Twin studies were the first studies used to provide data about heredity. There are two types of twin studies: first, those that measure the differences between identical twins raised apart, and second, those that measure the differences between identical and fraternal twins. (In scientific lingo identical twins are called MZ twins and fraternal twins DZ (Z=zygote – a fertilized egg cell). A “T” at the end means “raised together”, an “A” means raised apart. I avoid the terms here, using normal language.)
For any individual, their IQ test result is partly due to heredity, partly due to environment and culture, partly due to chance. There is no method at this time to determine exactly how much each factor is responsible (except for the chance fluctuation when taking a test twice). However, comparing identical twins it is possible to put some estimates on these factors.
IQs of Identical twins raised together or apart
The studies below show that adult identical twins raised apart only differ by 10 percent. That is, the different environments in the study only changed their variances explained by.10. (10 out of 100).
First, using data from charts and sources below, I note that when a person takes an IQ test twice, usually the second test is higher, but in any event the correlation is 97%. So, off the bat, we know that a non-heredity influence has a variance explained of 0.06 (1 – .94), due probably to chance. (The increase of IQ test results with repetition would apply to everyone repeating the test and would not affect the correlation.)
Second, I note that when adult (23+) identical twins both take an IQ test, their correlation is 92%, or a variance explained of .85. So we have a variance due to being a different person of 0.09. That cannot be heredity; it must be either chance effects in conception, the womb, or differences in their environment.
Third, the correlation of adult identical twins raised apart is 83%, for a variance explained of 0.69. So we have an additional variance explained of 0.16. This must be due to the different environments that the two twins are in.
Is this remaining variance explained of 0.69 for IQ test results all caused by heredity? I.e., is there any other variance explained that can intercede between when the chromosomes are selected and when the twins are born? (We have already accounted for effects after they are born.) Yes, there are chance effects affecting sperm and egg, and chance effects in the womb. I discuss this on the page about chance influences on IQ. Even with identical DNA and environment, IQ test results of an identical twin pair could be higher or lower.
it is important to keep in mind that the very low environment affects of identical twins raised apart do not rule out environmental effects between vastly different groups! There are two reasons: First, if one twin had a severe environmental effects, such as cholera, the twin pair is usually excluded from studies. Second, the separations and adoptions were in relatively similar cultures and environments. None of the studies, to my knowledge, compared separations where, typically, one twin grew up in suburban or rural Minnesota and the other lived in an inner-city family in Newark, for example, or downwind from factories in Mississippi.
Important Historical Note: Bouchard, at the University of Minnesota, made many contributions to the study of relatives’ correlations. His compilation of studies from May, 1981 (Source 1 ) is still used in many scholarly articles and trumpeted on the Internet, but he himself discredited most of its conclusions in 2013 — because the studies did not apply to adults. My look through recent papers discussing inheritance and environment surprisingly finds that some researchers are still not using data for adults. Bouchard’s analysis of similarity of identical twins raised apart, using data from The Minnesota Study of Twins. He found that if one identical twin of very low socio-economic status was adopted by parents of very high socio-economic status, they had a higher IQs by up to 12 points compared to their un-adopted twin. If the adopted twin came from a family of moderately low socio-economic status and was adopted by parents of moderately-high economic status, they had higher IQs by roughly 4 IQ points. (Source 2). But these results were for children.
IQ differences practically disappear when children reach adulthood.
ACE Analysis of heredity
The charts below use ACE analysis, but understanding ACE analysis is not necessary. It is sufficient to look at line A, the portion of IQ due to heredity.
But if you want to understand what the C and E are, here is the explanation: You will recall that when a population of identical twins take IQ tests, one identical twin’s results correlate highly with those of the other twin; and that this correlation may be converted to a variance explained. It is not usually labeled with a letter, but I will call this variance “Z”. E is influences that cause the IQ test results to vary, that is, it is 100% -Z. The portion of Z which comes from heredity is “A”. C is the influences on identical twin results that are not heredity , that is, Z – A. (I find the labels, used by scientists for decades, for C “shared environment” and E, “unshared environment”, to be completely misleading. I think they are mainly chance factors that affect both twins and which affect only one twin. See the page on this site on chance influences on IQ.)
An exhaustive analysis by Turkheimer, Haworth and others of 11,000 twins from studies of different ages in several developed nations, first published in 2009 and journal published in 2010, shows that heritability increases from childhood to the late teens, from about 40% of the variance explained in IQ in childhood to 66% of it by age 17 (a correlation of 81%). (Source 3) (The “E” in the graphic below of about 0.2 variance implies that young adult identical twin test scores correlated by 90%. As you see, the test scores of twins became more similar as they aged. The graphic does not take assortative mating into account. The authors say that they could not estimate its magnitude. However, as the authors point out, including assortative mating would not change the increasing magnitude of A (inherited IQ) with age [the slope of the line]. The little “bars” in the graphic show possible errors in the statistics – there is 95% confidence that the true value lies within the bars. (The childhood sample averaged age 9, the adolescents averaged 12, and the young adults, with ages from 14 to 35, averaged 17).
Increasing heritability and decreasing shared environment for general cognitive ability from childhood to adolescence to young adulthood. Lines show additive genetic (A), common environment (C) and non-shared environment (E) estimates from childhood to young adulthood. Error bars represent ±1 s.e. (bootstrapped with 10 000 samples). Confidence intervals (95%) for these parameter estimates can be found in Table 3.
A reanalysis of parenting-related influences on variations in teenage and young adult IQ that included 229 adopted children and over 5,000 non-adopted children, based upon data from the National Longitudinal Study of Adolescent Health (Source 4) , revealed “very little consistent variation in IQ in adolescents and “no consistent” variation in IQ in young adults due to family variables. This was also true for adopted identical twins. (Source 5)
Here is another graphic, the work of Bouchard, of increasing heredity with age, which the author calls the Wilson effect, after the person who first discovered it: (Source 6 ). Note that the heritable portion of IQ in one study continues increasing until age 26 (a variance explained of .85, a correlation of .92%).. Note that there is still a huge increase in the correlation between 17 and 23 years. This paper is IMO very valuable, because it shows that the IQ correlation of unrelated children raised together drops from .26 to nearly zero at adulthood. This again confirms that parenting and shared environmental influences fall to almost 0 for IQ in adults. Bouchard also states that the data developed by Wilson show that the similarity of IQ of fraternal twins drops to that of siblings in adulthood. Note also that the IQs of unrelated children raised together barely correlate.
The public disagreement about nature versus nurture arises in part because heredity only manifests fully in adults. For young children, family culture plays a greater role.
Many people are quite surprised that identical twins become more similar in intelligence as they reach adulthood. This occurs even though, during teenage years, children get into different environments and have different experiences, especially for identical twins separated at birth. The same phenomenon is also true for many other identical twin characteristics, such as height. Somehow identical twin IQs become more similar to each other and also to their parents, Out of curiosity, we might ask why: We know that IQ is linked to brain structure; and that brain structure changes during adolescence. An individual’s genes are inherited, but changes in gene expression occur with age (which is obvious to everyone at puberty). Thus, as twins reach adulthood, it should not be surprising that gene expression can affect IQ, causing identical twins to be more similar to each other and to their parents. Based upon my reading (in Sources 7 and 8) I speculate that everyone is born with a general childhood intellectual ability that increases until about age 10, and then that is gradually sculpted by a mainly genetically determined program as one ages.
Father — grown son IQ heredity
The degree of relatedness between father and sons and for siblings born apart is the same as between fraternal twins, that is, roughly one-half of the DNA, more or less, of the father is expressed in the son, The mother might have a high or low IQ. The IQs of mother father pairs has rarely been reported. Therefore father-son studies cannot prove much of anything.
For what it is worth, the largest father-son study, taken from Norwegian military records and published in 2008, used IQ test results of over 20,000 pairs of fathers and sons, each test taken at age 18. (All men must enroll in the military.) It found a correlation of only 38%, a variance explained of 14%. (Source 9). For all social groups, if t one father’s IQ was 10% higher than another, the son’s was likely to be higher by over 3%. The birth order of the son and the income of the family made almost no difference. The correlation was identical to that of a 1981 compilation of fathers and son cited above (though this compilation included children as well as adults).
Sibling IQ Heredity — including Fraternal Twins
Siblings and fraternal twins have much lower IQ correlations than identical twins. Data for adult fraternal twins is almost impossible to find. Historically a correlation of .6 was used for fraternal twins, based upon the 1981 Bouchard paper, which he later discredited, because in fact adult fraternal twins have a lower correlation. (The influence of family is replaced by a strong genetic influence, as discussed above.). The only study I have found, using USA national data of individuals recruited by telephone found a correlation for fraternal twins of only .30, which is substantially lower than that found for brothers in Norway. However this study also found a much lower correlation for adult identical twins of .57 These values held in all ethnic groups. Perhaps, in the USA, there are larger differences in health between siblings.
The best sibling data comes from the National Conscription Service in Norway between 1950 and 1985. The data is for young men between their 18th and 21st birthday. It would be expected that the correlations would increase by 10% when the conscripts reached age 23. The young men in the great majority of cases came from the same mother and father. The average correlation was .431, which is a variance explained of only 19%. Thus would suggest that for adult siblings the correlation would have been about .50 or a variance of 0.25.
Greater years of separation between the brothers caused the correlation to lower by .06 per year. The correlation for height of the brothers, which was initially .527, scarcely declined by years of separation of age. This suggest that cultural or environmental factors were responsible for the dropping correlation for IQ. The IQs of children from larger families (4 or more siblings) correlated more than those from smaller families (Source 10)
Calculating the size of heredity from IQ twin studies
I will summarize by saying that these calculations of IQ, widely used in the past and still cited today, are worthless!
Here is the worthless logic they used:
They took the identical twin heredity variance explained as 0.74 (they used data that included children), while the fraternal twins variance explained (in the old studies, but wrong) was only 0.36 (again data that included children). The difference in variances explained between the two types of twins, then, was 0.38. (See footnote 1 for reference published 1981). This difference was doubled to say that the amount of heredity was .76 (which is, stupidly, more than the .74). Scientists later adjusted this to account for assortative marriage and different data so that their estimates of the total amount due to heredity ranged from .6 to .8. However, as stated above, the data for fraternal twins used included children. With a variance for adult fraternal twins of only .25, the difference between identical and fraternal twins variance explained is .49, which would imply that intelligence was 98% inherited, which is far above what we know for identical twins raised apart. In my opinion, this calculation was never logical, because it left out chance effects, and mistakenly calculated the likelihood of siblings having the same genes.
Now if we just doubled the .25 that would suggest that heredity accounts for 50% of IQ test results, but this also leaves out chance effects. It is probable, then, that the contribution of half the same DNA is a variance is more than 0.25. To take this logic further, please refer to the page on chance influences on IQ on this site.
Assortative marriage increases the spread of IQ scores.
Compared to past centuries, marriages today are know to be more “assortative” for intelligence, meaning that men and women tend to marry persons of similar IQ. Studies in the mid to late 20th century put the correlation between the IQ of spouses at 40% (a variance explained of 0.16), (meaning, that their IQs were a little bit similar), but nowadays, more women are going to university and to graduate school, and many more are working in the professions, and fewer people live in small towns, so that the correlation between marriage partners’ IQs is now at least 55%, and several developed countries studies put it as high as 80%. (Interestingly, assortative mating based upon IQ is greater than for any other trait.) (Sources: 11 12) The following chart of USA marriage data, using education as a rough stand-in for IQ, makes the point ( Source 13, chart based upon Source 14). (“C” stands for college completion, “HS” for high school completion):
Most studies of Intelligence in the 20th Century did not mention assortative marriage, or if they mentioned it, they then ignored it. One study in 2010 estimated that the additional heredity variance explained due to assortative marriage is 0.11. (Source 15) This suggests that the correlations for siblings will be higher in future generations.
Some think that assortative marriage may be “breeding” super smart and super dumb individuals. Take flowers: When, in each generation, the top 20% of flowers by size are bred with each other, in only six generations almost all the flowers are much larger than those in the initial population. The opposite outcome occurs when the smallest 20% of the flowers are bred. (See the photo in Source 16).
What if inherited IQ causes an individual to change their environment?
Some critics argue that the genetic differences cause each twin to modify their environment. So, they argue, IQ tests results include an environmental component that lessens the hereditary values. This is a big area of analysis and contention in recent heredity studies. In this website, however, the issue is irrelevant: We are not considering innate intelligence, but only intelligence as measured by IQ tests. So if twins inherit traits that cause one or both to do comparably well or badly on IQ tests, regardless of whether these traits are intelligence or some other trait such as studiousness, it is still heredity. If having a higher IQ causes one to study more, so that the IQ test result goes higher, thus multiplying the direct effect of their genes, it is still caused by heredity.
Where does that leave us? Twin studies show that there is a maximum of 69 percent of IQ test results that is due to heredity, but they cannot show us the minimum, because of chance influences. Nor can they show us the effects of those raised in completely different environments or cultures.
Next page: https://iqscience.net/chance-influences/
- Familial Studies of Intelligence: A Review, Author(s): Thomas J. Bouchard and Matthew McGue Science, New Series, Vol. 212, No. 4498 (May 29, 1981), pp. 1055-1059 ↩︎
- http://people.virginia.edu/~ent3c/papers2/Articles%20for%20Online%20CV/(70)%20Turkheimer%20(1991).pdf. Individual and Group Differences in Adoption Studies of IQ, Psychologivcal Bulletin, 1991. Highly mathematical but see page 398 second column & 399. ↩︎
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2889158/, Eric Turkheimer, CMA Haworth et al, The heritability of general cognitive ability increases linearly from childhood to young adulthood ↩︎
- https://www.icpsr.umich.edu/icpsrweb/DSDR/studies/21600/versions/V21 ↩︎
- Kevin M. Beaver, Joseph A. Schwartz, Mohammed Said Al-Ghamdi, Ahmed Nezar Kobeisy, Curtis S. Dunkel, Dimitri van der Linden. A closer look at the role of parenting-related influences on verbal intelligence over the life course: Results from an adoption-based research design. Intelligence, 2014; 46: 179 DOI: 10.1016/j.intell.2014.06.002 ↩︎
- https://www.gwern.net/docs/iq/2013-bouchard.pdf The Wilson Effect: The Increase in Heritability of IQ With Age, Thomas J. Bouchard Jr. Twin Research and Human Genetics. 2013 ↩︎
- https://www.pbs.org/wgbh/pages/frontline/shows/teenbrain/interviews/giedd.html ↩︎
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3603574/ Development of Brain Structural Connectivity between Ages 12 and 30: A 4-Tesla Diffusion Imaging Study in 439 Adolescents and Adults Emily L. Dennis et al., Neuroimage 2013 ↩︎
- https://www.nber.org/papers/w14274.pdf Black et al., Like Father, Like Son? A note on the intergenerational transmission of IQ scores., National Bureau of Economic Research, 2008 ↩︎
- Sundet, Eriksen and Tambs,, Volume 19 number 9, 2008 Intelligence Correlations Between Brothers Decrease With Increasing Age Difference. ↩︎
- https://www.brookings.edu/blog/social-mobility-memos/2014/02/10/opposites-dont-attract-assortative-mating-and-social-mobility/. ↩︎
- https://www.ncbi.nlm.nih.gov/pubmed/22774442, Escoral & Martin-Buro, The role of personality and intelligence in assortative mating. ↩︎
- https://www.brookings.edu/blog/social-mobility-memos/2014/02/10/opposites-dont-attract-assortative-mating-and-social-mobility/ ↩︎
- https://www.nber.org/papers/w19829.pdf?new_window=1 (National Bureau of Economic Research, 2014) ↩︎
- Behav Genet. 2012 Mar;42(2):187-98. doi: 10.1007/s10519-011-9507-9. Epub 2011 Oct 4. Reconsidering the heritability of intelligence in adulthood: taking assortative mating and cultural transmission into account. Vinkhuyzen AA1, van der Sluis S, Maes HH, Posthuma D ↩︎
- https://www.nature.com/scitable/knowledge/library/the-breeder-s-equation-24204828 ↩︎