Gifted Issues Discussion homepage Forums GT Research genetic basis for intelligence

Charles Murray has a few words on the subject (note that his name is linked to an article he wrote):


He also makes the point that smarter kids tend to get higher scores on any test, be it an achievement test or a pop quiz, and that any mental test has some g-loading in it. Fair enough, but this doesn't mean that the scores on current SAT are a proxy for IQ. The range of scores doesn't even go to three standard deviations above the mean, so at best even if it isa proxy for IQ scores --- which I don't think it is --- the SAT has a very low ceiling at best.

It seems to me that when Hsu is adding SDs, he's referring to SDs in X, and not trying to imply that G and E have the same SD as X. Simply replace SD with 15 IQ points, and he's saying that someone with an IQ of 160 is more likely to have such a high IQ because they were bumped up 30 points by genetics, and 30 points by environmental factors, than it is that the person was bumped up 45 points by genetics and 15 points by environmental factors (because it is so rare to be bumped up 45 points by genetics). The fact that 15 points is actually sqrt(2) standard deviations in E or G isn't important.

I think the point was to use a theoretical example to explain a consideration to keep in mind when interpreting population statistics, nothing more.


Yes, that's true. That's what is implied when the heritability is said to be less than 1, which I think everyone agrees on.

Yes, I would agree to that as well. But we are discussing Hsu's post here and I did not see this mentioned anywhere, and that is clearly what I was responding to. Perhaps it is another assumption of his that the reader would understand this, but many people who might read this will not take that into account and be misled.

[quote=DAD22I think the point was to use a theoretical example to explain a consideration to keep in mind when interpreting population statistics, nothing more.

[/quote]

Perhaps, but I see little point in theoretical examples that can never be empirically tested. If you cannot provide valid, reliable measures of these variables, then what's the point? As a social scientist (not a physicist) I have to justify my assumptions and demonstrate the validity and reliability of my measures to my colleagues in order to be taken seriously. My point here was that in addition to the critiques of the model itself presented by pps, I question the likelihood that we could ever know what G & E are anyways, much less their standard deviations. So for example, I could just as easily create a "theoretical" model that says that genetics has nothing to do with intelligence and it is completely dependent on environment, if I can make any assumptions I want and not have to justify how I measure my variables. But if there is no way to disprove it, then it is not science.

What concerns me about things like this, is that people will read it and believe it to be true, just because a scientist said it. No matter how questionable his assumptions or models are.

The post you have been discussing:

http://infoproc.blogspot.com/2008/10/regression-to-mean.html

is only a pedagogical example and not meant to be realistic. Its purpose is to illustrate that if there is a non-genetic contribution to intelligence, which is not perfectly correlated between parent and child, then regression must follow even if the genetic component is perfectly transmitted. Taking G and E to have similar effect on IQ, and similar standard deviations, was only to make the mathematics easier. Even if GxE effects are significant, or parental E is somewhat correlated with offspring E, one still expects regression from a similar line of reasoning.

You can find more detailed analysis here:

http://duende.uoregon.edu/~hsu/blogfiles/midparent_offspring.pdf

Regarding environmental effects on IQ, you may want to read up on shared vs nonshared environmental effects in the behavior genetics literature:

http://infoproc.blogspot.com/2009/11/mystery-of-nonshared-environment.html

I've been thinking about this thread.

1. Are there any studies that involve intelligence (IQ-based) beyond immediate family members? I have a eg/pg DS6. My husband and I may be gifted, but we're not eg/pg. However, both of my parents had sisters who were eg and tested around IQ 160+ in 50s. Just wondering if there's any data on extended family members, genetics, and intelligence.

2. I've seen the quantitive approach to intelligence by criticized and wondered if anyone has come across what UDL (Universal Design for Learning) has been arguing:

Psychometrics from the time of Binet asserted that a single, underlying trait called "intelligence," which varied only in quantity or amount but not in underlying nature, accounted for academic achievement. UDL is based on new brain research (Rose & Meyer, 2002) that asserts that the intelligence of individuals differs qualitatively according to how three distinct neural networks interact [recognition, strategic, and affective networks]. Virtually infinite combinations of learning preferences emerge in an individual based on variation in the way they receive information, act upon it strategically, and engage in learning activities affectively. If human intelligence varied quantitatively according to a single factor, the logical implication would be that the more of this single attribute an individual possessed, the more of the curriculum they could absorb, and, therefore, the more material they should be able to access. However, from a UDL perspective, qualitative differences in the ways in which individuals learn implies that similar qualitative differences are possible and should be available in the curriculum itself so that each learner may approach it in the manner best suited to her or his preferred learning mode.

(http://aim.cast.org/learn/historyarchive/backgroundpapers/promise_of_udl/how_udl)

If we use UDL's approach to intelligence, then you have a much wider swath of the population who are gifted but in multiple ways - which better fits Howard Gardner's theory on multiple intelligences.

I often wonder if the genes that lead to a specific ability have implications for other abilities in the next generation. Are the children of gifted artists and musicians smarter on average than would be predicted by the IQs of their parents?

IQ tests are normed on the general population, including the mentally retarded. The SAT is taken only by students applying to college, who are on average smarter than high school juniors and seniors in general (and who in turn are smarter than the population of 16- and 17-year olds, which includes drop-outs). I saw a study once showing that the average SAT score on an unselected sample of high school seniors was considerably lower (maybe 100 points for math + verbal) than the average of students who sign up to take the test.

Therefore a +3SD score among SAT test takers corresponds to a higher Z-score when compared to the entire population of 16- and 17 year-olds.

You'll have to show me that study before I can comment on it. How old is it? How many students were in the sample? Did they only test students who did NOT sign up for the SAT as part of the unselected sample? Etc.

Interesting: because of the national mania to send everyone to college, more students are taking the SAT. Yet they don't seem to be doing terribly well. For example, according to the College Board, only 43% of 1.65 million seniors who took the SAT in 2011 got scores of 1550 or higher, which it considers to be the minimum score for success in college. (Note: I find the upbeat perkiness of that depressing press release to be rather odd.)

The US Dep't of Education says that there were 14.5 million public high school students in 2011, and an estimated 1.36 million more in private schools. This is just under 4 million high school seniors. So 41% of high school seniors took the SAT in 2011. Measuring very crudely, this number would include students with IQs all the way down to 103-104 (Notes: I'm not claiming everyone with an IQ>104 took the test or that no one with an IQ <103 took it. I also reckon that some of the best students took it as 10th or 11th graders, got super-high scores the first time, and didn't take it again).

Interestingly enough, by using my crude estimation system, roughly 60% of SAT takers would have had IQs of at least 115, which is the IQ benchmark for being able to succeed in college. This number fits nicely with performance on the test (only 43% of test takers got the College Board's college-success score). So I'm reconsidering my idea that the test isn't a good rough proxy for IQ. But big disclaimer: there could be other factors at work, and of course my estimates are very crude and I could be way off here. For example, the SAT tests your knowledge of information and you can study to improve your score. At least a portion of scores probably reflect course content and study habits.

So, yes, many of the SAT takers are smarter and the average IQ is almost certainly higher. But most of them are probably in the average range of ability.

And I stand by what I said about the ceiling of the test. The score distribution at the top of the math test is quite weird:. 1,250 students scored 790, yet eleven times that number (~13,800) scored 800? Huh? I don't get that. And the strangeness continues as you go down ~720-730.

The score distributions at the top were similar but less extreme for reading and writing.

(That same strange effect happens in the scores below but nearish to the average for all three tests.)

I'm not sure about lower down, but the weirdness at the very top looks like a granularity effect to me - if you miss one question, you still get an 800, but if you miss two questions, you get a 780, not a 790. There are a relatively small number of versions of the test, each of which has a slightly different score distribution associated with it. (When I took the SAT, I remember there was at least one version of the math test where it was impossible to get an 800 - everything right was still a 790.)