Talk:Joseph Greenberg
From Wikipedia, the free encyclopedia
Contents |
[edit] Greenberg & controversy
Scythian99 is quite right, I think, to make sure that readers will not get the impression that Greenberg's views on language comparison or on language families were mainstream. Indeed, he was one of the most controversial linguists of his century, and not easy to write a NPOV article about.
I've tried to be a bit more positive than Scythian99 without glossing over the intellectual scandals. My impression is that the early African work is indeed generally respected, and so were the early contributions to typological study. It was in his later work that Greenberg got into the biggest controversy, to the point of people questioning his competence.
Opus33 01:26, 2 Feb 2004 (UTC)
[edit] Discussing recent edit
Dear colleagues,
I've edited the article with the intent of making clearer that there really two objections to Greenberg's mass-comparison work:
- whether we can trust the basic method
- whether we can trust the data that Greenberg fed into the method
Both are important. For instance, if it turns out that only the second criticism is valid, we can hope that future scholars will be able to apply the method with greater scholarly acceptance. But both points should be made.
In connection with this, I've re-inserted User:Billposer's sentence:
- Many of them report errors in the majority of the items that they examined, and some even report that every form examined is erroneous.
which strikes me as very important and relevant in this connection. I hope this will not strike editor User:Jorge Stolfi as unfair. Cheers, Opus33 22:29, 2 May 2004 (UTC)
[edit] Unhappy about edits...
I an unhappy about the last round of edits chiefly becuse they moved the "data accuracy" criticisms into the section about the weaknesses of the methodology. The two issues are largely independent and should be addressed as such. Besides, I suppose that linguists have already tried to replicate Greenberg's analysis with corrected data, and it would be nice if someone reported on the results of such attempts.
Moreover, from the references given, it seems that that the "data accuracy" criticisms were directed mostly against the Amerind family work, whereas the Eurasian work is mostly rejected because of the methodology alone. (If that is not the case, then some references about errors in the Eurasian data are in order.)
Now, speaking as an outsider, I can appreciate (perhaps better than most linguists) the statistical and conceptual problems in Greenberg's method, and I cannot criticize those linguists who prudently reject it. Yet the strong feelings which I see in the debate, and the use of terms like "the community of historical linguists", make me think that the polemic may be fueled in no small part by academic socio-politics, rather than purely scientific issues.
That such "fuel" exists is undeniable. Linguists who have been teaching for decades that the bogus lexical comparisons of armchair linguists are worthless pseudo-science may be understandably upset at seing the idea being given academic status. Experts in the comparative method will undestandably feel a visceral antipathy for any development that threatens to make their expertise "old fashioned" if not obsolte, and make them followers rather than leaders. And so on.
These observations of course do not affect the valididity of the method and its criticisms, but weaken the significance of the "veredict" passed by "the community".
Having seen several such situations in my own field, I cannot avoid seeing parallels betweeh Greenberg's case and other "iconoclastic" revolutions in science, like Wegner's continental drift theory (which was strongly rejected by the "community" of geologists, before becoming a new dogma), Prusiner's prion theory (which took decades to be accepted), etc..
It is disturbing, for instance, that there seems to be little attempt by critics to separate (1) the data and methodology from (2) the proposed classification; on the contrary, it seems at times that some criticisms of (1) are chiefly motivated by "a priori" rejection of (2).
So in my opinion it is too early to pass judgement on the validity of Greenberg's method. At the very least, the article should explain the arguments for both sides, estimate the amount of support that each side has got, but leave the decision to the reader. IMHO, only a few decades from now we will be able to tell whether "mass lexical comparison" was another prion hypothesis, or another cold fusion affair.
All the best,
Jorge Stolfi 15:38, 3 May 2004 (UTC)
[edit] Reply for Jorge Stolfi
Hi Jorge,
I see your point, and would welcome further editing that made the article satisfactory from your point of view.
My priorities for the article are these:
- Maintain NPOV--hard though that may be to do!
- Make it clear that the "traditional historical linguists" are not just a bunch of fuddy-duddy humanities types poring over manuscripts. They've got their own method, which is very rigorous and has accumulated an excellent track record over the last fifteen decades or so.
- Keep separate the evaluation of the mass-comparison method from the evaluation of the data that Greenberg fed into it--I think we are agreed on this point, and it's only a matter of what gets put where. I now understand why you want to keep the Amerind and Asian discussions separate.
- Make clear that the quality of the data fed into the method was such that a lot of well-informed scholars were outraged (hence my restoring of User:Billposer's sentences). Alas, I don't know of any attempts to rerun the method with better data since Greenberg published his work--perhaps people have been deterred by the huge amount of labor that would be involved.
I would welcome some rewriting on your part that maintained NPOV as you see it, fixing the problems you mentioned while respecting the goals laid out above, with which I suspect you would agree.
One last bit: chucking all the stuff I put in about the "research community" I put in would be wise; that's just bullying and I regret having written it.
Yours very truly, Opus33 16:09, 3 May 2004 (UTC)
[edit] Explain move, requests for more info
OK, I have moved the data-quality criticisms back to the Amerind section. I have also kept User:Billposer's note that many critics found errors in most items that they checked, and also the "garbage in, garbage out" comment, which I believe accurately described the view that many linguists have of that work. (Was that phrase actually used in some of the articles?) Please check and fix as you may see fit.
Should we add that the method has yet to be re-run with corrected American Indian data?
Is it correct to say that the Eurasian work is rejected only because of the method, not because of the data?
Are there linguists who specifically disagree with Greenberg's trees and have proposed different (incompatible) groupings for those languages? Or is it only a choice between "Greenberg's tree" vs. "no tree"?
All the best,
Jorge Stolfi 18:21, 3 May 2004 (UTC)
[edit] Usher comment
Dear colleagues,
Jorge wrote,
- "Or is it only a choice between "Greenberg's tree" vs. "no tree"? "
Quite correct, at least in the Americas. One currently in-vogue counter-theory is that dozens of totally different migrant families exchanged their most basic vocabulary including personal pronouns, while passing through Beringia, even though no such phenomenon has been directly observed, and Beringia would have to have been a surprisingly lively and multi-ethnic place c. 12,000 or so b.p....
- "Are there linguists who specifically disagree with Greenberg's trees and have proposed different (incompatible) groupings for those languages?"
In Africa, Hal Fleming suggested changes (the recognition of Omotic) which were immediately accepted by Greenberg - just one region where provisional acceptance of JG's results, followed by vetting and modification, proved more fruitful than hostility and reflexive denial.
There have been many other minor changes to this classification, and also Gregerson whose joining of Niger-Congo and Nilo-Saharan into a large Kongo-Saharan family met with JG's general approval.
In the Americas, radical skepticism and just-so stories (see above), now fueled by the absurd theories of Johanna Nichols, have been the norm. No doubt there are errors to be found, but not this way.
In New Guinea, I have spent much of the last decade figuring out what is going on, and have arrived at a very substantially different (improved, actually!) classification with the benefit of piles of new data.
- "IMHO, only a few decades from now we will be able to tell whether "mass lexical comparison" was another prion hypothesis, or another cold fusion affair."
I'd suggest that some of the classifiations are right, some are wrong, some wrong in one way but right in another, etc. It is not as stark as prions or cold fusion. People deal with this in the way they do because to get involved and gain expertise is a life-long endeavor. What is needed is not to negate Greenberg's findings a priori - unless we'd like to start with Africa? - but to vet this work with the traditional techniques that are being promoted. If it is held that traditional methods don't work within this timeframe (I happen to disagree), that is a limitation of the traditional methods, not of Greenberg's.
-
- I was referring to the mass comparison method, not of the classifications. By the way, was the African tree also obtained with that method?
- Jorge Stolfi 15:28, 4 May 2004 (UTC)
Opus33 wrote,
- "...readers will not get the impression that Greenberg's views on language comparison or on language families were mainstream."
But some very salient views, such as his view on African classification - this is what, 30% of human languages? - are nearly universally accepted. His last work on Eurasiatic has been criticised for all sorts of reasons, but is actually not too far in its results from those of the Nostratic school, and as Jorge highlights, the alternative is no tree at all.
The main place where his work has been savaged is in the Americas, where splitters predominate - if one takes the current SIL classification at face value, by the age-area hypothesis, human language should have originated in the Americas!
But geneticists and physical anthropologists have confirmed the basic outlines of his classification. You are more likely to find problems by burying yourself in the data and suggesting alternative classifications than by negating the whole thing by a priori objections (which I note require a much lower time commitment than actually engaging the data in a better way!)
- "Many of them report errors in the majority of the items that they examined, and some even report that every form examined is erroneous."
Can you name a scholar who reported that "every form examined is erroneous", along with the language/list name Greenberg used, as well as the source cited by the objector? I will check the two lists against one another, and I will be extremely surprised if there is any list in which "every form is erroneous". This is just garbage other people have told you which you have not checked for yourself, but what if you discovered that your source hadn't checked it for themselves, either?
- "Make it clear that the "traditional historical linguists" are not just a bunch of fuddy-duddy humanities types poring over manuscripts. They've got their own method, which is very rigorous and has accumulated an excellent track record over the last fifteen decades or so."
Anything you'd care to cite in say, the last thirty years? I'd have a few suggestions myself, but I'm curious as to what you have in mind.
to all:
And for the sake of full disclosure, I knew Joe Greenberg, and I have lots of criticisms of/suggested improvements to his work and methodology, but he was definitely no charlatan. Further, it's clear to me that much of this criticism is just second-hand repetition of what you've heard from others.
If you're going to take the time to accuse a dead man of data fabrication, don't you think you should have checked the dataset yourself?
Someone asserted that the rate of linguistic change is so high that all genetic resemblances will be lost in random noise after a certain period of time - can anyone offer any proof that this has actually occured, and what this period of time must be?
Again, you are only repeating what you have heard from others, who did not know themselves.
Timothy_Usher
-
- You are presumably referring to a remark of mine. That is essentially a truism, not specific to languages but to any kind of "signal". Even the most fundamental features of a language can be eroded away in time, to the point that resemblances become are virtually undistinguishable from random coincidences and later borrowing. (The Indo-European case system has essentially disappeared from Romance languages and from English; the English plural switched from Germanic-like to Romance-like; and the second person pronoun "you" has almost completely disappeared from Brazilian Portuguese, over a couple of centuries). The big question is how long would that take. 10,000 seems to be beyond reach of the traditional method; I would rather not guess a limit for mass comparison, but given its reported failure in joining North and South Caucasian, it could be as short as 30,000 years.
-
- Jorge Stolfi 15:28, 4 May 2004 (UTC)
[edit] Why weakness section was rstored
I have restored large sections of my contributions that were summarily deleted, aparently without explanation (including a factual section on Greenberg's Eurasiatic work - the baby in the bathtub?)
Judging from what had been deleted, it seems that the revisors were unhappy about my "explanation" of Greenberg's method and especially its presumed "weaknesses". Now, I am not a linguist, so it is quite likely that I bungled my criticisms -- which I trust were perceived by Greenberg himself, and are largely based on statistics, something that I feel a bit more confident to write about. However, deleting the discussion of the method without providing an alternative can hardly be considered an improvement. Without such a discussion, the debate will be reduced to a my-daddy-is-stronger-than-your-daddy shouting match, and lay readers will not get anything useful out of the article. If you see mistakes, please fix them -- or just point them out, and I will be happy to fix them.
Ditto for the reference list: the proper response to a perceived anti-Greenberg bias is to add references to pro-Greenberg articles, not to suppress those against him.
As for errors of attribution, as long as we are discussing the method it does not matter who said what, or how the idea was originally presented: as in any other science, the method should be presented in the clearest possible terms, and discussed on its own merits. However I agree that this page is not the best forum for that, and perhaps we should create a separate mass lexical comparison page.
All the best,
Jorge Stolfi 14:38, 4 May 2004 (UTC)
[edit] agree on separating mass lexical comparison
- I agree on creating a separate mass lexical comparison page. Greenberg did a whole lot of work apart from mass comparison (his typology work, especially, is still very notable) and discussion of mass lexical comparison's merits and flaws belongs in its own page. - Mustafaa 16:40, 4 May 2004 (UTC)
[edit] removal of Eurasian section
Re the removal of the Eurasian section: This reference seems to be about it
- Greenberg, Joseph H. (2000) Indo-European and its Closest Relatives: the Eurasiatic Language Family – Volume I, Grammar. Stanford: Stanford University Press.
So what was wrong with that section?
Jorge Stolfi 15:28, 4 May 2004 (UTC)
-
- PS. Before I started editing this page, its contents was very critical of Greenberg's work; and later edits were all anti-Greenberg. Even though I tried to point out that the idea in itself is valid (just from statistical theory), I have generally tried to preserve the overall negative tone, trusting that it reflected the mainstream view. However, I now believe that the current text, especially the "weaknesses" section, comes out too negative and unfair. I still believe that the weaknesses are real and recognized even by "Greenbergians", but they certainly do not invalidate the idea.
-
- Jorge Stolfi 15:48, 4 May 2004 (UTC)
[edit] Usher comment #2
Billposer wrote,
- "Many of them report errors in the majority of the items that they examined, and some even report that every form examined is erroneous."
Who is the scholar who reported that "every form examined is erroneous"? To what Greenberg list was he referring, and what source did he cite?
Jorge Stolfi wrote,
- "The big question is how long would that take. 10,000 seems to be beyond reach of the traditional method; I would rather not guess a limit for mass comparison, but given its reported failure in joining North and South Caucasian, it could be as short as 30,000 years."
I'm not sure whose study you're referring to with Caucasian.
But you're right, the question is how long should this take. Greenberg addresses this in detail in (1987), and Sergei Starostin has some ideas on this you might find valuable - I am neither a statistician nor a glottochronologist,but I'm quite certain than 10,000 is not right, and even more certain that this has often been asserted, but never adequately demonstrated.
Even so, there is a crucial point here which is usually overlooked in this type of discussion: supposing the threshhold of randomness really is 10,000 years, this would only apply between two languages relative to one another. If we can reconstruct two protofamilies which are 5,000 years old, then we might be able to reconstruct their ancestor at 10,000, and so on.
But doesn't this apply only to the traditional comparative method? No. Mass root comparison takes a different route to exploting the same phenomenon, the transitivity of relationship. Languages may be classified relative to one another even if they share NO items in common, provided we can place them within a group which is then relatable to another group as a whole.
- "By the way, was the African tree also obtained with that method?"
Yes, it was, although with the benefit of a greater amount of preceding work to draw from than in the Americas.
- "Besides, I suppose that linguists have already tried to replicate Greenberg's analysis with corrected data, and it would be nice if someone reported on the results of such attempts."
Ha ha ha! Linguistics should be a science, but usually is not. To attempt to replicate his results would take lots and lots of time that is better spent denouncing him for thinking it possible!
There have been no such attempts that I know of, and if there are to be, they are more likely to come from his sympathizers than his detractors, who after all are saying that comparison at such depths is by definition a waste of time.
Timothy_Usher
[edit] how far back can we reach?
Timothy Usher asks what is the source for my comment "reported failure in joining North and South Caucasian". That was another unattributed statement which I read in some Wikipedia page on Caucasian linguistics (the line is now in languages of the Caucasus), and apparently refers to Greenberg's Eurasian work.
The estimate of 10,000 years for the traditional method is my guess based on the apparent reluctance or inability of mainstream linguists to propose trees that extend that far in the past. I take it that Nostratic is not considered mainstream either, is that right? What is the oldest family that is considered "mainstream"?
I gather that there is a major debate about when exactly were the Americas inhabited -- some say "no more than 12,000 years ago", others belive that some South American sites are older than that, perhaps 20,000. (A respected Brazilian archaeologist claims that there are several sites in Brazil which are older than 12,000 and show a very distinct population, more like the Pacific/Australian type than the Siberian/Mongolian one.) Anyway, the three waves recognized by Greenberg presumably came through Western Siberia. So we have three linguistic phyla in that small region that could not be connected by Greenberg's method. How much older could be their common root?
Jorge Stolfi 17:37, 5 May 2004 (UTC)
[edit] Caucasian
Hello Jorge,
The quote in the Caucasian article is,
"The South Caucasian and North Caucasian families are unrelated phyla even in Greenberg's deep classification of the world's languages."
This is a misinterpretation of Greenberg's position. Greenberg believed, as do I, that all natural human languages share a common historical origin, from which they subsequently evolved and diverged - evolution in the strict sense of "descent with modification".
Greenberg saw Kartvelian - and this is also supported by Nostraticists such as Sergei Starostin - as being closer to Eurasiatic than it is to N.Caucasian, just as English is more closely related to German than it is to Hindi, Albanian or French.
I don't know that he had any opinion as to whether the two Caucasian families were easily relatable (as opposed to related, which all languages are), but it's a moot point, since it would be more sensible to compare Caucasian with Eurasiatic as a whole, if one accepts this classification.
Supposing no descendent of Eurasiatic had survived, Kartvelian would still be placable in theory, although with less reliability, relative to the other languages of the world (Joe did have ideas about the outlines of such a classification, it's just that this was not the subject of his Eurasiatic book, as was erroneously suggested in earlier versions of this article). The difference would be the lack of a Kartvelo-Eurasiatic node, just as if English were the only surviving and attested Germanic language, it would be considered an independent branch of Indo-European.
Timothy_Usher
[edit] single origin or multiple origins?
Timothy, you say
- Greenberg believed, as do I, that all natural human languages share a common historical origin, from which they subsequently evolved and diverged - evolution in the strict sense of "descent with modification".
I happen to believe that too, but there are other possibilities.
Consider the biological evolution of modern Homo sapiens. The "seven Eves" popular theory notwithstanding, it is unlikely (indeed near impossible) that all modern humans descend from a single Adam and Eve pair. The species presumably evolved from a population of proto-humans whose genetic makeup drifted far enough from that of other populations to become a separate species. All the while the members of this population remained interfertile and so the whole population would have been substantially homogeneous genetically (numerically at least, 20-30 generations -- 500 to 1000 years -- seem more than enough to do that for a nomadic population.) The big question is how large was this initial population: one extreme camp thinks it was a small population in Africa which radiated only after evolving into modern humans, the other extreme says that the evolution happened over the entire world more or less simultaneously. The former view seems more popular, but the latter has not been ruled out AFAIK (genetic diffusion may be fast enough for that).
So the same thing may have happened with languages. One can imagine that, sometime after the necessary genes diffused over the globe, language was "invented" independently in different areas, with onomatopoeic lexicons and a (very rudimentary) random grammar. The result would be several proto-languages with no linguistic connection between them. Each of these proto-languages could have evolved into an independent family with sophisticated syntax, and perhaps two or more of those families managed to survive to this day.
My guess is that the "small tribe in Africa" is more likely for biological evolution, and that all languages of today are descendant of the language of that tribe; but it is only a guess.
Anyway, language evolution is fundamentally different from genetic evolution, in such a way that even if there were two or more unrelated linguistic phyla 200,000 years ago, is its extremely unlikely that more than one of them would have survived to this day. For this reason, mainly, I believe that all languages of the world have a common ancestor that is substantially more recent than the first human language — perhaps as early as 80,000 to 100,000 years ago
On the other hand, I see the simple tree model of language evolution that is assumed by historical linguists as beeing way too simplistic and useless beyond a few millenia. Languages don't just branch, they can mix, too. Saying that English is a Germanic language, or Persian is Indo-European, are attempts to arbitrarily binarize what should be a continuous measure: one should admit that English is X% Germanic, Y% Romance, an perhaps Z% Celtic (in grammar and morphology as well as in its lexicon). The "minor ingredients" can perhaps be ignored in studies spanning a few millennia, but for longer time periods they may cause the official "genetic" ancestor to become a completely irrelevant concept. Of course, if linguists were to admit that languages can mix, they would no longer be able to draw those neat trees that look so nice in their papers...
Jorge Stolfi 14:20, 6 May 2004 (UTC)
[edit] language mixing?
There are some cases of mixed languages that are pretty convincing on grammar and morphology as well as vocabulary (well, at least one: Michif language.) However, this seems pretty rare, and in the normal course of things pretty slow. As far as I know, only one "grammatical" morpheme in English has been borrowed: they, from a closely related Scandinavian language, whose absorption was no doubt facilitated by its similarity to the previous English form (though this depends what you define as grammar vs other components.) Lexicons mix a lot more easily. Of course, as you point out, even the small amount of basic-level mixing that exists will mess things up over a sufficient time scale; but I suspect that timescale is a lot longer than a few millenia (especially given that inter-communal communication has gone up, not down, in the interim, without creating all that many cases to compare.) It would be interesting to try and extrapolate statistics. - Mustafaa 18:27, 6 May 2004 (UTC)
[edit] American language families
Jorge wrote,
- "So we have three linguistic phyla in that small region that could not be connected by Greenberg's method. How much older could be their common root?"
See my comments on Caucasian above. It is not that they could not be connected, only they that a node containing all three of these families and only these three would be invalid - were this not so, only one American family would have been identified, with Na-Dene and Eskimo-Aleut merely its most distant members. Greenberg's refrained from doing this because it cannot be shown that the three American families are closer to one another than to families outside the Americas - in the case of Eskimo-Aleut this is especially clear, since, according to Greenberg and to most Nostraticists, closer to the Eurasiatic families than to Amerind, and was included in JG's Eurasiatic group. Na-Dene for its part has been placed within a larger Dene-Caucasian group - if this is correct, than the relationship between the three families of the Americas is a function of the relationship between Amerind, Eurasiatic and Na-Dene.
Timothy_Usher
[edit] Confusion
This interface is EXTREMELY confusing.
I cannot figure out if comments should be posted on the top of the page or the bottom. When one chooses "post comment", they wind up at the bottom, where several new omments have just appeared. I don't care either way, just wondering if there is some standard for this. However, when we post at the bottom, it takes us away from the menu items at the left, which is inconvenient if we wish to post our own comment.
Further, I can't figure out how to get my name and time/date stamp to post automatically.
Timothy_Usher
