Talk:Lewontin's Fallacy

From Wikipedia, the free encyclopedia

Contents

[edit] Conclusions in paper

Edwards seems to say more in the paper than that individuals can be classified, namely that the difference in phenotype can be arbitrarily large so the old statistic cannot be used to draw any inference about this ("It is not true .. that two random individuals from any one group are almost as different as any two random individuals from the entire world"). Is this a correct interpretation? If so it should perhaps be expanded in the article.

Filur 12:12, 10 July 2005 (UTC)

Before you consider expanding this article, would you express an opinion (below) on the question of whether a single paper than has only been cited four times is notable enough for an entire article in Wikipedia? Thanks. Guettarda 14:16, 10 July 2005 (UTC)
It's a fairly short paper. At some point it'd be easier just to provide a link (perhaps from the Edwards article) and allow readers to read it themselves. -Willmcw 21:34, July 10, 2005 (UTC)
I do not think the article should be about the paper, but about the fallancy. Many people (and Nature) makes the mistake of believing that because a loci to loci analysis shows that 85% of the variance is within populations groups (true) it follows that humans are more alike than dislike (undetermined). When the differences are systematic, as they are, the way genotype is mapped to phenotype will determine how much differences there will actually be. My guess is that the differences will actually be smaller than what is indicated from the loci-to-loci analysis since natural selections will push towards the same solutions, and most of the difference will be in neutral parts of the DNA. Filur 07:07, 11 July 2005 (UTC)

[edit] NPOV

I don't know anything about this paper, but any article about a paper that entirely consists of a section attacking it can't be NPOV. DJ Clayworth 14:40, 20 Jun 2005 (UTC)

66.214.170.174 wrote an extensive critique of Edwards with no citations. It is available in the prior versions for those who want to read it.Jim Bowery 02:15, 27 Jun 2005 (UTC)
I concur that there's a NPOV problem - the existence of this article, with this title, is non-neutral, and it should be merged into the appropriate general articles as discussed below. Danny Yee 12:48, 15 October 2005 (UTC)

[edit] Statistical methods

The statistical methods described therein are now being used in forensic analyses.[1]. This sentence is vague. Which statistical methods? I read the article and it seems to address a different issue: determining which race a person might belong to by examining certain sections of their DNA. Can you explain this assertion better, please? Thanks, -Willmcw July 1, 2005 00:43 (UTC)

Lewotin's fallacy is so basic that virtually any synergistic use of more than one locus to do statistical inference of ancestry can be classified as the statistical method described by Edwards. The point is that there have been a very large number of peer-reviwed scientific papers that could not have been done without implicitly flying in the face of Lewontin's rhetoric and these include papers published before Lewontin's paper, such as those cited by Edwards. Perhaps there should be a better link to DNAPrint's technical use of between-loci correlation in their estimate of admixture of geographic race. Here's a quote from a stock analyst's report[2]:

DNAPrint’s Platform Technology—SNPs and AIMs

DNAPrint genomics’ core technology encompasses proprietary gene mapping strategies and patented Ancestry Informative Markers (AIMs) developed through population-level quantitative genomics mathematical modeling. Genetic mapping is the process by which each gene locus is defined relative to another whose position is already known on the basis of the frequency of recombination (the random exchange of parts of two homologous chromosomes during meiosis) that is observed between the two loci. There are a number of relatively new methods for gene mapping as physical gene maps, which rely on cloning of genes and recognizing their effect at the cellular level, have only limited utility. Newer methods rely upon population and family studies which incorporate genetic linkage into mapping techniques.

[long excerpt snipped by -Willmcw July 1, 2005 20:25 (UTC)]

Jim Bowery 1 July 2005 11:13 (UTC)

Thanks. I don't see the similarity - they seem to be coming at it from opposite sides. Lewontin was saying that individuals within a race are more different than races are from each other. Edwards said that if you combine many factors that a racial correlation appears. DNAPrint seems to be saying that with a couple of genetic Loci they can determine a person's race. That is not saying that that person is more like other members of his race than races in general - they only care about the few genes which make the difference, not the thousands which are shared. Is there a way that we can re-write the sentence to say what specific statistical tools DNAPrint uses that Edwards used? Otherwise I think the sentence is out of place. Cheers, -Willmcw July 1, 2005 20:25 (UTC)


The fairly generic term "cluster analysis" is probably the correct one to use. DNAPrint has found a way of doing cluster analysis using fewer loci than competitors so they can offer the service more cheaply.[3] From Edwards' paper: "A cluster analysis will be necessary in order to uncover the groups, and a convenient criterion is again based on the analysis of variance as in the method introduced by Edwards and Cavalli-Sforza.(10) Here the preferred division into two clusters maximises the between-clusters sum of squares or, what is the same thing, minimises the sum of the within-clusters sums of squares... The entries of the half-matrix of pairwise distances will therefore divide into two groups with very little overlap, and it will be possible to identify the two clusters with a risk of misclassification which tends to zero as the number of loci increases." Jim Bowery 2 July 2005 12:57 (UTC)

[edit] Merge with Edwards

While this paper is interesting, I don't see how a single article which has been cited 7 times (ISI Web of Knowledge) really deserves an article of its own. Guettarda 2 July 2005 18:45 (UTC)

Is that all? In that case I'd agree. -Willmcw July 2, 2005 18:56 (UTC)


Well, it is only a 2003 paper so it may still have more of an impact, but it does strike me as at least premature to have a separate article on this paper.

1. Leroi AM, A family tree in every gene, JOURNAL OF GENETICS 84 (1): 3-6 APR 2005
Times Cited: 0

2. Jorgensen S, Mauricio R, Hybridization as a source of evolutionary novelty: leaf shape in a Hawaiian composite , GENETICA 123 (1-2): 171-179 FEB 2005
Times Cited: 0

3. Jorde LB, Wooding SP, Genetic variation, classification and 'race', NATURE GENETICS 36 (11): S28-S33 Suppl. S NOV 2004
Times Cited: 0

4. Mountain JL, Risch N, Assessing genetic contributions to phenotypic differences among 'racial' and 'ethnic' groups NATURE GENETICS 36 (11): S48-S53 Suppl. S NOV 2004
Times Cited: 2

5. Serre D, Paabo SP, Evidence for gradients of human genetic diversity within and among continents GENOME RESEARCH 14 (9): 1679-1685 SEP 2004
Times Cited: 8

6. Bamshad M, Wooding S, Salisbury BA, et al., Deconstructing the relationship between genetics and race , NATURE REVIEWS GENETICS 5 (8): 598-U2 AUG 2004
Times Cited: 20

7. Andreasen RO, The cladistic race concept: A defense, BIOLOGY & PHILOSOPHY 19 (3): 425-442 JUN 2004
Times Cited: 0

Guettarda 2 July 2005 19:26 (UTC)


The number of academic citations for a paper shouldn't be the sole criteria for deserving a separate article. Lewontin went over the head of his academic peers, directly to the public as well as crossing disciplinary lines via academic politics, when he promoted the idea that race as a taxonomic construct was invalid based on single locus Fst. Lewontin's position with respect to his peers regarding the "more variation within than between races" argument is very much like Gould's. Maynard Smith points out, more is at stake. Gould "is giving non-biologists a largely false picture of the state of evolutionary theory". This is basically the same attitude held toward Lewontin's public activism. Edwards' paper should be viewed as he says: "This article could, and perhaps should, have been written soon after 1974." He says 1974 rather than 1972, the date of Lewontin's academic paper due to the fact that 1974 was the date Lewontin went to the public with this misrepresentation of the state of evolutionary biology in his book: "The Genetic Basis for Evolutionary Change". Edwards is clearly concerned about the long-standing public miserading of the reality of racial taxonomy's solid foundations in evolutionary data and theory. The fact that his reach to the public is limited compared to Lewontin's should not be held against his paper. Jim Bowery 2 July 2005 20:54 (UTC)
But we don't judge potential importance, we report on actual impact. It's worth sticking into articles related to race, to Edwards, maybe to Lewontin...I'm not saying that the information is or isn't important, I'm just saying that, based on its impact to date (as judged by citations) I don't think we should have an article devoted to this one paper. Guettarda 2 July 2005 21:31 (UTC)


But Edwards' whole point is that his paper is not of academic importance -- his point is that this is very old news and none of Lewontin's peers took seriously, from a scientific stance, Lewontin's attack on race as a taxonomic construct. How could they? Aside from Lewontin's 1972 which merely reported single locus Fst, virtually the entire significance of Lewontin's fallacy is political rather than scientific. The real scientific importance is through the political impact on the funding of genetic research -- which is significant. So significant that virtually every honest researcher admits he has to weasel word anything he says about population structure. Jim Bowery 2 July 2005 22:11 (UTC)

[edit] This article is too opaque

The conclusion reported in this article is likely too opaque for the intended audience of Wikipedia. Now, ours is not the business to make an original and more accessible exposition ("explain it better") of the argument (which breaks WP:NOR and in some sense WP:NPOV), but we can build an argument on the basis of other people's expositions. Maybe we could use the following opinion piece from NYT, which does quite a good job:

Even the NY Times, that bastion of PC, is beginning to admit the truth about Lewontin’s error, albeit only in an OP-editorial. On March 14, 2005, ARMAND MARIE LEROI wrote specifically about Lewontin, in the aptly titled, A Family Tree in Every Gene:

“The [Lewontin’s] error is easily illustrated. If one were asked to judge the ancestry of 100 New Yorkers, one could look at the color of their skin. That would do much to single out the Europeans, but little to distinguish the Senegalese from the Solomon Islanders. The same is true for any other feature of our bodies. The shapes of our eyes, noses and skulls; the color of our eyes and our hair; the heaviness, height and hairiness of our bodies are all, individually, poor guides to ancestry. But this is not true when the features are taken together. Certain skin colors tend to go with certain kinds of eyes, noses, skulls and bodies. When we glance at a stranger's face we use those associations to infer what continent, or even what country, he or his ancestors came from - and we usually get it right. To put it more abstractly, human physical variation is correlated; and correlations contain information.

Genetic variants that aren't written on our faces, but that can be detected only in the genome, show similar correlations. It is these correlations that Dr. Lewontin seems to have ignored. In essence, he looked at one gene at a time and failed to see races. But if many - a few hundred - variable genes are considered simultaneously, then it is very easy to do so. Indeed, a 2002 study by scientists at the University of Southern California and Stanford showed that if a sample of people from around the world are sorted by computer into five groups on the basis of genetic similarity, the groups that emerge are native to Europe, East Asia, Africa, America and Australasia - more or less the major races of traditional anthropology.”

I found it here. I'm not saying we should quote it verbatim, but it is a good enough basis for a neutrally reported explanation of why Lewontin is wrong. Arbor 8 July 2005 12:58 (UTC)

Lewontin's point is valid inasmuch as "race" is biologically something like "subspecies", and showing that within-group variance exceeds between-group variance means that you cannot separate these groups statistically. If, on the other hand, you look ar covariance (not correlation) you can create clusters. But these covariances do not necessarily represent shared ancestry. If you look at gene trees you will see that Senegalese and Europeans are more closely related to each other than either group are to Solomon Islanders. This is old fashioned (and outdated) systematics.
The second point, about the USC/Stanford study needs to be cited - if you look through the archives of Talk:Race you will see a discussion of a 2004 or 2005 paper which "upheld" race...when it actually simply upheld the fact that selected groups within races are more related than other groups. Again, this sort of "type concept" is unrealistic because it pre-selects distinct groups. If you used a random sample of the people of the world, including the Indians and the Egyptians, the Tureg, the Kazakhs...and you could still find races, then you would have evidence of race. But if you pick small homogenous samples and demonstrate that they differ from other small homogenous samples, then you are only saying that your samples are different. It is possible to separate the Welsh from the English if you look at genetics, but that doesn't mean that they are separate races. Guettarda 8 July 2005 13:44 (UTC)

[edit] Can we widen the scope?

Hm... I think this article is necessary, not so much because Edwards' argument merits it in itself, but because Lewontin's argument permeates a large part of the public debate. Lewontin's argument is important, so it merits an article. However, from that point of view, the article cannot be called Lewontin's fallacy, because that is hardly a neutral way presenting the issue. Can we find another title for this article, and the both present Lewontin's claim, chronicles its public reception, and presents Edward's counterargument? For example, Lewontin's fallacy could be a section of such an article (so we can link to it from other articles if we want).

The only problem is that we aren't to invent neologisms in Wikipedia. Otherwise I would call the article Lewontin's argument or Lewontin's claim. Neither term exists, so that won't work. Suggestions? Other than Lewontin's argument against taxonomic validity of races within homo sapiens based on genetic variation? Arbor 12:53, 2 August 2005 (UTC)

We have a Contemporary views on race and Validity of human races. The information in this article is already in the latter article. I could see delving into the matter in one of those articles, not starting yet another article here. Having skimmed Edwards' paper I am not even convinced that he undercuts Lewontin in any substantive way - there's no data in the paper, merely a theoretical graph. PCA is not an hypothesis-testing tool, and it can never be better than the data. What Lewontin did - picking discrete "groups" of humans - maximises between-group variation, so when he says that within-group variation exceeds between-group variation, his arguments are solid. Now if you turn around and use a classification tool on these groups ("races") you will not be drawing valid inference. For that you need to sample the whole gradient, not the end points, if you want to say that race exists. Otherwise you are looking at artificial differences. Regardless, Edwards' paper hasn't got a drop of data in it. Guettarda 13:48, 2 August 2005 (UTC)
Good points. Should we simply merge this article in the Validity article? With a separate section named Lewontin's fallacy and a redirect? Or even without a separate section? (I don't think a vote for deletion will ever suceed, so a merge seems to be the only feasible way—maybe I'm wrong.) On a side note, Edwards is merely pointing out a rather trivial misconception in Lewontin's argument (at least in the eyes of a mathematician), so he needn't present any data at all. Arbor 19:51, 2 August 2005 (UTC)

[edit] Merge 2

Again, as has been discussed before, this paper is too thin to stand on its own. A paper which has been cited 7x really isn't a big enough deal to have its own article. Since the material is already in the race article (and others), I find it hard to justify this article. If someone wants to expand this, as Arbor suggested a while ago, it should be under a more neutral title. Guettarda 13:03, 15 October 2005 (UTC)

Again, since "Lewontin's Fallacy" addresses not Lewontin's 1972 paper, but rather Lewontin's subsequent books starting in 1974, none of which were peer-reviewed, it is not reasonable to assess the importance of "Lewontin's Fallacy" by a search for academic citations. If there is an article that needs rewriting it is the article on Lewontin since it emphasizes his early academic achievements -- which were relatively unimportant in his cultural impact compared to his "politics of racial science" books. If that article is rewritten to place appropriate emphasis on those books then it might be reasonable to move this article to the article about Lewontin's real importance. Jim Bowery 15:57, 15 October 2005 (UTC)
If this paper isn't scientifically important, but is part of a popular culture debate, then that's further argument against it having its own entry. It should be merged with the various entries on the "race debate". Which can discuss Lewontin's contributions to that debate as well. (I realise that some people here are obsessed with the "race issue", but we don't want to warp various biograhical entries to place a disproportionate emphasis on that.) Danny Yee 01:12, 16 October 2005 (UTC)
Its not scientifically important in the same sense that the denial that race is a valid taxonomic construct is a scientifically unimportant social fashion. Jim Bowery 02:17, 16 October 2005 (UTC)
The article isn't a priori too thin to stand on its own. It happens to be very short, but could be expanded. I don't recommend merging with race right now because I recently added ~50k of material to the race article, and it's still in need of strong editing for conciseness. It would be a mess to try to merge right now. --Rikurzhen 18:30, 15 October 2005 (UTC)
But merging with A.W.F. Edwards works. --Rikurzhen 19:22, 15 October 2005 (UTC)
I agree with the merge. Very few academic papers deserve articles of their own. -Willmcw 21:02, 15 October 2005 (UTC)
I support a merge, but not with A.W.F. Edwards. That would unbalance that entry drastically (I hardly think this paper is the culmination of Edward's life's work). A much better merge would be with validity of human races or one of the other "race debate" pages, since that debate is the context for this paper. Or maybe it could go in a stats entry like factor analysis (which might also guarantee more impartial editing). Danny Yee 01:02, 16 October 2005 (UTC)
The validity of human races article is, unfortunately, pretty crappy right now. We'd need to clean it up before we could merge in this material, else we'd just be making a bigger mess. I don't have any time for that kind of project, but I can help with the science or philosophy if someone wants to work in that. --Rikurzhen 01:24, 16 October 2005 (UTC)

[edit] Merge 3 (Significance)

Pinker writes here (emphasis is mine):

In March, developmental biologist Armand Leroi published an op-ed in the New York Times rebutting the conventional wisdom that race does not exist. (The conventional wisdom is coming to be known as Lewontin's Fallacy: that because most genes may be found in all human groups, the groups don't differ at all. But patterns of correlation among genes do differ between groups, and different clusters of correlated genes correspond well to the major races labeled by common sense. )

If he is right (and I understand that this is a self-fulfilling prophecy) then the current article deserves its own page. Dawkins uses the term as well in The Ancestor's Tale. I think that's notable enough. Arbor 14:20, 2 January 2006 (UTC)

  • I agree that Lewontin's fallacy deserves its own page because a number of writers other than Edwards have described it from somewhat different points of view, and different levels of technical detail. Merging it into Race would only make that poor entry even more confusing. DonSiano 16:02, 6 January 2006 (UTC)
  • As a student of Lewontin's I'm not especially fond of Edwards' choice of title, but nonetheless Edwards is entirely correct. It's important to note that the scientific questions asked by Lewontin and Edwards were different. Lewontin asked "What proportion of genetic variation (in the analysis of variance sense) in humans is within and among populations?" The answer is that ca. 85% is within populations, the rest among populations and races. This is the answer Lewontin gave, and it is entirely correct. Edwards asked "Can individual humans be assigned to races from genetic data?", or, alternatively, "Can human races be diagnosed (in the taxonomic sense)?" The answer is yes, they can. Edwards shows that his answer to his question is entirely compatible with Lewontin's answer to his (i.e. Lewontin's) question. A paper by Rosenberg et al. (2002. Science 298:2381–2385) clearly illustrates for a large data set the truth of both Lewontin and Edwards' answers to their respective questions. Lewontin goes on from his finding (with which Edwards entirely agrees), to argue further that this level of difference between races is not worthy of taxonomic recognition. Edwards doesn't actually express an opinion about whether human races should be recognized taxonomically, but does show that the 85/15 division of within/among population variation is no bar to doing so. Lewontin and Edwards agree on the moral equality of human beings; Edwards just doesn't want that moral equality to depend on any contingent facts of genetic similarity. Lewontin wouldn't want it to either, but regards the high genetic similarity among human races (which is much lower among races in some other species) as empirical reinforcement for his moral conclusion.
  • Parenthetically, of the two links concerning some commercial genotyping method, one was dead, and the other was to a poorly written patent application. The patent applicants do have some papers on the method in the peer-reviewed scientific literature. A reference to one or more of these might be more useful, but there are loads of such methods, and no claim by the authors of this one that it is any particular way inspired by or related to Edwards' critiques of Lewontin, so I deleted both. Also, Edwards discusses ordination more than cluster analysis, so I added that in.
  • As regards the merge with A.W.F. Edwards, I wouldn't normally endorse a page for a single paper, but a merge would, as Danny Yee correctly notes, drastically imbalance that article (Edwards' signal accomplishments are in the origination of quantitative phylogenetic methods and the use of likelihood in scientific inference, not this paper), and the evidence presented here by other editors shows that the term has leaked into the general culture, so I would somewhat reluctantly endorse keeping a separate page.-- MayerG 07:10, 13 January 2006 (UTC)

.

I have taken the liberty of writing a paragraph at the end of this article trying to succinctly summarize the different questions that Lewontin and Edwards are asking, and arguing that the Fallacy is or is not a fallacy depending on what question is being asked. (I am a student of the one and an old friend of the other so I'm trying not to get partisan). I would be happy to see my paragraph modified or withdrawn depending on the reaction here. I have no immediate opinions on the merging issue. Felsenst

[Copy from Talk:A.W.F._Edwards ] Oppose merger of the Edwards article and the Lewontin's Fallacy article. Edwards has been a pioneer in a wide variety of disciplines. That can and should be distinguished from his work on Lewontin's Fallacy, because the latter has taken on a life of its own in academia, genetics and popular politics and media. To ensure that Edwards' other work is given due attention, i.e., to prevent the "tail from wagging the dog", the Fallacy should be left as a stand-alone article. Lethiere 20:25, 29 June 2006 (UTC)

I agree. I've heard of Edwards on many occasions, without ever hearing of Lewontin's fallacy. Michael Hardy 21:51, 29 June 2006 (UTC)

Based on Pinker's summary ("The conventional wisdom is coming to be known as Lewontin's Fallacy")[4] and along the lines of Lethiere's above comment ("[Lewontin's Fallacy] has taken on a life of its own in academia, genetics and popular politics and media") a merge seems unnecessary. Does anyone have arguments against closing the merge proposal?--Nectar 22:35, 6 July 2006 (UTC)

[edit] Different questions?

I notice that Nectarflowed added a request for a reference to my assertion that "Whether or not the Fallacy is a fallacy depends on the question being asked." I have reverted this, as I don't think it needs a reference. Greg Mayer has made essentially the same point as I did in his contribution above. If you have a bunch of traits that each differ by a modest amount between two populations (say the differences of population mean are 0.1 within-population standard deviations) then with enough traits on an individual you can assign it almost unambiguously to its correct population. Does that mean the differences between the populations are "large"? It entirely depends on whether you want that statement to imply that the differences of a typical trait are large (they aren't) or you want it to imply that one can distinguish individuals from different populations using enough traits (you can). Felsenst 20:01, 25 August 2006 (UTC)

The problem is WP:NOR. I have no doubt that a reasonable argument can be made that they were asking different questions, but as far as I know, such an argument has never been published. Perhpas you or Dr. Mayer could write a response to Edwards, which we could then cite. --Rikurzhen 20:16, 25 August 2006 (UTC)
How about this one? Chakraborty, R. 1982. Allocation versus variation: The issue of genetic differences between human racial groups. American Naturalist 120: 403-404. The difference between these two ways of describing differences between groups were rather thoroughly hashed out in that journal in the late 1970s and early 1980s and don't need a new paper by Greg or me. Felsenst 21:20, 25 August 2006 (UTC)
Probably good enough. --Rikurzhen 21:30, 25 August 2006 (UTC)

[edit] Caricatures ... of what?

My "caricature" of Edwards's position was removed, leaving the caricature of Lewontin's position. I think we need to discuss this. If no one wants to do so, I will reinsert my caricature. If the ability to distinguish two populations by using multiple characters implies that "race is a valid taxonomic construct" and if we can distinguish Swedes and Norwegians using multiple genetic loci, why is that not relevant? Why does it mean that we don't get to call them different races? Felsenst 14:56, 20 November 2006 (UTC)

I agree with your point and the inclusion of your caricature. Let's put it back in Terry 04:43, 3 December 2006 (UTC)

[edit] Misleading article.

This article does not represent the scientific consensus on this topic. What it represents is largely the agenda of the Pioneer Fund and the American Enterprise Institute. There are any number of articles published in the last 5 years that specifically refute all the claims made in this article, most especially the claim that "genetic differentiation is greatest when defined on a continental basis". It's demonstrably false.

This article was written as a sock-puppet reference for a larger argument about the validity of race in genetics. On those grounds it should be rewritten or deleted, since its scope is both narrow and crafted with intention to deceive.

Insofar as this article is a synthesis of cherry-picked racist sources, it also qualifies as original research, and it should be deleted entirely. --76.222.55.170 (talk) 08:12, 7 June 2008 (UTC)