Thursday, December 22, 2011

Asian admissions, statistical prediction, and all that

This post addresses comments by Sineruse, David Versace, and others on earlier threads here, here and here.

I've been pretty busy during my visit to BGI so I kind of lost track of the conversation. But here are my thoughts.

I think it is possible that Asian "strength of application" overpredicts later career success*. There could be many reasons for this. For example, it could be that Asian hard work boosts test taking results and grades more than it does real world achievement. It could also be that tests and grades are fair, whereas Asians face a certain amount of race-related disadvantage later in their careers -- e.g., unconscious bias, lack of "ethnic affinity" networks, etc. (If you talk to highly trained Chinese scientists and businesspeople returning to China from the US, most will describe an uphill struggle for Asians in the US; this contrasts with glib statements from white Americans about how little anti-Asian bias there is in elite careers). Finally, Asians may have lower rates of sociopathy, which reduces their chances of making it to the top (close inspection suggests it is mostly sociopaths at the top ;-)

I don't think the evidence is overwhelming on this question. But if, say, Asians have a .5 SD advantage in g and 1 SD advantage in conscientiousness or work ethic, that might lead to a "fair" Ivy population representation which is less than 20% if by "fair" we mean: apportion slots based on future success odds. (An additional factor which is usually mistakenly ignored in numbers like +.5/+1 is the large offshore "reservoir" of Asians and the fact that some A-As are drawn from a very elite subgroup in their ancestral countries.) Note though, as emphasized by RKU, the current Ivy standards for what constitutes "success" may not be aligned with the real interests of the Nation. That is, the connection between money and power and actual value creation in our current system seems to have become quite weak of late. What is good for Harvard may not necessarily be what is good for the USA.

Having said all this, I think you are missing a key point -- perhaps because you think mainly in terms of (white) ethnic interests. Even if elite universities are acting in their narrow self-interest in assessing an outright Asian penalty to compensate for inflation of application strength, their methodology may violate the law. While some Asian application profiles overestimate later career prospects, universities should not be allowed to make generalizations based solely on race. This may not be a principle that you believe in, but it's an important one to me. If universities have some other way to correct for false signals in admissions profiles ("this kid scored high, but we know he's just a grind"), then fine. However, I suspect what is going on now is crudely (if perhaps subconsciously) race-based**.

* Something to keep in mind is that Harvard et al. would like to have influence abroad as well as at home, and Chinese ethnicity alumni are well placed to influence what will soon be the largest economy on the planet. Underperformance vs predictor in the US may be compensated by overperformance in the new reality of the coming century. Ask yourself why BGI was more willing to work with me than, say, the Sanger or Broad institutes might have been.

** An astute commenter asks why we should oppose race-based decision making, if there is real correlational information to be had from ethnicity. I offer two reasons: 1. this country has a bad record on race, and striving towards a race-blind society is worth some small sacrifices, 2. the evidence for genetic group differences is not conclusive and should be treated with great caution.


yulva said...

A little off topic but I wanted to share this article regarding the state of education in America:

Cal State campuses overwhelmed by remedial needs.

“Wracked with frustration over the state’s legions of unprepared high
school graduates, the California State University system next summer
will force freshmen with remedial needs to brush up on math or English
before arriving on campus. But many professors at the 23-campus
university, which has spent the past 13 years dismissing students who
fail remedial classes, doubt the Early Start program will do much to
help students unable to handle college math or English. . . . The
remedial numbers are staggering, given that the Cal State system admits
only freshmen who graduated in the top one-third of their high-school
class. About 27,300 freshmen in the 2010 entering class of about 42,700
needed remedial work in math, English or both.”

rogic said...

Take note, everyone: Obama voter Steve Hsu is a principled advocate of race-blind college admissions, rather than an Asian ethnic activist. If you couldn't tell by the number of blog posts and newspaper interviews he's devoted to arguing against affirmative action for so-called underrepresented minorities rather than merely in favor of increased representation for Asians, he will tell you so himself. 

tractal said...

That is all fine and well said, but all you have given us here is a moral axiom that Institutions ought not to generalize based solely on race. I can anticipate a lot of arguments against that principle, so I wonder if you could defend it later when you have time. 

Yan Shen said...

"But if, say, Asians have a .5 SD advantage in g..."

Steve, you're really stirring up the hornets nest tonight. Soon you'll have 20 angry white guys from iSteve, Mangans, Stormfront, etc telling us how that can't possibly be the case. You'll have also have David Versace adamantly insisting that white guys have a .5 SD advantage in alphaness and get laid far more often. ;)

Christopher Chang said...

Note his statement about violation of the law.

He explicitly leaves open the possibility that the law should be changed.  But in the present, he is correct that this practice may be illegal.

qc32p491993 said...

"If you talk to highly trained Chinese scientists and businesspeople returning to China from the US, most will describe an uphill struggle for Asians in the US"

Yeah...but if you talk to women / African-Americans / Hispanics / White, educated, middle-class men most will describe an uphill struggle for women / African-Americans / Hispanics / White, educated, middle-class men in the US.

Speaking of which: "Ask yourself why BGI was more willing to work with me than, say, the Sanger or Broad institutes might have been."

Are you suggesting you were the victim of ethnic discrimination, or just that a Chinese institute was readily drawn to you because of your ethnicity? Because if it's the former, a more parsimonious explanation would be that your research breaches certain taboos among educated, liberal whites. If it's the latter, the counterfactual claim about the Sanger and Broad institutes does no work.

Lastly: "their methodology may violate the law"
Equivalently, their methodology may not violate the law. Someone should probably look that up, I guess.

Your claim that "universities should not be allowed to make generalizations based solely on race" is, it seems to me, idiosyncratic. They already do, with good reason, and with legal justification: That's what diversity-based affirmative action in effect is.

Guy_Brodude said...

If you'll forgive a tautological, rather moronic-sounding comment: Harvard is Harvard. Harvard could switch to a purely meritocratic admissions system and be no worse for it. This is true, to a lesser extent, of Yale, Princeton and Stanford as well. They are the premiere universities of the most powerful, wealthiest nation on Earth, and they will always have a formidable cachet and brand name based on that.

The real problem is that these schools (HYP in particular) are absurdly small given their wealth, prestige and the strength of their pools of potential students. If they transitioned to a strictly meritocratic model, gradually increased size by a factor of 2-4 over the course of a decade, say, maybe increased the number of foreign students to compensate for the strain on financial aid, there would be no problems; the schools would, if anything, be even more diverse than they are now, and the bell curve, while not necessarily shifted, would certainly be compressed.

This will never happen, of course, but it's fun to speculate about.

tractal said...

An analysis of Verbal and Spacial IQ predictive power is badly needed. Certainly Mr. Rose is right in saying Asians do not have a simple G advantage over whites, but their spacial advantage is much greater than the very small and possibly non existent white verbal advantage. But there is really no data saying anything either way about verbal vs spacial and the greater white "breakthrough" potential some commentators suggest.

It is possible that the small verbal advantage or some other cultural elements underlie the "outstanding achievement" differential. But that hypothesis has to deal with many potential objections. First the "outstanding achievement" differential may not even exist. Calculating it fairly would be really hard, especially if you want to include humanities 'achievements' which may be victories more of style than substance (especially recently). And even if you did it, there is still the strong possibility that future cohorts will change the pattern. You need to account for the time gap between achievement and the Fields/Nobel etc.  

And even if you did that and found Asians lagging, you still have to confront all the environmental or cultural explanations. The only way to confirm the theory would be to draw some line between verbal IQ and breakthrough achievement and then quantify the Asian verbal gap and see how well it matches. Otherwise it is all speculative in my view. 

Kevin Rose said...

What a transparent attempt to shut down the debate by bullying people into accepting interpretations without questioning them. 

Hmmm, you know it occurs to me that when people have to resort to bullying/shaming tactics to force others to accept interpretations they are pretty insecure about those interpretations. Just saying.....

Get over it, dude. You cannot bully people into accepting interpretations that are clearly flawed, however much you want to. And why you would want to to begin with says volumes about your ability to handle objective truth, Yan Shen. But you are typically oblivious to what your comments reveal about you, lol. It is quite amusing. 


Yan Shen said...

Unfortunately, you commit the typical white nationalist fallacy. There are three major cognitive categories, math, verbal and spatial and not merely two, verbal and spatial. Relative to Europeans, East Asians are significantly higher in math and spatial and lag behind slightly in verbal.

Consider the following. The cognitive profile of men relative to women is fairly similar to that of East Asians relative to Europeans. See here.

"Males tend to score higher on visual-spatial and (starting in
middle-childhood) mathematical skills, while females tend to achieve
higher scores on several verbal scales."

Yan Shen said...

"One wonders, though, how much of that is sour grapes as arrogant
Chinese scientists discover to their dismay that their self-conception
is largely illusory as they bump up against actual white performance. I
have experienced this phenomenon first hand - Chinese people, especially outside the US, have a self-conception that is laughably out of sync with reality, and when it clashes with reality, it aint pretty."

This seems to contradict the study reference by Lynn here.

"This difference is also picked up in the United States in
performance on the Scholastic Aptitude Test (SAT), on which ethnic
Orientals invariably do better than Caucasians on the mathematics test
(largely a measure of general intelligence and visuospatial ability) but
less well than Caucasians on the verbal test (Wainer, 1988). A further
manifestation of the strong visuospatial and weak verbal abilities of
ethnic O1-iental Americans lies in their tendency to do well in
professions like science, architecture and engineering which call for
strong visuospatial abilities and poorly in law which calls for strong
verbal abilities. This pattern of occupational achievement has been well
documented by Weyl (1969, 1989) in his studies of the achievements of
the major American ethnic populations. His method involves the analysis
of the frequencies of ethnic names among those who have achieved
occupational distinction calculated in relation to their frequencies in
the general population. Thus he finds that common Chinese names like
Wong are greatly overrepresented in American Men and Women of Science,
as compared with their frequency in the general population, but under
represented in Who's Who in American Law. On the basis of this method he
constructs a performance co-efficient for which average achievement is
100. A co-efficient of 200 means that an ethnic group appears twice as
frequently in reference works of occupational distinction as would be
expected from its numbers in the total population, while a co-efficient
of 50 means that it appears half as often. In his first study he finds
that ethnic Chinese obtained performance co-efficients of 506 in
architecture, 308 in engineering and 438 in science but only 54 in law
(Weyl, 1969). His second study oil later data confirms this pattern for
the 1980s, when ethnic Chinese obtained a performance co-efficient for
science of 620, while for law their performance co-efficient was only

ben_g said...

I think there are relevant personality differences beyond sociopathy.  It seems like there are also differences in big picture creative thinking, verbal/social skills, competitiveness/agression, risk taking, and individualism.

ben_g said...

This isn't a moronic comment at all.  We can hope that the post-secondary education industry will become disrupted by technology and other forces, and the top tier (along with other sectors) will eventually become more meritocratic and able to serve the demand.

darklayersify said...

FWIW, Lubinski et. al 2006 "Tracking Exceptional Human Capital" found that children of immigrants were more likely to earn a six figure salary, become a tenure track prof, or have an MD than other highly gifted (top .01% of math or verbal ability on SAT at 12 or 13) individuals in their study and graduates from top 15 STEM doctorate programs. 

One particular trait of interest to overpredicting 'career' success is self-control. Some researchers have found that self-discipline is highly predictive of success in school. Other factors might swamp it later in life, and those might be relevant to educational strength versus later achievements. 

A second thought is that women perform a lot better in applications, college, etc. than in top positions through many areas of work. There are a lot of speculations about this situation, but one vital factor is life 'balance.' Benbow et. al 2000 found women performed about as well in education, despite their weaker abilities at the right tail. However, when the authors controlled for hours worked, income advantages from the males disappeared. Also, women have higher self-discipline, and that's highly predictive of coursework performance. Would anyone seriously argue that elite schools should admit fewer women because they are known to be less able and willing to work 60+ hours a week?   

tractal said...

I don't think this answers the question about V and breakthrough achievement. The question is just whether V partially underlies those achievements or not. Positing Terrence Tao again and again as a proof that high M is all you need isn't much of an argument. I could find just as many anecdotes on the other side I'm sure. 

It would be better to look at what data we have. All I have ever really seen on this was that study Professor Hsu mentioned where exceptional scientists and other academics were measured. If I recall, the verbal score for each discipline was very high. For theoretical physicists it was astronomical. 

Your argument that whites are to asians as women are to men also fails to show anything about V and breakthrough achievement. The argument I guess is supposed to say "look men do better than women at breakthrough achievement so M wins", but this ignores the narrower female g distribution. Even if the extreme"verbal is everything" hypothesis were true, men would still outperform women at the upper end. 

Also, the most parsimonious explanation of East Asian cognitive profile 1) is that something selected for visuospacial, and 2) visuospatial skill lends itself to math ability. The evidence for 2) seems decent, given a brief google search. 

Visuospatial Abilities and Mathematics: It has been suggested that certain visuospatial sex differencescontribute to observed sex differences in some aspectsof mathematical performance. For example, Casey, Nuttall,Pezaris, and Benbow (1995) found that the sex difference onthe Mathematics portion of the SAT (SAT-M) was eliminated inseveral samples when the effects of mental rotation ability werestatistically removed. This suggests that rotational skill maymediate the sex differences in certain high-level mathematicalabilities or, at the least, that these two abilities tend to covary

ben_g said...

Men and women probably differ for reasons unrelated to intelligence-- interests, competitiveness, etc.

Yan Shen said...

" All I have ever really seen on this was that study Professor Hsu
mentioned where exceptional scientists and other academics were
measured. If I recall, the verbal score for each discipline was very
high. For theoretical physicists it was astronomical. "

That study included social scientists and those in soft science fields. Also, the physicists were precluded from taking the mathematics exam, because it was deemed too easy for them. (You left out the observation by a previous commenter that the physicists were also very high in spatial ability.) Regardless, overall M scores were also fairly high. Furthermore, as commenter Ju Hyung Ahn has suggested, that sample could've been disproportionately Jewish, a group who are both high V and high M. The second point I make, highlighting that Jews and East Asians seem to be similarly over-represented on the IMO and amongst Intel Science Talent Search semi-finalists suggests that it may be high M that is the crucial element for success in mathematics and the hard sciences.

Let me posit the following. The people who are highest in verbal ability are not in fact pure mathematicians and theoretical physicists, but rather philosophers. While it is often stated that physicists on average are higher V than those in the social sciences, it seems to me that the comparison may be skewed by the fact that there is a much lower cognitive threshold the humanities and social sciences and therefore a much wider range of IQs in those fields as well, compared to say professional physicists. I would not be surprised if eminent individuals in other social science/humanities fields outside of philosophy were also high in verbal ability than theoretical physicists/mathematicians.

tractal said...

I put forth earlier that high M decent V is the cognitive profile of those who succeed in the mathematics and hard sciences. This, I've argued, is essentially the East Asian cognitive profile.
But Shen that isn't the important question here. No one doubts that the high M profile is a really good predictor of success in math and the hard sciences overall. The question is just whether outstanding original achievement in these fields is aided by very high V. Representation in high school math competitions is not the kind of evidence we need to evaluate that claim. 
The putative importance of V is totally compatible with philosophy being the most "V" loaded discipline. Philosophers are really smart. You seem to be attempting to make the point that "well if V is so important why aren't the high V guys in theoretical physics." But that line of argument confuses necessary with sufficient. High V could be necessary but not sufficient for success in theoretical physics or whatever. 
For instance, outstanding achievement in science might require or occur reliably at say 150 M and 165 V. Mostly I'm making those numbers up, but the point is that we cannot know a priori that they are false. There are many disciplines, and each probably has a different threshold. The point is that it is not necessarily a simple "G" threshold. And incidentally the mean verbal score in Roe's sample was 166. Remove the experimental physicists and it is probably over 170. That strongly suggests high verbal is important to this kind of achievement. Unfortunately we don't really have the same data for M because the M test was apparently too easy for the physicists. All we know is that the M mean was 154. 
In that context +4 verbal or whatever could be a greater asset in reaching the threshold than +10 spacial +6 math. I'm not saying I know that this is the case, I am saying that it could be the case and we need to know more. 


tractal said...

Sorry for making this so drawn out. I just want to make the point that it is possible that the simple G model of cognitive ability where you can stack rank G and determine future performance (ala the SAT model) may not actually predict elite elite achievement very well. 

Not that this matters very much for the argument about Asian admissions. Unless schools are really trying to maximize their Nobel laureates per seat. I'm sure they are in some sense, but that cannot be an important factor in too many admissions. More likely Asian SAT dock occurs for some combination of Asian prep renorming+not wanting their institutions to be too unrepresentative of the country. 

Yan Shen said...

"Dude, we are talking about top achievers, not mere representation."

Please learn to read more carefully Kevin.

" His method involves the analysis of the frequencies of ethnic names among those who have achieved OCCUPATIONAL DISTINCTION..."

Ju Hyung Ahn said...

I don't know where you get the idea that East Asians aren't innovators today.

List of Countries by Patents

I guess Europeans can sell their overpriced goods such as Chanel, Giorgio Armani, and Mercedes-Benz for now though.

tractal said...

Troll harder. Quantity of patents per se is almost meaningless. 

darklayersify said...

        SET tried to conduct studies on high ability populations--they had difficulty because of sample size issues. 

      Also, I wrote to Jonathan Wai, presently at Duke, about studies on genetic and hormonal factors that could contribute sex-typical cognitive abilities at the right tail. Few studies of high ability populations exist on hormones and genes, according to the review of Steve Ceci, Wendy Williams, and Susan Barnett in 2009. Wai mentioned Duke TIP working on a study of the genetics of high cognitive abilities, but results at that point (summer 2010) were inconclusive. Several American researchers contributed to a twin study of cognitive abilities at the 85%. I'm sure IP readers would laugh, but they chose that cut-off for sampling reasons.  

       Turkheimer & Halpern (2010) noted that no genes reliably linked to higher cognitive ability have been discovered, "despite extensive and well-documented efforts to find them."

        The TEDS in the UK is working hard on finding genes that contribute to variation in cognitive abilities.
        So, it might be a taboo at Broad and Sanger or a broad taboo that's manifested at Broad and Sanger. Western researchers have pursued the question, still. 

Ju Hyung Ahn said...

Yes, the same meaningless intellectual property WIPO bothered to recognize.  It must be just a coincidence that many of the top countries in the table happens to be global leaders in many industries.
Perhaps adding "i" to start off any product name is the type of ingenuity you're looking for.

tractal said...

Calm down bro. Patent numbers are just not a good proxy for intellectual achievement. No demarcation is made for patent importance, and corporate culture and legal systems probably have a lot to do with patent incentives. If national IQ was the primary cause of patent numbers, we would expect to see European nations patent share in rough proportion to their population relative to each other. We really don't see that. Its just not very good evidence. 

Ju Hyung Ahn said...

"Yet their technological prowess goes well beyond what these figures
suggest, so clearly the top layer of talent is somehow in some way, actually more talented than the top layer in countries with a higher average IQ (like, say, South Korea)."

I was responding to the statement above.  Again, dogmatic bloggers can throw out speculative statements without any relevant data, while any attempt to counter it is met with an outcry of "troll".  Perhaps you want to corroborate on this technological prowess for him then?  Notice East Asian nations (Japan, South Korea, and Taiwan) are pretty well represented in IT areas.  You are resorting to negative tactics without even presenting a barometer that says otherwise here; I find it very childish.

tractal said...

I mean come on. It would be like saying "Americans are more innovative than the Japanese" and then citing patent statistics. Of course the instant response would be "of course America has a lot of patents it is the economic center of innovation, troll." Which would be totally correct, but that is the kind of reason why simple patent numbers are not very informative.

Kevin Rose said...

Thanks, bud. Now let me know when you understand the difference between *occupational distinction* and *top achievers*. Thanks, champ. 

Heck, a mid level manager could be considered an example of *occupational distinction* depending on how you define terminology and what methodology you use. 

Ju Hyung Ahn said...

I was simply correcting his misguided view that Asians aren't there innovating.  He went on to juxtapose the top layers of East Asians and white people and somehow drew up the conclusion white people must be superior.  Perhaps my sarcastic comment about Europeans made it seem like I was insinuating East Asians are superior.  It wasn't my intention.  My intention was to ridicule the likes of Kevin Rose, who makes it seem like his gut feeling retort has any substance.

Matthew Carnegie said...

Richard Feyman is another high profile example of someone who was highly skewed towards M in favor of V. be honest, this does make me wonder about whether IQ tests really capture verbal ability very well.I mean, I've read Feynman's book and that's not low V communication. It doesn't really seem like decent V communication. It seems like high V communication. He isn't inarticulate. His ideas are clearly expressed and his arguments are beautifully structured. I mean "Feynman was a keen popularizer of physics through both books and lectures". That's "low V"? RKU's actually argued with Feynman, so I'd love to see his take on whether Feynman was low or decent V.(I didn't really get the sense of whether he was a visuo-spatial or verbal thinker from his writing: on the one hand, his visual analogies were lucid and clear, on the other he conflates mathematical syntax to verbal syntax and writes approvingly of the "modern approach" of just writing a lot of "little symbols" in sequential order (which is basically language like, although different).)On the other hand, if the IQ test does have a deficit in capturing true verbal ability, then that's a defecit that it is most parsimonious to assume hits everyone equally as it stands, and isn't a reason to assume boys are secretly as clever as girls (as much as the near equal representation of males at talky stuff would make us assume so) at V or Asians are secretly cleverer than Whites on V or Whites are secretly cleverer than Asians on V or whatever.

Matthew Carnegie said...


I found this during the last discussion we had about g and v/s in Asians

"We have a paper coming out (Fugard, Stewart, & Stenning, to appear) relating performance on the visuospatial Raven’s Advanced Progressive Matrices items, as determined by DeShon, Shah, and Weissbein (1995), to position on the sub-clinical autism spectrum, measured by the Autism-Spectrum Quotient.  The more you report having autistic traits, the better you are at the visuospatial items.  This result fits nicely with the enhanced perceptual processing theory of autism. It also provides more evidence that Raven’s matrices load highly on g because the test is a package of many kinds of intelligence test."

It would be cool to test Asians and Whites and see if the Asian advantage on Raven's is driven mainly or only by the spatial items. I'd bet it would be.

The Rushton link about a general supremacy in g-related items in Asians is interesting. There does seem to be only one cite for that, "Group differences and principal-component loadings in the Hawaii Family Study of Cognition: A test of the generality of “Spearman’s hypothesis”. I can't access to see what tests they used. If there was a spatial element in most test, then it might be like Raven's, driven by that effect.


I wouldn't worry too much about whether it was 0.5 stdv or 0.7 stdv or 0.3 stdv. They're all in the same neighbourhood and the exact level doesn't seem to add much prediction to country or ethnic group performance. When I last looked into this, spatial advantage (which Asians have more of) seems to drive math advantage really strongly (so not "limited significance). Steve is probably estimating slightly upward because he thinks that mathmatical advantages matter more in real world terms. I don't think it's meant unfairly or in an ethnic biased way - he'd probably state Mongolians are smarter than Chinese due to their spatial advantage as well (which would presumably drive a math advantage).

Interestingly, judging from the PISA data, the US and the UK tend to have about twice the relative advantage for boys at maths the average for OECD, very different from Asian countries where isn't really a gap. The US and UK, of course, aren't really particularly high performing countries, not compared to Asian countries where there isn't a big gap or the OECD average.

Given that spatial skill is involved in the boy-girl gap, we might have a situation where we in the USA and UK are up-biased in our understanding of how spatial skill contributes to math skills, possibly through skewed teaching (maybe we emphasise spatial strategies for maths too much?). 

Anonymous_IV said...

Why is it so important that universities not be allowed to generalize based on race?  This seems particularly inconsistent with the scientific evidence you've described here, which supports the biological reality of race and the clear statistical significance of some of its politically troublesome correlates.  Yes, race alone does not predict academic and life outcomes nearly well enough, and test results do much better.  But test results are still far from perfect.  If the scientific evidence says that test results plus race give significantly better predictions, why should universities wilfully avert their eyes?

Anonymous_IV said...

Alas HYP+ have no physical room to accommodate a doubling or quadrupling of the student body, and of the classroom, office, and lab space that would be needed for a commensurate increase of the faculty to maintain small class sizes etc.  Indeed, to the extent that top schools have top faculty, it would be impossible for all of them to significantly expand the faculty while retaining its quality.

Yan Shen said...

Interesting anecdote related to Albert Einstein.

"Visual thinking is also associated with being intellectually gifted.
Albert Einstein was a visual thinker who failed his high school language
requirement and relied on visual methods of study (Holton 1971-72). His
theory of relativity was based on visual imagery of moving boxcars and
riding on light beams. Einstein's family history includes a high
incidence of autism, dyslexia, food allergies, high intellectual
aptitude, and musical talent, and he himself had many autistic traits -
an astute reader can find evidence of them in Einstein and Einstein
(1987). Other great scientists such as Leonardo de Vinci, Faraday and
Maxwell were visual thinkers (West 1991)."

“These thoughts did not come in any
verbal formulation. I rarely think in words at all. A thought comes, and
I may try to express it in words afterward.”

Yan Shen said...

My post below addresses some of the issues you've brought up related to Feynman. It also covers many of the other preeminent physicists throughout history.

Ju Hyung Ahn said...

Yet again, you keep trying to indoctrinate people with your obnoxious propaganda without presenting any empirical datum.  I don't think your gut feeling retort will win over many readers here.  The thing is that the disparity between the number of resident filings per million population of Japan and South Korea to some of the leading European nations is of such magnitude (5-10 times higher) that it is hard to dismiss it by such simple rebuttal.  There's obviously some substance to patent data that they often get quoted in business papers as checklists to see which corporation is innovating.  Perhaps you should present some solid evidence to discredit East Asian patents as trivial and insignificant while European ones exude genuine ingenuity, since it isn't the case that Europeans are volunteering to relinquish their patent rights to what they deem more trivial.  From the recent Apple's patent war against its competitors to continual European criticism against Chinese firms for not observing their patent rights, the Europeans have been kin defenders of their intellectual property.

I would suggest presenting some form of empirical evidence the next time when you make some not-so-obvious speculative statements such as the following:

1) "I think a larger number of white people than Asians possess certain specific talents and abilities in a more intense form. In certain OTHER ways I think Asians are the ones who are *superior*."

2) "Yet their technological prowess goes well beyond what these figures
so clearly the top layer of talent is somehow in some way, actually
more talented than the top layer in countries with a higher average IQ
(like, say, South Korea)."

It isn't the case that you are making some metaphysical statements by stating the above.  The type of dogmatic banter you're engaging in is harmful to intellectual discourse.  You also need to make your points more clear.  As to the case of 1), what "certain specific talents and abilities" are you talking about?  I mean white people excel in sports such as American Football compare to East Asians.  Your statement is so vague that it doesn't point to anything definite about this world.

darklayersify said...

Yan Shen, anecdotes aren't data. Yes, it's noteworthy that you found those things about Einstein and Maxwell. However, it is not the same thing as a detailed statistical analysis of thousands of scientists, including a smattering of Nobel laureates, and finding that Asians perform as well as caucasians or that fewer awards relative to proportion of the PhD pool can be explained by discrimination or networks.  

Moreover, if you take a second look at Maxwell, he may have had relatively high verbal ability at some points, although markedly weaker than his spatial and mathematical ability. Per wikipedia:

"Maxwell was fascinated by geometry at an early age, rediscovering the regular polyhedron before any formal instruction.[21] Much of his talent however, went overlooked, and despite winning the school's scripture biography prize in his second year his academic work remained unnoticed[21] until, at the age of 13, he won the school's mathematical medal and first prize for both English and poetry."

He might have had language problems at some points, but he likely wasn't always 1 SD below the mean in verbal abilities. 

Also, if verbal abilities, are so unimportant, why did the theoretical physicists in the Roe sample average above 160 in  verbal ability? If that data had shown that say one half or one third of theoretical physicists scored below 130, that would have supported what you claim. However, the Roe data did not find that. It found verbal abilities were often 3 SDs above the mean in the eminent theoretical physical scientists. 

Perhaps, a substantial number of top performers in theoretical physics, are 160+ in verbal and 200+ in math or 145+ in verbal and 180+ in math. 

Instead of there being an optimal profile of high S, high Q, and decent v, there could be several optimal profiles that vary with subfields. Maybe, in some areas extremely high S and Q and decent v are best, whereas other areas benefit more from higher verbal ability. Michael Faraday was said to be less able mathematically, but high spatial abilities allowed him to contribute a great deal. Indeed, as strong as Einstein was with math, Hilbert once declared that " Every street urchin in our mathematical Göttingen knows more about four-dimensional geometry than  Einstein. Nevertheless, it was Einstein who did the work,  not the great mathematicians. " He also got a lot of help from Noether and Grossman on the math.  He might not have been the absolute top dawg in math ability in the extremely stratified circles he was engaged in. However, I'm sure there are other individuals with really high math ability and more modest spatial ability (Relative to M) that accomplished a great deal in their fields. 

And again, what about Murray Gell-Mann's success in linguistics, Newton's writings on theology, and how some top mathematicians contributed to philosophy for counter-anecdotes? If one in say 3 of top scientists in math and theoretical physics are like MGM and Newton with extreme M and more modest but very high V, that could diminish the theoretical advantage of Asians.

Yan Shen said...

"Yan Shen, anecdotes aren't data."

But when you start reading that many of the great physicists in history were described as primarily visual-thinkers and that some struggled with language, it becomes reasonable to infer that visual thinking is critical to the domain of physics.

Yan Shen said...

"Only 2 Asian kids from 1990-2011 actually won, though a lot of Asians in the top 3."

Do you honestly believe there is a significant difference in intellectual aptitude/creativity between the winner and the 2nd and 3rd place finishers?

RKU1 said...

This raises an issue relating to the sort of standardized ability tests we use which I've sometimes wondered about.  Perhaps some commenter more versed in the history of psychometric testing knows the answer to this.

My impression is that it's generally acknowledged that the three main principal subcomponents of intelligence are Verbal, Mathematical, and Spatial, and this has been known for a very long time.  Therefore, it seems a bit odd to me that the most widespread such tests used in America, the SAT/PSAT/ACT/GRE, has traditionally focused on just two of those subcomponents, namely V and M, and totally ignored S.  Wouldn't this tend to produce unfair results for individuals whose abilities skew most heavily in the "S" direction, namely East Asians (and to a much lesser extent Northern Europeans)?

Now, obviously, EAs already hugely outperform on the existing SAT, so including an additional, near-equal "S" subtest would merely further increase this gap, prompting even more severe racial quotas, and hence will never happen.  But since these tests were developed long before America contained a significant EA population, that's obviously not the original explanation for the absence.  I wonder if it might be because the analytical psychologists who originally developed these tests skewed heavily toward "V" and "M" and therefore decided those were the important things to measure.

Personally, I've always been very strong in "M" and "V" and much, much less so in "S", so I suppose I shouldn't complain too much about this strange bias...

Yan Shen said...

"Personally, I've always been very strong in "M" and "V" and much, much less so in "S""

That seems to accord well with what we know about the cognitive profile of people of your ethnicity. ;)

Yan Shen said...

Additional data point. Henri Poincare.

He excelled at linking together ideas, often in a visual fashion. Poincare's preference for visual thinking was characteristic of much of the mathematics he produced throughout his career. (Some historians have speculated that because of his poor eyesight, he was often unable to see what his professors wrote on the board, and as a result he had to create his own pictures in his head, thereby strengthening his visualization skills.

han said...


darklayersify said...

When you start reading that many of the great physicists in history were described as primarily visual-thinkers and that some struggled with language, it becomes reasonable to infer that visual thinking is critical to the domain of physics, far more so than verbal ability.

Various commenters here have suggested that V may be the crucial ingredient for high level success in the sciences. I decided to investigate the matter by attempting to find out more about the preeminent scientific figures throughout history. At least with respect to physics, I found absolutely nothing suggesting that V was critical to high level success. If anything, what was striking was the repeated emphasis on visual thinking as the primary mode of scientific discovery utilized by many of the great physicists, men such as Einstein, Maxwell, Feynman, Hawking, Faraday, etc, coupled with anecdotes here and there about how various giants in physics struggled with language."
First off, some commenters here might have suggested that verbal ability was the crucial ingredient for success in science, but that is not the argument that some of us are putting forth here. Some are not claiming that verbal ability is the decisive factor, or that one cannot become a successful researcher without verbal abilities 3, 4, or 5 SDs above the mean. 

While cognitive abilities clearly matter for success, several experts (Dean Simonton) would claim that they're enough. 

I am suggesting that verbal abilities might contribute, particularly in some areas and subfields. That, in turn, could be a small disadvantage among others for groups with lower verbal ability on average. I agree that mathematical and spatial abilities matter a lot more than verbal abilities for physics, math, and engineering. But verbal abilities could still contribute, particularly in some subfields. 
Also, other fields besides physics do exist, fwiw. Charles Darwin doesn't sound like a super duper high M. I obviously don't have the means to calculate exactly where his verbal abilities might have been, but he seemed to have strong verbal abilities.

"In the third week of January 1831 Charles sat his final exam. There were three days of written papers covering the Classics, the two Paley texts and John Locke's An Essay Concerning Human Understanding, then mathematics and physics. At the end of the week when the results were posted he was dazed and proud to have come 10th out of a pass list of 178 doing the ordinary degree. Charles shone in theology and scraped through in the other subjects. He was also exhausted and depressed, writing to Fox "I do not know why the degree should make one so miserable."

Darwin once said that he deeply regretted that he didn't study more advanced math, and a lot of people have praised Origin of Species as extremely well done. He might have had relatively high verbal ability.

EO Wilson has said that his math abilities aren't amazing and won two Pulitzer prizes for science writing, and he's done very well in evolutionary and organismal biology, becoming an eminent researcher.

Ernst Mayr derided mathematical approaches to evolution, so he might not have been a super math nerd.

Other biologists are a lot more into math: Haldane, Hardy and Weinberg, Robert May, John Maynard Smith, Fisher, Hamilton etc.

tractal said...

Also, the simplistic "visual+M=physics ability" hypothesis runs into a huge problem when we consider Ashkenazim theoretical physics success. And you still have not explained away the insane verbal scores in Roe's sample. 

Speculating about the cognitive abilities of people we never tested is not an adequate substitute for data. 

Soren Kay said...

"Finally, Asians may have lower rates of sociopathy, which reduces their
chances of making it to the top (close inspection suggests it is mostly
sociopaths at the top ;-)"

When I first read that my mind though "sociality" instead of "sociopathy"... and it's probably more accurate.

Yan Shen said...

"Also, the simplistic "visual+M=physics ability" hypothesis runs into a
huge problem when we consider Ashkenazim theoretical physics success."

Yet various preeminent Ashkenazi physicists such as Einstein and Feynman had well documented histories of being much weaker verbally and gave first hand testimony that they were primarily visual thinkers.

"Speculating about the cognitive abilities of people..."

What I have cited are the introspective testimonies of these preeminent individuals themselves. I would imagine that they know much better than anyone else what their primary method of thinking is.

ytrewq123 said...

"Yet again, you keep trying to indoctrinate people with your obnoxious propaganda without presenting any empirical datum. "
Empirical datum? There's tons of it if you go and look at research awards. It's interesting that Steve Hsu doesn't attempt to rebut  this part of Sineruse's post. Whether it's the Fields/Abel Prize/Cole Prize/Waterman Awards/Hertz Fellowships/Nobels/Fulkerson Prizes/ACM Doctoral Dissertation prizes/Morgan Prizes/Turing Awards/INFORMS Fellows/Knuth Prize winners/Nevanlinna Prize winners/Bocher Prize winners, etc. 
Some of these prizes have lots of young winners, so there's no lag effect that's yet to kick-in. The problem is that research ability is hard to quantify using a metric. It will necessarily remain "metaphysical," in the sense of being nebulous to define, but easy to observe. I don't know why you're getting so worked up about this. If it doesn't gel with your perception of Asian superiority, well, sorry.  I guess the only valid metrics are those that confirm your biases.   

Ju Hyung Ahn said...

The point was he kept making vague statements without pointing to any data.  Never in my posts did I say anything about East Asian superiority.  I think you're the one who's getting worked up here.

"Another good example is Israel, with an IQ of also about 94, and I have
read that even the Asheknazim in Israel have an IQ of a very modest 103.
Yet their technological prowess goes well beyond what these figures
suggest, so clearly the top layer of talent is somehow in some way, actually more talented than the top layer in countries with a higher average IQ (like, say, South Korea)"

Notice how the comparison was between Israel and South Korea. This obviously takes out contributions Jewish people made out of Israel. It is easy to see how more conducive research environment in the US or Europe worked favorably to Jewish people outside Israel than people within.  With Israel having the advantage of being influenced by many eminent Jewish scientists traveling between the US (or Europe) and Israel, I don't see by comparing sci/tech contribution done within Israel and South Korea vis-a-vis to see such a higher "top-layer" intelligence of Israelis to Koreans.

Next time, I'd suggest actually bringing out some relevant evidence to the topic.  Mindlessly listing data without making any coherent argument doesn't help your cause one bit.

ytrewq123 said...

"Never in my posts did I say anything about East Asian superiority.  I think you're the one who's getting worked up here."
Not referring to this particular post; just an impression I gathered from reading all the comments you've made on this blog.

Anyway, if the suggestion you're making about mindlessly bringing data to the table is directed at me, you are kind of missing the point of this exercise. When a preponderance of STEM majors are Asian, and when the top ranking feeder schools for STEM PhD programs in the US are Tsinghua and Peking, it's hard to make a disparate access to resources argument (I've not dis-aggregated E.Asians into their constituent nationalities, so I'm not sure about Kevin Rose's specific assertions, but the general thrust is not only plausible, but is certainly a reality in the context of research).  

"Mindlessly listing data without making any coherent argument doesn't help your cause one bit."

I don't have a dog in this fight. Nor do I  think Asians should be discriminated against when it comes to college admissions. Having thought about it I realize that there are enough mediocre people in the most elite undergrad programs in the country to make it pretty clear that there is no real selection for research ability ( other than for a small fraction of seats reserved for future academic superstars). I'd much rather have a bunch of real academic achievers in undergrad schools than have "well-rounded" lawyers in the making, though unfortunately it's likely that America will continue to induct glib mediocrities into its elite. 

ytrewq123 said...

You should probably withdraw your claim of having an IQ of 214 - it seems 
especially   preposterous given that a 125 IQ person took you down eons ago :)   

RKU1 said...

Huh???  I must say that the category of people who comment anonymously on HBD blogsites sometimes includes those who make the oddest accusations and provide the strangest sort of non-sequiturs ...

ytrewq123 said...

From Wiki: 

" Newspapers referred to Unz's candidacy as a Revenge of the Nerds and often quoted his claim of a 214 IQ."

The story, of course, comes from this: 

Note to self: Don't reply on Wikipedia.

Kevin Rose said...

Most Israeli technology companies are start ups by Israelis who graduated from a particular Army unit, if I understand this correctly. Steve posted on this a while ago. So yeah, they are Israeli, not American Jews. 

Israel and South Korea comparisons just provide a very stark way of illustrating my central point about the limits of what group averages in IQ tells us about the world. East Asian countries have the highest national IQ scores so I would have to choose one of them to illustrate my point best - it was not really a dig at South Korea.

Dont you see how fascinating this is though? Put aside your ethnic issues for a moment and just look at the world around and how it actually IS, not the results of some test on paper. We get all caught up in IQ numbers and which group has higher blah blah blah and yet right here in the real world, staring us in the face, are blatant examples of IQ numbers completely failing to predict extremely important metrics of intellectual performance. Perhaps THE most important metrics of intellectual performance are not predicted by group IQs. We are so fixated on measuring and quantifying that we simply IGNORE obvious facts that we do not yet know how to quantify or measure. I suppose this is one of the mental blocks high M are susceptible to ;) 

The REAL WORLD is far more complex, recalcitrant, and heterodox than anything suggested by test results. 

Now, looking at test results are important too, and it appears that even going by test results alone, if interpreted intelligently, Asians are not really being discriminated against

zk7 said...

"lack of "ethnic affinity" networks"

If you seriously believe that Chinese nationals, Chinese immigrants and Chinese Americans in the U.S. have less ethnic affinity networks than white Americans, than you are truly delusional.

Why should Chinese citizens have all the same rights and privileges in the U.S. as U.S. citizens?  Do U.S. citizens have the same privileges and rights as Chinese citizens in China?  Do Chinese organizations have non-Chinese in leadership positions?

Are you conflating Chinese Americans with Chinese nationals?  Three years in a boarding school does not make a Chinese parachute kid a Chinese American.

RKU1 said...

Well, I don't recall ever having claimed to be that Unz fellow, which is obviously just a trivial detail.

And anyway, if he supposedly has an ultra-high IQ while you seem to say I've revealed myself to be a total idiot, haven't you just refuted your own speculative hypothesis?

zk7 said...

"Nor do I  think Asians should be discriminated against when it comes to college admissions."

But do you, or any of the debaters, or the institutions making the decisions, distinguish between Asian Americans and Asians nationals.  Why are Americans, including Asian Americans, being required to compete with the entire world for American opportunities, and why are American institutions and American taxpayers being required to treat non-Americans as if they were Americans. 

Do Chinese nationals magically become Americans after 3 years in an American school, either a boarding school, or more often as a parachute kid in a public school that the kids parents have contributed nothing?    Do citizens and taxpayers of China somehow become Americans by sending there kids to school in the U.S.?

tractal said...

Now, looking at test results are important too, and it appears that even going by test results alone, if interpreted intelligently, Asians are not really being discriminated against. 

I think this is the most convincing argument brought up here for East Asian selective test interpretation. Differential Medical and Law Board passage rates per test scores is powerful evidence. There you can directly tie Asian test scores to an objective and important overprediction. 

Elsewhere it seems speculative to me. If it is true that Asian scores over predict in the upper upper end generally, it isn't the kind of truth you can easily measure. Imposing an admissions disadvantage when there is really no more evidence behind it than a broad heuristic overview of "achievement."

If Universities really want to go and investigate for Asian test score overprediction I guess they could. But we know it doesn't overpredict grades, so almost by necessity you would have to depend on more nebulous post grad criteria. And even if you did that I imagine that Asian's wouldn't obviously lag. By many measures I think the reverse would be true. 

From the reverse position, I hope you can imagine how this would feel. Asian SAT scores are perfectly predictive for GPA and general post-grad success. But then suddenly post-grad success is redefined as the tail's end end subset of achievement. It could easily feel like moving the goal posts. Ad hoc adjustments of achievement beyond the conventional objective metrics for this particular case is going to make people rightly sensitive. 

Yan Shen said...

Regarding current research, time lag and all that, it might be interesting to note that Thomson Reuters regularly publishes a list of top 10 papers in the fields of physics, chemistry, biology, and medicine. According to the site,  "These papers, published within the previous two years, have attracted a
notably high number of citations in brand-new journal articles indexed
by Thomson Reuters during a recent two-month period".

Now admittedly this is a small and crude data point. But, taking a look at the November-December 2011 Top 10 lists for





reveals a fair number of East Asian names, particularly in physics and chemistry and to a lesser extent biology.

I suspect thought that other data points like this exist. A snapshot of the current state of research would probably reveal a much different picture than the relevant historical analyses.

Sam H said...

As far as I am aware one topic that has not been discussed so far is the differences between ethnic groups as it relates to the measure of extroversion. Extroversion is influenced in part by testosterone and the sensitivity of those receptors which vary among the ethnic groups & races. 

Here is my generalized ranking for the trait 'extroversion,' looking at a limited number of groups:

N.E. Asians < Indians (tribal US) <S.E. Asians < Scandinavians < Eastern Europeans, Hispanics (tied) < S. Asians, Western Europeans (tied) S. Europeans < Jews < American Blacks

I believe that if all other traits are controlled for (ACT/SAT/IQ, SES of family, etc.) extroverted people tend to MAKE MORE MONEY and also tend to be more SOCIOPATHIC. Elite universities may not know about the latter but I think they believe in the former. There's no honest way for a university to measure extroversion because the tests are so easy to game so they look to extracurriculars. I have not read the Espenshade study but he probably took 
extracurriculars into consideration so the admit people probably just make inferences in part by race. 

Moreover perhaps they think Asians are less likely to make big donations if they hit it big, but I think that has been discussed some already. 

ytrewq123 said...

"haven't you just refuted your own speculative hypothesis?"

I'm stupid, so I'm not sure what you're referring to.   I don't recollect ever suggesting that it was my hypothesis that you have a 214 IQ. The number seems very very implausible, and is either a fabrication or the result of a very poorly designed test. I wouldn't hypothesize such an absurd IQ for anyone - dead or alive. Needless to say, such a hypothesis would be refuted easily given the additional information of  a not very high S score.  

That you made such a claim is the hypothesis, not that you have an IQ of 214. The don't-rely-on-Wikipedia part refers to your making that claim. I'll believe it if you said you didn't. 

Yan Shen said...

One reason why you sometimes see IQs of 200+ associated with people is that these are childhood ratio IQs, i.e. mental age/chronological age. Since most people stop progressing mentally after a certain age, the ratio IQ becomes meaningless for adults.

Ratio IQs are supposedly log-normally distributed, hence fat tails.

ytrewq123 said...

I'll accept that as a data point that runs against what I'm hypothesizing. Here are a couple of data points that support my hypothesis: the Godel Prize and the Fulkerson Prize, both based on the outstanding papers in the area over a period of time, both having been won by undergrads, are preponderantly Jewish/White with a few Indians/E.Asians thrown in.  

ytrewq123 said...

"But do you, or any of the debaters, or the institutions making the decisions, distinguish between Asian Americans and Asians nationals.  "

I meant Asian-Americans. Sorry. Will edit. 

and why are American institutions and American taxpayers being required to treat non-Americans as if they were Americans"
In what sense? Foreign students SUBSIDIZE Americans at the undergrad level by attending American institutions. At the grad level, they form cheap slave labor for faculty. I've seen faculty - American/E.Asian/Indian - not allowing cheap foreign labor aka PhD students to get their dissertations done till they've done a couple of extra years of work beyond what is needed for a PhD. America should welcome talented foreigners onto it's shores. They provide money or high-IQ labor. Both are desperately needed these days.   

ytrewq123 said...

Any idea what the parameter values are for the log-normal distribution? I'd like to see what Unz's purported IQ score corresponds to as a percentile on this distribution. 

ytrewq123 said...

Checked the Sloan data. The Asian split is 16% East Asian and 11% S. Asian. For Tech Review's 35, for this year, it's 17% E.Asians and 14% S.Asians. Are S.Asian Americans  discriminated against in the same fashion as E.
Asian Americans in college admissions? What percentage of students in elite undergrad programs are of S.Asian extraction (<5% would be my guess)?  

namae nanka said...

Why not?

namae nanka said...

I will be very hesitant to believe in a study that has Janet Hyde as one of its authors and involves maths and gender. (see la griffe, geoffrey falk's)
The validity of data they quote might not be in question, but they are more likely to include something that furthers their agenda and represent it that way.

Iamexpert said...

Is there actually any evidence that sociopathy correlates with success? The only studies ive seen suggests psychopaths are less successful.

Iamexpert said...

Is there actually any evidence that sociopathy correlates with success? The only studies ive seen suggests psychopaths are less successful.

Iamexpert said...

I think whites are more successful than East Asians of equivalent IQ because they have more testosterone and thus are more aggressive, competitive and ambitious analogous to why men are more successful than IQ equivalent women. I also think East Asians are more mentally stable which impedes creativity compared to whites of identical IQ given the link between Creativity and madness. I also think whites are less creative than IQ equivalent blacks for the same reason thus explaining the overrepresentation of blacks in the entertainment industry.

Iamexpert said...

Folks in this thread are arguing about whether East Asian superiority is a function if g or spatial ability. It is a function of both. The fact that East Aians excel on the Raven is because of g since the Raven has very little spatial component. Indeed i've known people with poor spatial ability who perform well on the raven because they verbalized the solution. The raven does not require the rotation of objects in the mind's eye, it is simply a test of abstract reasoning, analogical thinking and concept formation.

Further evidence of superior East Asian g comes from their performance on chronometric tests of complex reaction time. These tests have no verbal, spatial, or mathematical component but correlate 0.7 with IQ tests purely because they measure g.

Final evidence of superior East Asian g comes from their large brain size. Research finds that the more g loaded a cognitive ability, the higher its correlation with brain size.

Iamexpert said...

Folks on this thread are arguing whether math ability is a function of spatial aptitude or a separate talent. I think human intelligence has only 2 main parts: verbal and spatial. I think math ability is a good measure of overall IQ because it's simply a function of verbal ( algebra) and spatial (geometry). That's why both verbal Jews and spatial east Asians excel at math.

Iamexpert said...

Folks on this thread keep citing the Roe study but the IQ's in the roe study are statistically meaningless because the IQ test used was normed based on an old Binet test. The old Binet test gave ratio IQs which are difficult to interpret because both the standard deviation and the shape of the distribution vacillates across ages and across the type of ability being tested. So a verbal IQ of 160 at age 16 might be a one in 30,000 level achievement while a spatial IQ of 160 at age 8 might be only a one in 200 level achievement. The point is we have no idea precisely how gifted those scientists were at verbal, spatial or math talent.

Also we can expect a lot of brilliant physicists to have high verbal ability simply because g loaded abilities are highly correlated, so spatial geniuses are often verbal geniuses too. But this in no way implies that physics achievement involves verbal ability. Indeed the fact that some of the greatest physicists ever were verbally mediocre or even subnormal implies just the opposite.

darklayersify said...

I don't know about the testosterone explanation. Benbow et. al 2000 found that after controlling for hours worked, the gender-related variation in pay disappeared in their sample of males and females in the top 1% of math ability. 

darklayersify said...

This is an interesting perspective. It's certainly possible that what Roe found was just a statistical artifact without any casual weight, but it deserves noting that the physical scientists focused on experiments performed worse than the biologists and social scientists. The theoretical physicists in her sample performed better. Certainly, it's possible that's just an artifact of "g", but it would be nice to get confirmation through actual data. 

I suppose this could be accomplished with the GRE and Raven's. 

Iamexpert said...

But if high testosterone people work more hours because they're more driven, ambitious, competitive etc, then controlling for hours worked largely controls for the benefit of testosterone.

sineruse said...

None of the E.Asians on the Tech Review list attended a US high school, and the same is probably true for the Asians (and many of the non-Asians) on the Sloan list.   It is the ratio for (mainly East) Asian Americans and white Americans that is relevant to the college admissions discussions.  Foreign students have a harder admission track, but this is undisputed and nondiscriminatory.

sineruse said...

Is there a significant difference between making the US IMO team and getting a gold medal?  One could argue either way, but somehow the E.Asian Americans who accomplish the first have been less likely to accomplish the second (and far less likely to attain top results on the Putnam) than their white counterparts, just as the E.Asian Westinghouse/Intel/Siemens finalists and high finishers mysteriously are always unlucky enough to not make #1 if they make the top 3.   I mean this is just an incredible streak of bad luck to have the negative outcomes always land disproportionately on the US E Asian competitors for 15-20 years in a row.

With the science project competitions you could argue, not very convincingly, that the judging is discriminatory or that the use of recommendation letters and public presentations disadvantages the Asian competitors.   For blind-graded math exams (and the LSAT and medical board exams that I also mentioned) this explanation is not available.   Steve Hsu avoids this point when talking about the potential for discrimination to reduce (E)Asian numbers at the higher selections; the question is why the same thing happens in objective exam selections where discrimination is impossible.

Yan Shen said...

"somehow the E.Asian Americans who accomplish the first have been less likely to accomplish the second"

I find it hard to believe that you're still going on about the IMO. Given that East Asian Americans are probably over-represented by a factor of 9x, I have no idea what point you're trying to make.

sineruse said...

> Wouldn't this tend to produce unfair results for individuals whose
>skew most heavily in the "S" direction, namely East Asians

If a selection method (e.g., college admission, grad school exams such as LSAT, job hiring) underweights important abilities that a group, such as E.Asians, are advantaged on and that matter for later performance, then members of that group will overperform after the selection.  For example, if selections are based too heavily on "M" and "V" metrics, but as Yan Shen claims, it is "S" that is highly important, then E.Asians who make the cut also carry a hidden S superpower that, if S is important, will propel them farther than others who only have the required levels of M and V.  The proportion of E.Asian stars should be substantially higher than the proportion of E.Asians initially admitted, and if there is a "g" advantage this should only intensify the escalation of E.Asian numbers.

It's a nice idea, but the opposite phenomenon is the one universally observed.  E.Asian numbers decline as one goes down the pipeline.   It could be that S is not that important, or that it is underutilized in our society (but if proven to be useful, could be weighted more heavily), or that "S" itself is just a mantra recited by HBD fans and does not have much basis in the psychometric literature.

sineruse said...

>Unfortunately, you commit the typical white nationalist fallacy. There
are three major cognitive categories, math, verbal and spatial and not
merely two,
>verbal and spatial.

Please provide a reference for this claim.  According to this article, on the WAIS family of IQ tests there is a longstanding and robust pattern
of two principal components: "g" and verbal-spatial asymmetry (some multiple of V-IQ minus another multiple of S-IQ). As they point out, you
can report V and S subtest results separately, superficially providing three pieces of data, but it does not carry more information than "g" and the
V-S difference.  The "V" and "S" are not principal components like "g" (indeed, they cannot be; they have to correlate with "g").

>Relative to Europeans, East Asians are
significantly higher in math and spatial and lag behind slightly in

There may well be all sorts of tasks that show racial differences, some of them favoring E. Asians.  However, the division into "verbal" and "spatial" (and, especially, "mathematical") is only an intuitive one to make sense of the test questions.   The labelling of principal components or factors as V,M,S etc is derived from the degree to which different test items load on them but it could be that for a particular "spatial" questions the probability of answering them correctly loads more heavily on the "M" or "V" (i.e, "g") factors. 

Kevin Rose said...

*Thinking in images* is not the same as *spatial reasoning*. Poets frequently *think in images* (metaphors) yet this is clearly not spatial reasoning. When I say *your beauty is like the dawn* I am thinking visually and in images (though expressing it verbally).

Why do even the most basic concpets get distorted at the hands of Yan Shen?  

darklayersify said...

Iamexpert, this may be a complicated issue. I agree that spatial ability and possibly verbal ability contribute to math. However, I wonder if it's really that straightforward. In this paper, roughly seventy percent of the adolescents who scored in the top one percent of spatial ability were not in the top one percent of mathematical or verbal ability. These authors also found that kids with top 1% spatial ability, but not math or verbal ability, became STEM PhDs at This analysis also seems to question such a straightforward formulation: The purposes of this study were to determine whether mathematically gifted boys and girls were significantly different in cognitive abilities and personality factors and to determine whether these cognnve abilities and personality factors were accurate predictors of mathematical achievement. Mathematically gifted boys (n = 77) and girls (n = 62) completed a mathematics reasoning test (SAT-M), a verbal reasoning test (SAT-V), a measure of spatial ability (Group Embedded Figures Test), and personality test (California Psychological Inventory). They received mathematics achievement ratings depending on the number of summer courses completed. The findings indicated that boys not only have higher mathematical reasoning ability than girls, but alslo that this ability is the single best predictor of their mathematical achievement. Verbal ability is the best predictor of mathematical achievement for the girls. For both the girls and the boys, neither spatial reasoning ability nor personality factors significantly predicted mathematical achievement. Consider, also, this statistical analysis: 

sineruse said...

>Asian SAT scores are perfectly predictive for GPA and general post-grad success.

Espenshade-Chung-Radford's study, among others, found substantial overprediction of Asian college grades.  Their most interesting finding was that for prediction given SAT, HS GPA,  some socioeconomic indicators,  field of college major, athelete or legacy status, *and first year college class rank* there was still a negative Asian effect on class rank in the later years (i.e., in the more advanced courses).  The effect was larger than that of being an athlete or legacy and equivalent to more than the famous "140 SAT points" when quantified in the manner that so impressed US News, of holding all other factors equal.

highly_adequate said...

Your point about women has always struck me as an "interesting" one.

While there's tremendous talk and complaint among many about the seats in elite institutions set aside for underrepresented ethnicities and races, the great impact of Affirmative Action in most such settings derives from the compensations put in place for gender. Mostly, this is a matter of numbers: women are 50% of the population, whereas these other minorities don't add up to such a large figure (though it's not trivial, if one adds up African-Americans and Hispanics across the current US population).

But the problem with women is obvious: they simply don't turn out to be nearly as productive, on average, compared to males, even controlling for everything that would seem to be relevant, such as test scores and college or graduate GPA. Even if Affirmative Action strictly speaking were eliminated, and "Equal Opportunity" in a pure sense were enforced, removing any consideration of gender one way or the other, still women so chosen would be, quite predictably, less productive on average than the otherwise similar men with whom they compete for spots.

This is a real problem for society. Probably nowhere is this more obvious today than in the composition of physicians. It's well established that the current cohort of physicians, who are now quite closely approaching 50% women in representation, are not nearly as productive as were physicians in previous eras. And it's obvious as well what the major driver of that loss of productivity is: the far larger presence of women. It's not just that women are far more likely to take time off to have children and to raise them when young; they are, apparently, considerably less hard working and driven in all phases: they put less time in even when their children are older, and, I gather, retire earlier. The loss in precious resources to society, which invests great amounts of the irreplaceable time of highly trained professionals and of money to train physicians, is quite considerable; certainly if this loss could be corrected for without any other political or social consequences, it would be implemented immediately.

Yet there may well exist no predictive trait on which the relatively unproductive women could be culled out other than their gender itself.

The point is, the typical conservative position of race of gender blind admissions is not likely to come anywhere near resolving the problem here: unfortunately, it is almost certainly true that, as information in prediction, gender plays an ineliminable role; and it would not be surprising that likewise race and ethnicity plays such a role.

How to deal with these unhappy facts is, of course, an political and social hornets' nest.

highly_adequate said...

This strikes me as missing the point of what Roe actually did.

She didn't use the verbal measures from the Stanford Binet IQ test as her way of determining so-called verbal IQ. Rather, she had ETS devise a set of questions -- presumably very akin to questions like those on the current GRE verbal test -- and tested the scientists on those questions, and compared them to the scores of graduate students in the same fields (I believe). These graduate students were also administered the Stanford-Binet IQ test, so that the projections of verbal IQ were extrapolated to the scientists from those scores. Apparently, the median verbal IQ measured on the Stanford-Binet for those graduate students who went on to get a Ph.D. was 141.

Now a median 141 verbal IQ for a graduate student who gets a Ph.D in the sciences -- I assume from a top ranked institution like Harvard, since that's where Roe did most of her work -- does not sound implausible, even granting the imperfections in the IQ test.

What remains as a hard fact is that the scientists performed at the 166 range on the ETS verbal questions, whereas the graduate students only scored 141 on those tests. The projected IQ difference, which amounts to, let's say, 1.73 SD (assuming 15pts = 1 SD), is, likewise a pretty hard fact. Perhaps some adjustment needs to be made for both the imperfections in the Stanford Binet and the Flynn effect, but the tremendous advantage of an eminent scientist over the average Ph.D. student presumably at their own institutions remains, and attests to the apparent importance of verbal intelligence as at least a major correlate of great achievement in the sciences.

One can't help but think in the wake of this result that programs in science might do well to place considerably more emphasis on the GRE verbal than they currently do, if they are seeking to capture some of those students with the greatest potential for achievement.

Iamexpert said...

So what you're saying is the graduate students got a mean score of X on the ETS test and and a mean IQ of 141 on the old Binet, and the scientists got a mean score on the ETS of X + 1.73 SD, thus Roe assigned them an IQ of 141 + 1.73(15) = 166? That sounds reasonable, but it's hard to know what an IQ of 141 really means for adults taking the old Binet or for that matter what 1.73 SD means on the old Binet.

Marilyn vos savant for example I believe obtained an IQ of about 230 on this test by scoring a mental age of 23 at the age of 10 since 23/10 x 100 = 230

I believe the average adult has a mental age of 16 on this test so mental age of 23 might be a misnomer, it simply means that if intelligence continued to increase past 16, 23 is how old the brain would be when it typically achieved the level of intelligent Marilyn had at 10. These graduate students to have an IQ of 141 would have a mental age 41% higher than 16, and the scientists to have an IQ of 166 would have a mental age 66% higher than 16.

But how many SDs they are above the mean is a mystery because ratio IQ's have different SDs at different ages and on different tests. There was a famous East Asian 4 year old, his name escapes me who could do calculus at age 4. This implies a ratio IQ of above 400 since most Americans can't learn calculus until 16 (if they even can than). So with ratio IQs as high as 400 being possible, it suggests that the SD of ratio IQ's can be much greater than 15 and/or that the distribution is not normal, at least for certain g loaded skills.

tractal said...

All of your arguments are biased and only half valid. It is true that we can't know exactly how brilliant the candidates in Roe's study were, but we can know that they were more brilliant than the Ph D students, and also that they were in some cases brilliant in ways more complex than simple G. Anthropologists had really high V, lower M, for instance. Without reference to exactly how many SD they were up in verbal, that is evidence that V may be more important than M in anthropology. This kind of thing is trouble for the simple G view of ability. 

The same problem confronts the physicists. Why, if both experimental and theoretical physicists were brilliant and math, do we see the theoretical physicists only with really high V? Perhaps theoretical physics is simply more G loaded than experimental, and the theoretical guys simply had higher G which manifested as higher V. But that doesn't help us make sense of the outstanding M scores of experimental physicists. In some sense, the experimental physicists were performing because of exceptional ability. It is much easier to interpret this data as saying High V + High M is important for theoretical physics than it is to say theoretical physicists have simply higher G than experimental. 

Think about it from the reverse: the outstanding anthropologists and biologists have low M high V, experimental have high M lower V. You want to argue that really they all have just high G, with theoretical physicists having the most. But that doesn't explain why the relative M and V scores of different disciplines falling into intuitive patterns. The simple G model cannot explain why disciplines could have significant variation with respect to V vs M. 

2) you argue that math is the best general measure of G, because you think both V and S contribute to it. That is just a really weak argument. 

3) You argue the Raven is not spatial because your friend this one time had high V and did well on it. Ok. 

4) You then go on to speculate wildly about group differences in "success" which adds nothing to the discussion. 

tractal said...

Especially because the GRE-M is a laugh out loud joke of a test. 

Iamexpert said...

Anthropologists were higher in V than M partly because they had spent the last how many decades studying the verbally intensive field of anthropology. It seems reasonable to assume anthropologists have innate advantage in verbal over math , but we also know that the causation works in both directions. People who spend their entire lives in academia almost certainly have their verbal scores artificially inflated by the advanced vocab they are constantly exposed to. A career in academia is equivalent to decades of being coached for an ETS test. Similarly, the physicists have been coached in math skills thus invalidating math tests as a measure of their innate ability. It's like using a chess test to estimate the innate ability of chess champions. Probably not the best idea? Probably the only valid scores for any of these people was spatial since it's a measure of fluid ability and not crystallized achievement but because age was not controlled and normed for (the Flynn effect) and because all the scores were anchored using statistically sloppy ratio IQ's, we have no idea where they stand.

I think theoretical physicists had higher verbal IQ than experimental physicists because (1) theoretical physics is super abstract thus g loaded, so they do really well on all highly g loaded tests, and (2) theoretical physicists have much more practice expressing themselves verbally than experimental physicists. It's possible the verbal talent helped cause their theoretical physics eminence, but adding such a superfluous explanation violates Occam's razor.

Now Jensen claims that he has never seen a factor analysis where the raven had a significant loading on anything other than g and test specificity. It never seems to load on the spatial visualization factor. Now if you administered the raven with a strict time limit, then you might find spatial ability is important for parallel processing the problems rapidly, but with no time limit, it becomes purely a measure of abstract analogical concept formation. Indeed people with nonverbal learning disability often do quite well on the raven even when theyre impaired on other nonverbal tests.

Iamexpert said...

It sounds like the studies you cite have a lot of noise and contradict themselves. One study finds spatial ability predicts STEM careers, the other finds spatial ability does not predict math achievement. One study finds math aptitude predicts math achievement for boys but verbal ability predicts math achievement for girls. Taking it all together, I think the takeaway is that math skill is a measure of overall intelligence and thus involves a broad range of cognitive talents ranging from spatial to verbal.

I find it significant that the two most g loaded subtests on the WAIS-IV (figure weights and arithmetic) are also the two most math loaded subtests. None of the verbal subtests on the WAIS-IV are as g loaded as these two subtests.

tractal said...

But you are arguing that the V difference between experimental and theoretical physics is largely to do with the verbal practice theoretical physicists receive in their careers. It seems reasonable to say that some benefit to V could be derived, but if you are saying that the difference between the 125 V of the experimentals and 165 V of the theoretical physicists (yes, I know there are norming problems, but there is still a large  gap regardless) is due to training and practice then you are claiming that IQ tests don't really test innate faculties at all. Your position (theoreticals have high V but it doesn't help per se, they just learned it along the way) commits you to saying that you can "train" 2.5+ standard deviations, in which case IQ tests and IQ would be rather meaningless. 

highly_adequate said...

My point has nothing to do with ratio IQs, or even with the Stanford-Binet, now or then.

Presumably, Roe's most important result could have been obtained without any strict IQ testing at all; the Stanford-Binet scores, as best I can tell, were employed entirely to provide a score in terms of a standard understood quantity, IQ, that was also tied into a fairly standard measure of that quantity. The significant point is the number of standard deviations above the mean scored by the eminent scientists over the typical Ph.D. student in their own field. I don't know exactly how those ETS tests were normed to provide such measurements in SDs, but presumably it was done in a way essentially independent of any other measure of IQ.

I'd think it pretty obvious that such Ph.D. students are already 2 or 3 SDs out from the norm in terms of verbal intelligence -- where exactly might be determined by the best means available. But the further app. 1.7+ SDs in verbal intelligence for eminent scientists is an important finding.

Iamexpert said...

Well first the true scores on a properly normed test (sd 15) is probably something like a 150 VIQ for theoreticals and 123 VIQ for experimentals. I'm assuming the ratio IQ's resemble a lognormal distribution with an SD of 16 and conertibg them to a normal distribution with an SD of 15. Now subtract about 8 points from theoreticals because of the verbal practice they've had. I say deduct 8 points because a longitudinal study found that kids who stayed in high school gained 8 (verbal) IQ points on kids who dropped out, so let's assume experiments are analogous to high school dropouts and theoreticals are analogous to high school grads in terms of the extra academic practice they benefit from.

So now we've reduced the verbal gap to only about 1.3 SD (VIQ 142 vs 123)

Now what is the cause of this remaining verbal gap? Is innate verbal ability causing achievement in theoretical physics or is g causing both the theoretical achievement and the added verbal ability? One way to find out might be to look at their spatial scores. If the verbal gap is greater than the spatial gap, then the former hypothesis is probably more explanatory than the latter.

Iamexpert said...

Well there's no way to really know without seeing Roe's original data, but the most reasonable assumption is that the IQ equivalents she assigned had a log normal distribution with an SD of 16.

Converting this to modern IQ scales which have a normal distribution and an SD of 15, the V IQ 141 PhDs become about V IQ 135.

Meanwhile the V IQ 166 scientists become about V IQ 150.

So a gap of only 1 SD, but even this is inflated because the scientists have their V IQ inflated by their presumably advanced age. In the wechsler IQ test norms, I've noticed that middle aged people with identical vocabularies to twenty somethings obtain aged normed scores 0.33 SD lower.

So deduct 5 V IQ points from the scientists because they have had decades more lifetime to acquire vocabulary then the PhD students

So now the V IQ gap is only 0.66 SD and even this arguably vanishes when you consider that the added life experience of the scientists was spent in academia, so whatever increase in vocabulary that normally accompanies advanced age would be triple in such an educated environment.

RKU1 said...

Well, since this debate is still going, but the comment-thread structure has gotten unreadably complex, I think I'll add a new one.

Now I certainly don't have the apparent psychometric expertise of various other commenters here, so I'll have focus on some very crude empirical date.  Back when I was in high school and college, it was always pretty clear to me that the U.S. Math Olympiad winners I knew were among the very smartest people I encountered.  The other really smart people I knew tended to generally share this judgment.

But since all that was many years ago, I just now used Google to look up the names of the 2009, 2010, and 2011 U.S. winners, 36 in call (including duplicates).  Of those 36 names, 21 seemed to pretty clearly be East Asian (mostly Chinese), and my hand calculator tells me that's just under 60%.  Now unless I'm seriously deluded about American demographics, East Asians aren't 60% of the age cohort in question; in fact, they're probably under 5% (with whites being over 50%).  Doesn't that seem to indicate something?

By way of comparison, it looks like two of the 36 names are Jewish and one is South Asian.  Hmmm...  Almost all of the remaining 12 seem like they're probably Northern European (though admittedly it's possible that a few of these might actually be Jewish, Asian, or something else).  Hmmm...

Can someone please explain to me---in simple terms, so someone as ignorant and thick-headed as myself might possibly understand---how these facts are consistent with the endless claims made above that white Europeans are just as smart as East Asians and that Jews are considerably smarter?

Now admittedly, East Asians work very hard and are diligent students, but c'mon...  Presumably the reason that almost all the world's long-distance running records are held by East Africans is because...they train harder than everyone else.  And the reason that almost all the world's short-distance sprinting records are held by West Africans is...dedicated training.

In fact, I'm sure that if the Vietnamese in America trained really, really hard they could totally dominate professional basketball...

sineruse said...

Ron, I've discussed the data about competitions in various postings that you can find through the "activity" listing on Disqus.  Or search this thread and the preceding one for the words USAMO and Putnam.

The short story is that the Asian numbers are high by all measures compared to levels in the US population, but they steadily decline at the elite levels, from age 13 to 30+.  In the recent cohorts US E.Asians outnumber US whites by ratios of 3-4 to 1 and higher at the elite levels at grade 7 and 8 ("Mathcounts" national competition, where the Asian numbers are actually suppressed by the geography based selection of 5 contestants per state);  2 to 1 in USAMO qualifiers;  3:2 in USAMO winners; somewhat less in IMO gold medals (E.Asian contestant around 30-60% less likely to get gold after qualifying for US IMO); reversed to 1:2 or lower in Putnam Top 50 (Honorable mention);  1:3 in elite Putnam scores (top 24, 10, and especially top 5);  1:5 in superelite Putnam outcomes (multiple Putnam fellows); and exactly zero in Morgan Prize (#1 national math prize for undergraduates, based mostly on independent research).     This parade continues at graduate school and professional levels.

The end result is that relative to US population, E.Asians (almost) always do better than whites, but at all elite levels, Asian credentials overpredict Asian performance.  The exception is Indian-Americans who appear to overperform everyone at all levels, including Jews, who also overperform but not to the same degree.   I don't have any explanation of the Indian effect except the obvious one about the selection into Silicon Valley through IIT.

RKU1 said...

Well, I'm not familiar with the Morgan Prize or its details, but I do take Putnam winners very, very seriously.  In fact, by the most remarkable coincidence, back in my day they tended to overlap pretty heavily with Olympiad winners.  So I decided to look for myself at the winners of the last few years, and here's what I found:

2010: 3/5 East Asian = 60%

2009: 3/5 East Asian = 60%

2008: 3/5 East Asian = 60% 

2007: 2/6 East Asian = 40%

2006: 4/5 East Asian = 80%

So the average percentage of East Asian Putnam winners over these years is...just under 60%.  Gee, where have I seen a percentage like that before, hmmm, hmmm, I can't quite place it...  I know!  It turns out to be almost *exactly* the same percentage as for the Math Olympiad winners.  In fact, 21/36 is 57.7% and 15/26 = 58.3%, which is such an astonishing coincidence for tiny data-sets that I strongly suspect someone Upstairs is trying to make a point.

Meanwhile, I see exactly one single South Asian name in the last five years (though considering that he was a three-time winner, I grant he probably wasn't a total dimwit).  Not a single likely Jewish name (thereby underscoring the horrific anti-Semitism of America's mathematics community).  And eight Northern European ones, which amounts to 31% of the total.

Okay, so East Asians are 58% of the recent Olympiad winners and Northern Europeans are 33%.  Meanwhile, East Asians are 58% of the recent Putnam winners and Northern Europeans are 31%.  So what exactly again was that fascinating "steady decline theory" which is intended to explain this huge and seemingly mysterious discrepancy?

Since I myself was neither an Olympiad nor a Putnam winner, my math skills are clearly not of the highest, and it's perfectly possible that I mispunched my hand-calculator, so people should probably check my numbers for themselves.  And although I'm not so bad with words, my verbal skills are obviously massively inferior to those of our good friend "Sineruse", who has been able to come up with brilliant, complex philosophical verbiage conclusively demonstrating that 58% is totally, totally different than...58%!

There's sometimes been some HBD speculation floating around that whereas Mathematical and Spatial g-skills strongly correlate with value creation, Verbal ability is most strongly correlated with expertise at "value transference"...

sineruse said...

Ron, you miscounted both the Asian and non-Asian numbers.  Your supposed V skills did not include noticing the words "US" East Asian, "US" whites, East Asian-"American" and many explicit statements to that effect.

The Putnam is dominated by foreign students (i.e., multiple IMO gold medal winners from China, Romania, Russia, etc) these days.  I think the non-Asian imports outnumber the Asians but don't remember for sure, and it doesn't matter for purposes of the US analysis.   The "decline" pattern, as I have said many times, pertains to students who *attended US high schools* and are therefore *domestic admissions path* candidates for elite admission, which is the pool for which anti-Asian discrimination is asserted.  If you would like to separate the non-US students, we can compare the Asian and non-Asian numbers in the US pool we can see whether I counted correctly or not.   I posted counts for several years of the Putnam honorable mention and above, giving US and non-US origins of the contestants, in the CC boards. 

highly_adequate said...

This is a really confused argument.

Look, the calculation I made that the verbal IQ of eminent scientists was 1.73 SD above (actually, I realized that was incorrect -- it should have been 1.67 -- I entered the wrong number into the calculator) that of the graduate students may have to be reduced to 1.56 if the SD of IQ Roe used was 16, not 15. But how do possibly get down to 1 SD from that? I should think that the people at ETS who came up with the test for verbal intelligence knew well enough how to calculate a SD normed on the test results they had in hand; I don't see how one gets around the at least approximate figure of 1.56 SD.

And you apply all these supposed "corrections" for the eminent scientists. Honestly, you're just making stuff up there. Perhaps they have some advantage because of their age and position, but you have no reason to believe it adds up to the sorts of values you attribute to it. Among other things, vocabulary is only part of verbal intelligence as usually measured -- especially as measured by the SAT or GRE, the presumed prototype for the sorts of questions that might have appeared on this test. I've certainly never heard anything to support the idea that general verbal intelligence increases significantly beyond the mid twenties -- presumably the age of the graduate students in question.

In general, again, you're just making things up here to advance your preconceived point of view.

sineruse said...

FYI, here is the list of US high students who went on to win the Putnam competition since 1995 with Indian, E.Asian and white names separated.   In addition to the much lower Asian/white ratios than in the USAMO, you can see that the Asian representation declines further for the multiple winners.   Let us know if you have further objections to the idea of a decline pattern.   The causal mechanisms are essentially the same for underperformance by more mundane measures such as grades, graduate school results and fellowships, as long as outcomes are assessed relative to credentials.  In that respect the Putnam is just one illustration of the pattern where "discrimination" explanations are not available.

US White/Asian ratio of Putnam winners:

won 1+ times: 22 to 8 = 2.66 to 1

won 2+ times: 10 to 2= 5 to 1

won 3+ times:  5 to 1= 5 to 1

won 4+ times:  3 to 0 = ?? to 1

Kiran S. Kedlaya (Harvard)       1993, 1994, 1995
Davesh Maulik (Harvard)       1999
Arnav Tripathy (Harvard)       2007, 2008, 2009

Lenhard L. Ng (Harvard)       1993, 1994, 1995 
Stephen S. Wang (Harvard)       1996
George Lee, Jr. (Harvard)       2001
Ricky I. Liu (Harvard)       2005
Tiankai Liu (Harvard)       2005, 2006
Po-Ru Loh (Caltech)       2006
Alex (Lin) Zhai (Harvard)       2010
Bohua Zhan (MIT)       2008

Jeremy L. Bem (Cornell)       1994, 1996
Sergey V. Levin (Harvard)       1995
Robert D. Kleinberg (Cornell)       1996
Daniel K. Schepler (Wash. U StL)       1996, 1997 
Patrick K. Corn (Harvard)       1997
Mike L. Develin (Harvard)       1997, 1998
Ari M. Turner (Princeton)       1998
Nathan G. Curtis (Duke)       1998
Kevin D. Lacker (Duke)       1998, 2001
Christopher C. Mihelich (Harvard)       1999
Alexander B. Schwartz (Harvard)       2000, 2002
Gabriel D. Carroll (Harvard)       2000, 2001, 2002, 2003
Reid W. Barton (MIT)       2001, 2002, 2003, 2004
Melanie E. Wood (Duke)       2002
Daniel M. Kane (MIT)       2003, 2004, 2005, 2006
Aaron C. Pixton (Princeton)       2004, 2005, 2007
Oleg Golberg (MIT)       2005
Matthew M. Ince (MIT)       2005
Jason C. Bland (Caltech)       2007
Brian R. Lawrence (Caltech)       2007, 2008, 2010  (did not compete 2009)
William Johnson (U of Washington)       2009
Colin P. Sandon (MIT)       2010

Iamexpert said...

When you say that the scientists scored 1.56 SD above the PhDs, that's kind of ambiguous.

Here's what we know: on the ETS test, the PhDs had a mean score of A and an SD of B. On the Binet, the PhDs had a mean score of 141 with an SD of C.

The scientists scored D * B + A on the ETS test and thus were assigned an IQ of D * C + 141 = 166.

In other words, the mean ETS score of the PhDs was statistically equated to a ratio IQ of 141 and the mean ETS score of the scientists was statistically equated to a mean ratio IQ of 166.

If we assume ratio IQ's have an SD of 16, that's a 1.56 SD difference, but SDs become meaningless since ratio IQ's seem to depart from the Gaussian curve at such extremes. The only way to make them meaningful is to force them to fit the normal curve.

A simple method for doing so is suggested on a web page called "the statistical distribution of childhood ratio IQs" by John scoville. He assumes ratio IQ's follow a log normal distribution and thus forces them to fit the Gausian distribution. When this is done, the scientists go from having an IQ of +4.13 sigma (above the U.S. Mean) to about +3.33 sigma. The PhDs go from +2.56 sigma to about +2.33 sigma. That's how the gap between scientists and PhDs drops to only 1 SD. But you really do have to correct for age because on typical verbal tests (certainly the verbal tests on the WAIS-IV so presumably this test too) middle aged adults score 0.33 SD above twenty somethings, so about a third of the 1 SD edge the middle aged scientists had over the twenty something PhDs must be dismissed as age.

So the actual gap is only 0.66 SD or 10 verbal IQ points. Thats still pretty significant.

darklayersify said...

Well, I was merely responding to the idea that math ability is "
simply a function of verbal ( algebra) and spatial (geometry)" in your own words. I agree that math ability involves several cognitive skills or draws upon them and it's related to intelligence. I would hesitate to conflate the two as readily as you do, but I strongly agree that "g" matters a lot for math. 

I was questioning such a straightforward formulation of mathematical ability, as spatial and verbal. Spatial ability does predict STEM success, but that study also found that most of the top 1% in spatial ability is not in the top 1% of math ability. Although there is noise in the data, I think that (a) the fact that 70% of adolescents in the top 1% of spatial ability are not in the top 1% of math ability, (b) spatial ability was not a significant predictor in one the papers cited, and (c) the statistical analysis have enough signal collectively to cast doubt upon the idea that math ability is *simply* a function and verbal and spatial ability, although both contribute to mathematical ability. Other factors might matter too. For instance, Jonathan Wai has written that "Spatial ability can be best defined as the ability to “generate, retain, retrieve, and transform well-structured visual images.” There might be analytical and perceptual abilities  that are not creating, recalling, or transforming visual images or verbal ability per se that matter for mathematical abilities.  Spatial abilities are certainly extremely important in a lot of mathematical domains and open-ended mathematical and scientific problems. But I don't know that characterizing quantitative or mathematical ability as being a function of verbal and spatial ability, or strongly and simply emphasizing "g" is the best way to characterize the cognitive underpinning of mathematical understanding or quantitative aptitude. 

A mere note: sadly studies find that spatially talented youth are more likely than mathematically talented young people to underachieve with regards to education and employment. 

RKU1 said...

Well, it's certainly possible that things might have changed dramatically from back in my day when just a negligible fraction of my (quite numerous) East Asian undergrad classmates were non-US residents, and I indeed followed that mental assumption when examining the list of Putnam winners.  In any event, I have much better things to do with my time that trace down the precise autobiographical details of all the various East Asian, South Asian, and Northern European names on the list.  I suppose it's possible that a good fraction of China's absolutely best H.S. math students have decided to leave that country and attend American colleges, but I've never heard that claim floating around, and would appreciate seeing some evidence of that.  

And thanks for providing such an exhaustive list of multiple-Putnam winners for the last 15-odd years, but I don't see that it particularly strengthens your case.  The bottom line is what I determined in just five minutes with Google.  Over the last few years, roughly 60% of all U.S. Olympiad have had East Asian names and 30% have had Northern European ones.  And those percentages are *precisely* the same for Putnam winners.

Interestingly enough, your historical data seems to suggest something very different to my naive eyes.  I'd only bothered looking at the last half-dozen Putnam lists, but in now going back a little farther, I notice the East Asian percentages were considerably lower, probably averaging more like 10-20% of the total for the previous decade, which heavily impacts your 1994-2011 averages.  So the most striking pattern is that the last few years have witnessed a *gigantic* rise in East Asian performance on the Putnam, shifting from a mere presence to near total domination.  

I don't have a clue what the explanation might be.  Maybe Asians are getting smarter or whites are getting lazier or some complex mix thereof.  Or perhaps, just as you claim, the Chinese government has recently decided to start sending all their smartest students off to American colleges.  Since you seem to be some sort of fanatical researcher in this obscure subject-area, maybe you can figure out the answer and get back to the rest of us.

sineruse said...

>So the most striking pattern is that the last few years have witnessed a *gigantic* rise in East Asian
>performance on the Putnam, shifting from a mere presence to near total domination.

There has been a gigantic increase in US E.Asian *representation* in national math competitions, starting from elementary school.  It propagates to high school and college and graduate school, down the pipeline -- with considerable losses at each step (i.e., underperformance, even accounting for natural forms of attrition).   Some reasons:

1. More Asians (per capita) of high school age, many of them scientists' children. Recently there was an uptick of kids of the Tiananmen green card recipients.

2. Qualification became relatively easier. Number of USAMO qualifiers is 500 instead of 100.

3. USAMO and IMO became more trainable (and consequently more difficult in absolute terms). Problem books, databases and contest-focused math web sites have proliferated since 2000.

3b. As a result, starting early became more important for those wanting to reach the high echelons. Since Asians on average start their childrens' math education earlier and/or more intensively, there are relatively more Asians who have enough time to reach the ever-increasing training levels needed to reach USAMO and beyond.

4. Word of the association between olympiad success and elite college admission has reached the Chinese and Korean immigrant communities.

4b. As a result, dozens of math olympiad schools (some conduct classes in Chinese or Korean) and math olympiad camps (published attendance figure from the first year of a camp run by ex-US IMO coach: 80+ percent Asian) have opened in the past 5-8 years. A few are oriented toward Russian or Indian immigrants but the vast majority of olympiad schools are started by East Asian immigrants and pitched to East Asian parents.

4c. At the high end one could count Philips Exeter as a special program of this kind, producing one or more E.Asian winners of the USAMO or IMOeach year. Of the 20+ USAMO qualifications from Exeter in the past two years, 100 percent are Asian. The US IMO coach, Zuming Feng, is an instructor at Exeter, and apparently Asian parents are using Exeter as an olympiad school at a much higher rate than whites.

Equivalent factors don't operate after elementary and high school, and as a result one would expect some deterioration of the Asian percentages as time and selectivity levels increase along the academic "pipeline".  Which is exactly what happens.

Iamexpert said...

When you say that the scientists scored 1.56 SD above the PhDs, that's kind of ambiguous.

Here's what we know: on the ETS test, the PhDs had a mean score of A and an SD of B. On the Binet, the PhDs had a mean score of 141 with an SD of C.

The scientists scored D * B + A on the ETS test and thus were assigned an IQ of D * C + 141 = 166.

In other words, the mean ETS score of the PhDs was statistically equated to a ratio IQ of 141 and the mean ETS score of the scientists was statistically equated to a mean ratio IQ of 166.

If we assume ratio IQ's have an SD of 16, that's a 1.56 SD difference, but SDs become meaningless since ratio IQ's seem to depart from the Gaussian curve at such extremes. The only way to make them meaningful is to force them to fit the normal curve.

A simple method for doing so is suggested here:

Iamexpert said...

The web page is called "the statistical distribution of childhood ratio IQ scores" by John scoville. Bottom line is when you convert the ratio IQ's to deviation IQ's, the gap between the scientists and the PhDs becomes only 1 SD. I would reduce it further to 0.66 SD because on virtually every verbal subtest on the wechsler scales, middle aged people score 0.33 SD higher than twenty somethings, so the scientists almost certainly had their verbal scores inflated by advanced age.

sineruse said...

>I'd only bothered looking at the last half-dozen Putnam lists, but in
now going back a little farther,
> I notice the East Asian percentages
were considerably lower, probably averaging more like 10-20% of the
> for the previous decade, which heavily impacts your 1994-2011
>... the last few years have
witnessed a *gigantic* rise in East Asian performance on the Putnam,

re: the selection of time frame and the change in E.Asian numbers that Ron observed. 

The number of East Asian-*American* names on the high score list (Honorable Mention and above) has stayed about the same.  The huge growth in E.A names on the Putnam high score list has come overwhelmingly from foreign students and an expansion of the list by 20-30 students.

The relative US East Asian representation (the US EA/white ratio) did escalate, as one would expect from both the national demographics and the influx of Chinese PhD students starting in the late 1980's that begat the recent generation of USAMO competitors.  But in the upper ranks of the Putnam competition, most of the growth in E. Asian proportion comes from offshore, through increases in the importation of foreign students.  For instance, in the 2009 Putnam competition, the top 81 students (HM+) had the following origins:

20 foreign E.Asians  (IMO/IPhO medalists from China, Korea, Taiwan, Canada, Singapore)
11 US, E.Asian-American
11 foreign non-E.Asians (medalists from Canada, Thailand, Brazil, Romania, Moldova, Albania, 1 from India, etc)
34 US, white
5 US, Indian-American

By disregarding the distinction between domestic and foreign students one would see: 31 E. Asian names (3x higher), 6 Indian (20% higher), and 42 white (24% higher), plus 2 names from Thailand.  The ratio of E.Asian to other groups is more than doubled by adding international students.  This explains Ron's percentages.

Now, of course the 3:1 ratio of white to E.Asian-American competitors in 2009 was quite different from the USAMO qualifier and winner numbers in 2006-2009, which were the years in which the same students were in high school.  From figures compiled by a China Pride warrior when I had this discussion online a couple of years ago, the percentage of USAMO qualifiers in 2006-9 with E.Asian surnames were 50%, 49%, 55% and 57%, for total of 53% of all qualifications (933/1759).  If you disaggregate Indian, Thai etc names from the non-EA list the US E.Asian to "white" ratio is closer to 60:40, but let's underestimate it at 55:45.   The transition from high school to college performance metrics suppresses the  EA/W ratio by a factor of about (34/11)/(55/45) or 3.77.     Again, the chances for an E.Asian American math star in high school to progress to being a math star in college, are about 4 times lower than for a white American with the same high school performance.   You can of course consider 4-year blocks of results but in that case one can also count multiple qualifications, and as seen in the other posting, by this metric the disparities are higher.

For the choice of time period, there was a dry spell in the US E.Asian Putnam winners from 1997 to 2004.  The most favorable recent time frame for the E.Asian numbers begins in 1995, which was the last (and only) time a USEA has become a three-time winner.   The second most favorable starts in 2005 and I gave the more detailed breakdown for that period.  So if anything the numbers I presented are probably on the high side for the A/W ratio reductions.

Matthew Carnegie said...

Of course, an interesting angle with talking about visual thinkers is, I'm not sure whether we think in a natural spoken language at all.

The dominant hypothesis about our thoughts appears to be the "language of thought" hypothesis - that we don't speak in a natural spoken language, nor that our thoughts lack any language structure (they're just sort of visual or whatever), but that we have a "language of thought".

This being the case, I'd be unsurprised if a lot of "visual thinkers" are really more thinkers who are trying to communicate something for which natural languages are particularly ill adapted and so are more aware that they are not thinking in a natural language and might tend to extend the idea that they are not thinking in a natural language to the idea that they do not think in language at all.

Assuming people have a language of thought, it would seem like you could get a high functional verbal ability (ability to express concepts and form logical arguments in language) either through a) really high quality translation of an unexceptional language of thought into natural language or b) a language of thought that is just so high quality it shines through a not particularly refined translation mechanism.

I'd bet the b) people would think of themselves as visual thinkers though, while the a) people wouldn't.

Reactionary_Konkvistador said...

"What is good for Harvard may not necessarily be what is good for the USA."

I see little reason to mention this, because even if it is true, one
could never get the people of the USA to ever do anything about it, considering US general opinion has for over a century and a half consistently drifted in the direction of Ivy League opinion of a few decades ago, due to legislative measures, intellectual fashion trickling down the class divides, the Ivy's close connection to high brow mass media (think NYT) as well as plain old influence through mass education (which shifts each generation a bit closer to the values and biases held by those controlling education)... Now Ivy League opinion differs over time, it is however I think a safe bet that Ivy League opinion will always hold Ivy League interests at heart.

sddasasd said...

As old as this discussion is, and there being more than I can reply to, there's a few remarks here and there I noticed that went unremarked on that I'll take a show at:

"What is interesting about the Indian example is how it shows that IQ
averages for populations dont really mean all that much beyond a certain
point. Clearly the top level of Indians punches well above their
weight. Its the same thing for France. I think they have an average IQ
of 94 or so, yet no European country has contributed more to Western
civilization. Clearly the top layer of talent in France is somehow in
some way more talented than the top layer of talent, in say, oh I dont know, China,
where the average IQ is higher. Another good example is Israel, with an
IQ of also about 94, and I have read that even the Asheknazim in Israel
have an IQ of a very modest 103. Yet their technological prowess goes
well beyond what these figures suggest, so clearly the top layer of
talent is somehow in some way, actually more talented than the top layer in countries with a higher average IQ (like, say, South Korea).

Obviously IQ tests have some level of statistical value, but it is interesting to point out in how
many ways IQ tests convey only limited information about the real world
and fail to predict the realities of different populations and how they
perform. There are things we dont understand yet, which is why
real world performance is ALWAYS MORE SIGNIFICANT than peformance on a
TEST that we artifically devised ourselves based on theories and that we
dont fully understand the implications of yet. Unfortunately this does not paint the best picture of Asians."

You really shoot yourself in the foot when making this point by claiming something as stupid France having an IQ of 94- where did you get this figure? I'm going to guess it was either something you misread or one of Lynn's throwaway figures that you accepted uncritically. I don't think France's IQ is even that low after all of the immigration from africa they've seen, so it's more even more absurd they had an IQ that low in the past. Similar could be said of Israel, especially when you consider Lynn's bizarre framings of middle eastern IQ, and how he believes asiatic jews have IQ's in the low 90's or less.

sddasasd said...

God, I can't believe nobody ever commented on this histrionic, projecting piece of garbage. It really hits home how you uphold this as worse than the worst white nationalist websites, which would include, I imagine, Stormfront, which is filled with every sociopath, social outcast, and mental case you can imagine. This is basically your post:

"What's wrong with you east asians, why are you SO dishonest and intellectually insecure? I'm being dead serious when I've never seen anyone argue this badly for such an obviously untenable position! Why do you care so much that we're basically arguing you're largely a bunch of robotic, amoral study drones who get by on cramming, cheating and gaming the system?!?! This is the most stubborn, tenacious, and desperate racial advocacy I've ever seen- EVEN STORMFRONT IS BETTER THAN YOU GUYS!!!"

Get over yourself, you pathetic sperg.

sddasasd said...

So in reading many of these comments, I have to ask, how many of these studies cited on college exams and such differentiated between east asians and southeast asians?

Blog Archive