Steve Hsu writes:
I'll be giving a talk at Google tomorrow (Thursday August 18) at 5 pm. The slides are here. The video will probably be available on Google's TechTalk channel on YouTube.
The Cognitive Genomics Lab at BGI is using this talk to kick off our drive for US participants in our intelligence GWAS. More information at www.cog-genomics.org, including automatic qualifying standards for the study, which are set just above +3 SD. Participants will receive free genotyping and help with interpreting the results. (The functional part of the site should be live after August 18.)
Title: Genetics and Intelligence
Abstract: How do genes affect cognitive ability? I begin with a brief review of psychometric measurements of intelligence, introducing the idea of a "general factor" or IQ score. The main results concern the stability, validity (predictive power), and heritability of adult IQ. Next, I discuss ongoing Genome Wide Association Studies which investigate the genetic basis of intelligence. Due mainly to the rapidly decreasing cost of sequencing, it is likely that within the next 5-10 years we will identify genes which account for a significant fraction of total IQ variation.
We are currently seeking volunteers for a study of high cognitive ability. Participants will receive free genotyping.
Posts
57 comments:
Damn. 800M 760V minimum. I always knew I wasn't actually that smart, but this confirms it! Or is Steve Hsu just biased against us High V Low M folks?
Lol if I've suffered no cognitive decline I barely make the cutoff.
Judging by their "automatic" qualifications, their concept of intelligence is heavily biased toward mathematical ability. Does that really reflect the full breadth of intelligence?
I just barely qualify on the pre-recentered SAT. I guess they'll want transcripts from ... way back. Does anyone know whether this lab types different/more SNPs than 23andme?
You can combine the files to have more data.
Well Sailer, many of your readers could volunteer for the study.
Oh wait, they're looking for people +3SD in IQ, not -3 in IQ. Never mind, I misread.
>>Individuals with high IQ needed.
In that case, I can help with a control group!
My phenotype's all burnt out by age, booze, stress and modern idiocy. I swear there's a V8 genotype in there somewhere. But you'll need to take my word for it.
Gilbert Pinfold.
I'll be the first to out myself as not having standardized test scores high enough to participate.
For those who don't know, BGI is the Beijing Genomics Institute. The Chinese political environment is more open to this sort of science than the Western political environment. That's something to keep in mind when the MSM bashes China for lack of free speech or when it bashes Republicans for trying to block stem cell research.
I will be participating. This study has real promise to transform the situation. Once basic results are "out of the bag" it should raise enthusiasm for more work in China, and hopefully reduce the taboo around similar work in the U.S.
"their concept of intelligence is heavily biased toward mathematical ability. Does that really reflect the full breadth of intelligence?"
Wow, you completely misunderstand how correlation works.
So, in your case, yeah.
"BGI (formerly known as Beijing Genomics Institute) [has] the mission of supporting the development of science and technology..."
I'd like to participate, but a little voice in my head suggests that helping the Beijing Genomics Institute understand IQ in 2011 is a bit like helping the Kaiser Wilhelm Institute understand nuclear binding energies in 1936.
-Jack
Why isn't 740v/730m pre-recentering-and-with-a-tremendous-hangover-at-the-time okay? Something here is completely fucked-up. In my opinion.
I'll follow TGGP's lead. I'm a dolt according to this criteria, but then again my score on a standard IQ test, as well as my Verbal, puts me his supposed target range. Plus I know that with fairly lackadaisical practice sessions over two weeks, I radically increased my Math score-- though still not enough for this study. I think that the Math is more teachable than Verbal -- though maybe that is sour grapes.
Wow, you completely misunderstand how correlation works.
No, I'm assuming that a group of Computer Science PhD's don't necessarily capture the full range of genetic factors that in various combinations can result in high intelligence.
Hsu's automatic selection criteria are really strange:
Very selective on tests, super selective on math competition but completely lax on degree criteria: "PhD from top 5 ranked US program in physics, math, EE, theoretical computer science". The bulk of these folks will be +2SD only. And besides, how is top 5 rank defined? The laughable US News and World Report? - In EE that disqualifies Caltech and qualifies Georgia Tech. Does anyone have any doubts about IQ comparison between an average PhD from Caltech and that from Georgia tech?
I'll be the first to out myself as not having standardized test scores high enough to participate.
Neither do I, though my IQ is in the range he's looking for. Why not just accept IQ tests?
Hsu's automatic selection criteria are really strange
I second that opinion.
"PhD from top 5 ranked US program in physics, math, EE, theoretical computer science". The bulk of these folks will be +2SD only.
I highly doubt that the bulk of people at elite PhD programs in math/physics/comp sci at places like Princeton, Caltech, etc will be +2SD only. Your average undergraduate at a lower tier Ivy League is probably around +2 SD.
If you're interested in qualifying for the study read the comment thread on Hsu's blog. There are some specific answers there to questions raised here.
* Automatic just means automatic. You can still qualify with lower or different scores (e.g. LSAT or IQ test) but you won't know it until after you submit your information.
* In earlier posts Hsu has said that their Chinese sample is obtained via math, physics and informatics competitions in China. They may be trying for some kind of consistency in psychometric profiles, hence the math emphasis in their cutoffs.
* People admitted to *mediocre* PhD programs in the fields they list (physics, math, EE, theoretical CS) are about +2 SD. At the top 5 places the average is significantly higher and not that different from the GRE cutoffs they list.
- The "automatic selection criteria" are not exhaustive (especially when it comes to graduate degrees). There is a space to describe qualifications you have that are not explicitly included in the web form. If your scores come close to the automatic qualification cutoffs, your survey will be reviewed.
- "IQ tests" are actually less well-understood than the SAT/GRE, so it was decided not to ask everyone for their scores. However, we will note them; they just won't *automatically* qualify you.
- The formal research proposal notes that this study can be accurately framed as one of "mathematical ability". You can see this as the most practical way to get at the broader phenotype of "intelligence", or you can see this as fundamentally a study of something narrower. Neither interpretation is really wrong. We're not stopping anyone else from performing a more verbal-centric GWAS.
Damn. 800M 760V minimum. I always knew I wasn't actually that smart, but this confirms it! Or is Steve Hsu just biased against us High V Low M folks?
I believe there are a lot more high math scores than high verbal scores. Many more people score perfect on the math section than the verbal section every year.
"I believe there are a lot more high math scores than high verbal scores. Many more people score perfect on the math section than the verbal section every year."
Amongst white test takers, an 800 on either the M or the V is roughly at the 99th percentile amongst SAT test takers. The higher number of perfect SAT-M scores are due to Asian Americans.
See here.
http://professionals.collegeboard.com/profdownload/sat-percentile-ranks-by-gender-ethnicity-2010.pdf
For white Americans, 700, 750 and 800 on the M and V are identical in terms of percentiles.
700 for either was 94th percentile. 750 was 98th percentile. 800 was 99th percentile.
In other words, at least for white Americans, the SAT-M isn't an "easier" test than the SAT-V, since scoring a 700, 750, or 800 on either section puts you in the exact same percentile amongst white test takers.
According to Detterman and Frey's conversion of SAT score to IQ:
http://www.psychologicalscience.org/media/releases/2004/pr040329.cfm
an SAT score of 800M/760V converts to an IQ score of 128.521.
an SAT score of 800M/760V converts to an IQ score of 128.521.
Is it wrong of me to say "That's not particularity high"?
"No, I'm assuming that a group of Computer Science PhD's don't necessarily capture the full range of genetic factors that in various combinations can result in high intelligence."
Ah, so you didn't read the SAT qualifications. Or the GRE ones. Or ACT. Or the physics, math, and EE PhD path to qualification.
Well.
OK then.
"an SAT score of 800M/760V converts to an IQ score of 128.521.
"
Their conversion is highly flawed. Very few test takers score a 1560 combined on the SAT. It must be at least 99.5th percentile or higher amongst test takers, and this is already a self-selected group amongst the general population.
***Oh wait, they're looking for people +3SD in IQ, not -3 in IQ. Never mind, I misread.***
Steve,
Why don't you just ban Yan Shen? His repetitive trolling of your site is tedious and pathetic.
Their conversion is highly flawed. Very few test takers score a 1560 combined on the SAT. It must be at least 99.5th percentile or higher amongst test takers, and this is already a self-selected group amongst the general population.
That doesn't mean it's flawed.
Is it wrong of me to say "That's not particularity high"?
It might not seem that high, but it's about 2 standard deviations out in the white population. About 2.4% of the white population are at least that high.
The global mean IQ in 2011 is about 88.54. So an IQ of 128.521 is about 2 and 2/3 SDs out in the world population. About .3793% of the world population are at least that high.
Since East Asians have more of an advantage over other races in math-type intelligence than in verbal-type intelligence, using a group selected especially for the former as a model of high-g factor will make the Asians look even smarter, should these scientists examine how the smart genes they will discover are distributed among the races. Maybe the Asian researchers will confirm scientifically the racial hierarchy that Whites have been forbidden to consider, but with just the right technical bias to place themselves definitively rather than disputably at the top.
Dang, I am below the cut-offs by inches on every single criteria! 20 points below on the combined SAT, 1-3 points on ACT (don't remember my exact score anymore), and I have a MSc, not PhD, from a Top 5 school in one of their hard science fields.
It's a bit of a blow to the ego, but also delightful in how accurately these tests have got me pegged in terms of intellectual development, academic achievements, and life outcomes (All hail the prognostative powers of g!) Basically I've always punched a little above my weight when it comes to the hard sciences and math. In grade school I took one of those placement tests and got dumped in average track math as a result. By high school I figured I needed to be in honors to get into a good college, and so took geometry in summer school (geometry was the 1st year honors course, algebra the 1st year regular course) so that I could jump up to honors math my next year. I wouldn't say I at all struggled in the harder track, but as I progressed through my education it was obvious this was not my forte. Part of it was also that I was generally lazy (the idea of reading academic papers in my subject area NOT assigned as homework would have struck me as absurd), but even with more drive I think I would have barely scraped through a PhD program.
I never took an IQ test, but I always suspected it was in the 140's (IQ = pre-1994 SAT/10), and with the cut-off for this study at 3SD (145), this pretty much confirms it. Some popular IQ sites call this near-genius, but that is just an abuse of the term. Genius is Einstein/Feynman/Witten smart. Amy Chua's dad (someone pegged him at +5SD) is probably near-genius. +3SD, though, is only about average for a top-flight hard sciences graduate program. Being relatively humble and realistic about my own abilities, by junior high I started to suspect I maybe wasn't the smartest guy in the room. By high school I knew it, and by graduate school I knew I was only about average- better than maybe half, but worse than definitely 50%, and not just by a few degrees, but by entire orders when it came to the top-flight PhD candidates in my dept.
Steve once described getting lost in the details when it comes to really hard math or logic problems and that's been exactly my experience as well. I can juggle maybe 3 dimensions in my brain at once, but any more and my mind just starts to wander and I begin day-dreaming. I think for true near-geniuses (+4-5SD), the visuospatial centers of the brain are just so much more developed they can work with several more dimensions (or indepdent variables, etc.) at once than the rest of us.
So anyway, +3SD is good, but there will be dozens, if not hundreds or thousands, better than you when you get to a truly elite instutation (Caltech, Google, NASA, NSA). Someone should tell that to our smartest-guy-in-the-room commander-in-chief (+2SD max)!
It's a bit of a blow to the ego, but also delightful in how accurately these tests have got me pegged in terms of intellectual development, academic achievements, and life outcomes (All hail the prognostative powers of g!)
That's a shortcoming in this study. It would be interesting to discover the difference between the IQ140 guy with a physics Phd and the IQ140 guy stuck in some low skill administrative job. Or even the IQ140 guy on the unemployment line.
This is more a study of people with specific achievements than a study of people with high IQ. In other words, it's a study of a specific subset of the high IQ population. Which is why it tells you nothing about the prognostative powers of g.
It might not seem that high, but it's about 2 standard deviations out in the white population. About 2.4% of the white population are at least that high.
Yeah, but that's sort of a lot - two or three people out of every hundred.
Here's a brain-teaser at about the cut-off level of the study:
You have 10 bags full of coins; 9 of the bags contain only real coins, but 1 of the bags contains all counterfeit coins. A counterfeit coin is known to weigh 0.1 gram more than a real coin (say, it's mostly lead). How can you determine which bag holds the counterfeits using a scale and only ONE weighing?
See if you can get this in 30 minute or less w/o looking up the answer on the internet.
The problem is that I suspect the fine print says that the data, metadata and resulting findings are the propertary property of the research institute, Chinese funding agency or whatever for-profit spinout results.
For China, you probably don't even need to put such things in fine print anyway.
I also suspect any scientific findings will be shared only on a very limited basis to outside researchers. Anything of commercial value will end up protected behind very high intellectual property barriers.
Hint to avoid confusion on brain-teaser: it's not a scales that tells you which side is heavier, it's a scale that gives you a readout.
This study appeals to our vanity, and I'm vain enough to say I qualify. But I'm not at all sure I want to help the Chinese with genetic research, even if some of the results get into the intellectual public domain. In fact, I'm not even sure that I want to help American genetic science, or genetic science in general. Knowledge outstripping wisdom, and all that.
The problem is that I suspect the fine print says that the data, metadata and resulting findings are the propertary property of the research institute, Chinese funding agency or whatever for-profit spinout results.
For China, you probably don't even need to put such things in fine print anyway.
I also suspect any scientific findings will be shared only on a very limited basis to outside researchers. Anything of commercial value will end up protected behind very high intellectual property barriers.
I understand your wariness in light of China's mercantilist policies, but the organization behind this project, BGI, has an outstanding track record of cooperation with Western researchers.
greenrivervalleyman:
Here's the best I've ever seen (though I don't claim to've seen 'em all):
On 100 paper slips, write numbers of any size--in the entire range from negative infinity to infinity. They can vary: whole numbers, fractions, decimal fractions--all OK.
Put 'em all in a hat or box.
My job is to remove one at a time until I come upon one I believe to be the largest (or the smallest) number of the lot. If I, unknowingly, pass the number, I lose. If I misidentify one as the largest, I lose.
What odds might I need in order for this to be an even-steven proposition?
I solved the brainteaser in about 4 minutes. Incidentally, I have a B.S. in math from a top 50 public U and I am an underachiever.
You have 10 bags full of coins; 9 of the bags contain only real coins, but 1 of the bags contains all counterfeit coins. A counterfeit coin is known to weigh 0.1 gram more than a real coin (say, it's mostly lead). How can you determine which bag holds the counterfeits using a scale and only ONE weighing?
I knew this one, but I've seen similar brain-teasers in the past. To some extent intelligence can be learned, it's about looking at problems in certain ways. (To some extent, of course)
I also suspect any scientific findings will be shared only on a very limited basis to outside researchers. Anything of commercial value will end up protected behind very high intellectual property barriers.
I understand your wariness in light of China's mercantilist policies, but the organization behind this project, BGI, has an outstanding track record of cooperation with Western researchers.
Most BGI projects listed are basic science projects. Mapping the genetic basis for human intelligence is not only the ultimate commercial killer app, but the ultimate military weapon.
Even if BGI or the Chinese government itself does not immediately patent commercially valuable results, individual scientists and their VC backers will inevitably spin out.
These spliter scientists will start VC-backed startups that will privatize the enormous potential profits by incremental intellectual property legerdemain. These incremental IP patents will firewall off others from commercializaing whatever data that is shared publically.
Quite a clever startup stategy. In this case the sustained unfair competitive advantage of such startups is exploiting advanced Western genomic science that is limited to mostly arcane non-human applications or human applications of specific medical diseases.
A Chinese startup can uniquely exploit this arbitrage opportunity. Absorb and build on Western basic genomic science that by Western PC limitations, cannot advance into the most profitable application of human genomic enhancement.
Collaborate or simply have Westerners continue to do all the basic research and tool development while Chinese startups are uniquely poised to reap the most profitable commercial applications for humans.
Does having the intelligence to see through the moral and commercial problems with project qualify one to participate?
Ha, I was going to ask whether it was a scale -- actually I was going to state and assumption it was a scale. Also assuming the bags all have the same number of coins.
Finally, I think the answer also depends somewhat on the ambiguity in what you mean by 'weighing' -- but assuming this means just putting stuff *on* the scale, that answer is
1. Put all the bags on the scale, work out the average weight for each bag W/10, and then remove bags one by one, recording the difference in 'before' and 'after' weights each time, until you get to the bag where the differential weight between 'before' and 'after' excedes the average weight.
The SAT and Achievement Tests are normed at 500.
That's why you see so many scores close to 500.
600 = 1+ SD
700 = 2+ SD
800 = 3+ SD
Get it?
This results in no child getting below a 200; which to an idiot seems like twice as good as a 100 -- the kind of score they never get.
There is no 900 because statistics break down with such a small pool.
-----
The global IQ norms down in the low eighties. It is taboo to discuss it. The left end of the gene pool is the un-sampled end. It is vast.
Poor nutrition and cousin marriage = inbred idiots.
Afghanistan has millions of them, and they are still living by pre-Neolithic rule-sets.
Recent DNA discoveries have established that Neanderthal man hybridized with Homo sapiens at least 40,000 years ago. Consequently, 1-4% of non-African genes are Neanderthal!
Obviously, there are non-unique genes that were inherited by way of Neanderthals -- so their DNA is very much alive in us.
This crossbreeding is almost certainly the explanation for the explosion of new races: Paleo-Americans ( Maya/Sioux/Navaho/Inco..) Caucasians ( all of them ) Orientals ( all of them ) Polynesians ( all of them ) and a huge slug of modern South Asians )
Neanderthals lived for 300,000 years way up North -- so they are the reason we got so white in a hurry. Dark hair would be a staggering handicap up north -- one needs to be white to hunt in the snow. ( And Vitamin D )
The idea that Neanderthals were dummies is entirely contradicted by modern IQ scores. They are highest in non-African bloodlines. Neanderthals had notably big cranial volumes and elongated heads. Funny, my hat size is 4+ SD.
As for genius -- don't confuse it with Godhood.
130 = Near Genius
140 = Genius
Your Newtons & Feynmans display amazing free-ranging intellects able to see / remember / pull together facts and concepts to create profound insights.
Our IQ methodology can't capture that metric. Feynman's actual IQ test placed him below 130 !
Obviously that test was a failure -- by its constructors.
-----
Helping the Red Chinese to create super-men is not my idea of a wise course of action.
@stari_momak ~ That's not a solution because you are only allowed one reading. Also, please don't spoil it for others by publishing a correct solution.
greenrivervalleyman & Gene Berman ~ You fellas are making me feel smart. I solved the first one in 4 minutes (although the answer occurred to me almost immediately, I just took a few minutes to verify it). I solved the second on in a few seconds. All you need is some basic knowledge of set theory and probability.
And to think, I scored 1150 on the (old) SAT. I always thought that score was low, and what surprised me was I did better on the verbal than the math. I suppose that is because I tend to make minor computational errors, but I have a high capacity for abstract thinking. I would have taken it again but, being the underachiever that I am, I didn't care.
As for brainteasers, I prefer the more abstract kind, like this:
A monk climbs a mountain. He starts at 8 AM and reaches the summit at noon. He spends the night on the summit. The next morning, he leaves the summit at 8 AM and descends by the same route that he used the day before, reaching the bottom at noon. Prove that there is a time between 8 AM and noon at which the monk was at exactly the same spot on the mountain on both days.
You didn't say 'one reading', you said one weighing. To me, weighing is putting something on a scale. Certainly the 'correct' solution is more work, and thus less elegant, than mine.
stari_momak: The correct solution is less work than yours (requires even less math) and does not rely on each bag having the same number of coins. Also just for calibration purposes I have a measured IQ of 162 and figured it out in 10 seconds.
Good point on the fewer assumptions (same number of coins in the bag, which I also figured out).
I concede.
Jack, Loved your response. I'm happy to see that at least one 3 sigma person here isn't a moron about China. Funny how the very bright can be so stupid, no?
And yes, I'm sure that the Beijing Genomics Institute isn't keeping ANY of my personal identification information or handing it to any other organization in China. No, sir, on their honor, they are not doing that. They won't possibly cross reference it to the extensive files they already have on anyone who ever went to a top 5 grad school, or worked in a high tech firm, or a govt job with a clearance, or a defense contractor, or .. or ...
Sort of 'racy' piece on what Neanderthals brought to the human genome:
http://www.huffingtonpost.com/garret-loporto/surprising-way-your-neand_b_568455.html
Hint; first line:
> It may be our Neanderthal genes that are behind virtually all human progress.
Read it quick, I predict it'll get pulled, as it's full of racist (I say, reclaim the word!) implications.
You have 10 bags full of coins; 9 of the bags contain only real coins, but 1 of the bags contains all counterfeit coins. A counterfeit coin is known to weigh 0.1 gram more than a real coin (say, it's mostly lead). How can you determine which bag holds the counterfeits using a scale and only ONE weighing?
I knew this one, but I've seen similar brain-teasers in the past. To some extent intelligence can be learned, it's about looking at problems in certain ways. (To some extent, of course)"
Having given up, and thrown my iPad against the wall, I looked up the answer. Do I get a consolation point for understanding the solution within 30 Minutes of reading it? Seriously though, I can see how you could know this one because you'd seen it before. I'll be +3SD next time!
Gilbert P.
It's been a week since I signed in, and I still haven't received an email or any acknowledgment that my information was received.
Never mind.
The study is probably better off if it administers its own IQ test. That way the selection criterion is guaranteed uniform.
Post a Comment