Talk:Glottochronology

From Wikipedia, the free encyclopedia

Could somebody add dates for the paper references? They would give a time frame within which glotto was in vogue? -- Error

This article is dominated solely by the Starostin's line of arguments. Glottochronology is only founded on the observation of agreements in the amounts of changes in the basic vocabulary of some languages. This view, mechanistically carried over from the first days of 14-C dating, totally overlooks that a community of speakers is open to many changes, the reasons of which are completely irregular and unforeseeable. The reason of these misunderstandings is further that the described changes circle around some statistical peak, erroneously interpreted as a law of rate in these changes.HJJHolm 10:14, 8 January 2007 (UTC)

The article does explicitly mention that (a) glottochronology is far from being accepted by "mainstream" linguistics, whatever that might mean; (b) that glottochronology does not work in its original incarnation as presented by Swadesh; (c) that it is a method open to discussion and further testing. However, the "Starostin line of arguments", backed up by actual linguistic data, published in print and on the Web, has not been falsified on an empirical basis, which is why it deserves full representation in encyclopaedic sources. Expansions to the article, presented in separate chapters, are, of course, also welcome.Gstarst 12:30, 8 January 2007 (UTC)
As for language change, it does happen for many reasons and in different ways, but we are talking not about language change in general, but about a very particular kind of language change (basic lexicon), which has a reason every bit as explainable as the one behind 14-C decay.Gstarst 12:30, 8 January 2007 (UTC)
You have not explained a single bit! Counting is no explanation, and finding similar amounts of change is no explanation, either. Statistical distributions use to have one or more 'peaks'. And do you really pretend that reasons of language change make a difference and stop at start of basic word lists? You know better, but then I cannot follow this argument.HJJHolm 09:52, 8 January 2007 (UTC)
First, I agree that counting is no explanation. But counting is much better than an explanation, because counting is an objective procedure - it places you before statistical facts. Likewise, observing and postulating regular phonetic correspondences is much better than explaining them. To my knowledge, no one has ever proved why phonetic correspondences between languages have to be regular - assumptions have been made and various hypotheses put forth, but objective proof is still lacking. Yet that does not invalidate the fact that such correspondences do exist.
Counting simply yields the data-base. Before applying any formula, the stochastic properties have to be analyzed, as any serious textbook of statistics does in fact mention. Gray & Atkinson tried somewhat in this direction, glottochronologists, including S. Starostin, did not. Instead, GC is based - on the (sorry) primitive similarity principle that languages are more related the more features they share (see Holm 2003); - on the unhistoric belief into a constant rate of change. HJJHolm (talk) 17:56, 24 November 2007 (UTC)
The "primitive similarity principle" is one that works in every single case of historically attested language development, which is, of course, the only evidence that can be subjected to analysis of stochastic properties. And, once again, I stress that this is not a question of 'belief'. Given the agreeing results of the calculations in several geographically, historically and culturally distant areas of the world, the constant rate of change is more than a belief: a rational hypothesis based on nothing but objective facts.Gstarst (talk) 19:45, 24 November 2007 (UTC)
Nevertheless, constant rates CAN be rationalized, if not "proven" logically. Since language change happens all the time, it is inevitable that basic lexicon will change with time, just like any other part of the language. This is why there has not been detected a single case of the 100-wordlist remaining intact over a period of, say, 500 years. Yet since the lexicon forms the most communicatively important part of it, it is also inevitable that there should be a set threshold for the rate of change which cannot be surpassed without the population actually shifting to a new language. This is why there has not been detected a single case of the 100-wordlist losing more than, say, 15% over a period of 500 years. And even losses of that particular order are an exception - owed exclusively to intense borrowing from "superior" sources.
Where is the causality? That German had abundant loans from French in the nineteenth century, has comprehensible reasons. That these have been replaced after WW1, interrupted in the NAZI-time, and nearly completely after WW2, should also be comprehensible. All not in rates, but by external cultural influence. That Baltic languages have preserved much more of IE grammar, is due to the substratum of hunters an gatherers, contrasting with e.g. Hittite, where the substratum belonged to the - cultural much stronger - "advanced cultures"; again not due and by any inherent rate.HJJHolm (talk) 17:56, 24 November 2007 (UTC)
These arguments, I am afraid, have nothing to do with the one they are supposed to be directed against. I was explicitly not talking about the evolution of the cultural lexicon (the German-French situation; where are French loans in the German 100-wordlist?), nor about the evolution of grammar (actually, the Baltic vs. Hittite argument strikes me as exceptionally weak; Celtic languages have reduced Indo-European grammar to almost nothing, but where are the "advanced cultures" as their substratum? Besides, since I don't see much of Uralic grammar structure in Baltic, nor is there a lot of obvious Semitic and/or Caucasian influence on Hittite morphology, I am afraid there is no evidence whatsoever here apart from purely abstract speculation). The causality for basic lexicon change, however, comes from within the population, not from without (e. g. the small, but still inevitable drift in Icelandic, where there was no substratum to speak of).Gstarst (talk) 19:45, 24 November 2007 (UTC)
The question is: if there is no such thing as a constant rate, WHY don't we have situations like these? Why couldn't a language replace, say, 20% of the wordlist within a hundred years without any help from languages of their "conquerors" or "cultural dominators"? Why couldn't it stay unchanged completely for a thousand years? Why do similar figures keep cropping up for comparable chronological periods? Why discard the obvious? Are you ready to entirely dismiss these "statistical peaks" as something completely meaningless and irrelevant?
Without an example this is not very convincing.HJJHolm (talk) 17:56, 24 November 2007 (UTC)
Exactly - it is up to you to provide an example where a language has changed 20% of its basic vocabulary over a period of 100-200 years. I know of no such examples, and this works in favor of glottochronology. As for the similar figures, fine. From Old Japanese to Modern Japanese: 7 words, from Old High German to Modern German: 6 words (the time span is around 12 centuries for both). From Ancient Greek to Modern Greek: 20 words, from Old Chinese to Mandarin: 23 words (the time span is around 25-27 centuries for both). Lists of replacements can be found in Starostin's article in Renfrew, C., McMahon, A., & L. Trask, Eds. (2000).Gstarst (talk) 19:45, 24 November 2007 (UTC)
What you fail to take into account is that, while individual replacements ARE more or less "random" and unpredictable - no proponent of glottochronology would argue with that - it is not the individual replacement that matters, but the statistical total, which simply cannot exceed a certain point. Matching the necessity of basic lexicon change against the limit to basic lexicon change is exactly the thing that makes glottochronology valid. This is what the "statistical peak" is all about. If you don't like the word "law", let's not use the word "law". I can accept the word "tendency". Terminology is of little importance here anyway - it's all about a practical approach.
Please cancel this "to fail". Contrasting to glottochronologists, I HAVE analyzed the stochastic properties, your "total". However, there is neither a law nor tendency. Languages have changed in any time in any degrees. E.g. the "Golden Horde" switched in a few years from Mongolian to Turki. The Franks in northern France switched in short time into Middle French, not without influencing this themselves. It is tiring to bring examples here for historical events every historian knows well.
And what has this exactly to do with the subject at issue? We are talking evolution of language here, not loss of language. The two processes may occasionally intertwine in the case of creolization (which is a pain in the neck for linguistics as a whole, not just for lexicostatistics), but neither in the case of the Golden Horde, nor with the Franks do we see any creolization at all. Nevertheless, my question still stands: if "languages have changed in any time in any degrees", are you aware of any particular language that has changed, let's say, more than 20%, but less than 80% (to make sure we're not talking of a language loss situation) of its basic lexicon over a period of 200 years? If you are not, your statement should probably be retracted.Gstarst (talk) 19:45, 24 November 2007 (UTC)
Finally, it is obvious, of course, that language change does not begin and end with the 100 wordlist. But it is one area in which language change is empirically measurable. Relatively fast (in comparison with other languages) "collapses" of a language's phonology, morphology, and cultural lexicon are well-known and indisputable. "Collapses" of the basic lexicon are simply non-existent, not even in the case of Albanian or Brahui. Therefore, the oh-so-frequently seen statement "Glottochronology is wrong because language change occurs at different rates" does not make sense, since glottochronology does not contradict the latter statement.Gstarst 10:25, 8 January 2007 (UTC)
Sorry, I cannot follow this logic. Do you defend rates or not?HJJHolm (talk) 17:56, 24 November 2007 (UTC)
I defend the rate of constant change for a language's basic lexicon. The logic that I cannot follow is the "Battle of Hastings logic" that tries to substitute the primary idea of glottochronology with something the glottochronologists never suggested in the first place. It is useless to discuss the amount of French loanwords in English or the collapse of Indo-European morphology in Hittite under this heading because it has nothing to do with glottochronology. Gstarst (talk) 19:45, 24 November 2007 (UTC)
And before submitting any new article on the uselessness of glottochronology to any periodical (which will, most probably, simply rehash the same old arguments)...
You might notice below that my line of arguments is new. And not a mechanistic one as yours.HJJHolm 09:52, 8 January 2007 (UTC)
The major argument seems to be that "languages change due to socio-historical impact, which cannot be measured on a regular basis". This idea is anything BUT new. And as for accusations of being 'mechanistic' - well, if that's what we get for preferring to work with real language data and actual statistical calculations rather than philosophical theories, so be it. There's nothing wrong with that kind of 'mechanisticity'. Gstarst 10:25, 8 January 2007 (UTC)
Logic and historical knowledge is not philosophy.HJJHolm (talk) 17:56, 24 November 2007 (UTC)
This is actually one point that I agree upon one hundred percent. But it doesn't seem to be of much practical value in our current conversation.Gstarst (talk) 19:54, 24 November 2007 (UTC)


I would at least advise the author to try out a more practical approach himself. The more time one spends with 100-wordlists, the less banalities one is liable to pronounce about the method.Gstarst 09:19, 2 December 2006 (UTC)
Thanks for the "advice". The differences between social behaviour and natural sciences is not a "banality". And the JQL is not "any periodical" but a peer-reviewed specialized paper. It would be a fine task for Gstarst to correct the many errors in "Indo-European lexicostatistical data" (in fact the Dyen-list), mentioned at the end of the article.User: HJJHolm, 04 dec 2006.
Doing this on the basis of Indo-European languages alone is hardly enough; the important thing is to see how the average Indo-European data agrees with data from remote (and also verifiable by extra-linguistic evidence) families. I doubt that user HJJHolm has ever engaged in that kind of activity.
Sorry, I have around a dozen versions of that list here. Addionally there are lists of Marisa Lohr, Don Ringe, and many others.HJJHolm 09:52, 8 January 2007 (UTC)
A dozen Indo-European versions or a dozen versions for a dozen different language families? In the latter case, have you analysed them? Gstarst 21:58, 7 January 2007 (UTC)
I concentrate on lexicostatistical inference of genealogical relationships, what requires reliable lists of linguistic features.
Lists of linguistic features from what languages?Gstarst 10:25, 8 January 2007 (UTC)
Additionally, I, of course, know of no one today doubting the existence of differences between natural sciences and social behaviour, which is what the "banality" statement referred to. It is the existence of similarities between the two that glottochronology is trying to establish, which is a perfectly reasonable scientific hypothesis that should not be referred to as "nonsense" without a careful assessment of all its arguments.
Please tell me a single "similarity", in terms of functions and reasons, not of superficial resemblances of changes.HJJHolm (talk) 17:56, 24 November 2007 (UTC)
To my knowledge (and I am reasonably well versed in many specimens of the counter-glottochronology literature), such an assessment has not yet been done.
According to the original references here, you knew rather old articles. Only I added the Tischler 73, Nettle 99, McMahon 2005,
The original references are not mine. I only added Renfrew & McMahon 2000, which is hardly a "rather old" publication. I confess to not yet having seen McMahon 2005. However, I AM familiar with Renfrew & Forster 2006 (Phylogenetic methods and the prehistory of languages), which includes works by the McMahons and which, for some reason, you have not included - maybe because it includes articles by the accursed Dyen and Gray. Not a single "anti-glottochronology" work in there is dedicated to analysis of the data; some simply slag the method off with references to Bergsland & Vogt (at least those researchers used to back their accusations up with complete listing of data, a method so rarely seen today). Is McMahon 2005 at least partially dedicated to testing various formulae proposed by other people for glottochronology on the basis of adduced wordlists? Nor is there ANY practical assessment of glottochronology in Nettle 99. We are obviously speaking about different things. I care little for theoretical discussions on "why" glottochronology cannot work; I am much more interested in calibrating the formula on real wordlists.Gstarst 12:46, 8 January 2007 (UTC)
As for the many errors of the "Indo-European lexicostatistical data", many of them have been corrected already, although, of course, user HJJHolm cannot be blamed for not being aware of it.Gstarst 12:46, 5 December 2006 (UTC)
Please see above, what I mean by that list, and this has NOT been corrected, at least in the points admonished by Sheila Embleton.
What I meant was that the Dyen lists are far from the only Indo-European wordlists compiled. The Moscow school, for instance, uses its own, compiled by professional scholars of Indo-European linguistics. They are more extensive and generally better checked than the Dyen lists, although, of course, nothing is perfect. Gstarst 21:58, 7 January 2007 (UTC)

Contents

[edit] Glottochronology is only the best-known and least well-founded part of lexicostatistics

The article needs to distinguish between glottochronology and lexicostatistics. --Pfold 18:25, 6 April 2006 (UTC)

Very much so. If I have time at some point, I'll try to make that distinction. In fact, I would support having two articles. The relation of glottochronology to lexicostatistics is that of joy-riding to driving. For now, I've added McMahon and McMahon (2005) Language Classification by Numbers to the bibliography: they explicitly reject the notion that the two endeavours are the same. garik 13:27, 11 August 2006 (UTC)

Correct. We did not need M&M to find that out, but I am happy that they stressed that point, against well-known British handbooks (Crystal). I tried to embed this in the starting paragraph. User: HJJHolm, 04 dec 2006.

Well, I don't think they make any claim to having found it out - they cite various predecessors who make the same distinction. I mentioned it here since it was a book I'd read relatively recently - and it's good to have modern references. Anyway, I've created a Lexicostatistics stub. I hope someone will have time to do the article justice! Certainly it should never have redirected here. garik 17:59, 7 December 2006 (UTC)

[edit] (McWhorter page number needed)

I searched the book, but wasn't able to find this mentioned. Since this is a fairly general concept, we should either take out the reference or change it to another one. --CRGreathouse 05:41, 12 May 2006 (UTC)

The Bibliography is terribly outdated. I added some newer work on this field. User: HJJHolm 14:00 METm 25 may 2006

[edit] Quibbles

... his assumption has been demonstrated as the basic error, because it neglects the dependence of this percentage on three further factors.

Up to this point there is no mention of controversy. Does "basic error" mean the science is invalid? What three factors? This needs to be better introduced and clarified. —Michael Z. 2006-09-21 23:23 Z

[edit] Romance is a bad example

Thus, glottochronology shows Latin to have split into Romance vernaculars around the 1st-3rd centuries A.D., exactly when the process was supposedly taking place.

That is a bad proof. When Swadesh developed the technique, he calibrated it by setting the coefficients according to data from well-known families (the article should mention which ones). I very much doubt that he not tried Romance languages. Hence, this example has no predictive power. To proove the hypothesis, Glottochronology should be applied to a relation not tested by Swadesh in his calibrations and the results be compared with those from history, archeology or whatever. Remove the example or substitute it. --Error 18:57, 28 September 2006 (UTC)

I agree the example is not good. Well, glottochronology HAS been applied to a lot of territory not tested by Swadesh: Chinese, Japanese, Semitic, Turkic (I am only mentioning families where the data are at least partially verifiable through extra-linguistic means). No significant divergences here.Gstarst 09:19, 2 December 2006 (UTC)


[edit] "Modified Glottochronology"

By the way, the cited (late) S. Starostin is not at all famous between historical linguists for his exactness (Cf. e.g. Georg, Stefan. (2004). Review of S. Starostin, Etymological dictionary of the Altaic languages. Leiden: Brill, 2003/4. Diachronica XXI-2:447ff). User: HJJHolm, 04 dec 2006. S. Starostin then replied to Georg in the next issue of Diachronica, accusing Georg himself of being inexact.

Instead of making blanket statements based on second-hand sources, why not refer strictly to the data itself? At least consult Starostin's work on the subject?
Only very few readers would have the competence of arguing on that level.
Which presumably means that we should be arguing on the level of "linguist X says about linguist Y that linguist Y has made many errors" without any details or even direct relatedness to the subject. Excuse me if I say I find such blanket statements in rather bad taste. Gstarst 21:58, 7 January 2007 (UTC)
Also, speaking of exactness, there is a word or two to be said about the data of Bergsland & Vogt as well, whose "99% retention between Old Norse and Icelandic" really boils down to about three percent lower when Old Norse data are treated more discriminately - which, in this particular case, is very significant (99% would indicate that Icelandic sort of stopped dead in its tracks a thousand years ago, when in reality it has been evolving just like any other natural language).Gstarst 13:23, 5 December 2006 (UTC)
This is all at a level even specialized historical linguists would not agree in all points. Cf. some examples of Brett Kessler (2001)in his book The significance of word lists.
I agree - which is why I did not include concrete criticism of B&Vin the body of the article. But if one side of the debate is going to nitpick (without giving concrete examples), I don't see why the other side can't be allowed to engage in the same activity. It is rather hard to find a historical linguist who would always be correct and exact about everything, be it Bergsland, Starostin, Georg, or H. J. Holm - I think.Gstarst 21:58, 7 January 2007 (UTC)

[edit] Reverting to previous version

I have taken the liberty of reverting user HJJHolm's latest additions to my earlier version, not because I do not welcome further expansions and corrections of the article but rather because the last section of it was starting to look too much like an open discussion, with the word 'assessment' quite misleading (whose assessment? yours or mine, etc.?). Nevertheless, I have complete respect for user HJJHolm's views on the subject, and believe it correct to transfer these 'assessments' here in order to tackle them on the discussion page where they rightly belong.Gstarst 12:46, 5 December 2006 (UTC)

Exactly this 'basic' vocabulary was analyzed by H.Haarmann (1990) who demonstrated that even this may and has changed to great extents, with many examples of Albanian (which is a difficult field for not specialized linguists). Naturally the changes in the basic vocabulary of Albanian are fewer than in the cultural fields. But it is typical that Albanian changed 90% of its Indo-European heritage, while the neighbouring Greece did not.
The last two phrases refer to different things. It is good to see the phrase "naturally the changes in the basic vocabulary of Albanian are fewer than in the cultural fields", especially the word "naturally", since this shows that user HJJHolm is at least partially in agreement with the basic premises of glottochronology (a sharp distinction between 'basic' and 'cultural' vocabulary). However, already the next phrase refers to ALL vocabulary rather than the basic lexicon, or, to be sure, the 100-wordlist. This is a clear-cut case of substitution of values.Gstarst 14:10, 8 January 2007 (UTC)
In short: There is no break or "sharp distinction" between the 100 words assumed to be most stable, and all others. Instead, EVERY word has its own history and another probability of being replaced. This is no "substitution of values".HJJHolm 18:53, 7 January 2007 (UTC)
Well, I agree that real "sharp distinctions" are generally hard to find in language at all - there is always the matter of a "fringe territory", be it lexicon stability, difference between language/dialect or historical transition from one language state to another. Nevertheless, you do distinguish between "basic vocabulary" and "cultural fields" and admit to "natural" reasons for fewer vocabulary changes in the former. That's all we need to know. The rest is a matter of empiric calibration.Gstarst 21:58, 7 January 2007 (UTC)
Albanian did borrow around 18-20% of its 100-wordlist from other languages, which is, incidentally, about the same number as Brahui in Dravidian. So far we have no instances of a higher number of borrowings on the list, which is in itself a fact worth noting; this is presumably a natural threshold crossing of which essentially means shifting to another language altogether. However, as long as this threshold has not been crossed, the negative consequences of this for glottochronology are easily skipped by applying the "borrowing elimination" method, and in the case of Albanian, we, for the most part, know how to distinguish borrowings from non-borrowings.Gstarst 13:23, 5 December 2006 (UTC.)
English did not replace about 50 % of its originally Germanic vocabulary 'by time', but, as educated speakers of English know, by Norman dominance after the battle of Hastings, besides a long-lasting educational background of Latin. Though, within the Swadesh 100-wordlist there seems to be but one English item owed to effects of the Norman conquest ('mountain'), in the former 200-Word list (also employed by Isidor Dyen, and this, in turn, by Gray&Atkinson, Rexová et al., and MacMAhons 2002) there are much more: in the English list alone there are 12 loans from French (correctly coded), additionally at least 7 loans from Scandinavian (not detected by the author in his own mother tongue! Cf. S. Embleton 1986:125ff).
More substitution of values. The phrase "Though, within the Swadesh 100-wordlist there seems to be but one English item owed to effects of the Norman conquest ('mountain'), there are much more on closer inspection" would logically indicate that within the Swadesh 100-wordlist there are actually many more borrowings from French ('Norman') sources. Instead, we are immediately taken over to (a) the 200-wordlist rather than the 100 one, (b) borrowings NOT from French sources, but Scandinavian ones! What do Scandinavian borrowings have to do with the Battle of Hastings, which user HJJHolm was so eager to adduce before? Let us then at least wipe off the mention of that, no doubt, fabulous, but completely irrelevant historical event. I reiterate the following: the 100-wordlist contains (does not "seem to contain", but a much stricter "contains") but one borrowing from French. The 100-wordlist does further contain three borrowings from Scandinavian ('die', 'bark', 'skin'). This is not even remotely close to the Albanian figures, nor does the argument count at all with the introduction of the already mentioned "borrowings elimination" correction.Gstarst 13:23, 5 December 2006 (UTC)
at least. G.Starostin might notice that the Norman as well as the Danish or Old NOrse invasions or any other impact else reflect the same reasons for changing vocabulary, in particular beyond any rate! Sorry for the (now corrected flaw in my sentence).
Yes, of course, this is noticeable - and, reiterating it for the umpteenth time, is easily taken care of with the "borrowings correction".Gstarst 21:58, 7 January 2007 (UTC)
Assessment: All innovations can be looked upon as borrowings, be it from outside or inside a language family, external or internal strata (e.g. Celtic layer in English) or older stages, or more respected dialects. Very seldem an innovation is a free invention. Moreover: One must be an extremely specialized expert to detect these differences. S.Starostin is not at all generally known as such. In the often applied Dyen-List (linked at the end of that article) there have been several errors in English alone. It reveals the competence of its users, who not even detected these simple examples.
I do not see any need for this intentional confusion of terminology. For any historical linguist the primary differentiation between native replacement and borrowing from an external source is quite obvious.Gstarst 14:10, 8 January 2007 (UTC)
obviously not. Dyen did not even detect the state of french 'tête', ... Sorry that was a bad example: Dyen is correct, the famous Lyle Campbell (1998:184) is wrong.HJJHolm 16:54, 25 January 2007 (UTC)
My sentence did not refer to concrete cases, but to the general idea of differentiation between the two ways of replacement. The Dyen criticism here is a technical issue. Particular mistakes CAN be made, for instance, due to our incomplete knowledge of language history, and SHOULD be corrected as we unravel more of it. But that does not invalidate the principle itself, of course. By the way, Dyen's mistake is not quite clear to me here. What is the "state of french tête" if not a continuation of Vulgar Latin 'testa'?Gstarst 21:58, 7 January 2007 (UTC)
There is a clear-cut distinction between the replacement of Old English beorg 'mountain' by the respective French borrowing and, say, the replacement of Old English rēc 'smoke' by modern English 'smoke' from Old English smēoc. This has nothing to do with any Celtic layer and we have zero evidence as to the 'respected dialects' factor (and in any case, in order for the word to have been "borrowed" from a "respected dialect", the original form would have to have been natively replaced in the "respected dialect" itself). As for the statement 'very seldom an innovation is a free invention', it refers to philosophy more than empiric science.Gstarst 14:10, 8 January 2007 (UTC)
That you do not know, is of course no argument at all. BTW, both forms existed before and can still be traced in German dialects. Further: It is a commonplace in psycholinguistics, that forms are taken over from more respected dialects, or jargons, or languages. Do you really want references for that statement? Perhaps you can tell me a real "innovation"? Please note that I know the terminology.HJJHolm 09:39, 7 January 2007 (UTC)
That I do not know - in this case - means that you do not KNOW either, nor does any specialist or non-specialist in the field; what you can is only HYPOTHESIZE. "It is a commonplace in psycholinguistics, that forms are taken over from more respected dialects" - again, I can only reiterate: (a) does that really mean that ALL forms are taken over from more respected dialects, and if yes, do you have proof for that; and (b) if so, how do forms get replaced within the respected dialects themselves? Do they migrate from even more respected dialects? :) This "commonplace statement", true as it might be for SOME cases, is hardly answering the question "why do lexical replacements happen?", even though it pretends to be answering it. Glottochronology/lexicostatistics, on the other hand, does not pretend to be answering that question; it is merely constatating facts. It is not invoking the divine factor, either.Gstarst 14:10, 8 January 2007 (UTC)
Additionally, it is not clear why Starostin should be blamed for mistakes committed by Dyen (he does not even use the Dyen list in his calculations), and completely unclear how the Dyen mistakes tie in here in the first place since Dyen did not invent, subscribe to, or, in fact, know anything about the "borrowings elimination" correction when he was making his calculations. However, I do agree that differences between borrowings and native replacements are not always instantaneously recognizable. In many cases, one does have to go through the pains of consulting the standard etymological dictionaries of the respective families, usually available if said families have undergone at least a short period of etymological research. This is, however, not a very difficult task, and this condition has been specially mentioned in the description of Starostin's method.Gstarst 13:23, 5 December 2006 (UTC)
Nobody (!) blamed Starostin nor Swadesh for the Dyen list. But, as already mentioned, it has been used in recent glottochronological statements by Gray& Atkinson, Rexová et al., McMahon&McMahon 2002 (abandoned 2005), who all were and are unaware of the many mistakes...
Assessment: This is a circularity. Changes depend on the strength of socio-historical impacts, which in turn trigger reputation of peers, or any group of speakers. It would be most interesting to know which major socio-historical impact has triggered the switch from OE rēc to ME smoke.Gstarst 14:10, 8 January 2007 (UTC)
What else? An inherent urge? Or clock? See above.
The honest thing here, of course, would be to say that we don't know what did. In this case - and in thousands of others - "an inherent urge" is pretty much as good an answer as "a major socio-historical impact". Gstarst 14:10, 8 January 2007 (UTC)
Wrong: There is a 'tiny' but deciding difference: An 'inherent urge' might be regular (and this is exactly what you are forced to assume). Historical events never.
Regardless of which of these processes might be regular or irregular, there is no transcendental logical proof for either. And yes, you are right that I AM forced to regularity - but certainly not on the basis of speculative theorizing, but rather on the basis of empirically observed statistical figures. If the (chronologically quite similar) distance from Old English to modern English and the distance from Old Japanese to modern Japanese, in terms of the 100-wordlist, is the same (91-92% with the "borrowings correction" enhanced), I cannot NOT be forced to admit that this is not simply a matter of historical coincidence.Gstarst 10:00, 8 January 2007 (UTC)
We DO know that lexical replacements happen, and at an approximately stable rate (within the wordlist). We do not know WHY they happen. Particular reasons may be individual, including the ones you have mentioned and the ones I have mentioned. But the important thing is that individual reasons do average out - and this can be, and has been, demonstrated empirically.Gstarst 21:58, 7 January 2007 (UTC)
This statement is nothing but a hypothesis, based, for the most part, on intuition (this seems LIKELY to our minds) and partial induction (i. e. "SOME languages changes have been due to socio-historical impacts, thus ALL language changes should be ascribed to socio-historical impacts"). The FACT, as contrary to a hypothesis, is that (a) all languages, unless they are dead ones, are in a state of permanent change, regardless of their social or historical circumstances; (b) there are multiple occasions of languages changing in different ways despite seemingly sharing the same socio-historical conditions. In order for this statement to acquire credibility, a huge and systematic analysis of the world's linguistic legacy would have to be undertaken, rather than a chaotic mix of different, and frequently quite subjective, "examples" so frequently found in sociolinguistic works.Gstarst 14:10, 8 January 2007 (UTC)
Do not alway push against open doors. Nobody disputes that languages change, but never in a constant rate. And THIS is the point!
Allow me to retrace our steps. Your statement was: 'Languages change due to socio-historical impact'. My statement, with which you have agreed, was: 'Languages change constantly'.Gstarst 14:10, 8 January 2007 (UTC)
Perhaps "constantly" is misleading here, it would be more exact to state that "symplesiomorphies only decrease in number", what will not include the concept of constant rates.
If both of these statements were true, this would mean that languages are constantly under socio-historical impact, including even those that develop in relative or complete isolation. This is clearly not the case. Wouldn't that mean that there have to be other factors besides socio-historical ones?Gstarst 21:58, 7 January 2007 (UTC)
Additionally, the "word aging" argument has not been understood correctly by user HJJHolm - although I do take part of the responsibility upon myself, since it was not explained in details. Essentially, "word aging" refers to the process of the word in question acquiring secondary meanings and usages over the course of its existence in the language, with each new "acquisition" and broadening of its functions increasing the probability of its being replaced in its primary function. Nothing, really, to do with "socio-historical impact".Gstarst 14:10, 8 January 2007 (UTC)
Be sure I understood this, what does not mean that I agree with it. And you are not the person to teach us here what we have understood or not.
Well, I'm sorry. I already took the blame, but maybe I was wrong. It just did not occur to me that you could understand something that had not even been written about.Gstarst 21:58, 7 January 2007 (UTC)
YOU do not know that it has been written about. Sorry that I did not notice the source, unsuspecting of this discussion.
What has been written about?Gstarst 14:10, 8 January 2007 (UTC)
Assessment: Different stability is a commonplace; Generally spoken, high-frequently used vocabulary is more stable than seldom used features. But this has nothing to do with any rate.

But it certainly does. Stable vocabulary has a lower probability of being replaced, therefore, taking this into account necessarily corrects the rate. In fact, it is fairly productive (and has been done) to calculate the "average stability index" of each item on the 100-wordlist based on empirical data, and then use individual quotients for each word. Starostin has implemented this new (fairly complicated) method into the Starling system; surprisingly - or maybe not so surprisingly - the results are quite close to the ones based on a single quotient for the entire list.Gstarst 13:57, 5 December 2006 (UTC)

Sorry, probabilities have NOTHING to do with rates, neither have distributions. Let me give an example: You could compute, for any or every region in the world, compute how often it has been affected by a war in a given time. Than you can compute a rate for every region.
You could (and in some cases, that might even be meaningful), but the rates would obviously be different around the world. In the case of basic lexicon replacement, however, the rates are similar around the world. That's the big difference. (Sorry for the previous misunderstanding).Gstarst 14:10, 8 January 2007 (UTC)
Note that in the sciences, a single counter example falsifies a hypothesis; in spite of a hundred others seeming to support it.

I was actually taught different: if a counter example is found to rule A that can be demonstrated to have been caused by the superimposition of rule B, this only further corroborates rule A.Gstarst 14:10, 8 January 2007 (UTC)

Correct: "If".
Okay: "When".Gstarst 14:10, 8 January 2007 (UTC)
Actually, I assume that this sentence is calling for a complete abolition of any mathematical / statistic methods in linguistics as such. The entire field has yielded very few (quite a neglectable number, in fact) 'scientific hypotheses' that have not been counteracted by at least a single (usually more) counter examples. A classic case is the notion of regular correspondences between related languages. Time and time again these correspondences, no matter how many examples confirm them, yield exceptions from the rules, but I do not see linguists abandoning the very concept of "regular correspondences" as a result - even if the very idea of related languages HAVING to have regular correspondences should be every bit as alien to the "socio-historical impact" line of thinking as the idea of a constant rate of replacement for the basic lexicon, maybe even more so. Instead, alternate theories have sprung up to try and explain these exceptions (such as lexical diffusion, etc.) without trying to destroy the Neogrammarian model, but rather peacefully co-existing with it. Likewise, it would only be natural to try and look for alternate explanations for seeming exceptions from the glottochronological model - which does work fairly well on a lot of sources EVEN in the original Swadesh incarnation - instead of throwing the baby out together with the water.Gstarst 13:57, 5 December 2006 (UTC)
It is quite simple: The Neogrammarian model is an idealized one.
Which does not imply that it is useless, or that it should not continue to be the backbone of all scientific historical linguistics, want it or not. Glottochronology is also an idealized model; no one denies a certain margin of error within the calculations, or a set of particular problems on ultra-high or ultra-low chronological levels that arise due to that margin. Which does not imply that it is useless, either.Gstarst 14:10, 8 January 2007 (UTC)

PS. I also wish to specially mention that I will be watching this page further. I have no objections against putting in the words 'controversial' or 'not universally accepted', or detailed descriptions of factual critiques against glottochronology as a method, but, as a representative of the established Moscow school of comparative linguistics (which is a serious scientific school and not a 'collection of crackpots', as some would have it), I insist upon treating it as a scientific hypothesis, still open to discussion (such as the one carried out on this page), testing, modification, and use in argumenting one's position rather than a 'discarded pseudo-method of the past'.Gstarst 14:12, 5 December 2006 (UTC)

PS: Many of us appreciate very much your work for the Leiden IEED-project; but here it seems that George Starostin feels obliged to defend alone S. Starostin's work.HJJHolm 18:53, 7 January 2007 (UTC)
I would rather say that I am simply defending the basic premises of a scientific method, since we haven't even begun discussing any of S. Starostin's work - meaning, of course, the REAL work, with lists and numbers and formulae and stuff. Alternatively, I could also say that H. J. J. Holm feels obliged to falsify alone Swadesh and Starostin's work. And what's wrong with that? Who said we should be bringing reinforcements?Gstarst 12:46, 8 January 2007 (UTC)

As a sidenote, I do think that someone should expand the article by describing alternate approaches to glottochronology and lexicostatistics - J. L. Bender's methodics should be described, and the Gray/Atkinson proposals, as well. If I had more time on my hands and complete access to all relevant data, I'd do it, but I welcome anybody else to contribute.Gstarst 21:58, 7 January 2007 (UTC)

[edit] New revisions partially reverted

The following comment is specially for Hans Holm, whose latest series of revisions I have partially retained and partially dared to cancel:

It is very pleasant to see Prof. Holm take a somewhat more objective stand on the issue and add some important factual data, for which I and all the readers of Wikipaedia should be grateful. However, I still had no choice but to eliminate some of the recent changes, for the following reasons:

a) typos and grammatical errors (no one is immune to this, but the text SHOULD be readable);

b) discussion intonations (phrases like "Please note that this is...", etc.) sound like they belong on this discussion page, not on the main one;

c) the rewrite was obviously incomplete, leading to duplicate entries (e. g. the Bergsland & Vogt argument is first shown by me as overcome by Starostin's method, THEN reintroduced again in the critical section as a serious obstacle!);

d) argumentation that should have first been resolved on the talk page, where Prof. Holm is no longer responding. For instance (to quote two passages that I have omitted, although I have tried to incorporate the argumentation from the first one elsewhere):

Another line of arguments is represented by Haarmann (e.g. 1990), who demonstrated that there is no region of the vocabulary "safe" from being changed, as has been argued by glottochronologists, e.g. parts of the body, colour terms, numbers, or pronouns.

This statement is objectively wrong in two possible ways, depending on what is meant by "no region of the vocabulary safe from being changed". If this means that "some of the items on the 100 wordlist can never ever be replaced", then it is wrong, because no glottochronologist (at least, no SERIOUS glottochronologist who actually worked with specific lists) has ever said that. If this means "some of the items on the 100 wordlist can very very rarely be replaced", then, obviously, Haarmann could not have demonstrated that on his limited selection of examples; a much more global analysis would be in order.

For example, English did not replace about50 % of its originally Germanic vocabulary 'by time', but by Norman dominance after the battle of Hastings, besides a long-lasting educational background of Latin. Though this event changed only one or two percent of "the" (depends on the version) Swadesh 100-wordlist, there are six items changed already before by Viking influence. Please note that this is a difference in quantity, not in reasons and computability.

Apart from sounding like part of a discussion (see above), this also makes little sense AFTER the introduction of the Starostin argument. Viking or norman influence, the items on the list either remain as entities or are replaced by borrowings. When forms are borrowed from closely related languages - so that it is sometimes hard to distinguish a borrowing from an 'ancient' word without a good knowledge of the correspondences - this creates limited technical difficulties for the method, but does not invalidate its essence. And I already said that before.

Another example is Albanian, which changed 90% of its Indo-European heritage, and still about 75% in the Swadesh list, mainly by Roman dominance, later by South-Slavonian influence.

This phrase will lead the reader to believe that 75% in the Albanian Swadesh list have been replaced under the influence of (aka: by borrowings from) Latin and Slavic languages. This is false. The actual figure, just by looking at the wordlist, is closer to 15, at most 20%. On the other hand, Albanian and, for instance, Tokharian A, share about 20% common words on the list, which is more or less what we expect from them. This can hardly represent Tokharian influence on Albanian.Gstarst 10:50, 9 June 2007 (UTC)

I would be happy to see this Albanian word list referred to. Can You give me the source?HJJHolm (talk) 17:56, 24 November 2007 (UTC)
The wordlist we use for the calculations was compiled by Vladimir Oryol (author of the "Albanian Etymological Dictionary") and is in STARLING form; I will be happy to share, but you will need to download the software from our site (starling.rinet.ru). However, just about any existing wordlist for Albanian will do - e. g. the one found here: www.cs.rice.edu/~nakhleh/CPHL/ie-wordlist-07.pdf. (Note, please, that my figures refer to the 100-wordlist; for the 200-wordlist the figures may be slightly higher, since we are bringing in less stable lexemes).Gstarst (talk) 19:45, 24 November 2007 (UTC)
I would further be happy about any corrections to my not at all perfect english or any typos. Thank you.HJJHolm (talk) 17:56, 24 November 2007 (UTC)