Genetics and intelligence differences: five special findings

This article discusses five genetic findings that are specific to intelligence differences, including the increase in heritability from infancy to adulthood, high genetic correlations among diverse cognitive abilities, high assortative mating, the positive genetics of high intelligence, and the impact of intelligence on social epidemiology.

OPEN
EXPERT REVIEW
Genetics and intelligence differences: five special finding
R Plomin1 and IJ Deary2,3
Intelligence is a core construct in differential psychology and behavioural genetics,and should be so in cognitive neuroscience.It is
one of the best predictors of important life outcomes such as education,occupation,mentaland physicalhealth and illness,and
mortality.Intelligence is one of the most heritable behaviouraltraits.Here,we highlight five genetic findings that are specialto
intelligence differences and that have important implications for its genetic architecture and for gene-hunting expeditio(i) The
heritability of intelligence increases from about 20% in infancy to perhaps 80% in later adulthood.(ii) Intelligence captures genetic
effects on diverse cognitive and learning abilities,which correlate phenotypically about 0.30 on average but correlate genetica
about 0.60 or higher.(iii) Assortative mating is greater for intelligence (spouse correlations ~0.40) than for other behaviou
such as personality and psychopathology (~0.10) or physicaltraits such as height and weight (~0.20).Assortative mating pumps
additive genetic variance into the population every generation,contributing to the high narrow heritability (additive genetic
variance) of intelligence.(iv) Unlike psychiatric disorders,intelligence is normally distributed with a positive end of exceptional
performance that is a modelfor ‘positive genetics’.(v) Intelligence is associated with education and socialclass and broadens the
causalperspectives on how these three inter-correlated variables contribute to socialmobility,and health,illness and mortality
differences.These five findings arose primarily from twin studies.They are being confirmed by the first new quantitative genetic
technique in a century—Genome-wide Complex Trait Analysis (GCTA)—which estimates genetic influence using genom
genotypes in large samples of unrelated individuals.Comparing GCTA results to the results of twin studies reveals important
insights into the genetic architecture of intelligence that are relevant to attempts to narrow the ‘missing heritability’gap.
Molecular Psychiatry (2015) 20,98–108;doi:10.1038/mp.2014.105;published online 16 September 2014
INTRODUCTION
Nearly a century ago,intelligence wasthe firstbehaviouraltrait
studied using newly emerging quantitative genetic designs such as
twin and adoption studies.1–4 Such studies have consistently shown
thatgenetic influence on individualdifferencesin intelligence is
substantial.5,6Intelligence has become the target of molecular genetic
studies attempting to identify genes responsible for its heritability.
Here, we refrain from providing another general overview of the
genetics of intelligence. We begin by noting three regularities that
might almost be dubbed ‘laws’ from genetic research that apply to
many traits in the life sciences.The bulk of our review highlights
genetic findings that are specific to intelligence rather than these
generallaws.
THREE ‘LAWS’OF THE GENETICS OF COMPLEX TRAITS
(INCLUDING INTELLIGENCE)
All traits show significant genetic influence
Finding that differencesbetween individuals(traits,whether
assessed quantitativelyas a dimension orqualitativelyas a
diagnosis)are significantly heritable is so ubiquitous forbeha-
viouraltraitsthat it has been enshrined asthe first law of
behaviouralgenetics.7 Although the pervasiveness of this finding
makes it a commonplace observation,it should not be taken for
granted,especially in the behaviouralsciences,because this was
the battleground for nature-nurture wars until only a few decades
ago in psychiatry,8 even fewer decades ago in psychology,9 and
continuing today in some areas such as education.10,11
It might be
argued thatit is no longersurprising to demonstrate genetic
influence on a behaviouraltrait,and that it would be more
interesting to find a trait that shows no genetic influence.
No traits are 100% heritable
For some areas of behaviouralresearch—especially in psychiatry
—the pendulum has swung so farfrom a focus on nurture to
a focus on nature that it is important to highlight a second law
of genetics forcomplex traits and common disorders:All traits
show substantialenvironmentalinfluence,in that heritability
is not 100% for any trait.Acceptance ofthe importance of
both genetic and environmentalinfluences leads to interestin
the interplay between genesand environment,such as their
interaction(moderation)and correlation(mediation)in the
development of complex traits,Plomin et al.6 pp 105–127.
Heritability is caused by many genes of smalleffect
The first two laws come from quantitative genetic research,which
uses,for example,the twin method to assess the net contribution
of genetics to individualdifferences withoutknowledge ofthe
genetic architecture ofa trait, such asthe numberof genes
involved ortheir effectsizes.A third law has emerged from
molecular genetic research that attempts to identify specific ge
responsible forwidespread heritability,especially genome-wide
association (GWA) studies of the past few years:The heritability of
1King's College London,MRC Social,Genetic & Developmental Psychiatry Centre,Institute of Psychiatry,DeCrespigny Park,London,UK;2Department of Psychology,University of
Edinburgh,Edinburgh,UK and3Centre for Cognitive Ageing and Cognitive Epidemiology,University of Edinburgh,Edinburgh,UK.Correspondence:Professor R Plomin,King's
College London,MRC Social,Genetic & DevelopmentalPsychiatry Centre,PO80,Institute of Psychiatry,DeCrespigny Park,Denmark Hill,London SE5 8AF,UK.
E-mail:[e-mail]
Received 14 March 2014;revised 18 July 2014;accepted 22 July 2014;published online 16 September 2014
Molecular Psychiatry (2015) 20,98–108
© 2015 Macmillan Publishers LimitedAll rights reserved 1359-4184/15
www.nature.com/mp

traits is caused by many genes ofsmalleffect.12 This was the
premise ofquantitative genetic theory setout nearly a century
ago,13 but quantitative genetic methods themselves could not
shine much light on the distribution of the effect sizes of genes in
the population.For decades,the failure oflinkage analyses to
identify replicable linkagesto chromosomalregionscould be
interpreted as supportfor this hypothesis because linkage has
little powerto detectsmalleffectsizes.However,GWA studies
have made it clear that the largest effect sizes of associations are
very smallindeed.For example,we are aware ofalmostno
replicated genetic associations that account for more than 1 per
cent of the population variance ofquantitative traitssuch as
height and weight.Because GWA studies have adequate power to
detect such effect sizes,we can conclude that there are no larger
effectsizes,at leastfor the common single-nucleotide variants
that have been used in such studies to date.If the largest effect
sizes are so small,the smallest effect sizes must be infinitesimal,
which means that such associations willbe difficult to detect and
even more difficultto replicate.For example,the largestGWA
study ofintelligence differences,which included nearly 18 000
children,found no genome-wide significantassociations.The
largesteffect sizes accounted for0.2% of the varianceof
intelligence scores.14 Another recent GWA study ofa sample of
1500 children reported an association that accounted for 0.5% of
the variance of intelligence scores,15 but this association showed
no effect in the study of18 000 children (P = 0.73;Benyamin B,
personalcommunication).A GWA of educationalattainment—
which correlatesmoderately with intelligence—included more
than 125 000 individuals;the DNA variant with the largest effect
size accounted for 1% of the variance in years of education but the
variance explained wasonly 0.02% in a replication sample.16
‘Missing heritability’is the catch-phrase to describe the great gulf
between heritability and the variance explained by associations
with specific DNA variants.
Ratherthan reviewing evidence forthese generallaws in
relation to intelligence,our review focuses on five findings from
genetic research that are specific to intelligence.Because ofthe
controversy and confusion that continues to surround intelligence,
especially in the media and the generalscience literature,11 we
begin by brieflydiscussing the definition,measurementand
importance of intelligence.
WHAT IS INTELLIGENCE AND WHY IS IT IMPORTANT?
Although there are manytypes of cognitive abilitytests of
individualdifferences,they almost allcorrelate substantially and
positively; people with higher ability on one cognitive task tend to
have higher ability on all of the others.Intelligence(more
precisely,generalcognitive ability or g,as discovered and defined
by Spearman in 190417
) indexes this covariance,which accounts
for about40 percentof the totalvariance when a battery of
diverse cognitive tests is administered to a sample with a good
range ofcognitive ability.18,19As long as a battery ofcognitive
tests is diverse and reliable,a general‘factor’(often represented
by the first unrotated principalcomponent,which is not strictly a
factor,but thatis the terminology thatis often used)indexing
intelligence differences will emerge and correlate highly with such
factorsderived from other batteriesusing wholly different
cognitive tests.20 The generalintelligence component(factor)is
a universally found statisticalregularity,which means that some
have tried to provide an epithetfor what it might capture.
According to one view,the core of this general intelligence factor
is ‘the ability to reason,plan,solve problems,think abstractly,
comprehendcomplexideas, learn quickly,and learn from
experience’(Gottfredson etal.21 p.13;see also Deary22
). Intelli-
gence is atthe pinnacle ofthe hierarchicalmodelof cognitive
abilities that includes a middle levelof group factors,such as the
cognitive domains of verbal and spatial abilities and memory,and
a third level of specific tests and their associated narrow cognitiv
skills.18,23
Intelligence isimportantscientificallyand socially.Because
intelligence represents individualdifferences in brain processes
working in concertto solve problems,it is centralto systems
approachesto brain structure and function,24–26 and to the
conceptualisation ofhow diverse cognitive abilities decline with
age.27It is also one of the most stable behavioural traits, yielding a
correlation of 0.63 in a study of people tested at age 11 and then
again at age 79.28 Socially,intelligenceis one of the best
predictors ofkey outcomes such as education and occupational
status.29 People with higherintelligence tend to have better
mental and physical health and fewer illnesses throughout the lif
course,and longer lives.22,30
The rest of this review describes five genetic findings that are
specialto intelligence differences:dramatic increasesin herit-
ability during the life course,high genetic correlations among
diverse cognitive abilities,high assortative mating,the positive
genetics ofhigh intelligence and the impactof intelligence on
‘socialepidemiology’.Most ofthese findings are not new,31 but
highlighting these findings as specialfor intelligence isnovel.
Moreover,support for these findings has increased in recent years
from traditionalquantitative geneticresearch using the twin
design thatcompares identicaland fraternaltwins,and,impor-
tantly,from a new quantitative genetic method thatuses DNA
alone to estimate overallgenetic influence in large samples of
unrelated individuals.This method,which we will referto as
Genome-wide Complex Trait Analysis (GCTA),32–35is the first new
human quantitative genetic method in a century,and is described
in Box 1.
Heritability increases dramatically from infancy through adulthoo
despite genetic stability
It would be reasonable to assume thatas we go through life,
experiences—Shakespeare’s ‘whips and scorns oftime’—have a
cumulative effecton intelligence,perhapsoverwhelming early
geneticpredispositions.However,for intelligence,heritability
increases linearly,from (approximately) 20% in infancy to 40% in
adolescence,and to 60% in adulthood.Some evidence suggests
that heritabilitymight increase to asmuch as 80% in later
adulthood47 but then decline to about 60% after age 80.48
Most genetic research has been consistent with this dramatic
increase in heritability forintelligence in the early human life
course. Figure 1 shows the results of the first study to demonstra
significant increases in heritability in cross-sectional analyses of 1
000 twin pairs from childhood (~40%)to adolescence (~50%)to
young adulthood (~60%).49 The non-overlapping standard errors
across the three ages indicate that the increases in heritabilities
are significant.Although thesefindingshave been criticised
because they rely on cross-sectional comparisons (Mackintosh50p.
278),similarresults showing increases in heritability have been
found in longitudinaladoption studies51,52as well as in long-
itudinal twin studies from early to middle childhood53,54
and from
middle childhood to adolescence.55 Although GCTA can be used
to test this finding of increasing heritability across development,
the firsttwo attempts to do so using longitudinaldata did not
have sufficient power to detect the hypothesised age differences
in GCTA heritability.One study reported an increase in GCTA
heritability of intelligence from 0.26 (0.17 standard error) at age
to 0.45 (0.14)at age 12.56 Anotherstudy reported a decrease
in GCTA heritability from 0.48 (0.18)at age 11 to 0.28 (0.18)in
old age.46 Given the differencesin the ages tested in these
latter two studies,they are not directly comparable.As indicated
by the large standard errors,largerlongitudinalstudiesare
needed.
Why does the heritability ofintelligence increase so dramati-
cally from childhood to adulthood,as seen in twin studies? A clear
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
99
© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015),98 – 108

yet apparently contradictory finding constrains possible answers
to this question.Despite this greatincrease in heritability,the
same genes affect intelligence from age to age.For example,a
recent twin study reported a genetic correlation of 0.75 (standard
error = 0.08)from age 7 to age 12,despite increasing heritability
from 0.36 (0.03)to 0.49 (0.04)and despite mean changesin
brain structure and function from childhood to adolescence.55
GCTA analysesin the same study butusing unrelated indivi-
duals yielded a highly similargenetic correlation of0.73 (0.29)
from age 7 to age 12.Most strikingly,a 60-yearlongitudinal
study of intelligence,which was the first application of bivariate
GCTA,yielded a genetic correlation of0.62 (0.22)from age 11
to 69.46
Thus,the question becomes,why does the heritabilityof
intelligence increase during development despite strong genet
stabilityfrom age to age? That is, the same genes largely
affectintelligence across the life course and yetgenes account
for more variance as time goes by.Increasing heritability despite
genetic stability implies some contribution from what has been
called genetic amplification.57 This has recently been supported
in a meta-analysis of11 500 twin and sibling pairs with longi-
tudinal data on intelligence that found that a genetic amplificat
modelfit the data betterthan a modelin which new genetic
influencesarise with time.58 Genotype-environmentcorrelation
seems the most likely explanationin which small genetic
differences are magnified as children select,modify and create
environmentscorrelated with theirgeneticpropensities.This
active model of selected environments—incontrastto the
traditionalmodel of imposed environments—offersa general
paradigm for thinking about how genotypes become
phenotypes.59
Box 1 The first new quantitative genetic method in a century:Genome-wide Complex Trait Analysis (GCTA)
A new method for estimating genetic influence using DNA is a welcome addition to the armamentarium of quantitative geThe
significance of the method is that it can estimate the net effect of genetic influence using DNA of unrelated individuals ra
relying on familialresemblance in groups of specialfamily members such as monozygotic and dizygotic twins who differ in gene
relatedness.The method is often called GCTA,although its developers refer to it as Genomic-Relatedness-Matrix Restricted Maxi
Likelihood.32–35Other methods36 and modifications37,38are also emerging.39,40
Like other quantitative genetic designs such as the twin design,GCTA uses genetic similarity to predict phenotypic similarity.However,
instead of using genetic similarity from groups differing markedly in genetic similarity such as monozygotic and dizygoticGCTA
uses genetic similarity for each pair of unrelated individuals based on that pair’s overall similarity across hundreds of tho
nucleotide polymorphisms (SNPs) for thousands of individuals;each pair’s genetic similarity is then used to predict their phenotyp
similarity.Even remotely related pairs of individuals (genetic similarity greater than 0.025,which represents fifth-degree relatives) are
excluded so that chance genetic similarity is used as a random effect in a linear mixed model.The power of the method comes from
comparing not just two groups like monozygotic and dizygotic twins,but from the millions of pair-by-pair comparisons in samples o
thousands ofindividuals.In contrastto the twin design,which only requires a few hundred pairs oftwins to estimate moderate
heritability,GCTA requires samples ofthousands ofindividuals because the method attempts to extracta smallsignalof genetic
similarity from the noise of hundreds of thousands of SNPs.A handy power calculator is available,which underlines the large samples
needed for GCTA (http://spark.rstudio.com/ctgg/gctaPower/).
GCTA detects only those genetic effects tagged by the common SNPs (allele frequencies typically much greaterthan 1%)that
have untilrecently been incorporated in commercially available DNA arrays used in GWA studies.This limitation is changing as
exome arrays became available in 2013 thatincluded rare SNPs in ornearexomes (http://res.illumina.com/documents/products/
datasheets/datasheet_human_core_exome_beadchip.pdf);the limitation willbe lifted as whole-genome sequencing is more widely
used.In addition,GCTA is limited to detecting the additive effects ofSNPs;it cannotdetectgene–gene orgene–environment
interaction.Thus,GCTA heritability represents the upper limit for detection of SNP associations in GWA studies,which,like GCTA,are
limited to detecting additive effects of common SNPs.Conversely,GCTA heritability represents the lower limit for heritability estimat
in twin studies because twin studies can detect genetic influence due to DNA variants of any kind. In this way, the compa
GCTA and twin study estimates ofheritability reveals fundamentalinformation aboutthe genetic architecture ofcomplex traits,
including intelligence.
Similar to other complex traits,GCTA heritability estimates for intelligence are about half the heritability estimates from twin s6,41
This finding suggests that despite the modest yield so far from GWA studies of intelligence,14with sufficiently large samples, it should in
theory be possible to detect as much as half the heritability with the additive effects tagged by the common SNPs on cur
DNA arrays. The missing heritability gap between GCTA and twin studies is likely to be filled in part by less common DNA
willbe detected as whole-genome sequencing comes on line.42
The value of GCTA has been greatly increased by extending it beyond the univariate analysis of the variance of a single t
bivariate analysis of the covariance between two traits or the ‘same’trait at two ages;43,44
a recent approach is multivariate rather than
justbivariate.45 Bivariate GCTA was firstapplied to intelligence,yielding a high genetic correlation between intelligence scores in
childhood and old age,46 as described in the text along with other examples of bivariate GCTA.
Figure 1.A meta-analysis of11 000 pairs oftwins shows thatthe
heritability ofintelligence increasessignificantly from childhood
(age 9)to adolescence (age 12)and to young adulthood (age 17).
(Adapted from Haworth et al.49
).
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
100
Molecular Psychiatry (2015),98 – 108 © 2015 Macmillan Publishers Limited

Intelligence indexes generalgenetic effects across diverse
cognitive and learning abilities
Another special genetic feature of intelligenceis that its
differences are caused by genes that affect cognitive abilities as
diverse as,for example,spatialability,vocabulary,processing
speed, executive function and memory.Most of the genetic action
lies with these general(highly pleiotropic)effects,captured by
intelligence, rather than effects specific to each ability, leading to a
Generalist Genes Hypothesis.60This is a surprising finding because
very different neurocognitive processes appear to be involved in
such cognitive abilities.25 Although these genetic correlations put
intelligence at the pinnacle of the hierarchicalmodelof cognitive
abilities mentioned earlier,there is also genetic specificity that
builds the genetic architecture forthe restof the hierarchical
structure of group factors and specific tests.
In a meta-analysisof 322 studies,the average correlation
among individualdiverse cognitive tests is about0.3.18 Genetic
correlations among cognitive tests are typically greater than 0.6,
indicating that the same genes are responsible for the heritabil-
ities of these tests.60,61
Genetic correlations estimate the extent to
which genetic effects on one traitare correlated with genetic
effects on another trait independently of the heritabilities of the
two traits.They can be thought about roughly as the probability
that genes associated with one trait are also associated with the
other trait.Genetic correlationsare derived from the genetic
analysis of covariance between traits using the same quantitative
genetic methods used to analyse variance.6
These generalgeneticeffectspermeate notonly cognitive
abilities such as spatial and vocabulary that are used as part of th
assessmentof intelligence butalso extend to education-related
learning abilities such as reading and arithmetic.Figure 2 shows
the resultsof a multivariate genetic analysisof 14 teststhat
comprisefour distinct test batteries—intelligence,reading,
mathematicsand language—formore than 5000 pairsof 12-
year-old twins.62 The genetic correlations (and 95% confidence
intervals) between intelligence and learning abilities are uniforml
high:0.88 (0.84–0.92)with reading,0.86 (0.81–0.90)with mathe-
maticsand 0.91 with language(0.87–0.94).Weighting these
genetic correlations by the heritabilities of the latent factors, it ca
be shown thatabouttwo-thirds ofthe phenotypic correlations
between the factors can be explained genetically.One advantage
of using such latentfactorsis that they exclude uncorrelated
measurementerror.As a result,these genetic correlationsare
higherthan those found when uncorrected composite scores
rather than latent factors are analysed:0.66 (0.05 standard error)
for reading,0.73 (0.03)for mathematicsand 0.80 (0.06)for
language.63
The first attempts to use bivariate GCTA (see Box 1)to verify
these twin findings supportthe hypothesis ofgeneralgenetic
effects on broad cognitive and learning ability-related differences
The GCTA estimates ofgenetic correlation (and standard error)
between intelligence and learning abilities are highly similarto
the twin study estimates justmentioned forcomposite scores
uncorrected forerror:0.89 (0.26)for reading,0.74 (0.15)for
mathematicsand 0.81 (0.15)for language,estimatedfrom
Figure 2.Multivariate (common pathway)genetic analysis in which each latentvariable is indexed by three orfour tests and the twin
method is used to estimate additive genetic (A),shared (common) environmental (C) and nonshared environmental (E) contributions to the
variance and covariance among the latentvariables.Squares representmeasured traits;circles representlatentfactors.The lower tier
of arrows represents factor loadings;the second tierrepresents genetic and environmentalpath coefficients.The curved arrows atthe
top represent correlations between genetic and environmentallatent factors,although only the genetic correlations are shown here.(From
Davis et al.62
).
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
101
© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015),98 – 108

unrelated individuals from the same sample.63Within intelligence,
the major group factors of verbaland nonverbalability yielded a
genetic correlation of1.0 (0.32)in a bivariate GCTA in the same
sample.64 The high GCTA genetic correlation between verbaland
nonverbalbased on unrelated individualssupported the twin
study estimate of 0.60 (0.09) in the same study.
An important feature of bivariate GCTA is that it yields genetic
correlations similar to genetic correlations estimated from the twin
method,even though heritabilitiesare considerably lowerfor
GCTA than for twin estimates.In the study just mentioned,GCTA
heritabilities were consistently lower than twin heritabilities:0.35
vs 0.47 for intelligence,0.16 vs 0.59 for reading,0.32 vs 0.48 for
mathematics and 0.35 vs 0.41 forlanguage.As noted in Box 1,
GCTA heritability estimatesare limited to the additive effects
tagged by the common single nucleotide polymorphisms (SNPs)
used on DNA arrays (i.e., the direct effects of the SNPs on the array
and those variants with which they are in linkage disequilibrium);
GCTA heritability is lowered by imperfect tagging of causalSNPs.
As a result,GCTA heritability estimates are typically about half the
heritabilityestimatesfrom twin studies.This ‘missing GCTA
heritability’is due in part to non-additive effects and the effects
of rarerDNA variants.Why then are GCTA estimates ofgenetic
correlation so similar to twin study estimates? The likely reason is
that the GCTA estimate of the genetic correlation is derived from
the ratio between genetic covariance and the genetic variances
of the two traits.Because GCTA’sunderestimation ofgenetic
influence appliesto genetic covariance aswell as to genetic
variance,the ratio between genetic covariance and genetic vari-
ance cancels out this bias,leaving an unbiased GCTA estimate of
genetic correlation.63
This finding ofstrong genome-wide pleiotropy across diverse
cognitive and learning abilities,indexed by generalintelligence,
is a majorfinding aboutthe origins ofindividualdifferences in
intelligence.Nonetheless,this finding seems to have had little
impact in related fieldssuch as cognitiveneuroscienceor
experimentalcognitive psychology.We suggest that part ofthe
reason for this neglectis that these fieldsgenerallyignore
individualdifferences.65,66 Anotherreason mightbe that the
evidence forthis finding rested largely on the twin design,for
which there have alwaysbeen concernsabout some of its
assumptions;6 we judge that this willchange now that GCTA is
beginning to confirm the twin results.
This finding ofstrong genome-wide pleiotropy across diverse
cognitiveand learning abilitiesis compatiblewith multiple
neurocognitive models of causal pathways.The modularity model
of cognitive neuroscience might suggest that genetic correlations
among cognitive abilities are epiphenomenalin the sense that
multiple genetically independent brain mechanisms could affect
each ability,creating genetic correlations among abilities.How-
ever,the genetic principles of pleiotropy (each gene affects many
traits)and polygenicity (many genes affect each trait)lead us to
predict that generalist genes have their effects further upstream,
creating genetic correlations among brain structures and func-
tions,a prediction that supports a network view of brain structure
and function.25,67
In summary,multivariate genetic research—both from twin
studies and GCTA—suggests thatmostof the genetic action is
generalacross diverse cognitive abilities ratherthan specific to
each ability.Intelligenceis a good target for gene-hunting
because it indexes these generalist genes.
Assortative mating is greater for intelligence than for other traits
Although the phenotypic correlation between spouses, assortative
mating,might seem an esoteric topic,it has important implica-
tions for the geneticarchitectureof intelligence.Assortative
mating is fargreaterfor intelligence than formostothertraits.
For example,assortative mating is about 0.20 for height68 and for
weight,69 and about 0.10 for personality.70 For intelligence,
assortative mating is about 0.40.19,71
Moreover,verbal intelligence
shows greaterassortative mating (~0.50)than nonverbalintelli-
gence (~0.30),perhaps because it is easier to gauge someone’s
verbal ability such as vocabulary than their nonverbal intelligen
such as spatial ability.Assortative mating for intelligence is caused
by initialselection of a mate (assortment) rather than by couple
becoming moresimilarto each other after living together
(convergence).72,73 In part, spouses select each other for
intelligence on the basis ofeducation—spouses correlate about
0.60 for years ofeducation19
—which correlates about 0.45 with
intelligence.50 Assortative mating may be greaterthan it is for
intelligence for a few other traits such as social attitudes,smoking
and drinking,althoughthese traits might be affectedby
convergence.It should also be noted that not allof the genetic
variance forintelligence isadditive.For example,dominance,
which involves interaction among alleles at a locus,is indicated by
research showing inbreeding depression for intelligence.74 When
assortative mating is taken into account in variance componen
analysis,some evidence for nonadditivegenetic variance
emerges.73,75
The significance ofhigh assortative mating forintelligence is
that assortative mating forpolygenictraitsincreasesadditive
genetic variance.Additive genetic variance refers to the indepen-
denteffects ofalleles orloci that‘add up’,in contrastto non-
additive effects of dominance within a locus,and epistasis across
loci in which the effectsof alleles orloci interact.Assortative
mating ofparentsincreasesadditive genetic variance in their
offspring because offspring receive a random sampling of half
each parent’s genes and resemble their parents to the extent t
each allele shared with theirparentshas an average additive
effect.Because offspring inherit only one of each of the parents’
pairs of alleles,offspring differ from their parents for non-additive
interactions.
For example,if spousesmated randomlyin relationto
intelligence,highly intelligentwomen would be justas likely
to mate with men of low as high intelligence.Offspring of
the matingsof women of high intelligence and men oflow
intelligence would generally be of average intelligence.However,
because there is strong positive assortative mating,children with
highly intelligent mothers are also likely to have highly intellige
fathers,and the offspring themselvesare likelyto be more
intelligentthan average.The same thing happensfor less
intelligentparents.In this way, assortativemating increases
additive genetic variance in that the offspring differ more from
the average than they would if mating were random.The increase
in additive genetic variance can be substantialbecause its effects
accumulate generation aftergeneration untilan equilibrium is
reached.For example,if the heritabilityof intelligence with
random mating were0.40,the additivegeneticvarianceof
intelligence would increase by one-quarterat equilibrium given
assortative mating of0.40,Falconerand MacKay76 equation 5,
Table 10.6,p. 176.
The extra additive genetic variance for intelligence induced b
assortative mating is importantfor three genetic reasons.First,
parents share only additive genetic variance with their offsprin
so that genetic predictions from parent to offspring ought to be
greater for intelligence when polygenic scores,composite scores
based on associations of many loci with intelligence,are available.
Second,because GCTA hasso far been limited to detecting
additive genetic variance,GCTA heritability should be greater for
intelligence than for traits that show less assortative mating su
as personality.Some evidence supports this prediction in that
GCTA heritability estimates forpersonality appearto be much
lowerthan forintelligence,even taking into accountthe lower
twin-study heritability estimates forpersonality than forintelli-
gence.77–79 Moreover,GCTA heritability estimatesare greater,
although not significantlyso, for verbal than non-verbal
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
102
Molecular Psychiatry (2015),98 – 108 © 2015 Macmillan Publishers Limited

intelligence,41,80which is consistent with the greater assortative
mating forverbalthan non-verbalintelligence.Third,because
both GWA and GCTA are limited to detecting additive genetic
variance,the GCTA estimateof substantialadditivegenetic
influence on intelligence makes intelligence a good targetfor
GWA studies.
Two additionalpoints about assortative mating for intelligence
warrant mention.First,unlike inbreeding,which reduces hetero-
zygosity across the genome,assortative mating is trait specific—it
increases additive genetic variance (changing genotypic frequen-
cies but not allelic frequencies) only for genes associated with the
traitfor which mates assortand its genetically correlated traits.
Second,assortative mating induces a genetic correlation between
mates for a particular trait to the extent that the trait is heritable,
regardlessof whetherassortative mating isdriven by genetic
assortmentor by environmentalfactors such as propinquity.A
recent study using genome-wide genotypes showed that spouses
are more geneticallysimilarthan two individualschosen at
random.81 This DNA estimateof geneticsimilaritybetween
spouses is substantially less than assortative mating for education
levels,suggesting that assortative mating may be driven by ‘social
sorting processes in the marriage market’.81
Thinking positively:the genetics of high intelligence
Unlike psychiatric and otherdisorders,intelligence isnormally
distributed with a positive end of high performance as wellas a
problematicend of intellectualdisability.High intelligence is
responsible for exceptional performance in many societally valued
outcomes,as documented in long-term longitudinalstudies.82
Although many othertraits,such asthose related to athletic
performance, are also normally distributed, the importance of high
intelligence makes it especially interesting.Genetic exploration of
the positivetail of normallydistributedtraits is important
conceptually because itmoves away from the notion thatwe
are all the same genetically except for rogue mutations that cause
disorders,diseases and disabilities.
Quantitative genetic research on intelligence indicates that the
genetic causesof high intelligenceare quantitatively,not
qualitatively,different from the rest ofthe distribution.A recent
study of 11 000 twin pairsfound that the top 15% of the
intelligence distribution was just as heritable (0.50) as the rest of
the distribution (0.55).83 Most recently,in a study of 370 000
sibling pairs and 9000 twin pairs in Sweden from 3 million 18-year-
old males whose intelligence was assessed as part of compulsory
military service,not only was high intelligence (top 4%)justas
familialand heritable as the restof the distribution,a method
called DF extremesanalysissuggested thatthe same genetic
factors are at work.84 DF extremes analysis focuses on the genetic
causesof the average difference between an extreme group,
however defined,and quantitative trait scores for the population,
comparing the differentialregression to the population mean for
the co-twins ofidenticaland fraternaltwin probands.85 To the
extent that genetics is found to account for this average difference
(called ‘group’heritability),it implies that there is a high genetic
correlation between the extreme group and the quantitative
trait.60 In the Swedish study,DF extremes analysis showed that
genetics explained about half of the mean difference between the
high-intelligence group and the rest of the distribution,which was
similar to the traditionalheritability of individualdifferences and
implies strong genetic links between high intelligence and normal
variation in intelligence.
It is possible that scores more extreme than the top 4% of the
intelligencedistributionare aetiologicallydifferentfrom the
normaldistribution,which has been called the Genetic Disconti-
nuity Hypothesis.86 The mostpersuasive argumentfor genetic
discontinuity for extremely high intelligence was made by David
Lykken who noted that a key problem of genius is ‘its mysterious
irrepressibility and its ability to arise from the most unpromising
lineages and to flourish even in the meanestof circumstances’
(Lykken87 p. 29).Lykken87,88proposed that genius emerges from
unique combinations of genes;he referred to these higher-order
nonadditive (epistatic) interactions as emergenic.The emergenesis
hypothesis does not necessarily predict that different genes affec
high intelligence,but it does predict that genetic effects are non-
additive for high intelligence.The hallmark of an epistatic trait is
one for which identicaltwins are more than twice as similar as
fraternal twins.However,in the two twin studies described above,
high intelligence did notshow this pattern oftwin results and
model-fittinganalysesfound that all geneticinfluencewas
additive for high intelligence as wellas for the entire distribution
of intelligence.Althoughthese resultsdo not supportthe
Discontinuity Hypothesis,the studieswere limited to the top
15% and top 4% of the intelligence distribution,which is far short
of the extremes ofgenius,which Galton89 benchmarked as the
top 0.1%.
The aetiologyof high intelligenceis also interestingin
comparison to intellectualdisability.Similar to high intelligence,
most intellectualdisabilityis the low end of the normal
distribution ofintelligence.This has been shown mostrecently
in the Swedish conscriptsample mentioned above,with results
replicated in a similarly large conscript sample in Israel.90However,
extremely severe intellectual disability appears to be aetiological
distinct, as proposed by Lionel Penrose91in 1938 and confirmed in
the Swedish and Israelistudies.One criticalpiece ofevidence is
that siblings of persons with severe intellectualdisability have an
average intelligence quotient(IQ)near100 whereas siblings of
persons with mild intellectualdisability have an average IQ of
about85,aboutone standard deviation below the population
mean.The absence ofgenetic links between severe intellectual
disability and normalvariation in intelligence fits with current
moleculargeneticresearch thatfinds noninherited denovo
mutations associated with severe intellectualdisability.92
An hypothesis to integrate these genetic results for the low and
high ends of intelligenceis this: Normal developmentof
intelligence can be disrupted by any of many mutations including
non-inherited de novo mutations as well as prenatal and postnata
trauma,but high intelligence requires that everything works right,
including mostof the positive alleles and few ofthe negative
alleles associated with intelligence. This hypothesis is the rationa
for a recent genome-wide case–control association study for case
with extremelyhigh intelligence(IQ4150).84 However,one
study93 has found no associationbetweenrare SNPs and
intelligence in the normalrange of intelligence.In addition,
severalstudies have found no association between copy-number
variants,which are typically rare variants,and intelligence in the
normalrange,although such studiesmay have been under-
powered both in terms ofsample and difficulties in assessing
copy-number variants.94
Although the normalphenotypic distribution ofintelligence
makes it an obvious target for investigating the high as well as lo
extremes,the larger significance of positive genetics for psychia-
tric genetics is that polygenic scores created from GWA studies o
psychiatric disorders willbe normally distributed,which means
that there isa positive end with justas many people as the
negative end.This implies that at the level of DNA variation there
are no common disorders,only normally distributed quantitative
traits.95 It also raises the question of who these people are at the
positive end of the polygenic distribution of ‘risk’ for psychologica
and other traits.Are they merely individualsat low risk for
problems ordo they have specialpowers? Thinking positively
begins by thinking quantitatively—about ‘dimensions’rather than
‘disorders’and about genetic ‘variability’rather than genetic ‘risk’.
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
103
© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015),98 – 108

Intelligence brings (some) genetics to ‘social’epidemiology
It has long been known that intelligence,education and class are
correlated.The causesof these associationsand theirrelative
contribution to socialmobility is much disputed.96 Education and
socialclass are also well-established associates of health inequal-
ities, including all-cause mortality.30However, intelligence is a new
playerin health;its associationswith many health and illness
outcomes and all-cause and severalspecific causes ofmortality
have been discovered in the last decade or so.97
We shallexplain in this section that,akin to,but broader than
cognitive and learning abilities,intelligence shares genetic causes
with education and socialclass,which are touchstone ‘environ-
mental’variablesof diversesocial scientists.Major human
phenomena studied by these socialscientists are socialmobility
and health inequalities,which are unarguably important.They are
studied by sociologists,epidemiologists and economists.Finding
out why some people more than others make positive progress in
their social position through the life course,and why some people
are more prone to illnesses and early death have drafted in the
two favourite ‘environmental’social science variables of education
and social class.Education and parentalsocial class are predictors
of people’ssocial position in adulthood.98,99 Both, and the
person’s own adult socialclass,are associated with health,illness
and mortality:less educated people and those in less professional
jobs tend to die earlier.100–103
However,there is a third variable in
social mobility research,and a third variable in health inequalities
research:intelligence.104 Both education and socialclassare
substantially correlated with intelligence.29,61,105
Education and socialclass (which is indexed by occupation,or
income,or by the relative deprivation-affluence of where a person
lives)are often assumed to be indicators ofa person’s environ-
mentalinfluences,106 but they are correlated with intelligence,
which has a high heritability.Indeed,epidemiologists even use
height—shorter stature is associated with earlier mortality—as an
indicator of childhood social-environmentalinfluences,though it
has high heritability.For example,a recentsocialepidemiology
article described height ‘as a marker of early life insults’.107Here,
we emphasisethat it is an empiricalquestion ratherthan
something thatcan be assumed a priorias to whetherthe
three key variablesin social mobility and health inequalities
research—education,social class and intelligence—correlate
because of shared genetic and/or environmentalcauses.
Twin and family studies have shown thateducationalattain-
ment and socialclass are somewhat heritable.For example,the
pedigree-based estimates ofheritability (here as percentages of
phenotypic variance explained) in the Generation Scotland family-
based study ofover 20 000 people were 54% (s.e. = 2%)for
generalintelligence,41% (2%)for education and 71% (1%)for
social deprivationusing the Scottish Index of Multiple
Deprivation.108 The genetic correlationwas 0.65 (s.e. = 0.02)
between intelligence and education,0.40 (0.02)between intelli-
gence and deprivation and 0.48 (0.02)between education and
deprivation. An earlier report on a smaller sample (N46000) of the
same study found genetic correlations between intelligence and
being physically active outside work (0.25),fruitand vegetable
intake (0.23),ever smoking (0.45),smoke exposure (0.53)and
income (0.45),with high bivariate heritabilities for allof these.109
Another study identified over 2500 pairs of school-age twins from
population samples totalling over300 000 in England and the
Netherlands and found moderate to large genetic correlations and
bivariate heritability between intelligence and nationalexamina-
tion results in language,mathematics and science.61 Analyses of
older Danish twins found evidencefor genetic correlation
between cognitive ability and education and health.110,111
GCTA studieshave recentlyexplored theheritabilityand
genetic correlations of intelligence,education and socialclass.A
combined analysis ofSwedish and Australian unrelated subjects
(N ~ 11 500) used GCTA to provide an estimate of 22% (s.e. =
for the heritabilityof yearsin education and 25% (8%)for
attending college.16 In the Twins Early DevelopmentStudy for
3000 unrelated children,GCTA-based estimatesof heritability
were 21% (12%)for parentalsocialclassand 28% (17%)for
children’sIQ at age 7 and 32% (14%)at age 12.The GCTA-
estimated genetic correlation between parental social class and
was 1.00 (s.e. = 0.47)at age 7 and 0.66 (0.31)at age 12.56 GCTA-
based estimates of heritability on over 6500 unrelated people w
genome-wide SNP data in the Generation Scotland study were
29% (5%)for generalintelligence,21% (5%)for education and
18% (5%) for socialdeprivation.112The genetic correlations were
0.95 (0.13)for intelligenceand education,0.26 (0.16)for
intelligence and deprivation,and 0.45 (0.18)for education and
deprivation.Therefore,some of the variancein the social
scientists’key environmentalvariablescan be found in DNA
variation,some ofwhich is shared with the DNA variation that
causessome of people’sdifferencesin intelligence.Another
‘environmental’social science variable,height,shows a similar set
of findings in the Generation Scotland study sample.108The GCTA-
estimated heritability ofheightwas 58% (5%),its phenotypic
correlation with intelligence was 0.16,the GCTA-based genetic
correlation was 0.28 (0.09),and the bivariate heritability was 71%.
Bivariate GCTA-derived genetic correlations between intelligen
and health variables willrequire large numbers which are rare,as
yet.An analysis of data from the Swedish Twin Registry (N = 56
unrelated individuals) found GCTA-derived genetic correlations
0.13 (s.e. = 0.23)and 0.33 (s.e. = 0.33)between self-rated health
and,respectively,years in education and attending college16
).
The genetics of intelligence has a special place,therefore,in the
heretofore-named‘social’epidemiology.Indeed, these new
findings from twin/family-based and GCTA-based studies give a
corrective to the suggestion that ‘cognitive epidemiology’be re-
named ‘socialepidemiology’.Singh-Manoux’s113 suggestion was
partly made because epidemiologists preferred to use cognitiv
epidemiologyfor those studiesin which cognition wasthe
outcome,and so there was an objection to Deary and Batty’s
(2007)104definition,that is,‘the use of cognitive ability test scores
as risk factors for human health and disease outcomes,including
mortality’.Relevant to the genetic associations discussed in this
section was Singh-Manoux’s further discussion,
‘Given the association between intelligence and education,
extensively discussed by Deary and Johnson,106this definition
of cognitive epidemiology puts itsquarely in the domain of
socialepidemiology,a discipline concerned with the social
distribution of determinants of health.Location in this broader
church,ratherthan the micro-discipline ofcognitive epide-
miology,will avoid a narrow focus on intelligence that ignores
its associationswith markersof social position such as
education,income and occupation.’
One mightsay in reply thatthis conceptualisation ignores
possible genetic contributions to social/cognitive epidemiology
To sum up:there are genetic causes of some of the educational
and socialclass differences in the populations studied,and these
overlap with the geneticcausesof intelligencedifferences.
Intelligencegeneticsis specialhere, becauseit offers the
possibility offinding some ofthe connectionsbetween social
and medicaloutcomes,perhapsvia geneticcontributionsto
system integrity,allostaticload and the adoption of health-
promoting/reducing behaviours.114
FIVE SPECIAL FINDINGS AND POLYGENIC SCORES
These five specialfindingsabout the geneticsof intelligence
differences have emerged from traditionalquantitative genetic
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
104
Molecular Psychiatry (2015),98 – 108 © 2015 Macmillan Publishers Limited

research,primarily twin studies,and they are beginning to be
replicated using GCTA.However,nothing would advance the field
more than moving beyond GCTA to G,C, T, and A—thatis,
identifying specificDNA variantsthat contribute to the high
heritability of intelligence.As is the case for all complex traits and
common disorders in the life sciences,we now know that this will
be a difficult task.As discussed earlier,GWA studies have shown
that there are no large effect sizes in the population, which implies
that the heritability ofintelligence iscaused by thousandsof
DNA variants,many of these effects are likely to be infinitesimal or
even idiosyncratic.Nonetheless,GCTA has shown thatadditive
effects of common SNPs can theoretically account for at least half
of the heritability of intelligence,which means that a brute force
approach using everlargersamples willidentify some ofthese
genes.In addition,whole-genome sequencing willidentify DNA
variantsof any kind anywherein the genome,not just
common SNPs.
Associationsof smalleffectsize between DNA variantsand
intelligence can be summed acrossmultiple locito create a
polygenic score,which is analogousto aggregating itemsto
create a scale.Polygenic scores can aggregate a few candidate
SNPs orthousands ofSNPs across the genome,called genome-
wide polygenic scores (GPS),as described in Box 2.
Anticipatingthat GPS will be availablefor researchon
intelligence,we close by revisiting the five specialfindings about
genetics and intelligence,drawing hypotheses thatcan tested
using a GPS for intelligence,an exercise that we hope will help to
make the five specialfindings more concrete.
Heritability of intelligence increases dramatically from infancy
through adulthood despite genetic stability
GPS hypothesesfollow directly from the finding that the
heritability ofintelligence increases throughoutthe life course
despitestrong genetic stabilityfrom age to age: Variance
explained bya GPS should increasewith age, and a GPS
discovered atone age,adulthood forexample,is expected to
predict intelligence at other ages such as childhood.
Intelligence indexes generalgenetic effects across diverse
cognitive and learning abilities
A GPS hypothesis follows directly from finding strong genome-
wide pleiotropy across diverse cognitive abilities:A GPS thatis
discovered for any cognitive or learning ability should also predic
any other ability.Also,a GPS for intelligence should predict better
than a GPS forany othertrait.It has been suggested thata
pleiotropic GPS thatexplictly targets the substantialcovariance
among diverse cognitive and learning abilitieswill be even
betterthan a GPS based on a single composite measure of
intelligence.115
Assortative mating is greater for intelligence than for any other
trait
GPS supportfor the previoustwo hypothesesseemslikely
because preliminary GCTA results discussed above already provid
some supportfor these hypotheses.In the case ofassortative
mating,GPS could provide a noveltestof the extentto which
Box 2 Polygenic Scores
There are at least a dozen labels to denote such polygenic scores,most of which involve the word ‘risk’, such as polygenic risk scores, b
we prefer the term polygenic score because it makes more sense for quantitative traits like intelligence with positive as we
poles.115A polygenic score is created by adding genotypic values across loci. For example, for one locus (A) with two alleles
with the A2 allele associated with higherintelligence scores,additive genotypic values can be assigned forlocus A so thatA1A1
individuals = 0,A1A2 individuals = 1 and A2A2 individuals = 2.For each individual,these 0,1 and 2 additive genotypic values for locus A
can be added to those for locus B to create a polygenic score that varies from 0 to 4,and so on for dozens,hundreds or thousands of
loci.A refinement is to weight each lociby the strength of its association with intelligence.For example,100 associations that each
account for 0.1% of the variance of intelligence on average could together account for 10% of the variance.Their effects should add up
because DNA variants are uncorrelated unless they are very close together on a chromosome.Any loci could be aggregated to create a
polygenic score,such as candidate genes thought to be associated with a trait.However,the most productive use of polygenic scores is
to aggregate genotypic scores for DNA variants (usually SNPs) known to be associated with a trait.For example,much research on body
weight has used a polygenic score based on 32 SNPs that have shown replicated associations with body mass index,even though this
polygenic score only accounts for 2% of the heritability of body mass index.116
A more recent variant of polygenic scores goes beyond aggregating a few dozen individualSNPs associated with a trait to include
thousands of SNPs from GWA studies in a genome-wide polygenic score (GPS) that includes thousands of SNPs or even all S
array weighted by the strength of their association.115,117
The idea is that a GPS will be enriched for positive associations even thoug
the GPS willcertainly include false-positive associations.Although GPS can theoretically accountfor all the heritability shown in
GCTA,118 GWA studies ofintelligence and other traits have resulted in GPS that fallfar short ofGCTA estimates ofheritability.For
intelligence in childhood,a GPS derived from a discovery sample of more than 12 000 children accounted for about 2% of the v
in independent samples of more than 5000.14 For intelligence in adulthood,about 1% of the variance was explained by a GPS derived
from an adult sample of 3200 individuals and tested in an independent sample of 670 individuals,80 even though GCTA estimates of
heritability are about 30%.Using years of education rather than intelligence per se,the meta-analysis mentioned earlier with 125 000
individuals in a discovery sample yielded a GPS thataccounted for2 and 3% in two independentsamples.16 More variance in
intelligence is likely to be explained with GPS derived from larger samples,whole-genome sequencing and more novel strategies such
as using networks of functionally linked genes.119
Having a GPS forintelligence thatreliably accounts foreven as little as a few percentof the variance willenhance research on
intelligence.It will enable DNA analyses at the levelof individuals rather than families which can address the major questions of
quantitative genetic research such as developmental,multivariate and GE interplay issues mentioned earlier.Polygenic scores can be
used in the same way that candidate genes have been used.A neuroscientist might not find a polygenic score useful for investigating
molecular pathways between genes and behaviour through the brain,except perhaps to emphasise the need for a network approach
governed by pleiotropy (each gene affects many traits) and polygenicity (each trait is affected by many genes).A GPS for intelligence
would be like the other GPS (globalpositioning system)making it possible to triangulate on the genetics ofintelligence from all
domainsof the life sciences,for example,integrating research on the geneticsof intelligence from the genome,epigenome,
transcriptome,proteome and metabalome to the brain and behaviour without the need to assess intelligence.115
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
105
© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015),98 – 108

assortative mating forintelligence ismediated genetically by
correlating GPS between spouses.Another question that emerges
from previous genetic research is whether GPS assortative mating
is greater for verbalthan for nonverbalability.
Thinking positively:the genetics of high intelligence
Finding that the same genes affect high intelligence to the same
extentas the rest of the normaldistribution leadsto the
hypothesis thata GPS forintelligence from unselected samples
can also be used to predict high intelligence.
Intelligence brings (some) genetics to ‘social’epidemiology
Finding that,in twin and GCTA studies,the same genes influence
intelligence and socialepidemiologists’‘environmental’variables
of education,socialclass,and heightcan enlighten research in
health and socialinequalities.It leads to the hypothesis that GPS
scores for intelligence might contribute to health outcomes and
mortality,and that these might accountfor some of the
associations between education and class and mortality.
CONFLICT OF INTEREST
The authors declare no conflict of interest.
ACKNOWLEDGMENTS
RP is supported by a programme grant[G0901245]and a research professorship
[G19/2]from the MedicalResearch Counciland also a European Research Council
Advanced InvestigatorAward [295366].Ian Deary's work was undertaken in The
University ofEdinburgh Centre forCognitive Ageing and Cognitive Epidemiology,
partof the cross councilLifelong Health and Wellbeing Initiative [MR/K026992/1].
Funding from the Biotechnology and BiologicalSciences Research Council(BBSRC)
and MedicalResearch Council(MRC) is gratefully acknowledged.
REFERENCES
1 Burks B.The relative influence of nature and nurture upon mental development:
A comparative study on foster parent-foster child resemblance.In:Whipple GM
(ed).Whipple GMYearbook ofthe NationalSociety forthe Study ofEducation,
Part 1.Public SchoolPublishing Co:Bloomington,IL,1928,27,pp 219–316.
2 Freeman FN,HolzingerKJ, MitchellB. The influence ofenvironmenton the
intelligence, school achievement, and conduct of foster children. In: Whipple GM
(ed).Yearbook ofthe NationalSociety for the Study ofEducation,Part 1.Public
SchoolPublishing Co:Bloomington,IL,1928,27,pp 103–217.
3 Merriman C.The intellectualresemblance oftwins.PsychologicalMonographs
1924;33:1–57.
4 Theis SVS.How foster children turn out.State Charities Aid Association:New York,
1924.
5 Deary IJ,Johnson W,Houlihan LM.Genetic foundations of human intelligence.
Hum Genet 2009;126:215–232.
6 Plomin R,DeFries JC,Knopik VS,Neiderhiser JM.Behavioralgenetics,6th edn.
Worth Publishers:New York,2013.
7 Turkheimer E.Three laws ofbehavior genetics and what they mean.CurrDir
PsycholSci2000;9: 160–164.
8 McGuffin P.The past,presentand future ofpsychiatric genetics.In Bloch S,
Green SA, Holmes J (eds) Psychiatry: Past, present, and prospect. Oxford University
Press:Oxford,UK,2014 pp 22–44.
9 Pinker S.The blank slate: The modern denial of human nature.Penguin: New York,
2002.
10 Haworth CMA, Plomin R. Genetics and education: Towards a genetically sensitive
classroom.In Harris KR,Graham S,Urdan T (eds)The American Psychological
Association Handbook ofEducationalPsychology.APA:Washington,DC, 2011,
pp 529–559.
11 Check Hayden E.Ethics:Taboo genetics.Nature 2013;502:26–28.
12 Manolio TA, Collins FS,Cox NJ, Goldstein DB,Hindorff LA,Hunter DJet al. Finding
the missing heritability of complex diseases.Nature 2009;461:747–753.
13 Fisher RA.The correlation between relatives on the supposition ofMendelian
inheritance.Transactions of the RoyalSociety of Edinburgh 1918;52:399–433.
14 Benyamin B,Pourcain B,Davis OS,Davies G,HansellNK,Brion MJ et al.Child-
hood intelligence is heritable,highly polygenic and associated with FNBP1L.Mol
Psychiatry 2014;19:253–258.
15 Desrivieres S,Lourdusamy A,Tao C,Toro R,Jia T,Loth Eet al.Single nucleotide
polymorphism in the neuroplastin locus associates with corticalthickness and
intellectualability in adolescents.MolPsychiatry 11 February 2014.e-pub ahead
of print.doi:10.1038/mp.2013.197.
16 Rietveld CA,Medland SE,Derringer J,Yang J,Esko T,Martin NW et al.GWAS of
126 559 individualsidentifiesgeneticvariantsassociated with educational
attainment.Science 2013;340:1467–1471.
17 Spearman C.‘Generalintelligence’,objectively determined and measured.Am J
Psychol1904;15:201–292.
18 CarrollJB. Human cognitive abilities.Cambridge University Press:New York,
1993.
19 Jensen AR.The g factor: The science of mental ability. Praeger: Westport, CT, 19
20 Johnson W,Bouchard TJ,KruegerRF, McGue M,Gottesman II.Just one g:
Consistent results from three test batteries.Intelligence 2004;32:95–107.
21 Gottfredson LS.Mainstream science on intelligence:An editorialwith 52 sig-
natories,history and bibliography.Intelligence 1997;24:13–23.
22 Deary IJ.Intelligence.Curr Biol2013;23:R673–R676.
23 CarrollJB. Psychometrics,intelligence,and public policy.Intelligence 1997;24:
25–52.
24 Deary I.Looking down on human intelligence:From psychometrics to the brain.
Oxford University Press:Oxford,UK,2000.
25 Deary IJ,Penke L,Johnson W.The neuroscience ofhuman intelligence differ-
ences.Nat Rev Neurosci2010;11:201–211.
26 Duncan J.How intelligence happens.Yale University Press:New Haven,CT,2010.
27 Salthouse T.Major issues in cognitive aging.Oxford University Press:Oxford,UK,
2010.
28 Deary IJ,Whalley LJ,Lemmon H,Crawford JR,Starr JM.The stability of individual
differences in mentalability from childhood to old age:Follow-up of the 1932
Scottish MentalSurvey.Intelligence 2000;28:49–55.
29 Strenze T.Intelligence and socioeconomic success:A meta-analytic review of
longitudinalresearch.Intelligence 2007;35:401–426.
30 Deary IJ,Weiss A,Batty GD.Intelligence and personality as predictors of illness
and death:How researchersin differentialpsychology and chronic disease
epidemiology are collaborating to understand and address health inequalities
PsycholSciPublInterest 2010;11:53–79.
31 Plomin R.Genetics,genes,genomics and g.Mol Psychiatry 2003;8: 1–5.
32 Yang J,Benyamin B,McEvoy BP,Gordon S,Henders AK,Nyholt DR et al.Com-
mon SNPs explain a large proportion of the heritability for human height.Nat
Genet 2010;42:565–569.
33 Yang JA,Lee SH,Goddard ME,VisscherPM. GCTA:A tool for genome-wide
complex trait analysis.Am J Hum Genet 2011;88:76–82.
34 Yang J,Manolio TA,Pasquale LR,Boerwinkle E,Caporaso N,Cunningham JM
et al. Genomepartitioning ofgeneticvariation forcomplextraits using
common SNPs.Nat Genet 2011;43:519–525.
35 Yang J,Lee SH,Goddard ME,Visscher PM.Genome-wide complex trait analysis
(GCTA):Methods,data analyses,and interpretations.MethodsMol Biol 2013;
1019:215–236.
36 So HC,Li MX,Sham PC.Uncovering the totalheritability explained by alltrue
susceptibility variants in a genome-wide association study. Genet Epidemiol 2
35:447–456.
37 Gray A, Stewart I, Tenesa A. Advanced complex trait analysis. Bioinformatics
28:3134–3136.
38 Speed D,HemaniG, Johnson MR,Balding DJ.Improved heritability estimation
from genome-wide SNPs.Am J Hum Genet 2012;91:1011–1021.
39 Zaitlen N,Kraft P.Heritability in the genome-wide association era.Hum Genet
2012;131:1655–1664.
40 Zhou X,Carbonetto P,Stephens M.Polygenic modeling with Bayesian sparse
linear mixed models.PLoS Genet 2013;9: e1003264.
41 Plomin R,Haworth CMA,Meaburn EL,Price T,Wellcome TrustCase Control
Consortium 2,Davis OS.Common DNA markers can accountfor more than
half of the geneticinfluence on cognitive abilities.PsycholSci 2013;24:
562–568.
42 Vinkhuyzen AA,Wray NR,Yang J, Goddard ME,VisscherPM. Estimation and
partition ofheritability in human populationsusing whole-genome analysis
methods.Annu Rev Genet 2013;47:75–95.
43 Lee SH,Yang J,Goddard ME,VisscherPM, Wray NR.Estimation ofpleiotropy
between complex diseasesusing single-nucleotidepolymorphism-derived
genomic relationships and restricted maximum likelihood.Bioinformatics 2012;
28:2540–2542.
44 VisscherPM, HemaniG, Vinkhuyzen AAE,Chen G-B,Lee SH,Wray NR etal.
Statistical power to detect genetic (co)variance of complex traits using SNP da
in unrelated samples.PLoS Genet 2014;10:e1004269.
45 Zhou X,Stephens M.Efficient algorithms for multivariate linear mixed models in
genome-wide association studies.Nat Methods 2014;11:407–409.
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
106
Molecular Psychiatry (2015),98 – 108 © 2015 Macmillan Publishers Limited

46 Deary IJ,Yang J,Davies G,Harris SE,Tenesa A,Liewald D etal. Genetic con-
tributions to stability and change in intelligence from childhood to old age.
Nature 2012;482:212–215.
47 Panizzon MS,Vuoksimaa E,Spoon KM,Jacobson KC,Lyons MJ,Franz CE et al.
Genetic and environmentalinfluences on generalcognitive ability:Is g a valid
latent construct? Intelligence 2014;43:65–76.
48 Lee T,Henry JD,Trollor JN,Sachdev PS.Genetic influences on cognitive func-
tions in the elderly:A selective review of twin studies.Brain Res Brain Res Rev
2010;64:1–13.
49 Haworth CMA,Wright MJ,Luciano M,Martin NG,de Geus EJC,van Beijsterveldt
CEM etal. The heritability ofgeneralcognitive ability increases linearly from
childhood to young adulthood.Mol Psychiatry 2010;15:1112–1120.
50 Mackintosh N.IQ and human intelligence,2nd edn.Oxford University Press:
Oxford,UK,2011.
51 Plomin R, FulkerDW, Corley R,DeFriesJC. Nature,nurture and cognitive
development from 1 to 16 years:A parent-offspring adoption study.PsycholSci
1997;8: 442–447.
52 Skodak M,Skeels HM.A finalfollow-up on one hundred adopted children.J
Genet Psychol1949;75:84–125.
53 Davis OSP,Haworth CMA,Plomin R. Dramatic increase in heritability of cognitive
development from early to middle childhood:An 8-year longitudinalstudy of
8700 pairs of twins.PsycholSci2009;20:1301–1308.
54 Wilson RS.The Louisville Twin Study:Developmentalsynchronies in behavior.
Child Dev 1983;54:298–316.
55 TrzaskowskiM, Yang J,Visscher PM,Plomin R.DNA evidence for strong genetic
stabilityand increasing heritabilityof intelligence from age 7 to 12.Mol
Psychiatry 2014;19:380–384.
56 Trzaskowski M, Harlaar N, Arden R, Krapohl E, Rimfeld K, McMillan A et al. Genetic
influence on family socioeconomic status and children's intelligence.Intelligence
2014;42:83–88.
57 Plomin R.Development,genetics,and psychology.Erlbaum:Hillsdale,NJ, 1986.
58 Briley DA,Tucker-Drob EM.Explaining the increasing heritability ofcognitive
ability across development:A meta-analysis of longitudinaltwin and adoption
studies.PsycholSci2013;24:1704–1713.
59 Plomin R. Genetics and experience: The interplay between nature and nurture . Sage
Publications Inc.:Thousand Oaks,CA,1994.
60 Plomin R,Kovas Y.Generalist genes and learning disabilities.PsycholBull2005;
131:592–617.
61 Calvin CM,Deary IJ,Webbink D,Smith P,Fernandes C,Lee SH et al.Multivariate
genetic analyses of cognition and academic achievement from two population
samples of 174 000 and 166 000 school children. Behav Genet 2012;42: 699–710.
62 Davis OSP,Haworth CMA,Plomin R.Learning abilities and disabilities:Generalist
genes in early adolescence.Cogn Neuropsychiatry 2009;14:312–331.
63 TrzaskowskiM, Davis OS,DeFries JC,Yang J,Visscher PM,Plomin R.DNA evi-
dence forstrong genome-wide pleiotropy ofcognitive and learning abilities.
Behav Genet 2013;43:267–273.
64 TrzaskowskiM, Shakeshaft NG,Plomin R.Intelligence indexes generalist genes
for cognitive abilities.Intelligence 2013;41:560–565.
65 Baddeley A.Working memory:Theories,models,and controversies.Annu Rev
Psychol2012;63:1–29.
66 Giedd JN, Rapoport JL. Structural MRI of pediatric brain development: What have
we learned and where are we going? Neuron 2010;67:728–734.
67 Kovas Y,Plomin R.Generalistgenes:Implications forthe cognitive sciences.
Trends Cogn Sci2006;10:198–203.
68 Keller MC,Garver-Apgar CE,Wright MJ,Martin NG,Corley RP,Stallings MC et al.
The genetic correlation between heightand IQ:Shared genes orassortative
mating? PLoS Genet 2013;9: e1003451.
69 Whitaker KL, Jarvis MJ, Beeken RJ, Boniface D, Wardle J. Comparing maternal and
paternalintergenerationaltransmission ofobesity risk in a large population-
based sample.Am J Clin Nutr 2010;91:1560–1567.
70 Vandenberg SG. Assortative mating, or who marries whom? Behav Genet 1972; 2:
127–157.
71 Bouchard TJ Jr., McGue M. Familial studies of intelligence: A review. Science 1981;
212:1055–1059.
72 Mascie-Taylor CGN.Spouse similarity for IQ and personality and convergence.
Behav Genet 1989;19:223–227.
73 Vinkhuyzen AAE,van der Sluis S,Maes HHM,Posthuma D.Reconsidering the
heritability of intelligence in adulthood:Taking assortative mating and cultural
transmission into account.Behav Genet 2012;42:187–198.
74 BashiJ. Effectsof inbreeding on cognitive performance.Nature 1977;266:
440–442.
75 Chipuer HM,Rovine MJ,Plomin R.LISREL modeling:Genetic and environmental
influences on IQ revisited.Intelligence 1990;14:11–29.
76 Falconer DS,MacKay TFC.Introduction to quantitative genetics,Vol4. Longman:
Harlow,UK,1996.
77 Rietveld CA,Cesarini D, Benjamin DJ,Koellinger PD,De Neve JE,Tiemeier H et al.
Molecular genetics and subjective well-being.Proc NatlAcad SciUSA 2013;110:
9692–9697.
78 TrzaskowskiM, Dale PS,Plomin R.No genetic influence for childhood behavior
problemsfrom DNA analysis.J Am Acad Child Adolesc Psychiatry 2013;52:
1048–1056.
79 Vinkhuyzen AAE,Pedersen NL,Yang J,Lee SH,Magnusson PKE,Iacono WG et al.
Common SNPs explain some of the variation in the personality dimensions of
neuroticism and extraversion.TranslPsychiatry 2012;2: e102.
80 Davies G,Tenesa A,Payton A,Yang J,Harris SE,Liewald D et al.Genome-wide
association studies establish thathuman intelligence is highly heritable and
polygenic.Mol Psychiatry 2011;16:996–1005.
81 Domingue BW,FletcherJ, Conley D,Boardman JD.Genetic and educational
assortative mating among USadults.Proc Natl Acad Sci USA 2014;111:
7996–8000.
82 Kell HJ,Lubinski D,Benbow CP.Who rises to the top? Early indicators.Psychol Sci
2013;24:648–659.
83 Haworth CMA,Wright MJ,Martin NW,Martin NG,Boomsma DI,Bartels Met al.A
twin study ofthe genetics ofhigh cognitive ability selected from 11 000 twin
pairs in six studies from four countries.Behav Genet 2009;39:359–370.
84 Plomin R,Simpson MA,Cederlöf M,Lichtenstein P.The genetics of high cognitive
abilities.ISIR:Melbourne,Australia,2013 Available at http://www.isironline.org/
wp-content/uploads/2014/05/program2013.pdf;Accessed 19 May 2014.
85 DeFries JC,FulkerDW. Multiple regression analysis oftwin data:Etiology of
deviant scores versus individualdifferences.Acta Genet Med Gemellol1988;37:
205–216.
86 PetrillSA,Kovas Y,Hart SA,Thompson LA,Plomin R.The genetic and environ-
mentaletiology ofhigh math performance in 10-year-old twins.Behav Genet
2009;39:371–379.
87 Lykken DT. The genetics of genius. In Steptoe A (ed) Genius and the mind: Studi
of creativity and temperamentin the historicalrecord.Oxford University Press:
New York,1998 pp 15–37.
88 Lykken DT.The mechanism of emergenesis.Genes Brain Behav 2006;5:306–310.
89 Galton F.Heredity genius:An enquiry into itslaws and consequences.World:
Cleveland,OH,1869.
90 Plomin R,CederlöfM, Lichtenstein P.Low IQ and mild mentalretardation are
heritable but severe mental retardation is not: A genetic analysis of 740 000 sibl
and 18 000 twins.ISIR:Melbourne,Australia,2013 Available athttp://www.isir
online.org/wp-content/uploads/2014/05/program2013.pdf;Accessed19 May
2014.
91 Penrose LS.A clinicaland genetic study of1280 casesof mentaldefect.HM
Stationery Office:London,UK,1938.
92 Ellison JW,Rosenfeld JA,Shaffer LG.Genetic basis of intellectualdisability.Annu
Rev Med 2013;64:441–450.
93 Marioni RE, Penke L, Davies G, Huffman JE, Hayward C, Deary IJ. The total burden
of rare,non-synonymous exome genetic variants is not associated with child-
hood or late-life cognitive ability.Proc R Soc Lond B Biol Sci 2014;281:20140117.
94 Kirkpatrick RM,McGue M,Iacono WG,Miller MB, Basu S,Pankratz N.Low-
frequency copy-number variants and generalcognitive ability:No evidence of
association.Intelligence 2014;42:98–106.
95 Plomin R,Haworth CMA,Davis OSP.Common disorders are quantitative traits.
Nat Rev Genet 2009;10:872–878.
96 Deary IJ,Taylor MD,Hart CL,Wilson V,Smith GD,Blane D et al.Intergenerational
socialmobility and mid-life status attainment:Influences ofchildhood intelli-
gence,childhood socialfactors,and education.Intelligence 2005;33:455–472.
97 Calvin CM, Deary IJ, Fenton C, Roberts BA, Der G, Leckenby N et al. Intelligence i
youth and all-cause-mortality:Systematic review with meta-analysis.Int J Epi-
demiol2011;40:626–644.
98 Erikson R,Goldthorpe JH.Has socialmobility in Britain decreased? Reconciling
divergent findings on income and class mobility.Br J Sociol2010;61:211–230.
99 Saunders P.Socialmobility myths.Civitas:London,UK,2010.
100 Chandola T,Clarke P,Morris JN,Blane D.Pathways between education and
health:A causalmodelling approach.J R Stat Soc SerA StatSoc 2006;169:
337–359.
101 Davey Smith G, Hart C, Hole D, MacKinnon P, Gillis C, Watt G et al. Education an
occupational social class: Which is the more important indicator of mortality risk
J EpidemiolCommunity Health 1998;52:153–160.
102 GalobardesB, Lynch JW, Smith GD.Is the association between childhood
socioeconomic circumstances and cause-specific mortality established? Update
of a systematic review.J EpidemiolCommunity Health 2008;62:387–390.
103 Lawlor DA,Sterne JAC,Tynelius P,Davey Smith G,Rasmussen F.Association of
childhood socioeconomic position with cause-specific mortality in a prospective
record linkage studyof 1 839 384 individuals.Am J Epidemiol2006;164:
907–915.
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
107
© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015),98 – 108

104 Deary IJ,Batty GD.Cognitive epidemiology.J EpidemiolCommunity Health 2007;
61:378–384.
105 Deary IJ,Strand S,Smith P,Fernandes C.Intelligence and educationalachieve-
ment.Intelligence 2007;35:13–21.
106 Deary IJ,Johnson W.Intelligence and education:Causalperceptions drive ana-
lytic processes and therefore conclusions.Int J Epidemiol2010;39:1362–1369.
107 Gray L,Davey Smith G,McConnachie A,WattGC, HartCL, Upton MN etal.
Parentalheight in relation to offspring coronary heartdisease:Examining
transgenerationalinfluences on health using the Westof Scotland Midspan
Family Study.Int J Epidemiol2012;41:1776–1785.
108 MarioniRE,Batty GD,Hayward C,Kerr SM,CampbellA, Hocking LJ et al.Com-
mon genetic variants explain the majority of the correlation between height and
intelligence:The Generation Scotland Study.Intelligence 2014;44:91–96.
109 Luciano M,Batty GD,McGilchrist M,Linksted P,Fitzpatrick B,Jackson C et al.
Shared genetic aetiology between cognitive ability and cardiovascular disease
risk factors:Generation Scotland's Scottish family health study.Intelligence 2010;
38:304–313.
110 Johnson W,Deary IJ,McGue M,Christensen K.Genetic and environmentallinks
between cognitive and physical functions in old age.J Gerontol B Psychol Sci Soc
Sci2009;64:65–72.
111 Johnson W,Deary IJ,McGue M,Christensen K.Genetic and environmental
transactions linking cognitive ability,physicalfitness,and education in late life.
PsycholAging 2009;24:48–62.
112 Marioni RE, Davies G, Hayward C, Liewald D, Kerr SM, Campbell A et al. Molecular
genetic contributions to socioeconomic status and intelligence.Behav Genet
2014;44:26–32.
113 Singh-Manoux A. Commentary: Is it time to redefine cognitive epidemiology? Int
J Epidemiol2010;39:1369–1371.
114 Deary IJ.Looking for 'system integrity'in cognitive epidemiology.Gerontology
2012;58:545–553.
115 Plomin R,Simpson M.The future ofgenomics for developmentalists.Dev Psy-
chopathol2013;25:1263–1278.
116 Speliotes EK,Willer CJ,Berndt SI,Monda KL,Thorleifsson G,Jackson AU et al.
Association analyses of249 796 individuals reveal18 new lociassociated with
body mass index.Nat Genet 2010;42:937–948.
117 Dudbridge F.Power and predictive accuracy of polygenic risk scores.PLoS Genet
2013;9:e1003348.
118 Wray NR,Lee SH,Mehta D,Vinkhuyzen AAE,Dudbridge F,Middeldorp CM.
Polygenic methods and theirapplication to psychiatric disorders and related
traits.J Child PsycholPsychiatry advance online publication,1 August2014;
doi:10.1111/jcpp.12295 (e-pub ahead of print).
119 HillWD,Davies G,van de Lagemaat LN,Christoforou A,MarioniRE,Fernandes
CPD etal. Human cognitive ability is influenced by genetic variation in com-
ponents ofpostsynaptic signalling complexes assembled by NMDA receptors
and MAGUK proteins.TranslPsychiatry 2014;4: e341.
120 Lykken DT.Research with twins:The concept of emergenesis.Psychophysiology
1982;19:361–373.
This work is licensed undera Creative Commons Attribution 3.0
Unported License.The images or other third party materialin this
article are included in the article’sCreative Commonslicense,unlessindicated
otherwise in the credit line; if the material is not included under the Creative Com
license, users will need to obtain permission from the license holder to reproduce
material.To view a copy of this license,visit http://creativecommons.org/licenses/by/
3.0/
Genetics and intelligence differences:five specialfindings
R Plomin and IJ Deary
108
Molecular Psychiatry (2015),98 – 108 © 2015 Macmillan Publishers Limited