About the SA Blog Network

Beautiful Minds

Beautiful Minds

Insights into intelligence, creativity, and the mind
Beautiful Minds Home

The Heritability of Intelligence: Not What You Think

The views expressed are those of the author and are not necessarily those of Scientific American.

Email   PrintPrint

One of the longest standing assumptions about the nature of human intelligence has just been seriously challenged.

According to the traditional “investment” theory, intelligence can be classified into two main categories: fluid and crystallized. Differences in fluid intelligence are thought to reflect novel, on-the-spot reasoning, whereas differences in crystallized intelligence are thought to reflect previously acquired knowledge and skills. According to this theory, crystallized intelligence develops through the investment of fluid intelligence in a particular body of knowledge.

As far as genetics is concerned, this story has a very clear prediction: In the general population– in which people differ in their educational experiences– the heritability of crystallized intelligence is expected to be lower than the heritability of fluid intelligence. This traditional theory assumes that fluid intelligence is heavily influenced by genes and relatively fixed, whereas crystallized intelligence is more heavily dependent on acquired skills and learning opportunities.

But is this story really true?

In a new studyKees-Jan Kan and colleagues analyzed the results of 23 independent twin studies conducted with representative samples, yielding a total sample of 7,852 people. They investigated how heritability coefficients vary across specific cognitive abilities. Importantly, they assessed the “Cultural load” of various cognitive abilities by taking the average percentage of test items that were adjusted when the test was adapted for use in 13 different countries.

For instance, here is the cultural load of the Wechsler Intelligence Test subtests:

They discovered two main findings. First, in samples of both adults and children, they found that the greater the cultural load, the greater the test was associated with IQ:*

This finding is actually quite striking, and suggests that the extent to which a test of cognitive ability correlates with IQ is the extent to which it reflects societal demands, not cognitive demands.

Second, in adults, the researchers found that the higher the heritability of the cognitive test, the more the test depended on culture. The effects were medium-to-large, and statistically significant:

As you can see above, highly culturally loaded tests such as Vocabulary, Spelling, and Information had relatively high heritability coefficients, and were also highly related to IQ. As the researchers note, this finding “demands explanation”, since it’s inconsistent with the traditional investment story. What’s going on?

Why did the most culturally-loaded tests have the highest heritability coefficients?

One possibility is that Western society is a homogenous learning environment– school systems are all the same. Everyone has the same educational experiences.  The only thing that varies is cognitive ability. Right. Not likely.

The next possibility is that the traditional investment theory is correct, and crystallized intelligence (e.g., vocabulary, general knowledge) is more cognitively demanding than solving the most complex abstract reasoning tests. For this to be true, tests such as vocabulary would have to depend more on IQ than fluid intelligence. Seems unlikely. It’s not clear why tests such as vocabulary would have a higher cognitive demand than tests that are less culturally-loaded, but more cognitively complex (e.g., tests of abstract reasoning). Also, this theory doesn’t provide an explanation for why the heritability of IQ increases linearly from childhood to young adulthood.

Instead, the best explanation may require abandoning some long held assumptions in the field. The researchers argue that their findings are best understood in terms of genotype-environment covariance, in which cognitive abilities and knowledge dynamically feed off each other. Those with a proclivity to engage in cognitive complexity will tend to seek out intellectually demanding environments. As they develop higher levels of cognitive ability, they will also tend to achieve relatively higher levels of knowledge. More knowledge will make it more likely that they will eventually end up in more cognitively demanding environments, which will facilitate the development of an even wider range of knowledge and skills. According to Kees-Jan Kan and colleagues, societal demands influence the development and interaction of multiple cognitive abilities and knowledge, thus causing positive correlations among each other, and giving rise to the general intelligence factor.

To be clear: these findings do not mean that differences in intelligence are entirely determined by culture. Numerous researchers have found that the structure of cognitive abilities is strongly influenced by genes (although we haven’t the foggiest idea which genes are reliably important). What these findings do suggest is that there is a much greater role of culture, education, and experience in the development of intelligence than mainstream theories of intelligence have assumed. Behavioral genetics researchers– who parse out genetic and environmental sources of variation– have often operated on the assumption that genotype and environment are independent and do not covary. These findings suggests they very much do.

There’s one more really important implication of these findings, which I’d be remiss if I didn’t mention.

Black-White Differences in IQ Test Scores

In his analysis of the US Army data, the British psychometrician Charles Spearman noticed that the more a test correlated with IQ, the larger the black-white difference on that test. Years later, Arthur Jensen came up with a full-fledged theory he referred to as “Spearman’s hypothesis: the magnitude of the black-white differences on tests of cognitive ability are directly proportional to the test’s correlation with IQ. In a controversial paper in 2005, Jensen teamed up with J. Philippe Rushton to make the case that this proves that black-white differences must be genetic in origin.

But these recent findings by Kees-Jan Kan and colleagues suggest just the opposite: The bigger the difference in cognitive ability between blacks and whites, the more the difference is determined by cultural influences.**

As Kees-Jan Kan and colleagues note, their findings “shed new light on the long-standing nature-versus-nurture debate.” Of course, this study is not the last word on this topic. There certainly needs to be much more research looking at the crucial role of genotype-environment covariance in the development of cognitive ability.

But at the very least, these findings should make you think twice about the meaning of the phrase “heritability of intelligence.” Instead of an index of how “genetic” an IQ test is, it’s more likely that in Western society– where learning opportunities differ so drastically from each other– heritability is telling you just how much the test is influenced by culture.

© 2013 Scott Barry Kaufman, All Rights Reserved

* Throughout this post, whenever I use the phrase “IQ”, I am referring to the general intelligence factor: technically defined as the first factor derived from a factor analysis of a diverse battery of cognitive tests, representing a diverse sample of the general population, explaining the largest source of variance in the dataset (typically around 50 percent of the variance).

** For data showing that Black-White differences in cognitive ability are largest on the highly culture-dependent tests, I highly recommend reading Chapter 4 of Kees-Jan Kan’s doctoral dissertation, “The Nature of Nurture: The Role of Gene-Environment Interplay in the Development of Intelligence.”

Acknowledgement: thanks to Rogier Kievit for bringing the article to my attention, and to Kees-Jan Kan for his kind assistance reviewing an earlier draft of this post.

image credit:

Scott Barry Kaufman About the Author: Scott Barry Kaufman is Scientific Director of The Imagination Institute in the Positive Psychology Center at the University of Pennsylvania. Follow on Twitter @sbkaufman.

The views expressed are those of the author and are not necessarily those of Scientific American.

Rights & Permissions

Comments 29 Comments

Add Comment
  1. 1. neuroecology 11:56 am 10/17/2013

    I wish that they had plotted these (maybe in 3d?) vs cultural load. It looks like there’s is little heritability vs cultural load correlation, but there is between cultural load/%var and heritability/%var. Do they talk about this/what do you think that means?

    Link to this
  2. 2. Adam_Smith 12:09 pm 10/17/2013

    Hanging around smart people will make you smarter — at least, so I have long suspected. I sometimes see individuals avoiding association with people they see as smarter than themselves because they don’t want to feel dumb. That is a big mistake. The more you feel dumb, the more likely you are learning.

    Link to this
  3. 3. Noone 2:29 pm 10/17/2013

    Why would “abstract reasoning” not also be “culturally loaded”? If you grow up in a simple agricultural economy where you follow the lead of others rote, where are your opportunities for such?

    Link to this
  4. 4. anjely 3:39 pm 10/17/2013

    Completely… you can train yourself to think outside the box… or you can train yourself just to remember rote facts, the more you do one, the worse you will become at the other. Kind of like how the current method of school teaching in America seems to overwhelming kill the creativity that initially flourishes during childhood.

    Link to this
  5. 5. jayjacobus 4:07 pm 10/17/2013

    There are inferior students and there are superior students. Most likely the superior students are well fed, motivated and mentally healthy. They do not necessarily have intelligent parents.

    To me that doesn’t sound right. After all, evolution favors the survival of the fittest. Could inferior students be the fittest? The evolution of intelligence must be genetic. Nurture plays a role but not the primary one.

    Link to this
  6. 6. rkipling 5:20 pm 10/17/2013

    I have thought for a long time that cultural influence could be a significant factor. As genetically similar as all living humans are, it didn’t make sense to me that racial IQ differences, with the spread reported by some studies, were genetically driven.

    There are those who say all cultures are equal. Like the LSU professor who sat next to me on a flight and explained knowingly and somewhat patronizingly that the Somali tradition of having the smallest children eat last (if there was anything left) was as valid as our traditions. “Well”, I said, “wouldn’t malnutrition adversely affect their physical and mental development?” His demeanor changed. He said he would bring up that point with his students.

    I contend that some cultures are more equal than others. This study seems to say that too.

    Link to this
  7. 7. NeverConvex 5:52 pm 10/17/2013

    Glad to see genotype-environment covariance getting more press as an important factor in heritability studies, but I think it’s kind of misleading to make it out to be a totally new route for the field. I mean, 12 years ago Dickens and Flynn proposed more-or-less exactly the same explanation as the present authors for IQ heritability / reliability / etc data before:

    “In 2001, Dickens and Flynn presented a model for resolving several contradictory findings regarding IQ. They argue that the measure “heritability” includes both a direct effect of the genotype on IQ and also indirect effects such that the genotype changes the environment, thereby affecting IQ. That is, those with a greater IQ tend to seek stimulating environments that further increase IQ. These reciprocal effects result in gene environment correlation.” ( )

    Link to this
  8. 8. mgtorres 7:18 pm 10/17/2013

    The concept of “cultural load” does not seem to accurately encompass the complexity of societal demands or the realities of access to educational or learning opportunities. There are very far jumps in logic made in this study that do not take into account knowledge from disciplines that study culture and its impacts.

    Link to this
  9. 9. mathari 8:04 pm 10/17/2013

    We are about 5 years away from having the ability to scan the brain (and other organs) so we can determine function, capacity and damage. Such a scan will make the use of intelligence testing obsolete. In combination with GWAS, we will be able to resolve many questions which have been answered incorrectly based upon faulty assumptions. We are skipping so many steps in this process and making huge leaps of faith by assuming things that should not be assumed.

    Link to this
  10. 10. LSDcasualty 9:01 pm 10/17/2013

    “such as vocabulary would have to depend more on IQ than fluid intelligence”

    For your consideration, I suspect you have heard similar before.

    Vocabulary must be defined by the essential ability of mental strategic pattern formation, the negation of a core component of memory relative language is there required to be compensated by smoother and more active direct momentary accuracy within the given structured conversation as related to temporal understanding of objective at the time.

    The issue is that at certain levels the strategic apparatus of a successful eventuality within a structured and fluid conversation might be completely void of potential successful conclusion, to avert this one can learn language in a new way the same as one learns to type, muscle memory effect in the mind is equivalent to word play even in the simplest of eventualities, this is why those who drive their intellect at time of decay of crystallisation can conform to assorted formations of smooth and flowing charactor outside the usual scope of normal accepted and trained persona. If one loses the basics of language it is not impossible to generate a new formation whether it crystallises or not, I do not know how relevant this is to you, I just point it out for those who find language is failing them due to harder memory formation, exercise in the correct manner could conceivably avert the onset of complete lack of will to be a part of society. I am of the way of thinking that IQ is not related entirely to hard defined heritability of core knowledge which is evident as memory. It may just be that the memory flawed who know this to some extent fade or decide to seek silence more.

    Link to this
  11. 11. LSDcasualty 9:51 pm 10/17/2013

    I had to contemplate what I wrote again at the request of an egghead.

    ‘They say perhaps it is related still to crystallization’.

    So if this were the truth the only way I could conceivably be construed as correct is if there is something different about myself.
    Now, momentary conversation as defined as a flow which exists on top of an absent subset of standard order can be generated in the moment without really knowing the words, sometimes this is what people with speech problems have done, they do not recognise they actually have an asset they do not use, the ability to decay out of pattern, decaying out of pattern can be a beautiful thing if one presents it as such with the will of ones mind, one only has to find the correct phase of ‘state’ if you will, in order to begin generating words, it is how abstract formations of hip hop and rap and techno lyricists do the hot step between words, you can rebuild with this and use it, despite no evidence it existed is present within you.

    ‘So your discussing learning’
    No, I am dicussing generation, learning need not be involved, except ‘who you are’.

    ‘This would be construed as silence though’

    ‘If you do not have the words you cannot begin’

    ‘If you cannot learn’

    YES YOU CAN, none crystal formation can hold it too, it just begins constantly until you learn what beginning is and if you can rap intermission with a fine tuning, you can lower it down to communication easily, it only requires that you do not lose faith. ‘So you do learn then’
    Not in the area your suggesting, by denouncing me as wrong. Fluid intellect is therefore alternate adaptation in terms of a second learning which is none related and as one loses the memory space previously used, one gains something else, The death of the energetic construct in terms of corruption is inaccurate, the damage generally associated with such a thing is deemed a loss in total, however, if fluid intellect is adaptive to acquire part of the energy or space involved or not, there is still a secondary set of beneficial factors to consider, speaking of path width and energy flow for one, damage to memory is often considered ‘due to cell death’ this may be inaccurate a great deal from where I am sat, though clearly neuroscientists have a better grasp than myself.

    Apologies for the double post, it seems that hope comes in minor and major forms.

    Link to this
  12. 12. LSDcasualty 10:22 pm 10/17/2013

    My thoughts since I am told its all a bit negative.

    If genes are, in my laymen’s terms: limited to design patterns within electrochemical carbon based intellect as a fluid dynamic which becomes constituted at a micro level by diminishing levels of density based cohesion towards a fluid state which exists almost exempt of pattern flow, then the answer to the riddle is one of two things:

    1. Worst case scenario repaired at tail end for purposes of general clarity.
    Electro voltaic balance includes atomic decay as a variable we ignore at our peril, the purpose of synthetic constituents to oneself being introduced in a matrix designed to hold a life form for the purposes of learning and existing as a construct with choice outside of flow towards end or mass in terms of connected energetic synergy with levels of power beyond or spiritually above oneself on the plane of peaceful thought pattern harmony is to be oneself for as long as one can achieve such a thing, that decay occurs can be re-buffered to remove only the none aware still, given energy constructs are and should always be (up to a point in universal multiplicity) fundamentally growing more than evacuating, even evacuation itself can be a notion which amounts only to connective value of relevance to extents which amount to a functional state of relative completion as a being here on Earth and perhaps far beyond.

    I will think on 2 later maybe.

    Feel free mods to take this and bin it, all three, just please use what you can if needed.

    Sorry once again, I am hyper.

    Link to this
  13. 13. rkipling 10:35 am 10/18/2013

    I thought LSD was out of style? Guess not.

    Link to this
  14. 14. rkipling 10:36 am 10/18/2013

    It appears to be used now as dressing for word salad.

    Link to this
  15. 15. Noone 2:09 pm 10/18/2013

    I’ve found (warning: anecdote here) that people EQUALLY adept at hard science, social science, technical writing and fiction-writing (including poetry) SEEM more intelligent than those who excel profoundly in only one area. Does SEEMING more intelligent have any meaning?

    Link to this
  16. 16. Mariam FM 6:07 pm 10/18/2013

    This is amazing. The topic of intelligence is always a tricky subject but factoring the cultural load pushes away all the invalid junk on IQ tests stating the cognitive ability of “Blacks” are lower to their “White” counter-part. When I hear this, when you state this in a class (especially in this day and age where there is larger population of the “minority” in college) it will come across in the wrong sense.

    It’s no mystery to why we see such lower scores compared to white counterparts. It’s a distinct different environment, a different cultural influence. It near impossible to compare “white culture” if there is one to “black culture”. Now, some food for thought, who are the people researching into the topic of IQ differences, is there a mean view point for the goal of the studies? And backtracking, there are so many confounding variables.

    I know from a few neuroscience classes, every infant (born without any abnormalities and a nurturing environment) has the capacity to become an intellectual or at least be somewhat “smart” when the brain development is fostered. If an infant lives in a violent environment, brain connections are lost.

    Link to this
  17. 17. Chuckiscool 1:22 pm 10/19/2013

    “But these recent findings by Kees-Jan Kan and colleagues suggest just the opposite…”

    Scott, Kan et al.’s analyses don’t suggest the above. See table 4.2 of Kan’s thesis. That table clearly shows that the magnitudes of the BW gaps correlate BOTH with g-loadings and with cultural-loadings. As such, there is no showing “just the opposite [of Spearman’s hypothesis]“. While the findings are consistent with Spearman’s hypothesis, they are just inconsistent with the view that “cultural load” is not positively correlated with g-load.

    As for the index of “cultural loading” used, my colleague discussed some concerns on the blog “Human Varieties”. See here. Kan et al.’s “cultural load” variable didn’t correlate with many seemingly cultural effects such as the Flynn effect, adoptive effects, shared environmental effect, and unshared environmental effect. It only seemed to robustly correlate with genetic g effects. Moreover, it ended up correlating with Gf-loading (as did the BW gap). We find these patterns of associations to be curious for a supposed index of cultural influence. We suspect that Kan et al.’s “cultural load” strongly indexes g not because of some elaborate COV(GE) model but simply because cultural loaded subtests happen to be broader measures of intelligence.

    Link to this
  18. 18. Chuckiscool 1:37 pm 10/19/2013

    Mariam FM said: “It’s no mystery to why we see such lower scores compared to white counterparts.”

    If differences between groups were driven by factors which were unique to one or the other group, one would not expect to find measure equivalence which indicates that individual scores are unconditioned on group membership (Meredith and Millsap, 1992). But this is exactly what is found between the specific groups in question. So, yes, there is a mystery. Or, at least, the differences between the specific groups mentioned are understandable in terms of the factors which induce differences within groups (genetic and environmental).

    Link to this
  19. 19. CherryBombSim 8:47 pm 10/19/2013

    So we have a trait that is determined by a large number of genes (not necessarily interacting with each other additively), modified by the environment iteratively and non-linearly over a person’s life. Mathematically, this is deep water.

    Link to this
  20. 20. jsweck 7:23 am 10/20/2013

    It’s not surprising that people are finding poor correlations between genetics and mind. If minds are software systems, then even if you duplicate the hardware perfectly (genetically or otherwise), you will not develop a similar mind.
    Software means information in memory. A mind is accumulated information over time from the creation of the organism. So everyone (and any subset of the population), will have different problem solving abilities. There is no special kind of knowledge that you get from genetics – there is only the stuff you learn.
    This is why babies are not hired as judges. In other words, it’s understood in our culture that the younger you are the less problem solving you can do. This is because all problem solving is anchored in software systems. Less software means less intelligence.

    Link to this
  21. 21. David Marjanović 8:51 am 10/20/2013

    To me that doesn’t sound right. After all, evolution favors the survival of the fittest. Could inferior students be the fittest? The evolution of intelligence must be genetic. Nurture plays a role but not the primary one.

    You’re doing this science thing wrong.

    You need to fit the theory to the data, not the data to the theory.

    What you’re really making here is an argument from personal incredulity: “I can’t imagine it’s true, so it’s false”. Evolution is complicated, it often has seemingly contradictory effects.

    Link to this
  22. 22. KeesJanKan 11:27 am 10/21/2013


    In our model, cultural load is not expected to relate to the Flynn-effect, at least not necessarily so (see below). Also, in this model, a correlation between subtests cultural load and subtest complexity can be positive. No problem or conflict or whatsoever: If society put high demands on complex reasoning, complex reasoning will show high g-loadings and high heritability.

    A relation between cultural load and subtest complexity is a ‘g/Gf-Gc theory thing’. g/Gf-Gc theory states that g loading relate to subtest complexity and also that the completion of culturally loaded tests such as vocabulary is not cognitively demanding (and there they have a point). This is not to say that the ACQUISITION of the knowledge that is required to complete such tests is not cognitively demanding. Substantial g loadings of culture loaded tests can thus be explained through the investment hypothesis (without investment hypothesis, g theory would have a serious problem). So, there are two mechanisms at work that ’cause’ differences in g-loadings: subtest complexity and (let’s call it) ‘historical’ complexity.

    If I needed to make a prediction, I would say that Flynn effects would correlate with INCREASES/CHANGES in societal demands. If society strengthens the demand on abstract reasoning and visuospatial abilities, for example, I would expect an increase in those abilities (and actually also in their g-loadings and heritability). Note that such increase doesn’t say anything about changes in the demands of verbal knowledge.

    In this the research of the sources of the Flynn-effect, Structural Equation Modeling is a better tool however. I mean, better than the calculation of correlations between g loadings and other correlations. That’s also why we maintain that our GE-covariance hypotheses (i.e.: 1. the contribution of GE-covariance to heritability differs across cognitive abilities and 2., more specifically, the contribution of GE-covariance to heritability is higher for culture dependent skills than for culture fair abilities) still need to be tested.

    We state that without incorporating GE cov, empirical findings as ours (and other findings like the increase of the heritability of intelligence) are very difficult to explain, especially when one doesn not consider the role of education and experience (hence when one wouldn’t adapt the investment hypothesis). We also state that a real biological g can be incorporated in our model, but that this is not necessary (it just would make the model less parsimoneous). Our model and g theory do not contradict each other, for example. The only thing is that the combined relations among cultural load, g-loading, and heritability do not follow from g theory, so these combined relations require explanation. In other words there is a gap between theory and empirical findings and this gap needs to be filled. THAT is the challenge for every theory of intelligence, including the dominant theory, i.e. g theory.

    A good way of thinking about our results is the following: If heritability of Fullscale IQ (FSIQ) is 80% and (2*)GE covariance would account for 30% of the variance, then assigning people randomly to environments (rather than on the basis of cognitive outcome measures) this would yield a drop in heritability of FSIQ from to 50%. And if we would assign on the basis of cognitive outcome measures so that (2*)GE cov would contribute -30%, heritability would drop to 20%.

    Link to this
  23. 23. dreadedhill 11:55 pm 10/22/2013

    Think about the inverse of one of the statements in the article:

    Perhaps IQ tests are more a measure of vocabulary than fluid intelligence.

    Link to this
  24. 24. Chuckiscool 3:58 pm 10/23/2013



    The main problem with your study is that you did not establish the construct validity of your “cultural-loading” (let us call this cl(a)-loading) as a measure of “within population heterogeneity of exposure to prerequisite information needed for correct responses to subtest questions” (let us call this cl(b)-loading). If you had shown that cl(b)-loadings did in fact correlate with g-loadings that would have been interesting, indeed. But you did no such thing. Instead, you merely insisted that tests that involved more cultural elements, that is, tests with higher cl(a)-loadings, would be more cl(b)-loaded because, quote, “We also doubt that our Western society creates a homogeneous learning environment….” (You actually make a double inference here. You assume that cl(a) loadings highly positively correlate with cl(b) loadings and, moreover, that the relation between cl(a) loadings and g-loadings is mediated by cl(b) loadings.) Now, imagine that there is no or is trivial COV(GE) with respect to IQ. (Whether or not there is will have to be determined by a meta-analysis of multivariate genetic analyses that specifically assess the magnitude of active, passive, and reactive COV(GE).) If there is no or is trivial COV(GE), then one could reasonably take the difference between MZ twins as an index of cl(b), since, with the effects of genes and prenatal factors controlled for, the differences between these twins, assuming no COV(GE), could reasonably be said to index the “heterogeneity of exposure to prerequisite information” within a population. Your results would then show that cl(b), so indexed, negatively correlated with cl(a) and thus that your “cultural load” as a measure of cl(b) was invalid. The point is that you have to assume what you seek to prove in order to defend the construct validity of your “cultural load” in face of the common sense interpretation that MZ twin differences index cl(b). Since you don’t provide evidence that cl(a)-loadings correlate with cl(b)-loadings (or that cl(b)-loadings mediate the correlation between cl(a)-loadings and g-loadings), all we have is your doubting that learning environments are more homogeneous with regards to what is measured by highly cl(a) loaded subtests than with what is measured by low cl(a) loaded subtests. Now, I would tend to think that “Western society creates a homogeneous learning environment” more so with regards to the information that is measured by more cl(a)-loaded tests than with regards to the information that is measured by less cl(a)-loaded tests for the simple reason that the “schools and school systems” etc., homogenize environments somewhat and also happen to teach more cl(a)-loaded information than not. But, apparently, you and your esteemed colleagues came to the exact opposite conclusion.

    You go onto say: “We state that without incorporating GE cov, empirical findings as ours (and other findings like the increase of the heritability of intelligence) are very difficult to explain….”

    You need better arguments. As noted before, the increasing heritability of g with age can be explained just as well in terms of genetic amplification as defined by Plomin (1987). More generally, all sorts of traits that presumably do not involve extensive COV(GE) show increasing heritability e.g., height. So increasing heritability does not make for a good stand alone argument.

    You continue: “A relation between cultural load and subtest complexity is a ‘g/Gf-Gc theory thing’. g/Gf-Gc theory states that g loading relate to subtest complexity and also that the completion of culturally loaded tests such as vocabulary is not cognitively demanding (and there they have a point)….”

    “g theory” only needs the investment hypothesis to the extent that it retains the Gf-Gc model. If you dispense with the Gf-Gc model and adopt another, you don’t have the same problem. It’s funny that proponents of other models critique the Gc-Gf model in part because Gf is not more heritable than Gc (e.g., Johnson and Bouchard, 2004), but you, instead, grant the Gc-Gf model and criticize heritability estimates! That said, there is something to the Gf-Gc model, so there is something of interest here.

    Now, as for the Flynn effect, the point is that you can take the secular changes in subtests scores as a sort of alternative index of “cultural load” (in the sense of cl(a)). You can then set up a reductio ad absurdum. Your cross-cultural cultural loadings negatively correlate with cross-temporal cultural loadings. In a sense, tests that are the least fair for individuals across space are the most fair across time. One could then argue that the Flynn effect loadings better measure cl(b)-loading, right? So, the Flynn effect is greatest on those tests that have the lowest cl(a) loadings (your cultural load) because these tests have the highest cl(b) loadings, the lowest g-loadings, and the lowest heritability’s. Very parsimonious. Not necessarily an argument that I would make, but…. it could be made.

    Now, I am not saying that your results are meaningless or that they are uninteresting. And I don’t think that the problem was with the analysis. That was very creative. Rather, it was with the leap from cl(a) to cl(b). I am open to be persuaded that cl(a) well indexes cl(b) — but you have to do better than conjecture e.g., “we also doubt…”

    Link to this
  25. 25. KeesJanKan 8:33 am 10/25/2013


    You confuse g and g-loading. Whereas ‘mediated by g’ denotes something meaningful (and testable with Structural Equation Modeling) ‘mediated by g-loading’ is meaningless. g-loading is not an interindivual variable but a test characteristic. The point of our paper is that we found an effect that was not predicted, and is not explained by, g theory. It thus deserves an explanation.

    Our operationalization of cultural load may be a point of disussion/criticism, but it is clear that vocabulary is highest heritable. In g-Gf/Gc theory a vocabulary test is clearly a highly culturally loaded test (which makes sense also for the non-g theorist). Most importantly, g theory might explain why vocabulary is highly heritable, but not why is is the HIGHEST heritable.

    Introduce the following variables:
    g (the biological, but unknown variable in g theory)
    F (fluid intelligence)
    C (crystallized intelligence, a result of fluid intelligence, (cultural dependent) education and other influences)
    Edu (education)
    and a scaling constant k (not really nessecary)

    F = g + ξf (in Jensen’s view the residual ξf is 0, because he considers g and fluid intelligence to be one and the same variable, but I’ll leave this an open question)

    C = k*F + Edu + ξc (investment hypothesis)

    Note that if there is no gene-environment covariance heritability of educational attainment will be 0. Note that if education is assumed homogeneous, variance in Edu equals/approaches 0 (and covariance between fluid intelligence and education will be absent anyway). Also residual variance is by definition independent from g, so there are no covariance terms that contribute to variance in crystallized intelligence.

    σ(C) = σ(k*F) + σ(Edu) + σ(ξc)

    Heritability of crystallized intelligence is thus a weighted sum of the heritability of education (which is 0 under the assumptions above), the heritability of the residual and the heritability of fluid intelligence and. In g theory the residual is less heritable than fluid intelligence (this is assumed in order to explain the positive correlation between g-loading and heritability coefficients of fluid tasks). This implies the heritability of crystallized intelligence is expected not to exceed the heritability of fluid intelligence. If environmental variance approaches 0 and heritabiity of the residual wouldn’t differ too much from the heritability of g, the heritability of crystallized intelligence approaches that of fluid intelligence. If the heritability of the residual is assumed higher than the heritability of g crystallized intelligence can be higher than heritability of fluid, but then 1) the assumption that g is main source of genetic differences is discarded and hence 2) g theory does not provide an explanation of the correlation between g-loading and heritability anymore. This would run against our findings.

    Another thing: “We also doubt that our Western society creates a homogeneous learning environment….” that referred to the situation in which others would want to ‘explain’ the effects as resulting from a homogenous environment and while they do not adhere to g theory. When adhering to g theory, homogeneous environment does not provide any explanation of the findings (see above). It is preety clear learning environment is indeed not homogeneous, for one thing because educational attainment is cnosiderable heritable, which implies there is variance in aducational attainment (it also implies there is GE covariance present).

    Link to this
  26. 26. KeesJanKan 8:40 am 10/25/2013

    And I forgot one important last remark:
    I devoted an entire chapter to the interpretation of crystallized inteligence. I don’t confuse factor models with developmental models, for example, as you do here.

    Link to this
  27. 27. KeesJanKan 8:46 am 10/25/2013

    Oh forgive me the misspelling of neccesary (I didn’t grew up in an English speaking country)

    Link to this
  28. 28. KeesJanKan 8:49 am 10/25/2013

    Oh darn, now I made a grammatical error (must be in my genes right?)

    Link to this
  29. 29. Francois Amsallem 11:47 am 12/17/2013

    The interesting point is one of a much stronger correlation, as empirically admitted to date, between culture and inheritability. It’s a bomb in the garden of a lot of the so called post modern theories and their derivative ideologies, as gender or multiculturalism. I will not just wait and see, I will read and think, there is enough material to motivate free thinkers to use their mind and start a broader revisionism of popular wisdoms. Is there enough here tho for the gentleman’s favorite intellectual sport of high flying pseudo-scientific paradigms shooting game.

    Link to this

Add a Comment
You must sign in or register as a member to submit a comment.

More from Scientific American


Get All-Access Digital + Print >


Email this Article