Bomb#20
Contributor
- Joined
- Sep 27, 2004
- Messages
- 8,213
- Location
- California
- Gender
- It's a free country.
- Basic Beliefs
- Rationalism
No, we really don't. "Said it's a phenotype group" does not equal "discussion of a phenotype group". It was some of the other things he said in his discussion of a phenotype group that show he thinks it's also a genotype group. On the other hand, if the rival contention you made is something you're no longer supporting, that would appear to be agreement.Yeah, I got that point. And it's true. So what? It doesn't provide any logical support for your contention. All it shows is that we cannot infer, from only the fact that Mayr said it's a phenotype group, that he also thinks it's a genotype group.
So.. the reason why I'm disagreeing with you that Myr's discussion of a phenotype group shows that it is a genetic group is because I don't agree that it does show that, not because I'm supporting some rival contention. We now appear to agree.
Huh? Are you suggesting that if somebody proposed a human taxonomy in which people were grouped by the presence or absence of alcohol resistance rather than by complexes of many correlated characteristics, this would be accepted by biologists as valid taxonomy?A race isn't just any old phenotypical grouping; it's a phenotypical grouping with the additional properties of differing taxonomically from other populations...
So are many phenotypical groupings. Alcohol resistance, for example.
"If", you say. So does that mean you're desperately clutching at straws, or do you think that's actually what happened? Do you believe we know which genes are responsible for most of the many visible anatomical differences anthropologists use to judge race? For instance, do you think we know what genes cause an epicanthic fold? And do you think, what's more, that we already knew about all those genes over 20 years ago, when the Human Genome Project was barely underway? And out of all the biochemical markers Cavalli-Sforza et al. could have studied, do you imagine they picked the particular ones that drive visible phenotypical race characteristics, and forgot to mention it?In short, the computer recognized what is almost the exact category called "Caucasoid" in traditional pre-PC physical anthropology. How did this happen, if there's no such thing as a biological Caucasian? Coincidence?
Quite possibly, yes. If the 150 markers being used are the same ones that drive the phenotypical characteristics that drove the comparison in the first place, then all that's being measured is the same phenotype grouping that was initially selected.
In any event, what HGHG says about the particular 42 populations and 120 markers they chose is that they collected all the published marker frequency data they could lay their hands on -- data about hundreds of markers and hundreds of population groups -- and then they eliminated populations for which too few genes had been reported and eliminated genes for which too few populations had been reported; and in some cases they grouped two neighboring populations together; all with the goal of driving the problem of holes in the Gene-by-Population matrix down to a manageable level. About a quarter of their chosen alleles were from white blood cell antigens alone -- lots of data about those genes' ethnic correlations is available, because it's been been studied up the wazoo, because they're what determine tissue compatibility for transplants.
Not in itself, that's true. You also have to run tree construction algorithms over and over with random variations in which genes you include and check whether the results are repeatable. The closest the Caucasoid group came to not showing genetic coherence is that the Berbers of North Africa sometimes got grouped with Negroids instead of with the rest of the Caucasoids -- with a probability of 2%.Similarly, if you look into the pariwise comparisons, what you're seeing are between groups measures and intragroups measures. While it's tempting to simply line them up and compare the relative strength, that's not in itself a demosntration of genetic coherence.
But that could only happen if you applied the formula only to carefully chosen genetic markers that you know in advance control the specific phenotype features you sorted by, along the lines of your speculation above. It's not going to happen if you just run the formula on whatever markers you have available.Social constructs exist only in the brains of human beings. Does the Fixation Index formula mathematically read their minds?
It could easily be, yes. For example, if you create an entirely arbitrary phenotype group, then you're going to find some genetic markers that they have in common, and which other groups do not have in common. In that sense, the formula is reading your mind, because all it's doing is confirming that within the group you've a priori chosen as signfiicant, there are features you can use to carve out some genetic similarities, that establish the group you've chosen as significant. All it's doing in that case is reinforcing your intiial assumption.
Are you accusing me of writing something four years ago, to some third party, in an FRDB thread I don't recall you participating in, that is retroactively an attempt to misrepresent your current views?Anyone claiming "Caucasian" is a social construct resulting from drawing arbitrary lines across a featureless terrain of human genetic variation has an awful lot of explaining to do.
Yes, such as 'why are you made of straw?', and 'how did you come to repalce the people who actually disagree with me'.
If the position you are arguing for when you say race is not a genetic grouping is substantively different from the opinions the race denialists in that thread were advocating, feel free to expand upon your own position, draw my attention to the significant differences between your view and my summary of their views, and tell us why there's no prima facie incompatibility between races not being genetic groupings and the facts reported in HGHG.
Of a prediction? Certainly. I predict that Obama is heterozygous for a -33T->C SNP in the erythroid promoter region of the FYB allele. If you mean an example of a prediction we can test here on TFT, no, of course I can't, since we don't have access to any individual's gene sequence. If you want to test the predictive power of race here on TFT, we'll have to do it with predictions about statistical characteristics of populations.Ok, so you reckon race has predictive power? In the case of individuals? Can you give an example?
Do you have an argument for that opinion, or is it just something you assume is true because you picked it up by osmosis from ideologically compatible people? If the latter, is it an opinion you're committed to, or is it an opinion you'd toss on the junk heap if you saw their beautiful theory slain by an ugly fact?And similarly, testing genes in 150 dimensions* does not tell us much about the variations on the other few million.
Well, they're not random at the galactic level of grouping into spiral arms instead of amorphous clouds; but they're pretty random on the hundred-odd-light-year scale where we can resolve individual stars with the naked eye. The different stars in, say, Ursa Major, are at wildly different distances and got within a few degrees of each other in the sky by very different histories from very different origin points. In contrast, a typical Irishman looks more like a typical Italian than like a typical Korean because the Irishman and the Italian are Nth cousins, the Irishman and the Korean are Mth cousins, and N is a good deal smaller than M. Between N and M generations back, the Irishman's and the Italian's ancestors shared a history they did not share with the Korean's ancestors.If you repeat your multivariate analysis in a different space, with a different hundred axes measuring a different hundred genes and phenotypical features, you'll get pretty much the same clusters, because there's only one history of who was making babies with whom, and that history created both clustering patterns.
Not following this, I'm afraid. Stars aren't random, and also only have one history.
Approximately the same, yes. If you include enough neutral and weakly selected genes to average out the statistical peculiarities every gene has, then the results will be repeatable. That is the whole point of modern taxonomy: repeatability is what makes it science. Did you think biologists switched from classifying organisms by subjective similarity to classifying them by common descent because Darwin made common descent fashionable?Are you saying that it doesn't matter which genes you track, you'll get the same results? Because that's not a claim I've seen in your sources...
My banner headline?!? That's AA's banner headline. My banner headline, if that's what you want to call it, is merely that what Mayr wrote doesn't reinforce genetic race critics' contentions. Races being real is something I only argued for because you kept pressing me to. It's a question I think belongs in the natural science forum, where the expected standards of evidence are higher.The point I'm making is that it isn't enough to show that there is some sort of genetic group out there, but rather you need one that would make racial distinctions valid at least in theory, or else your banner headline (races are real) is at best highly misleading.
I'm not following what you mean by a race having features intrinsic to the group and features extrinsic to the group. Is mostly having a C at a spot in the FYB gene where most other people have a T an "intrinsic" feature of the "Negroid" group? Is being widely perceived by Americans as "probably better than me at math" an extrinsic feature of the "Mongoloid" group? (An unfortunate name if there ever was one!)What happened is his data showed "Caucasoid", "Negroid" and "Mongoloid" correspond to natural features of the human gene pool. That looks to me like a relationship to races as the term is commonly understood. If that doesn't look to you like a relationship to races as the term is commonly understood, why doesn't it look that way to you?
Because for a 'race' to be a useful and valid signifier, it needs quite a few characteristics. It needs to be identifiable, which we have. It needs to have features intrinsic to the group, and features extrinsic to the group, such that membership of the group is a highly reliable predictor of the presence and absence of various traits outside of those used to identify the group in the first place, and so on.
Whatever intrinsic and extrinsic mean, there are millions of genetic features outside of those used to identify the groups in the first place that membership in any of the groups I listed is a highly reliable predictor of the presence and absence of. However, very few of these traits are as simple to describe as the above-mentioned T->C variation. The great majority of predictable traits are statistical ensembles of many genes. If you'd like, you can go discover one yourself.
Suppose you want to make a novel prediction about somebody's genes, based on his membership in the Caucasoid group. You can do it pretty easily using Yale medical school's online database of gene frequencies, which is freely available to the public. All you need to do is click the "Sort by: #Pop" button to get the genes with the most data to the top of the chart, generate a set of random numbers, use them to select a bunch of well-sampled genes, and look up their frequencies in various populations. So if your random number generator says "1", you'd click on the first site, rs1801133, and click on "Frequency Display Formats: Table". Add up the numbers and get the average frequency of having an A instead of a G at that site across the Caucasoid populations, across the Mongoloid populations, and across the Negroid populations. (Eyeballing it, it looks like about 25% among Caucasoids, 35% among Mongoloids and 5% among Negroids. In a perfect world you'd weight the numbers by the size of the respective populations, to take into account that a random Caucasoid is way more likely to be French than Irish; but don't bother -- you won't need that level of precision to make reliable predictions.)
Now take a blood sample and find out if the guy's DNA has an A or a G at that site. Then calculate the "self-information", assuming he's in each of the three groups. (That's the logarithm of the reciprocal of the probability of the observation.) Then subtract the number from the group he's actually in from the other two numbers. So if his DNA has an A there, you get (log(1/0.35) bits - log(1/0.25)) bits, subtracting the Caucasoid number from the Mongolid number, and (log(1/0.05) bits - log(1/0.25)) bits, subtracting the Caucasoid number from the Negroid number. Conversely, if his DNA has a G there, you get (log(1/0.65) - log(1/0.75) bits and (log(1/0.95) - log(1/0.75) bits, respectively. Repeat this procedure over and over, on many randomly chosen genes, and add up the results. The prediction is that summing these formulas across lots of genes will give you positive numbers. This is because it takes more bits of information to describe how the guy's DNA is different from an average Mongoloid's or from an average Negroid's than the number of bits it takes to describe how his DNA is different from an average Caucasoid's.
So what are these formulas actually saying about the individual's genetic makeup? Instead of making a prediction about a single DNA base, what you're making is a prediction about a weighted average of what's in his DNA at many sites. In other words, this sort of prediction is rather like saying "Among these particular 100 sites, since this guy is a Caucasoid he probably has more than 32 C bases at these sites, a condition that's rare in a Mongoloid or Negroid.". But it's a more general class of prediction than that, since it lets you make a prediction about sums of As, Ts, Cs and Gs instead of sums of one type of base at a time, and since it doesn't require you to give every site equal weight toward the sum. A genetic trait like this is not, of course, a particularly interesting trait. But then interesting isn't what we were going for -- it's just an existence proof. Yes, membership of the group is a highly reliable predictor of the presence and absence of various traits outside of those used to identify the group in the first place. You can discover as many millions of such predictable genetic traits as you can generate sets of random numbers.
Whether self-identification matches biology is going to depend on all manner of cultural influences. In the U.S. context a guy who's 3/4 Caucasoid and 1/4 Negroid may well self-identify as African-American even though there's nothing in biology to support treating "African-American" as if it were a dominant gene. But that's an atypical case -- commercial genetic ancestry tests usually back up people's self-identification.Because some of the men on the street are so ignorant they think an Australian Aborigine is a "Negro"? That is not a good reason.
Depends on what you're using it for. It is if you're using self-identifcation as your measurement criteria, or if you're advocating public policy.
As for public policy, what does that have to do with the question? If enough people are discriminating against Jews for it to be a problem then public policy needs to deal with that, regardless of whether the Jews are a race.
Hey, English is full of nouns with "big" or "small" in their common-usage meanings, and there's rarely any hard-and-fast rule for how big something has to be to qualify. How big does a hill have to be before it's reasonable to call it a mountain? Such is life in an analog world with digital languages. The consequence of this for taxonomy is that biology is full of "splitter-joiner" arguments that are just disputes over words, not over facts. Don't be surprised if someday we decide the differences between Homo and Pan are too small to be "genus-level", so we rename chimpanzees "Homo troglodytes".4) These races are pretty much the conventional races, as the term was commonly understood, by educated people, before ideological race-denialism spread disinformation about the topic;
Well hang on there. You've just defined race through 1)-3) as being genetic clustering with no reference to size other than 'suitably large'. So how big is a 'race', and why is it that size? Tradition? Or is there more to it?
The point is, it's the tree structure of relations among subdivisions of H. sapiens that's scientific fact. Which of the levels in that tree we call "races", "sub-races" or "groups of related races" is just a matter of where people feel like drawing the line between hills and mountains -- there's no right answer to that type of question. In pre-postmodern anthropology, sometimes the American Indians were called a race and sometimes they were called a subgroup within the Mongoloid race. What difference does it make? A rose by any other name and all that. What's fact rather than mere terminology is that the American Indians form a natural division within the Northeast Asian branch of the human gene pool.