Science, Society, Sustainability
The ISIS website is archived by the British Library as UK national documentary heritage ISIS members area log in ISIS facebook page ISIS twitter page ISIS youtube channel ISIS vimeo channel

ISIS Report 30/01/12

No Genes for Intelligence

Extensive genome-wide scans have failed to find a single gene for intelligence; instead, environment and maternal effects may account for most, if not all correlation among relatives, while identical twins diverge genetically and epigenetically throughout life Dr. Mae-Wan Ho

A fully referenced version of this article is posted on ISIS members website and is otherwise available for download here

Please circulate widely and repost, but you must give the URL of the original and preserve all the links back to articles on our website

The heritability of intelligence or IQ (intelligence quotient, see Box) has been hotly debated for decades. The most recent round of exchange was provoked by The Bell Curve of Richard Hernstein and Charles Murray published in 1994.  The book argued that IQ tests are an accurate measure of intelligence, that IQ is a strong predictor of academic and career achievement, that it is highly heritable and little influenced by the environment, and most controversially, racial differences in IQ are likely due to the genes. Consequently, the authors were sceptical about the ability of public policy initiatives to have much impact on IQ or IQ-related outcomes [1].

What is IQ?

IQ, intelligence quotient, is a score resulting from one of several standardized tests designed to assess intelligence. Modern IQ tests were constructed to have a mean score of 100 and a standard deviation (SD) of 15. Approximately 95 % of the population have scores within two SDs of the mean i.e., between 70 and 130.

Quite apart from the fierce debate over the heritability of IQ or intelligence, the claim that IQ assesses intelligence and the validity of any single measure of intelligence are both strongly contested.

The Bell Curve sold 300 000 copies and attracted a great deal of uncritical media attention. The American Psychological Association commissioned a report from a panel of experts rebutting its main claims published in 1996. Now 16 years later, a second report has been issued to take account of the many new findings, including the following.

  • Almost no genetic variants have been discovered that are consistently associated with variation in IQ in the normal range
  • The heritability of IQ varies significantly by social class
  • The importance of the environment for IQ is established by the 12 point to 18 point increase in IQ when children are adopted from working-class to middle-class homes
  • Even when improvements in IQ produced by the most effective early childhood interventions fail to persist, there can be very marked effects on academic achievements and life outcomes
  • In most developed countries studied, gains on IQ tests have continued and are beginning in the developing world
  • The IQ gap between Blacks and Whites has been reduced by 0.33 SD in recent years

Revolution in the heartland of genetics

The report does not quite capture the revolution breaking out in the heartland of genetics.

Simply put, there are no genes for intelligence in the human genome, however you choose to define intelligence; and there has been much contention on that alone. For example, recent research shows that non-intellectual factors such as test motivation can increase IQ scores by an average of 0.64 SD, with larger effects for individuals with lower baseline IQ scores [2].   

More fundamentally, the heritability of IQ estimated in the conventional model is now widely seen as deeply flawed. Heritability – the component of population variation (variance) attributed to genes – has been inflated by gene interactions, gene-environment interactions and other nonlinear effects, in the same way that the heritabilities for common diseases have been inflated (see [3] Mystery of Missing Heritability Solved? SiS 53). Not only that, the heritability estimated from resemblance (correlation, covariance) between twins and siblings could be due to shared environments especially maternal environments.

Even more seriously, the classical Mendelian inheritance on which all estimates of heritability depends has been severely compromised by pervasive epigenetic (and environmental cultural) inheritance. Epigenetic and cultural inheritances often go together, resulting in correlations between relatives that have been erroneously attributed to shared genes. On the other hand, epigenetic variations due to individual experiences, and somatic mutations from a host of DNA marking and changing processes, make even monozygotic twins diverge genetically from each other to substantial degrees. These observations strike at the very core of the conventional genetic determinist paradigm.

The elusive IQ genes

The hunt for IQ genes has been inspired by the large heritability estimated in conventional biometrical models based on correlations between twins and other biological relatives (see [3].  But the results so far have been disappointing to say the least.

A genome-wide association study (GWAS) on 7 000 subjects published in 2008 found only 6 genetic markers (SNPs, single nucleotide polymorphisms) associated with cognitive ability, and only one of those remained statistically significant on further tests. Together, the 6 markers explained barely 1 % of the variance in general cognitive ability [4]. Recently, the association between 12 specific SNPs and ‘general intelligence’ factor g was put to test in an attempt to replicate the associations found in earlier studies, but only 1 SNP remained significant. The researchers conclude that [5] “most reported genetic associations with general intelligence are probably false positives.”  

As in the case of common disease traits [2], IQ or intelligence is plagued by the problem of “missing heritability”. Even the heritability of human height, estimated at ~90 %, failed to turn up common variants contributing more than 0.5 cm; and the set of 180 height-associated SNPs identified by the most comprehensive meta-analysis (on pooled data from many studies) only explains about 10 % of the population variance.

The usual explanation for the missing heritability is that it is difficult to detect genetic variants with a small effect. In the case of intelligence, much is made of the findings in a new study led by researchers at Edinburgh University in the UK, which claims to [6] “establish that human intelligence is highly heritable and polygenic”. The group first used data from five different GWAS and failed to identify any individual marker associated with either ‘crystallized’ or ‘fluid’ intelligence. (Crystallized intelligence is the individual’s store of knowledge about the nature of the world and learned operations such as arithmetic that can be drawn upon to solve problems; while fluid intelligence is the ability to solve novel problems that depend relatively little on stored knowledge as well as the ability to learn). They then applied a new method that tests the cumulative effects of all the SNPs, essentially by calculating the overall genetic similarity between each pair of individuals in a sample, and then correlating this genetic similarity with phenotypic similarity (in IQ) across all the pairs. The result is that all the ~550 000 SNPs together could jointly explain 40 % of the variation in crystallized intelligence and 51 % of the variation in fluid intelligence.  This exercise sounds more like a counsel of despair than a solution to the problem, and the result certainly does not offer any useful predictive information.

Other researchers are tackling the problem at the more fundamental level of the heritability estimates.

Maternal environment accounts for much of heritability

One of the first rebuttal to The Bell Curve came from Bernard Devlin and colleagues at University of Pittsburgh Pennsylvania in the United States in a paper published in Nature in 1997 [7]. They showed that covariance (correlation) between relatives may not be due only to genes, but also to shared environment, especially maternal environment, which is not taken into account in conventional models. In a meta-analysis of 212 previous studies supplemented with twin studies published after 1981, Devlin and colleagues showed that an alternative model with two maternal womb environments, one for twins - both monozygotic and (MZ) and dizygotic (DZ) - and another for siblings, fit the data much better.

Maternal effects, often assumed to be negligible, account for 20 % of the covariance between twins and 5 % between siblings, thereby correspondingly reducing the effects of genes, so the two measures of heritability were both less than 50 %: the broad and narrow sense heritability were 48 % and 34 % respectively.

The shared maternal environment may explain the striking correlation between the IQs of twins, especially adult twins reared apart.  It also accounts for age-effects: an apparent increase in heritability with age. Devlin and colleagues pointed out that cultural inheritance and interaction between genes and environment may also be at work to boost the apparent heritability of intelligence.

There is substantial brain growth in utero, and the brain has 70 % of its final mass within a year of birth.  IQ is known to be affected by prenatal environment: it is positively correlated with birth weight. Twins usually weigh less than singletons, and score on average 4-7 points lower on IQ tests.

Devlin and colleagues rejected Herrstein and Murray’s conclusion; instead, they believed that “Interventions aimed at improving the prenatal environment could lead to a significant increase in the population’s IQ.”

Devlin and colleagues may well have underestimated the shared maternal environment for MZ twins, which in addition to sharing the same womb as for DZ twins, usually share the same placenta, and more importantly, originate from the same egg with common cytoplasmic components, including mitochondrial DNA and transcripts and gene products that control early embryonic development [8]. Common cytoplasmic effects will be expected to further reduce heritability estimates.

Virtual twin studies and rearing environment

Nancy Segal and colleagues at California State University Fullerton in the United States have pioneered the study of behaviour in ‘virtual twins’ (VTs): same-age, unrelated siblings reared together since infancy. VTs replicate the rearing environment of twins but without the genetic relatedness, thereby enabling direct assessment of shared environmental effects on behaviour. Virtual twins are created in adoption, in which infants were adopted before one year of age; the unrelated sibling differing by less than 9 months in age, attend the same school grade, the pair being free of adverse birth events, and at least 4 years old. The foster homes are predominantly upper-middle class.

In an updated analysis of IQ data based on a sample of 142 VT pairs, the VTs mean IQ score was 105.83 (SD = 13.37) and correlation between VTs is 0.28 (p < .001), showing a substantial contribution of rearing environment during infancy [9].

The mean IQ score of the biological siblings exceeded that of the adopted siblings and when the paired data for members of 49 adopted-biological pairs were examined, biological children scored 113.08 (SD = 14.64), whereas adopted children scored 105.67 (SD 12.53), a difference of 7.41 points. 

Significantly, there was greater similarity in IQ scores between adopted-biological than adopted-adopted pairs, resulting in correlations of 0.47 vs 0.1. Similar results have been found by other research groups, suggesting that the environmental stimulation from a high-IQ biological child may also enhance adopted sibling’s IQs. The IQ correlation of the adopted-biological pairs (0.47, p < .001) approaches that of DZ twins (0.46) and full siblings (0.47) reported by others.

The research of Segal’s group that Devlin’s group together makes it highly likely that common rearing environment during infancy and maternal effects could account for most, if not all the heritability in IQ that has been attributed to genes.

Socioeconomic status and IQ

The higher IQ scores of the biological children relative to the adopted children observed by Segal’s team [9] are not surprising in view of the predominance of upper middle class parents in the study, whereas adopted children are predominantly from parents of lower socioeconomic status (SES). Studies dating to the 1990s have shown that adopted children typically score 12 points or higher than siblings left with birth parents or children adopted by lower SES parents [1]. A meta-analysis carried out in 2005 found an average effect of adoption of 18 points when extremely deprived institutional settings were included in the comparison [10].

What correlates SES with IQ? There are marked differences beginning in infancy, between the environment of higher SES families and lower SES families in factors that are likely to influence intellectual growth, including nutritional status, which will be considered in the next article in this series ( [11] How to Improve the Brain Power & Health of a Nation, SiS 53). A study published in 1995 [12] showed that by the age of three, a child of professional parents would have heard 30 million words spoken, while a child of working-class parents would have heard 20 million words, and a child of unemployed African American mothers would have heard 10 million words. The child of professional parents received six encouragements for every reprimand, the child of working-class parents received two encouragements per reprimand, and the child of unemployed African American mothers received two reprimands for every encouragement. These findings were extended using HOME technique (Home Observation for Measurement of the Environment) (see [1]). HOME researchers assess family environments for the amount of intellectual stimulation: how much parents talk to the child, how much access to books, magazines, newspaper, and computers, how much the parents read to the child, how many learning experiences outside the home (trips to museums, visits to friends); degree of warmth of parents versus punitive behaviour towards the child, etc. Very substantial association was found between HOME scores and IQ scores: a 1 SD difference in summed HOME scores is associated with a 9 point difference in IQ. These studies do not separate genetic and environmental contributions, but as the authors of the new report of the American Psychological Association commented [1]: “It is almost surely the case, however, that a substantial fraction of the IQ advantage is due to the environments independent of the genes associated with them.” That is because of the knowledge that adoption adds 12-18 points to the IQ of unrelated children, who are usually from lower SES backgrounds.

Shared environmental effect on IQ applies not just to children. According to a review of six well-conducted studies, share environment effect in adulthood is about 0.16 on average.

Consequently, most if not all twin studies, especially studies of adults, overestimate heritability of IQ, especially as lower SES individuals are difficult to recruit to laboratories and testing sites.

Epigenetic and cultural inheritance

Human population geneticist Marcus Feldman and his colleagues Stanford University California in the US and University of Aarhus in Denmark noted that current models are strictly based on Mendelian genetics, failing to consider non-Mendelian epigenetic modifications of genes in response to environmental states (see [13] Epigenetic Inheritance - What Genes Remember and other articles in the series, SiS 41, also [14, 15] Mismatch of RNA to DNA Widespread, and How Food Affects Genes, SiS 53). These epigenetic modifications usually do not involve DNA base sequence alterations and are hence not detected in SNP scans; they are also independent of SNP variations. Instead, they involve chemical markings of DNA or histone proteins that bind to DNA, or other mechanisms that change the state of expression of certain genes. Epigenetic modifications are often passed on to subsequent generations.    

Feldman and colleagues addressed the problem at the level of biometrical models by extending the models to include epigenetic and environmental effects [16].

They found that variation in epigenetic state and environmental state can result in highly heritable phenotypes through a combination of epigenetic and environmental inheritance. These two inheritance processes together can produce familial covariances (correlations) significantly greater than those predicted by models of purely epigenetic inheritance, and similar to those expected from genetic effects.  The results suggest that epigenetic variation, inherited both directly, and through shared environmental effects, may make a key contribution to the missing heritability.

In other words, epigenetic and environmental effects working together can account for practically all the variation now attributed to the genes.

Chief among environmental effects is cultural inheritance. Feldman and colleagues referred to the aggregation of the disease Kuru in families of the Fore tribe of Papua New Guinea due to the ingestion of a prion protein (infectious protein agent) during funeral rituals in which dead relatives or close acquaintances are consumed. This case of purely cultural inheritance was originally mistaken for a genetic disorder because of high disease correlations between relatives.

A well-studied case of environmental epigenetic inheritance is the mother’s licking-grooming of offspring in mice, which induces epigenetic changes in the offspring, influencing its response to stress as adults, and perpetuates the maternal behaviour in her female offspring (see [17] Caring Mothers Strike Fatal Blow against Genetic Determinism, SiS 41). This results in highly-correlated behaviour between mother and offspring in both maternal behaviour and response to stress as adults, even though the epigenetic modifications are erased during early embryogenesis. Consequently, cross-adoption between mothers with high and low licking-grooming behaviour will break the biological mother-offspring correlations in a single generation.

Identical twins are not genetically identical

As noted above, biometrical genetics models are based on classical Mendelian inheritance, in which genes are immune to direct or predictable environmental influence and passed on unchanged to the next generation except for rare random mutations.  The old paradigm has been discredited at least as far back as the late 1970s (see [18, 19] Genetic Engineering Dream or Nightmare and Living with the Fluid Genome, ISIS publications), long before the Human Genome Project was conceived.

In the post-genomics era, an increasing number of geneticists have begun to take notice of non-Mendelian inheritance and its invalidation of the basic tenets of biometrical genetics.

In a paper about to be published in Behavioral and Brain Sciences, Evan Charney at Duke University speaks of a ‘paradigm shift’ in the science of genetics. He points to recent discoveries of numerous processes that create extensive mutations in genome sequences and structure, as well as epigenetic modifications, which are completely at odds with the Mendelian model of inheritance underpinning heritability estimates [8]. Individuals do not have genes that are immutable throughout life, nor do they have the same genes in every cell of the body. He highlights retrotransposons – jumping genes that replicate and integrate themselves into different sites in the genome – which alter the sequence and state of activity of many genes; copy number variation and chromosomal abnormalities (aneuploidy) similarly, occur frequently in somatic cells as well as germ cells, both as part of normal development and in response to noxious environmental stimuli. Different tissues show distinctly different propensity for change; brain cells being especially prone to such modifications.  These add to the already large repertoire of epigenetic modifications that modify genes in response to environmental stimuli (see [11, 13-15, 18, 19] most notably in the brain [20] Rewriting the Genetic Text in Human Brain Development, SiS 41).

The fundamental assumption of twin studies - that monozygotic twins share 100 % of their genes – is demonstrably false. MZ twins differ, to begin with, in the mitochondrial DNA (mtDNA) complement allocated in cell division of the original oocyte that generated the twins. The oocyte may have had different sets of mtDNA, a condition referred to as heteroplasmy. MZ twins diverge substantially in epigenetic modifications as well as retrotransposition, copy number variations and aneuploidy throughout life. Although the numerous processes that alter genomes occur in normal development, perhaps as part of ‘natural genetic engineering’ [17, 18], the same processes are known to be involved in many behavioural, psychiatric, and neurodegenerative diseases, leaving us in no doubt that they have phenotypic consequences [8].   

In addition, stochastic nonlinear developmental changes account for substantial divergence in the activities of different brain regions between twins [21].

To conclude

No genes for intelligence can be found in the human genome. Instead, common environments, including maternal and rearing environments, along with epigenetic and cultural inheritance create substantial correlations between genetically unrelated individuals, while even ‘identical twins’ diverge genetically and epigenetically throughout life.

The fundamentally circular causation between genes and environment makes it futile to separate genetic from environmental contributions to development (see [22] Development and Evolution Revisited, ISIS scientific publication).  Consequently, we must redouble all efforts at appropriate interventions to improve the mental and physical wellbeing of the nation ([22, 23] How to Improve the Brain Power & Health of a Nation, SiS 53).