulebound, it is only a statistical phenomenon. Which somehow reduces the value of this insight in practical situations where documentation is weak. 

> 
> > comparing languages that presumably split up 8-10,000 years ago simply
> >makes no sense, ... If there are similarities, they are most likely due to
> >chance.
> 
> This prescisely is Ruhlen's "IE myth". It is just as unproven as a
> *general* theory, as a theory of a Eurasiatic common language at c. 10,000
> BCE plus. Not to speak of Eve's language. The footwork  has still to be
> done for *all* theories mentioned just now.

In principle I agree with this, at least up to a point: the itchy word is of course "proof". Having seen the word "proof" turn up in a large number of contexts where it doesn't really belong, I would prefer to turn to the concept of probability. Some things can't be proven without proper historical documents, and since there are no historical documents to help us here, we are forced to rely on probabilities. Now, here is a piece of elementary logic: probabilities NEVER prove anything. If I say "there is a 99% probability that A = B", then I have NOT proved that A = B. I have just given a probability, although a very high one. In practical life, we would act upon such a probability, in fact we usually rely upon far worse probabilities when making informed choices or guesses. The "general theory" cannot be proven.  But common sense may still uphold it. 
 
> But I do not think  that this is impossible: one just has to use the
> standard methods of comp. ling. and pay attemtion not only to sound
> correspondences but also to some UNIQUE forms in grammar.
> The most obvious case, in Nostratic, perhaps is the acc. -m in IE,
> Finno-Ugrian (-m), Altaic (Mong. -b), Japanese (-w) etc. (Same sound
> correspondences in other words, such as in "I/me"), or the possessive
> genetive in -n- in the same languages.  Just for starters. ---
> Such grammatical elements, *if* they fit established sound laws, can be as
> little due to chance as the same correspondences of these forms *within* IE.
> Similar sentinent by one of the long range comparativists, I. Peiros,  in
> the Shevoroshkin festschrift  (ed. I. Hegedus, et al. , JIES 22, 1997)

The interesting question here is of course if such similarities are due to area linguistics or genetics. And they MAY still be due to chance.... :-)
 
> Cooperation between various fields is of the essence. Yet, maybe, I should
> not criticize Sergent, as I am finalizing a similar undertaking, a
> re-combination of old and new articles on the subject...

I think such cooperation is a prerequisite for making progress, and before we have learned to do it properly, we will all screw up sooner or later. Probably sooner.  But it is worth trying.

Best regards,

Lars Martin Fosse