Talk:Race and intelligence/Archive 30

more about PF
Do we have any independent secondary sources not associated with the Pioneer Fund that advocate the pro-hereditarian stance? --JereKrischel 08:01, 29 August 2006 (UTC)
 * Charles Murray. --Rikurzhen 08:05, 29 August 2006 (UTC)


 * Nearly all the research that Murray and Herrnstein relied on for their central claims about race and IQ was funded by the Pioneer Fund, described by the London Sunday Telegraph (3/12/89) as a "neo-Nazi organization closely integrated with the far right in American politics." The fund's mission is to promote eugenics, a philosophy that maintains that "genetically unfit" individuals or races are a threat to society. - http://www.fair.org/index.php?page=1271. Sounds like a pretty close association - do you have others without that link? Mostly curious, of course - given the controversey in the field, I wouldn't blackball every last bit of research connected to the Pioneer Fund in the article, since otherwise we wouldn't have very much at all it seems.  --JereKrischel 08:17, 29 August 2006 (UTC)


 * PF has funded a great amount of research. The connection of PF to Murray is that he cites the work, as would anyone who is looking to summarize the field. You have no justification to "blackball" any PF researchers published scholarly work. Such a justification would need to rely on some aspect of policy described above (or elsewhere). In fact, no distinction is made in thes scholarly literature between PF funded and not-PF funded research.


 * This gets us back to the discussion about PF and "bias" from above. As per the long thread above, you have mischaracterized claims about the effects PF on research. The test is in how PF research is treated by the reseacher's peers -- not what outsiders and nonscientist think. The opinion is well summarized by Sternberg in the Skeptic magazine interview that was linked previous: PF doesn't matter when evaluating the science. --Rikurzhen 08:44, 29 August 2006 (UTC)

Cultural, genetic, or environmental?
Seeing as we've snagged again on a Jencks/Phillips quote, can I suggest renaming the section "cultural, genetic or environmental?" Jencks, and the concept of "labeling bias", denies the genetic as primary or significant, without resorting to saying it is "cultural"...although we mention the word environmental, perhaps we should have a new sub-section regarding environmental factors not related to culture? --JereKrischel 09:20, 29 August 2006 (UTC)
 * I think Explanations is the best headline. No reason to force a false dichotomy. Arbor 11:02, 29 August 2006 (UTC)


 * Explanations is fine, but the organization into two competing sets of hypotheses is probably necessary. Re: "labeling bias" -- as far as I can tell, this is a neologism that originated with and is only used by Jencks. It's not clear what the word refers to that would have been discussed by other researchers (i.e., it may only represent an opinion held by Jencks), so developing it into a major part of this summary style article is probably inappropriate (see section above). --Rikurzhen 17:04, 29 August 2006 (UTC)

issues
this is what I wrote to Arbor to summarize the current debate:

Gould:


 * Issue 1: Whether SJ Gould's MMoM is a reliable source for expert opinion about intelligence research.
 * Issue 2: Whether the existence of Gould's criticisms is noteworth and sufficient to require the demotion of research findings that would otherwise be treated as "fact" because they are otherwise generally accepted.


 * Conflict: whether Neisser's (1997) [rejoined for the APA report] confirmation that Rushton and Lynn are right about B-W-EA diffs in brain size is sufficient to establish this difference as fact. Notably, Neisser is adversarial to Rushton in the following paragraph.

Race:


 * Issue: Whether we can make a "necessary assumption" about race in order to describe research which assumes a working defintion.

--Rikurzhen 07:25, 29 August 2006 (UTC)


 * Issue 1 - I would argue that MMoM is just as criticizable as any Pioneer Fund research;


 * Issue 2 - Gould is not the sole source of criticism; presenting opinion as opinion is not "demotion"; I agree that narrow constructions are "generally accepted", but without proper context they get twisted into generalized "facts" that are patently false;


 * Issue 3 - there is no reason to try to establish this as "fact", simply report their findings;


 * Issue 4 - it depends on what your "necessary assumptions" are, and your working definition, otherwise, research being cited with one particular definition, combined with research being cited with a different definition, can be combined into a reasonable conclusion that is misleading and false.


 * --JereKrischel 08:22, 29 August 2006 (UTC)
 * Arg! I wrote a long reply and lost it. Here I go again, possibly much more concise:
 * Issue 1—Gould is clearly fringe. I don't even remember him making any pronouncements of the form "Experts agree that…", but even if he does, his work has been met with derision among psychologists. I have never seen any claim of the form "MMoM represents mainstream science"—quite the contrary, even form those who like the book. On the other hand, there is no published claim (by anybody) of the form "The WSJ report does not represent mainstream science", and only very narrow criticism of the APA report. So I think we are in the happy situation that it is very easy—even for amateurs—to get good and hard support for deciding that "mainstream science" accepts as fact. Gould certainly is fringe, and it's quite easy to verify that claim. Arbor 08:47, 29 August 2006 (UTC)
 * Issue 2—As per above, Gould about intelligence should be treated as a creationist about evolution. It's a noteworthy POV that needs to be reported, but it should not change our presentation of the scholarly debate. If there was a large group of psychologists (say, 20%) who agreed with Gould that would be a different matter. Sternberg is a good example of somebody who has a notable minority POV who whould be treated as serious. Gould, on the other hand, is fringe. Arbor 08:47, 29 August 2006 (UTC)
 * Issue 3—It's a clear fact. However, when presented in this context it needs to mention take Neisser's reservation. In fact, brain size needs to be presented since it appears in many a major surveys (including Gould's book!), so by not writing about brain size we are willfully neglecting an important issue, and we cannot do that. That being said, the current paragraph is not great. Arbor 08:47, 29 August 2006 (UTC)
 * Race assumption—That seems to be a textbook application of WP policy. I still don't like it and never have as I have said many times now. Maybe more about that below. Arbor 08:47, 29 August 2006 (UTC)


 * Arbor, do you have a for example case of Gould being "fringe"? I'd like to understand better whether or not it is specific claims he makes that are fringe, or if you're trying to assert that everything he has ever said about intelligence is discounted.  I think we may be crossing wires on that point - what is it that is being attributed to Gould that specifically is discounted as "fringe"?  I'm sure you won't be able to list everything, but one or two example would be appreciated! --JereKrischel 08:56, 29 August 2006 (UTC)


 * The Mismeasure of Man includes some examples. --Rikurzhen 08:59, 29 August 2006 (UTC)
 * Actually, I tried to move some of them to Wikiquote (which is linked from that page). The MMoM page currently is an unseemly collection of quotations pro et contra, and we should fight that. Arbor 09:09, 29 August 2006 (UTC)


 * Is there a specific example on The Mismeasure of Man page that you think relates to the R&I article? Most of the quotes seem fairly general and frankly, emotional in content...I don't think you're trying to say he's fringe because people have called him names.  I guess if you had something like, "he believes the world is flat", or "he believes the universe was created by a spaghetti monster", rather than "he misrepresented my research and is a terrible scientist and distorts things by using old data", I could better understand what it is about Gould you think is fringe.  Is it just his assertion that the genetic argument is always 100% genetic argument, and that he doesn't address the partial-genetic argument?  --JereKrischel 09:19, 29 August 2006 (UTC)
 * No. I am arguing he is fringe because Hans Eysenck calls the book "a paleontologist's distorted view of what psychologists think, untutored in even the most elementary facts of the science." You would need to find a different quote from someone respectable that says something along the lines of "Gould gives an honest and complete presentation of intelligence research". I don't think such a quote exists, even in spirit. Instead all praise for Gould praises his opinion. None of the praise is for the way he presents the mainstream opinion. That's a big difference. Symmetrically, all the attacks on APA and WSJ is because of the opinions expressed there, but nobody says they aren't honest and complete presentations of the mainstream. That's an easy way for amateurs like you and me to find out whom you can trust in the issue of representing the mainstream. Of course, you can still chose to side with either opinion. Arbor 09:38, 29 August 2006 (UTC)

Main Diagram needs more information.
According to the text the normal distribution is based on: "U.S. test subjects from 1981 (the most recent, large-scale, published adult IQ scores)"

I go to the reference and find this:

"The sample included 1, 880 adults stratified according to sex and age (equal numbers of males and females within nine groups), race, occupation, urban-rural residence, geographic region, and education. There were 1, 664 whites, 192 blacks, and 24 from other nonwhite groups."

Since when was a sample size of less than 24 a 'large-scale' test. How on earth does such a nice normal distribution get drawn for the Asian and Hispanic groups given that they have 24 individuals between them. The 192 blacks: how were they selected? What do other researchers say about how representational they are of the U.S. population?

192 is a small sample whatever way you cut it. 'Stratified by sex and age'. So 91 blacks of each sex, and about let's say, 20 for each decade of age/group. Surely, there must be other researchers who have commented about these tiny numbers.

This diagram is featured very prominently on the front, but there doesn't seem to be any indication that the numbers are so small. In fact where does the nice smooth normal distribution itself come from? Someone stick the figures into a calculator and extract the SD and mean and go 'hey presto, here's a normal curve'

I think a diagram of a point plot of all 192 results would be useful.

The WAIS-R test? What do researchers say about it?

"the causes and meaning of the different average scores for these groups are debated" I doubt there is much debate about the causes and meanings of the differences between the Hispanics and the Asians with a sample size of about 12 each divided in 9 groups.

It doesn't seem to make sense that this diagram represents this topic. Macgruder 11:04, 29 August 2006 (UTC)
 * The diagram is concordant with lots and lots and lots of studies, many of them linked in the footnotes. It is our best shot at a visually appealing, honest presentation of the fact that (1) there are differences in the averages and (2) the curves overlap significantly. here are some slides from Linda G's course on intelligence that use the same idea: use a picture to present most of the facts quickly. I think that's very good exposition; I'd wish more science articles were written that way. In the footnotes, and in the later "correct" figure in the text there is more data, including a warning about the Hispanic and Asian data from the 1981 study. Arbor 11:17, 29 August 2006 (UTC)


 * Then you need a section detailing how these figures were compiled, and a rewording of the diagram explanation. Any statistician will tell you that you can't just take a whole load of studies and add them together. Any number of the 'lots and lots and lots' of studies could be flawed, have varying methodologies of selection, suffer from selection bias, etc. What we need is the single (or more) published paper that took all these (or enough of these) studies to produce that diagram with the appropriate footnote of course. After all any test that 'showed' 60% or 70% of the Australian Aboriginal population to be mentally retarded would seem to have a flaw or two I would guess. I have no idea from reading the article the methodology of selecting the 'black' sample, say. Macgruder 16:53, 30 August 2006 (UTC)
 * Nevertheless, that's what the author of the world-wide IQ studies compilation did: he took a whole lot of studies and added them together, with the assumption that if there were enough studies, biases and errors would eventually cancel each other out. And yes, several of us here dissent with this line of reasoning.--Ramdrake 18:01, 30 August 2006 (UTC)
 * You can pick your favorite review article that includes averages for all 4 groups and draw a set of curves. It will look like this set. --Rikurzhen 18:13, 30 August 2006 (UTC)
 * I'm not concerned with just averages. Standard deviations are important. Particularly sampling errors for small populations. It's important to know exactly how this figure is arrived at, and how the populations are selected. There must be dissenting opinions about how these figures were compiled. I doubt it would look like this set even if I did get the 4 averages, because you can't draw curves from averages alone - only spikes. If there are opinions that the diagram itself is totally flawed because of these issues it needs to be alluded to. As Ramdrake points out the author of a world-wide IQ studies compilation is only one study, and there is dissension to the line of reasoning that biases would count each other out (quite how anyone would think that would be interesting to know as it seems ludicrous to me - for example if many of the Asians in these studies from overseas are Japanese/Hong Kong/Singapore then the bias towards Western style education is simply compounded rather than cancelled out.) . Diagrams - especially ones that are used as 'the face' of the debate take on an element of factuality external to their footnotes which incidently don't appear many places where the diagram does. As an aside, I don't need to look at any studies to know that the diagram is flawed. The diagram has a curve for Hispanics - a group that there is no definition for. If I saw an economic diagram that had the GDP of Atlantis, I would know that is flawed too. Macgruder 09:31, 31 August 2006 (UTC)
 * And, finally, not only is the information correct (in the sense that it is concordant with all major studies), the way to present these results in this form is not our own idea. (See the linked teaching materials.) Non-original presentation of accepted data. Informative. Accessible. And purposefully picked to deny the false implication that "All Blacks are stupider than Whites". The only way I can see that Macgruder has a point is that the caption currently tries to give too many references. What this figure needs is not more information—it needs less information in the sense that it should make clear that this is a rough sketch. In the body text, and in the caption of a later figure, we need the proper (non-normalised) curves and a reference to hard and fast data—with variance caveats and whatnot. And let my (as usual) take this place to voice my displeasure with including the Hispanic curve, which is opaque to us non-USers and constantly opens the article to "But Hispanics is not a race, it's an ethnic group!" sermons. I think this article should concentrate on the B-W-A gap, just like the Rusthon–Jensen survey. Arbor 18:48, 30 August 2006 (UTC)


 * I see your point about the "Hispanic" curve. I think now I would prefer if it said "Latino", since that's a more common self-referrential term. However, the "But Hispanics is not a race, it's an ethnic group!" sermons could be useful if it could get the point across that "race" is not a fixed at some definite number, but rather a complex concept. I recognize that it probably doesn't get this point across. The higher average of Hispanic/Latino populations compared to Black populations in the U.S. is a somewhat important point. I grant it is not necessarily top priority. Of course, anyone is welcome to draft up better looking lead images for consideration. I've tried and failed several times to make any real improvement. --Rikurzhen 20:55, 30 August 2006 (UTC)


 * Why is the higher average of a mixed ethnic group like Hispanics an important point? What about Hispanic blacks?  How do they get counted in the graph? Or any mixed "race" person, for that matter?  I think that there is certainly some simple, important context that should be included: 1) averages are not controlled for other factors, 2) categories are by self-identification, and gloss over any mixed "race" people.  We should at the very least include that kind of information in our list of assumptions (that is to say, this work is predicated on the concept of ignoring mixed "race"). --JereKrischel 21:55, 30 August 2006 (UTC)


 * Re: annotating the graph -- Aren't these things self-evident? Re: mixed race -- There's no assumption of ignoring mixed race. It's part of many experimental designs (see for example the Minnesota Transracial Adoption Study, but not something that get's top billing when describing averages because most people self identify as beloning to only a single race. --Rikurzhen 22:05, 30 August 2006 (UTC)


 * BTW --- Arbor's main point is correct. The less detail and pretension to absolute accuracy, the better. The conclusion that there are sig. diffs. in the average IQ of groups is without doubt, as is the finding of considerably overlap. This is all that needs to be conveyed in an initial figure. Anyone who wants to dig deeper should read more than the lead block. --Rikurzhen 21:08, 30 August 2006 (UTC)


 * I think it might be more useful to have data that was controlled for socio-economic status and environmental factors - as it is, I think the existing graph exaggerates the differences between the averages. Perhaps a note regarding the lack of control for other variables would help? --JereKrischel 21:55, 30 August 2006 (UTC)


 * The averages are what they are, be it due to environment, genetics, or divine intervention. The SES controlling graph, which is only available for Blacks and Whites, is found later in the article. --Rikurzhen 22:05, 30 August 2006 (UTC)


 * All very important context, of course. As it stands, I think the graph strongly implies the difference is due to genetics.  Maybe there is some way to make the context clearer? --JereKrischel 22:33, 30 August 2006 (UTC)


 * [i assume you are still talking about the lead figure] I'm not sure how you would arrive at that conclusion if you read either the caption or the lead. --Rikurzhen 05:46, 31 August 2006 (UTC)


 * Supplementing what Arbor wrote... The Black and White IQ distributions are very well studied, and this drawing is fairly accurate by showing a 1 SD gap. The Black IQ SD is less than 15, but it would be impractical and unimportant to show that difference here. The Hispanic mean IQ is somewhere between the Black and White mean IQs, but probably closer to the Black mean IQ. The Asian mean IQ is somewhere above the White mean IQ, but nationally representative samples don't include enough people of Asian ancestry to get a precise number. The largest sample I know, from the NLSY-79, gives a mean of 106 (n=42), but because there is a verbal/non-verbal skew in the IQ scores of Asians compared to Whites, the true difference in g may be masked. --Rikurzhen 16:58, 29 August 2006 (UTC)


 * So at least two of the curves (Hispanic and Asian) are not precise? Sounds sketchy to include these rough guesses if they aren't as well studied as the BW gap...maybe we could make the lines fuzzy to illustrate the uncertainty in the graph? --JereKrischel 21:59, 30 August 2006 (UTC)


 * Few things in psychology are as well studied as the BW IQ gap (e.g. Roth 2001 had n = 6 million). It's all relative. You should mock up figures for whatever ideas you have. --Rikurzhen 22:05, 30 August 2006 (UTC)
 * Yes, but this study is a clear example of the bias that runs through the article.
 * Article: "Roth et al. 2001 found that the recent U.S. Black-White gap in g is 1.1 sd, similar to characterization of the historical U.S. Black-White gap."
 * Study itself: "However the 1 standard deviation summary of group differences fails to capture many of the complexities in estimating ethnic group differences in employment settings. For example, the results indicate that job complexity, the use of within job vs across job study design, focus on applicant vs incumbent samples, and the exact construct of interest are important moderators of standardized group differences. In many instances, standardized group differences are less than 1 standard deviation."
 * It's precisely the omission of these kinds of pertinent issues where the authors of the papers themselves note the problems of the measurements that leads myself and others to conclude that this article is inherently biased. Did the author of the above line in the Wikipedia (a)not notice Roth's major qualification of the results, (b)notice it but decide to leave it out? I'd certainly like to know, and how many other studies referenced in this article are glossed over in such a way.Macgruder 10:42, 31 August 2006 (UTC)


 * The sentence is an argument made in the on-going Flynn, R&J, Murray debate. I forget which one made it in that fashion, but I recall all three cite Roth and agree that the conclusion of that paper is a 1.1 SD gap and that the historical BW gap is around 1.1 SD. --- The reason why it's not important, however, is that the gap is measured in terms of g, which is extracted away from the considerations that Roth is mentioning there. It is the gap in g that concerns Flynn et al. --Rikurzhen 17:03, 31 August 2006 (UTC)


 * I was noticing also that you've made all curves of equal height, which seems like misleading information - shouldn't they be of the proper height in proportion to the population size? Maybe that would help - does the information you used to build the graph as is include population size?  Or did they not do proportional testing? (that is to say, if one group was twice the size of another, make sure they tested twice as many people for that group to get a representative sample) --JereKrischel 22:30, 30 August 2006 (UTC)


 * If the curves were proportional to the U.S. population, the Black, Hispanic, and Asian curves would be small bumps under a tall White curve. This would be informative for understanding the stucture of IQ in the U.S. population, but the point about overlapping distributions would be lost. [p.s. pending discussion on Jencks above] --Rikurzhen 05:46, 31 August 2006 (UTC)

To restate the point. The footer of the diagram clearly states that it is the result of one particular study (even after Arbor's update). For the sake of clarity it is a 'best shot at a visually appealing, honest presentation of the fact...'. But that study has no facts for hispanics or Asians unless statistics has changed such that you can make a meaningful statement about 2 billion people with a sample size of 12. Yes, it is true that you display a fact with a clear diagram, but we are not talking about a fact here.

That diagram shows the following: 'Blacks have lower IQs'. (IQ on one axis, label on another) This is not a fact. It should have: 'US blacks score lower on IQ tests'. These are 2 very different things. We could say the second is a fact if we accept it is possible to say someone is 'black' or 'white'. Macgruder 10:40, 31 August 2006 (UTC)
 * Macgruder, you are constantly conflating your own opinion about the validity of this research with your helpful comments about how this should be presented. Maybe it would be more constructive if you separated the two. The possibility of self-identifying as Black or White (while disgusting to me as a non-American) is beyond doubt. I am told US citizens to this all the time, and these data are routinely stored. That's why we have so much US data about different statistical trivia (income, SAT scores, Army admission, crime, etc.) by race, and very little from other countries. (For example, in Sweden, where I now live, it is forbidden to store anybody's race/ethnicity/whatever. "School performance by biogeographic ancestry" in Sweden is simply not available. Politically, I think that is how it should be. But that does not make the US data go away.)  Arbor 11:08, 31 August 2006 (UTC)
 * Just because Americans do so, doesn't mean that it has a meaningful correlation with the worldwide 'race' of blacks. Anyway, my main point was 'IQ' versus 'IQ tests'. 'IQ' is an absolute, 'IQ test' is an observed result open to interpretation and sampling errors. There are facts about IQ tests scores which we can state, but not directly about IQs. For example:
 * Xs have low IQ scores. Researches has discovered this is because of errors in the way the samples were chosen.
 * Xs have low IQs. [We cannot allude to the fact that perhaps this is due to sampling errors]
 * Regarding the validity of the research. It's hard to tell whether what this article says refers to valid research, because of issues as I pointed out above of mistatement or omission. (See my comment above about Roth's research). This is why I don't trust this article. Not because I don't respect the research but when I do take the trouble to look at the researchers, what they say and what is written here turns out to be different. My honest feeling is that their is a bias in this article and whenever someone tries to balance it they are overwhelmed by one or two editors and after a week or two of trying they just give up and go away.


 * In the langauge of Jensen (1998) "IQ" is a measurement, "IQ test" is a vehicle to produce the measurement, and "g" is the construct being measured. Per this and the above, your concerns appear to stem from a partial misunderstanding of the research you are reading. It's best to start with review articles, rather than jumping into the data without context. --Rikurzhen 18:57, 31 August 2006 (UTC)
 * And the retort to that, would be your concerns appear to stem from a misunderstanding of statistics. 'g' might be the construct being measured but with seemlingly small sample sizes combined with test results being simply lumped together, you cannot go from the 'test results show this' to 'the truth is this'.

You can read the cummulative IQ distributions for the four groups from page 43 of Gottfredson (2004)PDF --Rikurzhen 17:46, 31 August 2006 (UTC)

3 year olds and fallacy
"Responding to such concerns, Dickens and Flynn 2001 have proposed a solution which rests on genotype-environment correlation, hypothesizing that small initial differences in environment cause feedback effects which magnify into large IQ differences.[73] Such differences would need to develop before age 3, when the Black-White IQ gap can be first detected."

1) I think the 3-year olds IQ difference (reading the Flynn article) is a controversial result.

2) you cannot use a controversial result to make factual statement like : "Such differences would need to develop before age 3, when the Black-White IQ gap can be first detected."

3) where is the citation for this statement: 'Such differences would need to develop before age 3, when the Black-White IQ gap can be first detected.'

The 3-years olds issue feels to me like 'weasel-words'. How many large scale studies have been made on 3-year olds? Taking controversial studies on 3-year olds and then applying their results as truth to other parts of the article. Macgruder 11:33, 31 August 2006 (UTC)
 * I did not know that the 3-year-old results were debated. Interesting. That being said, the sentence you point to indeed needs a source, otherwise that line of argument is indeed a clear example of Original Research (independently of whether Peoples et al. or Dickens and Flynn are right or not.) Arbor 12:17, 31 August 2006 (UTC)
 * Found it! Rushton–Jensen make exactly this point in their PPP survey (p. 270). Peer-reviewed and all. Should be kosher for WP. Arbor 12:22, 31 August 2006 (UTC)