On the Stability of IQ

On a recent post about genes and IQ, commenter JL objected to my statement that the brain is a uniquely plastic organ, and that humanity’s astoundingly extended childhood appears to exist to maximize that plasticity. More specifically, JL objected to my conclusion that given the brains plasticity, we would expect intelligence (and all the other factors included in IQ testing) be incredibly responsive to the environment.

While IQ is highly malleable in principle, in practise it is one of the more stable human characteristics across the lifetime.

The nice thing about disagreements like this is that we can look at the data (what data there is).

I suggested that JL was overestimating the degree of individual stability represented by a high correlation between earlier and later tests. Amusingly, both JL and Bryan Pesta (who will likely show up on the comments here to attempt to convince the world I have no idea what I’m talking about by saying I have no idea what I’m talking about without actually demonstrating it–dude needs a new hobby) referred me to the data set I was looking at when I made the comment. Only I was looking at the paper that included a visualization of the data (pdf).

The following is a scatterplot of IQ scores of Scottish people tested as schoolchildren in 1932 and again much later in life.

Stability of IQ, Age 11 to Age 77

The center teal line is where scores would fall if they were the same on both tests. The lines to either of side of those represent a difference of 15 IQ points, which is the standard deviation for an IQ test.

As you can see, there is significant deviation from the age-11 scores. There are people whose scores would have registered as mentally retarded at age 11 who registered as average intelligence at age 80. Former near-geniuses are now average. Even excluding the outlier, those of average IQ score in 1932 (red line) span three standard deviations of IQ scores decades later.

It’s also worth noting that starting this sort of comparison at age 11 cuts off much of the more extreme variability that is seen in early childhood. And that, my friends, is what “one of the more stable human characteristics across the lifetime” looks like across much of the lifetime.

{advertisement}
On the Stability of IQ
{advertisement}

92 thoughts on “On the Stability of IQ

  1. 2

    Is it just my eyes playing tricks or do IQ scores on average seem to drop? Are there data using ‘younger’ older people (80 would seem to include a lot of conflating factors due to aging, although maybe that’s point).

  2. 4

    Not that I disagree with you, but your argument is less than convincing.

    “Even excluding the outlier, those of average IQ score in 1932 (red line) span three standard deviations of IQ scores decades later.”

    This is pretty much a tautology. That’s exactly what the standard deviation measures, after all. Regardless of the slope of that line, I would expect to see deviations from that line of about 2 to 3 standard deviations.

    Consider the explanation that IQ is, in fact, completely constant, and we are simply observing the measurement error associated with the testing instrument. This graph cannot really help you distinguish between that scenario, and the scenario where the testing instrument is completely accurate, and the people are what are changing.

  3. 5

    I’ve taken dozens of IQ tests. My lowest ever score was 89, and the highest 176. So I’m either borderline subnormal, or a super-genius.

    But then, I’m not dumb enough to think something as complex and badly defined as intelligence could be measured by a glorified word-association game.

  4. 6

    First visit here. As such I hope my remarks do not give offense, as I question the notion expressed in the post that the included chart supports any conception at all about the variance or invariance of IQ over time. Please be aware that I profess no expertise in statistics and offer this comment with complete willingness to be corrected where I am mistaken.

    First, what exactly is IQ and what should the tests for it entail? For instance, components of IQ might be: memory, knowledge, creativity, etc. So without knowing the particulars of the tests administered at ages 11 and 80, it is quite possible that a different emphasis on which aspect of intelligence comprises the test would reasonably allow an unchanged intelligence to test differently at the different age points.

    Second, an earlier commenter noticed the tendency for the later test to show lower scores. This is most easily seen by counting the data points below the bottom green line and above the top green line. There are twice as many of the former. It was correctly noted that senility onset at that age point is adequate explanation, so that IQ variability as explanation is not necessarily indicated.

    Finally, careful observation will note that the densest collection of data points is within half of a standard deviation of the central green line, which if it tended to show anything, would tend to support constancy rather than variability of IQ (whatever that is).

    All of does not negate the fact that effective intelligence can vary in one’s lifetime, but my take would be it would tend to stay the same unless one took interest in things and invested time in increasing one’s knowledge and capabilities.

    Like I said. All subject to revision – as what I have written may aid others in clarifying their thoughts on this, I am willing to have my thoughts critiqued also.

    And finally, thanks for the blog and stimulating thinking.

  5. 8

    Mike, there is some evidence of a decline in later years, and the researchers who collected this data applied a correction factor for that which raised the correlation on this data set from .63 (what you see here) to .73. I also have some concerns that the highest scores seem to plateau at 120 for the adults. That may be a “feature” of the scoring for people of that age.

    There are other long-term (multi-decade) data sets that range in correlation from .41 (age 4 to age 42) to around .8 (ages 17-19 to ages 35-50). Roughly speaking, the younger your starting age or shorter your interval between tests, the higher your correlation, but it is a small group of studies.

  6. 9

    MadMax, if your scores for a group that previously scored at the mean are distributed in a way that looks (statistically speaking) very much like the population as a whole, then your original average scores don’t give you much predictive power over time.

    exrelayman, this stability was offered as an indication that the variability in IQ is determined genetically. Your point that it would be expected to be stable as long as the environment was stable is pretty much exactly the point.

    To the extent that we see large changes in IQ scores over time (and this graph shows that we do, upward as well as downward), we can be pretty sure that this variation over time is not genetic. The stability that we do see may be attributable to a stable environment (because those do exist) or it may be attributable to genetics, but this kind of graph can’t differentiate between the two. It can only tell us that something that is almost certainly not genetics can cause some wide variation between early and later scores.

  7. 10

    @ Madmax: Uh, wait a minute. If IQ tests have a margin of error similar to the standard deviation from the mean in the population, what’s the fucking point? That means that random error in the test could completely account for the distribution in measured IQ without having to posit any difference at all between members of the population. If the tests are that bad, how do we know we’re really measuring anything? If this graph doesn’t show that IQ changes, it shows that IQ doesn’t exist.

    Anyway, margin of error on the test is checkable. If it was just the test, the data Stephanie quotes in comment #8 would be no different from what she displayed in the OP, and that’s not the case.

    @ Lou Jost: Not really. The graph shows that IQ rarely changes by more than 15 points. But 15 points is similar to the average IQ difference between two people. In other words, if the second test 70 years later was accidentally given to the wrong person, the measured IQ would still rarely change by more than 15 points.

    The standard statistician trick for analyzing data like this is to compare them with what’s called the “null hypothesis.” In this case, that means, what if IQ at 80 has nothing to do with IQ at 11? In that case, we could look at the envelope formed by straight horizontal lines at 85 and 115 – since the null hypothesis says we can’t use the first test to predict anything, we assume an individual’s IQ at age 80 is probably about average. How many dots are in that region? Almost as many as the one shown above. So our data shows us something a little different from what we’d expect if IQ were completely unstable over time. But not very different.

  8. 11

    I can’t get at the original source (link seems to be broken), so I’m speculating somewhat, but if the correlation is that high, then I would say that the population at the mean value of X is quite differentiable from the overall population. You are correct in saying that the predictive power is not great, but the effect is certainly there.

    I think there is some confusion here since there are at least 4 things on that graph that could be labeled “standard deviation.” Do you have any information on the size of the measurement error associated with the IQ test? (Variability if you repeatedly test people with the same IQ?)

  9. 12

    but my take would be it would tend to stay the same unless one took interest in things and invested time in increasing one’s knowledge and capabilities.

    Hmmm

    I’ve always viewed the brain to be comparable to a muscle in it’s use and potential. Train it right and develop the right sort of routine for it and it will continue to improve. Do the same thing again and again and it will plateau. Neglect it and it will diminish.

    Of course with how lousy IQ tests are it may not be measuring ‘intelligence’ at all, just how rusty you are on certain subjects. Your ability to gain, retain and process new information may be staying the same or rapidly changing.

  10. 13

    Robert B and Madmax, I understood Stephanie’s “15 points is the standard deviation of an IQ test” to mean that this is the variability sue to measuring error. In other words, if you tested the same person many times, at the same age, teh scores would have a standard deviation of +/- 15 points. Under that interpretation, Robert, my statement is correct. The data almost all fall into the envelope predicted by the null hypothesis that IQ does not change with age.

    Stephanie, can you tell us what you meant by your statement that the standard deviation of an IQ test is 15 points? I do not think Robert’s interpretation is right.

  11. 15

    Robert B, actually your interpretation of the +/-15% is ruled out by the graph itself. The standard deviation of that cloud of points along either the x- or y-axes is much greater than +/-15 points.

  12. 16

    @ Lou: Nope, 15 points is the standard deviation of an individual from the mean for the population. It’s not, as you propose, the standard deviation of a single result for one person from the mean of all tests on that individual person. Actually, the IQ scale is calibrated, for each test, so that the mean score is 100 and the standard deviation is 15. Stephanie didn’t say that, that’s background knowledge I had, but you can check it. The same article says that the margin of error when repeatedly measuring the same person is about three points. (That also answers a question MadMax asked.)

    Remember, we are not looking at repeated measurements of the same thing, but at a large number of different people measured twice each. One hopes, if the IQ test is worth anything at all, that differences in measured IQ reflect differences between measured individuals. The spread along the X and Y axes ought to mean that the rightmost and topmost dots represent smarter people than the leftmost and lowermost dots. Assuming the test is worth something, the spread is due to difference between people, not random error in the test.

  13. 17

    PS: Remember that we expect to find about 65% of data points within one standard deviation of the mean. When you look at a sizable data set, you’re usually going to see a range of three standard deviations on either side of the mean – 99.5% of the points should be within that range.

  14. 20

    I appreciate the clarification about the standard deviation of the IQ tests, Robert. Actually, a +/-15 pt std dev due to measurement error would have made IQ tests pretty nearly worthless.

  15. 21

    Yeah, that looks to be a correlation of about .65. Considering the test-retest gap here is 69 years, that’s incredibly impressive (consider also the sophistication of an IQ test 70 years ago, as compared with today’s measures).

    Correlations ignore mean differences on x and y, so something like the Flynn effect would not affect reliability here (i.e., the person’s rank on the IQ test at age 11 correlated with his/her rank at age 80).

    And again reliability coefficients are variance explained. 70% of variance in IQ at age 80 is explained by IQ at age 11, yet you see this graph as compelling evidence of instability?

    Do you think the people who died before time 2– if included here– would increase or decrease the reliability shown above?

  16. 22

    Stephanie, comment 8: The correction factor was used to deal with range restriction– the correction had absolutely nothing to do with lower IQ scores later in life (i.e., mean differences on x and y). Stats 101?

    Also, I’ve asked this before: The brain is plastic surely. But, not all brains are equally plastic (individual differences in plasticity exist). What if IQ determines plasticity? Seems plausible, too, if indeed IQ tests measure the speed and efficiency with which brains process info.

    Null hypothesis comment: The NH is kind of pointless when looking at correlational data. Statistical significance here is irrelevant (unless N is tiny); what matters is variance explained . The answer is 70%.

  17. 23

    Finally, re similarity of environments for people in this study:

    The reliability would actually likely go up if the study included people of vastly different environments. IQ scores would be more variable; tending to increase correlation coefficients, and it’s likely IQ at both times would be strongly correlated with those in extreme environments (they’d have to be, unless despite extreme poverty as a kid, Joe managed genius level IQ in old age; and despite extreme privilege as a kid, Susy’s life experience made her an idiot).

    I guess my hobby is exposing the ignorance of those so sure of their intellect that they believe their off the cuff insights: (a) never occurred to 100s of scientists in field, and (b) single-handedly show that people who have devoted their lives to studying this issue are so moronically inept, that clever bloggers can take the whole field down by misinterpreting a scatter plot.

  18. 24

    @ Bryan: Was .65 meant to be the r^2, or the r? At .65, it makes a difference. Curse that broken link, or I’d look for myself. Either way, it would mean that the correlation is measurable but far from perfect. It makes it clear that stable factors like genetics can’t be the whole story with IQ, though it also makes it clear that there are stable contributing factors.

    And I understood the axis label “MHT IQ” to mean that MHT referred to a specific IQ test that was used in both tests, which means that advancements in testing wouldn’t matter. That would be good procedure, anyway, but again I can’t read the paper to check if I’m right.

    I think people dying could only reduce the sample size or at worst introduce a systematic error. For it to affect the correlation, average lifespan would have to be linked not to IQ, but to the stability of IQ, which doesn’t seem plausible.

  19. 25

    Thanks Robert.

    Reliability co-efficients are fairly unique as far as correlations go. The raw correlation represents variance explained (i.e., you do not square reliabilities to get at variance explained).

    I still submit that reliability is a separate issue from genes versus environment. The question posed by the blogger: Is IQ stable. The answer is a strong yes, given the 69 years between re-testing (and a large literature showing impressive reliability measured in different ways– internal consistency, e.g.).

    What causes the stability is a separate, but interesting, empirical question.

    ***

    IQ predicts mortality (imperfectly). Dead people couldn’t be tested (validly anyway) at time 2. Not including these people then artificially restricts reliability.

    I’d need to read up on whether iq stability predicts mortality. I’m thinking about terminal drop, where among old people, short-term changes in IQ moderately strongly predict death by the next wave of testing. So, I’m not sure.

    But, suppose IQ at age 11 correlates .30 with age of death. Could you explain how not including dead people (when studying 80 year old survivors longitudinally) would have zero effect on reliability (I’m not seeing it).

  20. 26

    I guess my hobby is exposing the ignorance of those so sure of their intellect that they believe their off the cuff insights:

    Cool story, bro.

  21. 27

    Bryan, in your comment #23, that is a worthy hobby, but sometimes people from outside a field do find deep and long-held errors, because they took a novel perspective. Write me (my name, all one word, at yahoo.com) for some examples. I would enjoy passing them through your debunking filter.

  22. 28

    Robert, the name of the pdf has spaces in it, which means the link doesn’t want to resolve when copied. Plug “The impact of childhood intelligence on later life: following up the Scottish mental surveys of 1932 and 1947” into Google Scholar for the paper with the scatterplot, and “The stability of individual differences in mental ability from childhood to old age: Follow-up of the 1932 Scottish Mental Survey” for the original.

  23. 29

    It’s really sweet how others unfamiliar wih your body of blog-work also think your sneering condescension is a “cool story”, don’t you think, Dr Pesta? The way you treat Stephanie also contrasts extraordinarily well with the way you treat Robert.

  24. 31

    Bryan:

    I still submit that reliability is a separate issue from genes versus environment. The question posed by the blogger: Is IQ stable. The answer is a strong yes, given the 69 years between re-testing (and a large literature showing impressive reliability measured in different ways– internal consistency, e.g.).

    What causes the stability is a separate, but interesting, empirical question.

    I’m used to physical, not social, sciences, but I wouldn’t call that correlation “strong.” And if you admit the 69 years between tests as a factor, you’re accepting Stephanie’s entire thesis for the post, which is that an individual’s measured IQ can vary a great deal over time. Genetics isn’t really central to her point, it’s just an example of a factor that does not vary over time and therefore cannot completely determine IQ. (It’s an example perhaps more relevant than some, because of the way IQ gets used in discussions of racial politics.)

    But, suppose IQ at age 11 correlates .30 with age of death. Could you explain how not including dead people (when studying 80 year old survivors longitudinally) would have zero effect on reliability (I’m not seeing it).

    It would be a straight-up shift in the mean IQ, wouldn’t it? Take out a bunch of high-IQ or low-IQ individuals, whichever group hypothetically dies faster, and it’s just going to shift the population we’re looking at. It would reduce the sample size, naturally, with the universal effect that has on the certainty of any statistical result. But we wouldn’t expect that to degrade the correlation. It would be like taking the graph above and losing, at random, 50% of the points to the left of x=90. (Or, if you like, below y=180-x). It gives you less data to work with, but it doesn’t make the fit any tighter or looser.

    The terminal drop thing is interesting, though – that sounds like an effect of the very kind I thought that was implausible. How fast do they die after IQ becomes unstable? What percent of 80-year-old Scotsfolk would we expect to be in the middle of terminal drop when tested? Oh, and should I assume from the name that the instability is all downward? Because that would be a differently-shaped effect than what we’re seeing here.

    As an aside, though – I’m loath to abandon a reasoned debate, but in between your first post and my response to it you got suddenly abusive toward Stephanie, and I’m not really comfortable with it. I appreciate that you and she are in the midst of an ongoing disagreement, but accelerating vitriol toward the host does not make a good environment for intellectual discussion, at least not for me. If you’re going to talk to both of us at once, could I ask that you keep the two conversations at roughly the same emotional tone?

  25. 32

    Robert, the vitriol goes both ways. I’ll spare your sanity and not include examples. But, point taken. Without knowing our histories, my tone indeed comes across as dickish here. So, apologies to readers and Stephanie (but not JT, just because he’s Canadian;).

    Correlations can be translated to effect sizes (these directly answer the question of “how strong”): From wiki: .10 is small; .30 is medium; .50 is large. .70 variance explained (implying a correlation of .84) is massive (my word– Wiki doesn’t give adjectives for correlations greater than .50).

    ***

    IQ scores do indeed vary over time, but that’s not what the correlation here measures: Instead, reliability is the consistency of the rank-ordering of people on IQ at time 1 and 2. IQ scores go up and down over 60+ years, but how smart you are relative to your peers is incredibly stable. My claim is no variable in social science is more stable than IQ, and the graph above is just one example supporting that inference.

    ***

    To get strong correlations, one needs:

    1) lots of people average on both x and y.
    2) the rest of the dots all fall either on the bottom left or top right (with few to no dots in the other quadrants).

    Note in the graph the lack of dots top left and bottom right.

    Anything that restricts the variance on x or y (like low-iq people dying) artificially reduces the true correlation between x and y.

    If IQ indeed predicts death, then we’re under-representing potential dots in the bottom left quadrant (without the correlation, the dots would be equally likely to appear bottom left or top left). That’s range restriction, again resulting in the under-estimation of reliability.

    Mean differences on X and Y have zero effect on correlation. In fact, correlations first standardize the mean difference to zero, and then look at co-variance.

    re: terminal drop. The people in the graph were all alive at time 2, so this graph can’t address the issue. If there was a time 3 testing (say one year later) than TD would exist if people with dots below the regression line were more likely dead at time 3, relative to people above.

  26. 33

    Ooop, Robert:

    Yeah, the exact same test was used at both times. But many here are especially skeptical of old IQ tests.

    To the extent modern IQ tests are more construct valid, using a crappier test to measure stability would once again be an under estimate.

  27. 34

    What I don’t get is this – all IQ tests I’ve seen so far are collections of math and logic puzzles, word association games and language. All of these things can be learned. How exactly do you correct test scores for education and experience from previous tests? Unless what you want to measure is skill in solving typical IQ tests and then call that ‘intelligence’.

  28. 35

    This still does not address the question of what exactly IQ measures… If you accept that it measures some reified “intelligent quotient” that is actually a decent measure of intelligence, then you have to accept that African-Americans are less intelligent. I certainly do not. Also, the test’s roots are in eugenicist projects for proving the superiority of Aryans.

    Why is this concept still taken seriously?

  29. 36

    Bryan:

    IQ scores do indeed vary over time, but that’s not what the correlation here measures: Instead, reliability is the consistency of the rank-ordering of people on IQ at time 1 and 2. IQ scores go up and down over 60+ years, but how smart you are relative to your peers is incredibly stable. My claim is no variable in social science is more stable than IQ, and the graph above is just one example supporting that inference.

    I just don’t see this graph establishing that. A significant number of those dots – maybe 10% by my eyeball estimate – aren’t even going to be in the same quartile on both axes. If this is the most stable variable in social science, that says more about social science than it does about IQ.

    You know what, maybe some context would help. I was trained in physics. I’m used to scatter plots fitting their theoretical curves like this, or this. When you say “massive correlation” I expect to see something about like the second link there (in terms of how close the points are to the curve); that’s so clearly not what we’re looking at that I now doubt you could have meant what I thought you meant.

    It could very well be that your field just has more noise in its data than mine. I’d be unsurprised if social science data just never gets as tight as I’ve been trained to want data to be. Pions might just be intrinsically more predictable than people.

    But what that would mean, basically, is that everything about people can change, that all differences between human beings are not perfectly stable but vary significantly over time. This may seem obvious to you, since you’re an expert, but laymen don’t necessarily know that, or don’t necessarily know that when you use phrases like “incredibly stable” you mean in comparison to other social science variables that are even less stable.

    Could that what’s going on here? Could we (or at least I) be misunderstanding you to mean a much stronger claim than you’re actually trying to make? If so, I very much doubt I’m the only one. I think a lot of people would hear you say “incredibly stable” and expect to see IQ staying within, say, five points or so for your whole life, barring extraordinary circumstances like illness or injury.

  30. 37

    I would like to suggest that age 11 is too late to start to test for stability over time. So much happens during early childhood that can enrich a child’s life, and thus IQ, that there can be wild fluctuations during that time. Further, given that IQ is a ratio where one part of it is age, as you get older you must get smarter or you simply fall behind. Given all that, there are simply some bulbs that burn brighter and some that burn dimmer, lifelong.

  31. JL
    38

    The plot represents a raw correlation of .63. This, however, underestimates the true stability. Corrections for range restriction and unreliability raise the stability coefficient to around .80. I’m baffled that anyone could regard that study as evidence against the idea that IQ scores are stable across the lifetime.

    When I said that “While IQ is highly malleable in principle, in practise it is one of the more stable human characteristics across the lifetime”, I did not mean that a particular individual’s IQ cannot change substantially. Many environmental (and genetic, too) factors can obviously greatly change IQs. When I speak of stability, I am referring to the empirical fact established in longitudinal studies that IQ measured many years or decades apart shows high stability for most people. This means that IQ tests, far from being arbitrary, actually tap into some pretty stable characteristics of human beings.

    this stability was offered as an indication that the variability in IQ is determined genetically.

    No one claimed that. That’s your own canard. Stability and heritability are separate issues, and heritability could be high even if stability was low, for example. The actual relationship between stability and heritability is an empirical question to be elucidated by research.

    It’s also worth noting that starting this sort of comparison at age 11 cuts off much of the more extreme variability that is seen in early childhood. And that, my friends, is what “one of the more stable human characteristics across the lifetime” looks like across much of the lifetime.

    One the other hand, at age 11 IQ is less stable than in adulthood, so starting at age 11 underestimates the stability that exists after childhood.

    One longitudinal study kept retesting a sample of people from toddlers to adulthood. What they found is that averaging the results of tests taken over a few years enables highly accurate predictions of adult IQs at a relatively early age:

    age range | correlation with adult score

    42,48,54 months | .55
    5,6,7 years | .85
    8,9,10 years | .87
    11,12,13 years | .95
    14,15,16 years | .95

    This means that childhood IQs are much less chaotic than Stephanie suggests.

    Some further points about Deary’s study:

    * The scatterplot is age-corrected, so you can’t directly compare raw scores at ages 11 and 80. You can only look at differences between the two ages. If you look at raw scores, 80-year-olds actually score higher than 11-year-olds – this is because intelligence is not yet fully developed at age 11.

    * The Flynn effect is irrelevant here. It changes scores between generations, not within them.

    * The test used in the study primarily tests verbal intelligence. Verbal intelligence is crystallized intelligence which shows less age-related decline than fluid intelligence. A Raven’s matrix type test would probably show less stability over 69 years.

  32. 39

    One the other hand, at age 11 IQ is less stable than in adulthood, so starting at age 11 underestimates the stability that exists after childhood.

    Eh?

    If IQ really is as stable as your arguing it is through out the lifetime of an individual why would you have to wait for it to stabilize? Doesn’t that undermine your point?

    This means that childhood IQs are much less chaotic than Stephanie suggests.

    What would you call extreme variability?

  33. 40

    julian, you misunderstand. The point is not to actually make some kind of coherent argument about what conclusions can or cannot be reasonably drawn from IQ scores, particularly on the individual level, which is where they are used. The point is to suggest that I don’t know what I’m talking about, so no one listens when I make arguments about what conclusions can or cannot be reasonably drawn from IQ scores.

  34. JL
    41

    If IQ really is as stable as your arguing it is through out the lifetime of an individual why would you have to wait for it to stabilize?

    IQ is stable once it has developed fully. Compare it to height. Adult stature can be predicted only imperfectly from childhood stature, but that does not mean that once you’re fully grown your height, compared to others in your cohort, will fluctuate to the extent it did when you were a child.

    The point is not to actually make some kind of coherent argument about what conclusions can or cannot be reasonably drawn from IQ scores, particularly on the individual level, which is where they are used. The point is to suggest that I don’t know what I’m talking about, so no one listens when I make arguments about what conclusions can or cannot be reasonably drawn from IQ scores.

    Nonsense. I have presented detailed arguments with evidence refuting your views. If you don’t understand them, I can elaborate.

    IQ scores can be and are frequently used as independent variables in sociological and, increasinly, epidemiological research, similarly to how one might use variables such as SES or education. IQ is relevant also on the population level, not just the individual level.

  35. 42

    Ender –

    I posted this to the post this one was spawned by, in response to your comment there:

    I would just like to note that IQ testing isn’t actually useless, it’s just useless in terms of what a lot of people consider it useful for. The tests used in IQ testing are integrated into occupational psych assessments that help determine how best to help kids who are underperforming in school or who have specific disabilities that interfere with their ability to function normally. I would go as far as to say they are critical elements of such assessments and that such assessments are extremely useful in determining the best course of action in terms of treating mental illness, disability and general academic dysfunction. It isn’t going to make everything perfect, but it does go a long ways towards helping kids reach their best potential.

    That said, IQ tests as they stand alone aren’t used in occupational assessments. They take the various elements of IQ tests and mix them into several parts of those assessments. They aren’t the whole of occupational assessments either, just important parts. This only applies to comprehensive assessments. Typically these run between 6-12 hours, over several sessions.

  36. 43

    IQ is stable once it has developed fully.

    That still seems circular to me. In order to establish the predictability of someone’s IQ throughout their life you’re excluding the ages when it varies to wildly.

    Compare it to height.

    I’m sorry but, how do you know the comparison to height is valid? What makes the comparison to height (which is relatively constant throughout life) legitimate but a comparison to weight (which is also relatively constant throughout life but for very different reasons) invalid?

    Sorry if these questions seem naive (this isn’t an area I know a lot about).

  37. 44

    JL, what positive arguments about IQ have you made?

    You haven’t bothered to understand my views in order to refute them. I mention genes having an effect on environment, listing a number of ways in which this happens, and you go on a multi-comment tear focusing on location only and suggesting I don’t understand that any individual can be genetically divergent from the local group and how location–just one factor that can covary with genes–isn’t an issue for a particular study. At no point have you actually dealt with gene-environment interaction, just said I don’t understand a study. Nor have you dealt with the critique of heritability estimates. Instead you’ve simply repeated what you said.

    You and Bryan just enjoy each other and pick around the edges of everything, though. Have fun telling me that something I’m not saying is wrong, wrong, wrong! I’m done it for now.

    Oh, and Bryan, yes, I misremembered which attenuation they corrected for in my haste to get an answer to Mike. I’d considered several possibilities between reading the abstract and getting my hands on the pdf. It was indeed the attenuation of the range of responses from the first study, not the apparent attenuation seen in the upper range of responses in the second test and not the attenuation of performance in the older group. But speaking poorly from memory has what to do with my understanding of statistics? For that matter, what does it have to do with the original post?

  38. 45

    Robert:

    Hard scientists usually scratch their heads when they see social scientists claim 70% of the variance explained is incredible (I remember a colleague dancing through the hallways in glee when he found a meta-analytic effect size of r = .20 for something he was studying! That’s 4% of the variance explained…).

    Even worse, since we’re talking specifically about reliability here, .70 represents just the minimum “accepted” level for using a test in research (and no commercial test would tolerate reliabilities less than .90).

    I still say the graph shows incredible stability. Imagine lining up kids based on IQ at age 11, and then testing them again at age 80 to form a new line. You’ll see slight shifts in the line (43rd place is now 52nd, while 16th is now 10th) but no dramatic departures.

    There is no little johnie in the graph who started with a low IQ but raised himself substantially through determination, and/or an enriched environment / quality education. There appears to be 1– maybe 2– little suzies who initially scored well, but suffered a hard life, with a correspondingly much lower IQ in old age. The majority of the dots seem remarkably stable (despite 69 years! And, despite the test not being the best measure of g, and despite measurement error).

    The authors call the data set a “national treasure” in the discussion section. They seem excited too.

    ***

    re: practical utility and sloppy correlations: Suppose we had a .70 correlation between IQ and job performance. Assume 100 people apply, and half are qualified (but we can’t look at any one person and determine whether he/she is qualified). We need to hire 10 people.

    Without a valid predictor of job performance, 1 out of every 2 hires would be unqualified (selection accuracy here would equal the base rate). With, however, the parameters I made up above, accuracy would increase to 95%! This is why I think psych testing in general is the biggest contribution social science has made to humanity (fwiw, the real correlation between IQ and JP is about .5, which would increase accuracy to only 84%).

    ***

    Stephanie: You are obviously high IQ / well read. My problem is not your intellect, but your seeming (imo) unwillingness to accept that (a) perhaps some level of expertise is needed to make sense of a 100-year old scientific (albeit social-scientific) discipline, (b) blank slate is flatly wrong (so too is 100% genetic determinism).

  39. JL
    46

    That still seems circular to me. In order to establish the predictability of someone’s IQ throughout their life you’re excluding the ages when it varies to wildly.

    IQ varies more in childhood because, among other reasons, children develop at different rates. As I pointed out above, more reliable estimates of children’s eventual cognitive ability can be obtained by averaging several test results taken over several years. Related to this, the heritability of IQ increases linearly with age.

    Now, if you think it’s circular to argue that height and IQ are stable once they have fully developed, then so be it. It’s more of a semantic or philosophical argument.

    I’m sorry but, how do you know the comparison to height is valid? What makes the comparison to height (which is relatively constant throughout life) legitimate but a comparison to weight (which is also relatively constant throughout life but for very different reasons) invalid?

    I never said that a comparison to weight is invalid. If it displays similar patterns of development and stability as height and IQ, then it’s valid.

  40. 49

    We finally get to the nub of your long-standing antipathy for Stephanie, Dr. Pesta — you imagine she’s argued for blank slate. Have you tried, you know, ASKING HER if that’s what she believes, or has argued for, ever? Is it because in your zeal for defending your ideas, you think she’s arguing against your pet beliefs that IQ is mostly genetic and that the various “races” will show a great deal of that difference?

  41. 51

    Keep reading, Bryan.

    What is in dispute is the likelihood that genes will be found that account for any significant fraction of the variability found in human intelligence and whether the current literature on the topic is sufficient to predict that. Here is where disagreement with Thompson comes into play. He has published a number of papers with “genetics” in the title (“Genetic influences on brain structure,” “Genetics of brain structure and intelligence,” “Genetics of brain fiber architecture and intellectual performance”) that involve no genetic testing whatsoever.

    Instead, these studies rely on degree of relatedness (usually between identical and fraternal twins) as a measure of shared genes. This sounds reasonable, and to a degree it is. However, unless researchers can measure or control for the way genes unrelated to intelligence interact with the environment, these studies can’t tell us how much variation in brain structure is due to shared genes that code for intelligence and shared genes that code for something else, such as illness that limits time in school. Until these studies are designed to look for genetic influences in addition to environmental influences, these studies are useless for their intended purpose.

    Sounds to me like she’s criticizing the structure of the studies he used to argue for primarily genetics.

  42. 55

    Are you saying you don’t believe intelligence is mostly genetic and that the various human races will vary as a result, Bryan? Because in every single thread I’ve seen you take on Stephanie and Greg Laden directly, that appears to be — to me, a layman — the position you’ve argued almost exclusively. Perhaps I’ve just been misreading you for several years. Wouldn’t that be a huge shame!

  43. 58

    DuWayne –

    Yeah I get that it could be useful for educational assessments, but this (and the earlier) post is not about using it to measure childhood development, as it is about the stability through time of the measurements…

    And it flirts with some of the ways that the test has been badly misused. Like using the term “near-genius” and whatnot.

    I hate to nitpick about the post that wasn’t written (it really isn’t fair to the author) but a historical critique of the REASONS people keep trying link IQ to genetics would be useful. It has been one of the major goals of the IQ tests (and their predecessors, the Binet tests) since the eugenicist inception of the tests starting around 1900 with the same sorts of people who were practicing racial craniometry.

    The fact is, is that the popular understanding of IQ is badly flawed, representing “intelligence” as a quantity is a reification fallacy, IQ tests have historically been used for nasty political agendas, and this background should be stated up front in discussions of adult tests and alleged links to genetics.

  44. 60

    Hi Stephanie. I’ll have to check out your other posts. Looks like I have more reading to do! And like I said previously, my critique of these two is not really fair. Especially as your first two posts were about the issue I was carping on… Silly me.

    I guess what always gets to me, is that I see IQ discussions as bearing more than a little resemblance to theology. There are a lot of complex concepts and analyses being tossed about without much evidence for the utility (or existence) of the underlying concept.

  45. JL
    61

    JL, what positive arguments about IQ have you made?
    You haven’t bothered to understand my views in order to refute them.

    For the benefit of others, let’s review the debate. In the original post, you attempted to refute a new study on IQ heritability using the following argument:

    As for the strong correlation between genetic similarity and similarity in IQ? I’m not remotely disappointed in the finding. It’s exactly what we would expect if those who were genetically similar shared a large portion of their environment as well. They do, in fact, both because families and immigrant groups tend to cluster in similar environments and because a number of things more directly encoded in genes (illness, skin color, etc.) have an effect on the environment in which intelligence is developed.
    I’m only disappointed in those who think that this is somehow conclusive evidence of subtle genetic effects on IQ.

    In my comment, I pointed out that you had misunderstood the study design (there were no immigrants, non-whites, or relatives in the sample, just unrelated ethnic Swedes), and that your argument against the study was thus invalid.

    You replied to this by abandoning your original argument, and arguing instead that the study’s results were due to non-relatives living in the same locality sharing both their genes and their environment, with the latter being the causal factor behind their similar IQs. Replying to this, I, firstly, pointed out that for your argument to make sense, there would have to be unrealistically high correlations between genes and geography and between geography and IQ. Secondly, I noted that you presuppose the existence of extremely potent environmental influences regardless of the fact that no evidence for such effects exists. Thirdly, I explained that even if your argument made sense, the sampling process of the Swedish study precluded the possibility of there being many people from the same locality in the sample. Thus your new argument against the study was also invalid.
    Moving on, in your next comment you quoted some random snippet from a paper I had cited, supposedly to refute my argument. I still have no idea why you quoted it. Then you repeated your earlier argument. In my reply I pointed out the absurdity of your argument (Swedes are not a collection of heavily inbred subpopulations with severely limited IQ ranges), and again underscored the study design which you had completely disregarded.

    There was also some discussion about whether the study’s heritability estimate was a lower or upper bound. I explained why it’s a lower bound. Then there was debate about IQ stability, but I won’t review it here. My disagreements with you concerning stability can be read in this thread.

    So, I have argued, among other things, that this particular study (along with others) shows that IQ is strongly heritable, and that you have misinterpreted the study and made irrelevant criticisms of it. I’m still interested in hearing how you would explain the actual results of the study, not your own misinterpretation thereof. To wit, if the very strong correlation between IQs and genetic similarity of unrelated people from the same ethnic group raised apart is not due to genes influencing IQ, how would you explain it? Note also that the earlier study by Davies et al. that used the same heritability estimation method found not only that overall genetic similarity predicted IQ similarity, but that the lengths of chromosomes were related to IQ differences, with longer chromosomes explaining more of IQ variation than shorter chromosomes.

  46. 62

    JL, you have indeed not bothered to understand my original argument. You still don’t seem to understand that heritability is actually estimated by reference to environmental factors as well as to shared genetics. Or perhaps you do, but you assume that because this is Sweden, you somehow don’t have to worry about gene-environment interactions or that cultural transmission frequently follows the same paths as genetic transmission.

    You always have to worry about environmental confounds–unless you control for them or unless you understand gene-environment interactions so well that you can definitively say, “Yep, we’ve got that covered.” Good luck, by the way, on being able to say that. We know they exist, but we’re far from having them quantified. Still, you got hung up on the specific examples given and assumed they covered or were meant to cover everything. No, they don’t.

    Note the words, “conclusive evidence,” in the bit you cite. They’re critical. That is, in fact, all I’m saying, all I say at any point about heritability research. There is a tendency to point to relatively simple correlations that fail to consider environment as a factor (despite significant research establishing strong influences from environmental factors) and say, “See? There must be some kind of genetic influence.”

    No, there doesn’t necessarily have to be some kind of genetic influence. It’s possible that there is, and there is plenty of research that doesn’t rule it out. There is, however, none that has done the hard work to establish that conclusion over another. And frankly, as long as the genetic and familial research is allowed to go on as though environment isn’t a strong influence without criticism, we’re going to keep seeing research that doesn’t even try to tease the role of genetics and environment apart.

  47. 63

    Bryan:

    There is no little johnie in the graph who started with a low IQ but raised himself substantially through determination, and/or an enriched environment / quality education. There appears to be 1– maybe 2– little suzies who initially scored well, but suffered a hard life, with a correspondingly much lower IQ in old age. The majority of the dots seem remarkably stable (despite 69 years! And, despite the test not being the best measure of g, and despite measurement error).

    … Are we reading the same graph? I count four people who started below 70 and ended above 90 (like third to thirtieth percentile). Some poor kid started at 50 and jumped up to 80, and someone else went from just above 90 to just below 120 – that was probably a jump of over 50 percentiles.

    There are definite signs of stability, sure. The majority of points, yes, are about where you’d expect them to be. And I suppose the difference between you calling the stability “remarkable” and me calling it “limited” is just a matter of perspective. But there’s no denying that there is some variation and it’s significantly larger than even an old-fashioned test’s margin of error would be. (I imagine. If you actually have a number on that margin of error, feel free to share.) This graph says that IQ is a strong tendency, not a fixed destiny.

    And once again, if the time between the tests matters even though the test hasn’t changed, then you are admitting that there are factors that can change IQ over time.

    And since mechanisms are being discussed only in the most general terms (nothing more specific than stable vs. variable factors) your comment about enrichment and so on strikes me as a cheap shot and a strawman attack. But since you bring it up, if you’ve got this variation from perfect stability, and you can’t account for all of it with random error, there must be some mechanism causing that variation. Even if the variation is much smaller than in most social science variables, it’s there and it must have at least one cause. How can you look at any proposed mechanism – say, education – and say “no, that isn’t it”?

    (Do the points showing improved IQ, those above y=x, seem to cluster between x=100 and x=110? Isn’t that right where mainstream education focuses its efforts? Just some idle speculation, but it might be worth someone’s time to crunch numbers and see if that’s a real cluster.)

  48. 64

    Ender –

    The reason I am inclined to mention it’s usefulness, is precisely because IQ testing is so misunderstood. I have spent rather a lot of time on the topic, because I have an interest in eugenics. I get very irritated with silly little people who like to spout the sorts of assertions that Stephanie is responding to here, because they not only tout IQ as being something it is not – they effectively try to justify eugenics. All the while “apologetically” claiming they are just going where the science leads, “isn’t it unfortunate it leads of here.”

    While I am not going to claim that every asshat who likes to tout this trope of being a eugenicist, I will comfortably assert that they are fucking vile little people who are all too comfortable making unsupported assertions about the intelligence of people who frighten them. Unfortunately, eugenicists or not, they provide political fodder that has real world consequences.

  49. 66

    RB:

    I admit to just now noticing perhaps the biggest outlier in the graph (the dot at the bottom of the red band) but I still see the graph as showing remarkable stability.

    If you’re interested, here’s a much simplified version of the paper (the one linked by Stephanie is a follow up, and is considerably more complicated and dry):

    http://www.abdn.ac.uk/~psy086/dept/pdfs/Intelligence_2000_Stability_IQ.pdf

    Since our perspectives differ, maybe we shouldn’t use subjective adjectives to describe what the graph looks like, and instead agree that 70% of the variance in IQ at age 80 is explained by knowing one’s IQ at age 11 (on the other hand, the authors also use the phrase “remarkable stability” to describe the correlation, but they don’t seem to accept that it’s variance explained / versus that .70 needs to be squared. I need to look into the latter issue).

    So, this leaves us with the observation that at least 30% of the variance in IQ at age 80 cannot be explained by IQ at age 11. I can only speculate as to why:

    1. Different life experiences / environments after age 11.
    2. Measurement error.
    3. “State” differences (in motivation or whatever) across testing times (this is a subset of 2).
    4. Construct invalidity (a fluid measure of IQ, or g derived factorially, would offer a more powerful test of stability). For example, the Raven’s (a superior measure of IQ) doesn’t correlate overwhelmingly with this test.
    5. ?

    ***

    I don’t really understand your clustering comment. To me, the graph just shows that this sample was (for whatever reason) smarter than average. I think more dots above (or below) the regression line would offer evidence of something like education affecting IQ (note, that Stephanie’s middle blue line can’t be the regression line—you can visualize it though. I see an intercept of about 50, and a steeper slope).

    Fwiw, I am not a eugenicist. I’m quite comfortable knowing that many are smarter than I, and many have bigger vitas too. This is my area of study. It fascinates me. I won’t apologize for that. I’d rather know reality than be a PC ostrich (especially considering the nexus of variables that correlate strongly with IQ).

    B

  50. 67

    Bryan:

    Now that sounds reasonable. I haven’t actually read the math, but I’m happy to agree that, barring arithmetic errors, IQ at age 80 is 70% (or 49%?) explained by IQ at age 11. I’m also happy to agree that one possible explanation for the other 30% (or rather, some of it – I’m sure some of it is testing error, though I really doubt that all of it is) might be differences in environment or life experiences. Or that some of it might be construct invalidity – that this IQ test is including some irrelevant things in what it measures, which affect the results in weird ways.

    … PC ostrich? Did you… did you just insert a political correctness slam against a third party in the middle of a conversation about statistics? I mean, okay, I saw you get not-so-subtly accused of racism a few comments back, and that must be upsetting. But… “stop being so PC” is not the response to a false accusation of racism. It’s the response to a true accusation of racism. What the hell, dude? “I am not a eugenicist” is not really very reassuring now that I think about it.

  51. 68

    Can I join in the conversation? There’s some research apart from Deary’s that’s worth mentioning, about the malleability of teen and adult intelligence, and there’s a bunch of misconceptions that could be cleared. To quote the author of the paper under discussion, ‘…it is interesting to reflect on how many other areas of psychological research attract such attention from journalists and psychologists from outside their fields.’ (Deary, 2011). Sometimesan outider may make a contribution, but usually, they are struck by something that was discussed 60 years earlier, is part of the background, but they think it’s novel.

    1. There is a well-articulated theory of the components and structure of intelligence – CHC theory, and the Moray House Tests fit into it. As they are vocabulary and verbal reasoning tests, they are expected to be the most stable tests of cognitive functioning over time.
    2. The Flynn effect would be expected to have some effect, depending on scoring. The children in this study, retested at age 80, would be expected to do substantally better than the 80-year-olds tested in 1937. However, as they are all int he same cohort (range of birth years), that’s not going to explain any variance within the group. The Flynn effect is also much strnger on measures of reasoning skill rather than knowledge – perhaps teaching critical thinking and algebra is having an effect. This will have little effect on the MHT scores.
    3. There is substantial interest in ways to increase cognitive functioning, and preserve skills, through later adulthood, effectively boosting IQ. There’s an excellent review by Christopher Hertzog (Enrichment Effect on Adult Cognitive Development, 2009), summarizing the literature on exercise (a strong intervention to boost skills), cognitive enrichment, and social engagement, as well as some of the neural underpinnings behind such changes (and some biological risk factors, such as the APOE4 allele).
    4. Another study, that I don’t think Stephanie has mentioned here or in her earlier post, recently came out in Nature, titled Verbal and non-verbal intelligence changes in the teenage brain. This also found substantial overall similarity over time (correlation of .8 for verbal skills over three years, .59 for visuo-spatial skills over the sme period), but changes of up to 20 points for individuals. MRI data showed corresponding changes in brain structure, and the individuals have described some changes that may have precipitated these changes in skills and brain structure (becoming interested in a particular topic and working to master it, for example).

    Yes IQ can be malleable for individuals, even though group data may not show a lot of change.

    But don’t just take my word for it (or decide I’m some internet crank – I could be a poodle who likes to type) – look at the actual research.

    Deary, I. (2011). Intelligence. Annual Review of Psychology, 2012, 453-482.

    Hertzog,C., Kramer, A., Wilson, R. & Lindenberger, U. (2009). Enrichment Effects on Adult Cognitive Development: Can the Functional Capacity of Older Adults Be Preserved and Enhanced? Psychological Science in the Public Interest, 9, 1-65.

    Ramsden, S., Richardson, F., Josse1, G., Thomas, M., Ellis, C., Shakeshaft, C., Seghier, M. & Price, C. (2011). Verbal and non-verbal intelligence changes in the teenage brain. Nature, Published online 19 October 2011. doi:10.1038/nature10514

  52. 69

    Robert:

    I concur with all comments before the PC paragraph.

    I may doth protest too much, but you would have to search pesta / laden / iq and then wade through 12 or so threads, and over (probably) 1,000 comments, to see why I defend myself now the way I do.

    I don’t recommend it, unless you’re bored at work (thereby getting paid to suffer through all this).

  53. 70

    Also:

    My argument is that .70 is variance explained. I know that calculating reliability with parallel tests is one scenario where the correlation is not squared. The same test at different times surely meets the requirement of “parallel test”, so my best guess is that .70 should not be squared.

    Would be interested in a more authoritative answer.

  54. JL
    71

    JL, you have indeed not bothered to understand my original argument.

    Stephanie, do you read my posts at all? I have patiently tried to explain why gene-environment correlations of the sort you have proposed cannot account for the strong genotype-phenotype associations found in the Swedish study. One of the great things about the study’s methodology is precisely that G-E correlations cannot explain the results. That’s the reason why the authors are so confident of the existence of strong genetic effects.

    Contrary to what you claim, traditional behavioral genetic family studies have for decades tried to find the sorts of environmental effects you propose. However, it has turned out that the magnitudes of these effects vary from tiny to non-existent. This has not discouraged critics like you who have postulated that there are unknown, undetected effects of enormous power that explain, for example, why MZ twins are more similar than DZ twins, or why MZA twins are so similar.

    However, those olds arguments are not applicable to studies where the subjects are unrelated and do not share environments to which they are quasi-randomly assigned. Why does genetic similarity explain a minimum of 47% of IQ variance in the sample if genes are not the causal factor? You have not answered this question. Hand-waving about unspecified environmental factors does nothing to refute the results of the study. Please propose specific mechanisms that could account for the results.

    Note the words, “conclusive evidence,” in the bit you cite. They’re critical. That is, in fact, all I’m saying, all I say at any point about heritability research.

    Actually, previously in this discussion you were were adamant that the results could be explained by simple G-E correlations. Now you seem to be more agnostic. That’s progress.

  55. 72

    Contrary to what you claim, traditional behavioral genetic family studies have for decades tried to find the sorts of environmental effects you propose. However, it has turned out that the magnitudes of these effects vary from tiny to non-existent.

    Citations for these “tiny to non-existent” effects?

  56. JL
    73

    Sesardic, “Making Sense of Heritability”, pp. 113-115 contains a concise summary of different ways that have been used (with scarce results) to test for G-E correlations in pedigree studies. There are further references therein. However, I’m not going to get bogged down into that, because those sorts of correlations are irrelevant when it comes to the studies like those by Davies et al. and Chabris et al.

  57. 74

    Yes, you’re absolutely right, JL. The discussion of the equal environment assumption in Sesardic is really neither here nor there. The variables under discussion involve personality traits and don’t really have much to do with this particular line of investigation.

    What you want are studies that actually measure the environmental variables already known to correlate with the dependent variable. Did you have any citations for broad-population studies that have actually done that?

  58. JL
    75

    You are avoiding the subject at hand, Stephanie. I’m not getting into a discussion of G-E correlations unless you show how they are relevant to the study we’re talking about here.

  59. 76

    I think everyone, including Stephanie, seems to be missing one important point — certain people have entirely too much of their identities wrapped up in IQ scores, which among other things explains why some people are convinced The Bell Curve is actually scientific. The idea that IQ isn’t constant just messes them up and makes them stompy and whiny. All the data in the world won’t change that.

  60. JL
    77

    which among other things explains why some people are convinced The Bell Curve is actually scientific

    Yes, that must be why leading experts in intelligence research and behavioral genetics thought that The Bell Curve is firmly supported by mainstream science.

    Let’s see, BrianX. Based on the results of a hundred years’ worth of research using the best methods in social science, researchers think that IQ is an important trait. On the other hand, you disagree with them on the grounds of some lazy armchair psychologizing. Yes, Brian, it’s true: you’re a moron.

  61. 78

    JL, one more comment insulting someone here on the basis of spurious “best methods in social science” gets you booted. I put up with you doing it to me long enough to find out what you thought your argument was, but you seem to have run that out. Saying we don’t have to control for environmental variables shown to affect IQ in order to call our correlations “heritability” because we’ve already done that, then deciding that, no, you don’t need to back that assertion up with any citations because it should just somehow be obvious that environmental variables have no effect? Yeah, not impressive.

    Contribute more information and less attitude, or go away.

  62. 80

    JL –

    I am rather curious what “the best methods in social science” would entail. Especially as they have apparently been used for a hundred years.

  63. JL
    83

    JL, one more comment insulting someone here on the basis of spurious “best methods in social science” gets you booted.

    BrianX posted a comment that was purely a personal attack with no factual or rational basis. Such a comment is moronic, and I pointed this out to him. Why is it that you seem to have no problem with his comment, only mine?

    Saying we don’t have to control for environmental variables shown to affect IQ in order to call our correlations “heritability” because we’ve already done that, then deciding that, no, you don’t need to back that assertion up with any citations because it should just somehow be obvious that environmental variables have no effect?

    Behavioral genetics is all about devising ways to separate genetic and environmental causes. Criticisms of behavioral genetics center on attempts to disprove that behavioral genetic methods succeed at providing such separation. However, for these criticisms to be noteworthy, they must be specific. For example, critics of the classic twin design have specifically attacked the equal environments assumption that the method relies on. Similarly, if you want to make a case against these new molecular genetic studies of IQ heritability, you must point out specific flaws that you think they have. Otherwise you have proved nothing.

    I am rather curious what “the best methods in social science” would entail. Especially as they have apparently been used for a hundred years.

    The methods have been developed over a period of a hundred years. Methods developed mainly for the purposes of IQ research (e.g. classical test theory, IRT, factor analysis, various forms of SEM) have colonized many other disciplines, too, because of their superiority, whereas behavioral genetic methods are superior because they can shed light on both genetic and environmental sources of human behavioral variation, unlike, say, traditional sociological analysis.

  64. 84

    JL –

    I’m not going to accuse you of being a moron, because I don’t actually know for sure that you are, but mostly because I have a lot of respect for our host. I will mostly assume you’re rather ignorant of just how powerful the research tools you are talking about actually are. They are certainly useful in certain contexts. But the behavioral genetics tools are not particularly useful for testing what you think they do.

    The brain is both a lot more malleable than we once thought and we are shaped by cues far more subtle than we ever imagined possible – especially in infancy. This understanding tends to bring doubt to the assumptions made, based on such studies. While it by no means definitively determines they are wrong, it absolutely reasserts the questions they were assumed to have answered.

    The only way we will ever be able to untangle this web, is to dramatically increase our understanding of genetics and neurobiology. We will need to not only determine the gene sequences that govern intellect, but also to determine exactly what they do. I have little doubt that eventually we might get there, but we are not there now. Given our modern understanding of human cognition and neurodevelopment, there is simply no use in making firm assumptions about the genetics of intelligence. About the best we can say is that genetics are likely to play some role, while environment plays some role as well.

    Pretending any sort of certainty either way, would just be silly and irresponsible.

  65. JL
    85

    DuWayne, if you want to reject the results of the sorts of studies we’ve been discussing, you will have to do better than “oh, it’s so terribly complex, surely we can’t know anything before we know everything blah blah blah”. That’s the classic method of creationists, climate change denialists, and other charlatans.

    The purpose of the biometric approach of behavioral genetics is NOT to find out how genetic and non-genetic influences play out in the development of an individual. Behavioral genetics is about the causes of variation at the population level, not in individuals. There are always different levels of analysis in science. Behavioral genetics can be useful when entangling the developmental process in individuals, too, but heritability analyses are not dependent on our knowledge of the developmental process.

  66. 86

    JL –

    DuWayne, if you want to reject the results of the sorts of studies we’ve been discussing, you will have to do better than “oh, it’s so terribly complex, surely we can’t know anything before we know everything blah blah blah”. That’s the classic method of creationists, climate change denialists, and other charlatans.

    I did better than that and if you weren’t an ignorant dumbass, you would realize that.

    What we have learned about the malleability of the brain in the past five years has shown that the tools used in those studies are absolutely useless. And to understand the role genetics has in neurodevelopment, we must understand how it generally plays out on an individual level. Unfortunately, we don’t have the tools to do that.

    I am absolutely *not* claiming it is too fucking complex. If you understand English on more than a rudimentary level, I would suggest you re-read my comment. You will note that nowhere in my comment, did I say anything about it being too complex. I spoke of our ignorance and a lack of tools to alleviate that ignorance. I did so because I am pretty fucking certain that we will eventually develop those tools. Once those tools are developed, this shit won’t be any more complicated than understanding how and why lightening happens. Not everyone will understand it, but it will be easily understood by people who choose to learn about it.

    The bastion of creationists is the need to grasp tightly to outdated, discredited ideas. Between us, the person engaging in that behavior would be you. I would suggest you spend some time learning about neuroplasticity and it’s implications for the discussion of genetics versus environment, before you start trying to argue as though you have the foggiest fucking clue what you’re talking about. I actually have paid rather a lot of attention, because it is integral to the research I am going to be doing.

    To be very clear, I want the tools to make these determinations to exist. I wish they did because they would make my research a hell of a lot easier. Not in the context of some idiotic measure of some types of intelligence, to be sure – I couldn’t possibly care less about that. But generally understanding the role that genetics actually plays in cognition is extremely important to me, because it would make the treatment of mental illness a hell of a lot easier.

    So you’re going to have to do a hell of a lot better than further than trying to dismiss me because you either didn’t read or couldn’t comprehend what I actually wrote.

  67. 87

    DuWayne, not to discourage you, but you are talking to the person who told me that my statement that research looking into the role of genetics in IQ testing needs to account for the environmental variables already known to correlate with IQ because culture is frequently passed along in tandem with genes wasn’t a specific enough criticism to take seriously. Just sayin’.

  68. 89

    What if our IQ depended on how much cake and cream we fed the pixies in our brains, Bryan?

    And where have I ruled out a role for genetics based on anything, much less plasticity? Note that this has been asked multiple times before, and you’re still not answering.

  69. 91

    I actually very specifically and clearly stated that plasticity *doesn’t* rule out genetics, merely that it requires that the assumptions about genetics versus environment be reassessed.

Comments are closed.