Never has the nature vs. nurture question been more hotly debated than over the question of intelligence. To what degree are individual differences in IQ due to environment or heredity? Traditionally, the heritability of intelligence—the extent to which differences in intelligence are due to differences in genes—has been calculated by testing the intelligence of closely related individuals, such as twins, and comparing the results to tests of unrelated people. It has been much harder to examine the human genome directly in order to determine the extent to which genetic variation is associated with IQ differences.
However, our understanding of the genetics of intelligence is progressing quickly. Less than 10 years ago, a few studies had found gene variants associated with intelligence in certain population samples, but the results could not be replicated when the same tests were done on other populations. Today, on the basis of a survey of thousands of genes, we can assign someone a score that correlates moderately with performance on standardized tests which line up extremely well with IQ scores.
The genetics of personality has progressed less quickly than that of intelligence, but is also advancing, and a paper released this month takes us one step closer to solving two problems that still plague both fields.
First, there is the issue of “missing heritability.” This refers to a consistent discrepancy between the two kinds of heritability estimates. Studies of twins and adopted relatives consistently report higher heritability estimates than do studies that survey the genomes of unrelated individuals and predict trait similarity based on gene variants. Twin and adoption studies suggest that intelligence is between 50 and 80 percent heritable, while personality traits such as neuroticism and extraversion are between 35 percent and 50 percent heritable. Genome sequencing studies, by contrast, have typically estimated the heritability of intelligence at between 20 percent and 50 percent, and the heritability of personality to be between 0 percent and 15 percent. Those who believe that intelligence is largely influenced by environment argue that these lower figures discredit traditional estimates of heritability.
Secondly, there is the problem of how there can be substantial genetic variation at all in traits that contribute to evolutionary fitness. Heritability and evolution go together naturally, but if variations in intelligence and personality lead to differences in fertility—as they surely do—then we might expect evolution to have removed all significant genetic variation that does not contribute to the optimal values of these traits. Instead, we see lots of variation in these traits even within the same populations. How, evolutionarily, could this happen?
The concept of mutational load offers one possible explanation for both problems. Mutational load refers to the number of fitness-reducing mutations in a person’s genome. When these mutations have a large effect, selection removes them quickly, but mutations with small harmful effects remain in a population for many generations. For instance, a new mutation that occurs in a single individual and that decreases evolutionary fitness by 0.1 percent will last an average of 14 generations and afflict roughly 1,000 people before it is eliminated from the gene pool. Individuals end up differing in how many damaging mutations they have inherited and this may be one cause of genetic differences in mental traits. Because new random mutations happen each generation, this is one mechanism by which differences in a trait can remain heritable even in the face of natural selection which should have eliminated these differences.
Such mutations could also explain why sequencing studies produce lower heritability estimates than twin studies do. What normally happens in sequencing studies is that a fraction of the known polymorphisms (gene variations) present in the human genome are sequenced, and similarity in this sample of polymorphisms is used as a proxy for total genetic similarity. This method works well when accounting for variance in a trait caused by gene variants that are common in the population and that are either at the exact spots of the genome that are sampled and sequenced or are predictable based on a person’s gene variants at those spots. However, when similarity in gene variants causally related to a trait only poorly correlates with similarity in gene variants that are actually sampled, then sequenced genetic similarity of just this portion of the genome becomes a poor proxy for the kind of genetic similarity that actually matters.
This is what we might expect if slightly harmful mutations are causing people to differ in a trait. Suppose that a given random mutation is present only in one in 1,000 people, and it is located near, but not on, one of the sequenced sites of people’s genomes. The vast majority of a population—which would have every version of that sequenced part of the genome—will not have this mutation. Thus two individuals who have the same gene variant at the sequenced gene will not be distinguished from each other—even if one has the mutation—because the mutation will not have been sequenced.
Based on this reasoning, several recent studies have looked at whether differences in mutations can predict differences in intelligence. Of six such studies, only one found that intelligence test scores significantly and negatively correlated with levels of mutational load. Four found a negative association that was statistically insignificant. This lack of statistical significance means that people with more mutations did tend to be less intelligent, but the effect was so small that it might have arisen by chance in this sample. One study found that more mutations had an insignificant but positive relationship with intelligence. Finally, one study found that mutational load significantly predicts educational attainment, which is a well-known correlate of intelligence, such that the more mutations a person had the fewer years of school he tended to complete.
|Study||Sample Size||Effect of mutations on intelligence|
|Yeo et al., (2011)||74||Negative, significant|
|MacLeod et al., (2012)||3,210||Negative, not significant|
|McRae et al., (2013)||800||Negative, not significant|
|Bagshaw et al., (2013)||723||Positive, not significant|
|Marioni et al., (2014)||1,596||Negative, not significant|
|Kikpatrick et al., (2014)||6,199||Negative, not significant|
|Ganna et al., (2016)||14,133||Negative, significant|
The non-significant findings do not discredit this approach as much they may appear to. For one thing, those studies used small samples. They may therefore repeatedly be finding insignificant effects because the real effect is too small to measure reliably with small samples. Furthermore, all of the above studies have been able to look at only a few types of the mutations that can occur in the genome. So, in addition to the sample size problem, it may be that the actual effect is too weak to measure accurately when looking at only a few mutation types. Both difficulties can contribute to statistically insignificant findings.
A new paper by Professor David Hill and colleagues overcame these limitations by examining both personality and intelligence in a large sample of roughly 20,000 people and using a method that allowed them to capture variation caused by all types of mutations. Professor Hill’s team was able to do this by looking at genetic similarity among not only unrelated people but also members of the same extended family. Looking at extended families is helpful because even people as distantly related as 4th cousins are much more likely than strangers to share rare mutations that they have inherited from their shared ancestors. Because of this, genetic similarity within families is a far better measure of similarity in mutational load than is the genetic similarity of unrelated people. Professor Hill and his colleagues used these data to separate individual variation in IQ and personality caused by differences in common gene variants in the general population on the one hand, and differences due to rare mutations that are common only within families on the other.
They also controlled for environmental effects. Prof. Hill and his colleagues measured the degree to which being part of the same family, being siblings, or being a couple predicted similarity in IQ and personality. This was necessary because living in the same household, as families and couples do, means having a very similar environment, which can influence traits in non-genetic ways.
The results for intelligence were promising: General genetic similarity accounted for 23 percent of variability in IQ, family-specific genetic similarity explained another 31 percent of the variation, 9 percent was due to the shared environment of siblings (that is, was not genetic), and 22 percent was predictable by people being a couple. Professor Hill’s findings showed that, in his sample, 54 percent of variation in intelligence could be predicted based on genes alone, and most of this variation was due to family-specific genetic effects. These figures are higher than previous gene-based studies because Prof. Hill was far better able to account for the effects of mutations. Importantly, this gene-based heritability estimate is close to the estimates for intelligence that result from twin and adoption studies.
Even this heritability estimate may be low. Being a couple was one of the variables meant to control for environmental similarity, on the assumption that living in the same household makes couples more similar to each other. However, married couples tend to be more genetically similar than average. Therefore, this variable probably captured trait similarity due to both genes and the environment and both these components were subtracted from the study’s heritability estimates, meaning that the heritability figures produced are probably too low.
The results for personality were less promising. For neuroticism, 11 percent of variation was attributable to general genetic similarity, and 19 percent was due to family-specific genetic similarity. For extraversion, 13 percent of variation could be accounted for by general genetic similarity, while family-specific genetic effects had no significant effect, and the family environment explained only 9 percent of variability. These weaker results may be due to gene-gene interactions in which the effect of one gene is dependent on the version of another gene that a person has. It may be that gene-gene interaction plays a larger role in the genetics of personality, and such effects are extremely difficult to capture in studies like this.
Returning to intelligence, it is worth noting that deleterious mutations with small effects may also be involved in racial differences in intelligence. The psychologist Phillipe Rushton showed that the more inbreeding lowers a given cognitive ability, the more blacks and whites tended to differ in that ability. Inbreeding depression is caused by the heightened probability that damaging recessive mutations will be expressed when people mate with family members. Thus, the correlation between the black-white IQ gap sizes on these subtests and the degree of inbreeding depression implies that racial differences in rates of mutation may be a causal factor in racial intelligence differences.
By taking into consideration individual differences in mutation load, genomic analysis is arriving at heritability estimates of intelligence that come close to those based on kinship studies. Once the genetic and mutation patterns associated with intelligence become clearly established, it should be possible to test for a person’s genetic potential for intelligence through direct genetic assessment rather than relying on standard IQ tests.
We still have a lot to learn about the genetic underpinning of the heritability of traits such as intelligence and personality, but we are learning more each year. It can be expected that this research will pave the way for repeated demonstrations of our ability fully to account for the heritability of IQ using genes alone, which would silence the blank slatists—those who insist that intelligence is overwhelmingly influenced by environment—once and for all.